The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Anna Kovner, James Vickery, and Lily Zhou
Do Big Banks Have Lower
Operating Costs?
l
Concern that some banks remain “too big to
fail” has prompted many calls for limits on
bank holding company (BHC) size.
l
But such limits could have adverse effects if
they were to undercut the economies of scale
associated with large banking firms.
l
Reasoning that scale economies may be
achieved in part through lower operating
costs, the authors of this study examine
the relationship between BHC size and
noninterest expense.
l
Their analysis, which considers these costs at a
finer level of detail than in past studies, reveals
a robust negative relationship between
BHC size and scaled noninterest expenses,
including employee compensation, information
technology, and corporate overhead costs.
l
The results suggest that limits on BHC size
may, in fact, increase the cost of providing
banking services—a drawback that must be
weighed against the potential financial stability
benefits of limiting firm size.
Anna Kovner is a research officer, James Vickery a senior economist, and
Lily Zhou a former senior research analyst at the Federal Reserve Bank of
New York.
Correspondence: james.vickery@ny.frb.org
1. Introduction
T
he largest U.S. banking firms have grown significantly over
time, their expansion driven by a combination of merger
activity and organic growth. In 1991, the four largest U.S. bank
holding companies (BHCs) held combined assets equivalent
to 9 percent of gross domestic product (GDP). Today, the four
largest firms’ assets represent 50 percent of GDP, and six BHCs
control assets exceeding 4 percent of GDP. Despite recent
financial reforms, there is still widespread concern that large
banking firms remain “too big to fail”—that is, policymakers
would be reluctant to permit the failure of one or more of the
largest firms because of fears about contagion or damage to the
broader economy (see, for example, Bernanke [2013]).
A growing number of market observers advocate shrinking the size of the largest banking firms in order to limit the
problem of too-big-to-fail. The most direct approach would be
to simply impose a firm cap on the size of assets or liabilities;
for example, Johnson and Kwak (2010) propose a size limit of
4 percent of nominal GDP. An alternative would be to impose
levies or progressively higher capital requirements on large
banking firms to encourage them to shed assets.
Would such policies impose any real costs on the economy?
A number of recent academic papers suggest that the answer
may be “yes” because of the presence of economies of scale
in banking. Scale economies imply that the cost of producing
an additional unit of output (for example, a loan) falls as the
The authors thank Peter Olson for outstanding research assistance and Gara
Afonso, Jan Groen, Joseph Hughes, Donald Morgan, an anonymous referee,
and workshop participants at the Federal Reserve Bank of New York for helpful comments and suggestions. The views expressed in this article are those of
the authors and do not necessarily reflect the position of the Federal Reserve
Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
1
quantity of production increases. A number of papers find
evidence of scale economies even among the largest banking
firms (Hughes and Mester 2013; Wheelock and Wilson 2012;
Feng and Serletis 2010). Taken at face value, this research
implies that the introduction of limits on bank size would
impose deadweight economic costs by increasing the cost of
providing banking services.
We contribute to this line of research by studying the
relationship between size and components of noninterest
expense (NIE), with the goal of shedding light on the sources
of scale economies in banking. NIE includes a wide variety of
operating costs incurred by banking firms: examples include
employee compensation and benefits, information technology,
legal fees, consulting, postage and stationery, directors' fees,
and expenses associated with buildings and other fixed assets.
Our hypothesis is that lower operating costs may be a source of
scale economies for large BHCs, because large firms can spread
overhead such as information technology, accounting, advertising, and management over a larger asset or revenue base.
Our analysis therefore tests for an inverse relationship between
BHC size and scaled measures of different components of NIE.
One novel contribution of this paper is to make use of
detailed noninterest expense information provided by U.S.
banking firms in the memoranda of their quarterly regulatory
FR Y-9C filings. The Y-9C reports contain detailed consolidated
financial statements and other data for U.S. BHCs (see Section 3
for details). Since 2001, about 35 percent of total noninterest expense is classified in the Y-9C as part of a broad “other
noninterest expense” category. For the period 2008 to 2012, we
disaggregate this line item into nine author-defined categories,
using memoranda information from Schedule HI of the Y-9C.
In part, this involved manually classifying about 5,500 individual “write-in” text fields reported by individual BHCs. To our
knowledge, ours is the first paper to make use of these data.
We start by estimating the relationship between bank holding company size (measured by the natural logarithm of total
assets) and total noninterest expense scaled by net operating
revenue, assets, or risk-weighted assets. We find a statistically
and economically significant negative relationship between
BHC size and these NIE ratios, robust to the expense measure
or set of controls used. Quantitatively, a 10 percent increase in
assets is associated with a 0.3 to 0.6 percent decline in noninterest expense scaled by income or assets, depending on the specification. In dollar terms, our estimates imply that for a BHC of
mean size, an additional $1 billion in assets reduces noninterest
expense by $1 million to $2 million per year, relative to a base
case in which operating cost ratios are unrelated to size.1
These results hold across the size distribution of banking
firms, and over different parts of our sample period. We find
no evidence that these lower operating costs flatten out above
some particular size threshold. The point estimate of the slope
of the relationship steepens, if anything, although the statistical uncertainty associated with the estimate becomes larger
owing to the small sample.
The relationship between size and the NIE ratio is negative
for each of the three main components of noninterest expense
reported in BHC regulatory filings: employee compensation,
premises and fixed asset expenses, and other noninterest
expense. Using our novel by-hand classification of other NIE
into nine subcomponents, however, we find significant variation
in the size-expense relationship among the subcomponents. The
inverse relationship between size and expense is particularly
pronounced for corporate overhead (for example, accounting,
printing, and postage); information technology (IT) and data
processing; legal fees; other financial services; and directors’
fees and other compensation. In contrast, large BHCs spend
proportionately more on consulting and advisory services than
do smaller firms, relative to revenue or assets. Large BHCs also
incur proportionately higher expenses relating to amortization
and impairment of goodwill and other intangible assets.
Overall, our results are consistent with the presence of
scale economies in banking, as found in recent academic
literature (for example, Wheelock and Wilson [2012]; Hughes
and Mester [2013]; Feng and Serletis [2010]) and industry
research (Clearing House Association 2011). In particular, our
findings suggest that these scale economies stem in part from
an operating cost advantage of large BHCs in areas such as
employee compensation, information technology, and corporate overhead expenses.
We emphasize that a number of caveats apply to our
results. First, our estimates represent reduced-form statistical
correlations; caution should be exercised in drawing a causal
interpretation from them. Although our regressions control
for a wide range of BHC characteristics, firm size may still be
correlated with omitted variables that are also associated with
lower expenses, such as the quality of management. This caveat
also seems to apply more generally to the existing literature on
scale economies in banking.
Second, our results may also reflect factors other than
scale economies. One possibility, closely related to scale economies but conceptually distinct, is that large firms operate
closer to their production frontier on average; that is, they
have greater X-efficiency (see Section 2 for a discussion).2
2
1
For details of this calculation, see Appendix B, available as a separate file at
http://www.newyorkfed.org/research/epr/2014/1403kovn_appendixB.pdf. The
appendix was omitted from the main document because of space constraints.
2
Do Big Banks Have Lower Operating Costs?
Our analysis does not attempt to separate the effects of X-efficiency from
those of scale economies. We note, however, that Hughes et al. (2001) and
Hughes and Mester (2013) find that estimated scale economies are larger for
more efficient banks than for less efficient ones, controlling for size.
Another possibility is that large banking firms have greater bargaining power vis-à-vis their suppliers and employees. If cost
differences are due only to bargaining power effects, then limiting the size of the largest BHCs would not necessarily generate
deadweight economic costs, although it might instead reallocate
rents to employees or suppliers. An additional possibility is that
our results are influenced by too-big-to-fail subsidies for large
BHCs. Our prior is that such subsidies would be more likely
to be manifested as a lower cost of funds for large firms, or a
more leveraged capital structure, than as lower operating costs.
However, it is still possible that a too-big-to-fail banking firm
could respond by reducing expenditures on functions such as
information technology or risk management; these would show
up as part of noninterest expense.
These caveats aside, our results and those of related
research suggest that imposing size limits on banking firms
is unlikely to be a free lunch. For example, taking our
estimates at face value, a back-of-the-envelope calculation
implies that limiting BHC size to no more than 4 percent
of GDP would increase total industry noninterest expense
by $2 billion to $4 billion per quarter.3 Limiting the size
of banking firms could still be an appropriate policy goal,
but only if the benefits of doing so exceeded the attendant
reductions in scale efficiencies.
A second contribution of this article is to present new evidence on other determinants of BHC operating costs. In particular, we find that proxies for organizational complexity (for
example, the number of distinct legal entities controlled by the
BHC), as well as measures of the diversity of business activities, are robustly correlated with higher expense ratios. This
result appears consistent with prior research on the diversification discount in banking (for example, Goetz, Laeven, and
Levine [2013]). A third contribution is to present new stylized
facts about the composition of noninterest expense, based
on our data collection efforts. For example, we document the
large share of NIE that is composed of corporate overhead,
investment technology and data processing, consulting and
advisory services, and legal expenses.
The remainder of the article proceeds as follows: Section 2
presents background and reviews the literature on economies
of scale in banking. Section 3 describes the data, discusses
our method for classifying other noninterest expense, and
presents descriptive statistics. Section 4 presents multivariate analysis of the relationship between size and noninterest
expense ratios. Section 5 studies components of noninterest
expense. Section 6 summarizes our findings.
3
Details of this calculation are presented in Appendix B, http://www
.newyorkfed.org/research/epr/2014/1403kovn_appendixB.pdf.
2. Background and Literature
Our analysis is closely related to academic literature on
scale economies and organizational efficiency in banking.
In a microeconomic production model, the cost function
traces out the relationship between output and the minimum
total cost required to produce that output, for a given set of
input prices. A firm exhibits economies of scale if minimum
cost increases less than proportionately with output—for
example, if the firm could double its output by less than doubling its costs, holding input prices fixed.
A large literature empirically estimates the cost function for
banks and/or BHCs, and tests for the presence of scale economies by measuring whether the elasticity of total costs with
respect to output is greater than, equal to, or less than unity
(indicating diseconomies of scale, constant returns to scale, or
economies of scale, respectively).
The earliest studies of scale economies in banking (for
example, Benston [1972]), estimated during an era when U.S.
banking organizations were on average much smaller than
today, found evidence of modest economies of scale. Subsequent research, using more flexible cost functions, found that
these scale economies were limited to small banks (for example,
Benston, Hanweck, and Humphrey [1982] and Peristiani
[1997]; see also Berger and Humphrey [1994] for a survey).
More recent research, however, has found evidence of scale
economies even among the class of large banks and bank
holding companies. Examples include Wheelock and Wilson
(2012), Hughes and Mester (2013), Feng and Serletis (2010),
and Hughes et al. (2001). This departure from earlier findings
reflects greater statistical power, attributable to the use of
larger datasets with many more observations for large banking
firms, as well as the evolution of empirical techniques. For
example, Wheelock and Wilson (2012) estimate a nonparametric cost function rather than the typical parametric
translog function estimated in earlier literature, while Hughes
and Mester (2013) and Hughes et al. (2001) endogenize bank
risk and capital structure decisions. The difference in time
periods may also play a role (for example, the greater use
of information technology may have changed the extent to
which scale economies are present).
The theoretical derivation of the cost function assumes
that the bank maximizes profits, or equivalently, minimizes
costs for any given level of output. A related body of literature
on bank efficiency, however, finds evidence of surprisingly
large cost differences between otherwise similar banks. These
differences are viewed as evidence of X-inefficiencies, that is,
firms operating inside their production possibilities frontier
because of agency conflicts, management problems, or other
inefficiencies (DeYoung 1998; Berger, Hunter, and Timme
1993; Berger and Humphrey 1991).
FRBNY Economic Policy Review / December 2014
3
Rather than analyzing total scale economies or X-efficiency, this paper instead presents disaggregated evidence
on the relationship between firm size and detailed components of noninterest expense. We have in mind the idea that
operational and technological efficiencies related to size are
likely to show up in the data in the form of lower operating
costs in areas such as information technology and corporate
overhead (for example, accounting and human resources)
because large BHCs are able to spread the fixed component of
these costs over a broader revenue or asset base. Our goal is to
shed additional light on the mechanisms driving differences
in efficiency between small and large firms. We note that our
empirical finding that large BHCs have lower average operating costs could be driven by the presence of scale economies
in the production of banking services, higher average X-efficiency for large firms, or both. For some categories of NIE, it
could also be possible that lower costs for larger banking firms
not only reflect technological efficiencies, but also greater bargaining power relative to suppliers, customers, or employees.
Our analysis is related to recent research by the Clearing House
(2011) that uses proprietary management information systems
data from a number of large banks to estimate product-specific
scale curves in seven areas: online bill payment, debit cards,
credit cards, wire transfers, automated clearing house, check
processing, and trade processing. The Clearing House finds
that in each of these areas, unit costs are decreasing in production volume, a conclusion that suggests the presence of fixed
costs or other technological benefits of size. The economies
of scale associated with these seven services are estimated to
total $10 billion to $25 billion per year.
Although our approach is similar in some respects to the
analysis by the Clearing House, we make use of data from
audited regulatory filings, rather than internal management
information system data, and study components that together
sum up to total noninterest expense, rather than just a subset of NIE (the seven items studied by the Clearing House
together cover only 7 to 10 percent of NIE). We also study
the entire cross-section of BHCs, while the Clearing House
sample consists of only six firms.
Our approach is related to the literature on banking
mergers that uses accounting variables to estimate the effects
of mergers on operating performance. Kwan and Wilcox
(2002) find evidence that bank mergers reduced operating
costs, although more so for the early 1990s than the late 1980s.
Cornett, McNutt, and Tehranian (2006) examine different
measures of efficiency improvements for large mergers, and
find evidence for cost-efficiency improvements in addition to
other revenue improvements. Hannan and Pilloff (2006) show
that cost-efficient banks tend to acquire relatively inefficient
targets. Using German banking data, Niepmann (2013) finds
4
Do Big Banks Have Lower Operating Costs?
a negative correlation between size and scaled operating
costs—a result consistent with our findings for U.S. firms.
Davies and Tracey (2014) argue that standard estimates of
scale economies for large banks are influenced by too-big-tofail (TBTF) subsidies, and that scale economies are no longer
present after controlling for TBTF factors. Hughes and Mester
(2013) dispute this conclusion, arguing that the cost function
used by Davies and Tracey is misspecified. One potential
advantage of our focus on noninterest expense is that operating costs (for example, information technology, printing,
postage, and advertising) may be relatively more likely to
reflect technological features of the firm’s production process
than any distortions due to TBTF. Instead, TBTF seems most
likely to affect the firm’s funding costs and capital structure. It
seems difficult, however, to rule out the possibility that TBTF
subsidies may affect our results or those of previous literature.
3. Data and Descriptive Statistics
Our analysis is based on quarterly FR Y-9C regulatory data
filed by U.S. bank holding companies. The Y-9C filings
include detailed balance sheet and income data, as well as
information about loan performance, derivatives, off-balance-sheet activities, and other aspects of BHC operations.
Data are reported on a consolidated basis, incorporating
both bank and nonbank subsidiaries controlled by the
BHC (see Avraham, Selvaggi, and Vickery [2012] for more
details). Our analysis considers only “top-tier” BHCs—that
is, the ultimate parent U.S. entity. Our sample includes toptier U.S. BHCs with a foreign parent, although it excludes
“stand-alone” commercial banks that are not owned by a
BHC, and BHCs that are too small to file the Y-9C (the Y-9C
reporting threshold varies over time, but is currently $500
million). Our sample excludes investment banks, thrifts, and
other types of financial institutions, unless those firms are
owned by a commercial BHC.
Noninterest expense is reported in the consolidated Y-9C
income statement (Schedule HI), broken down into five
categories. Note that noninterest expense does not include
loan losses due to defaults, trading losses, gains and losses on
owned securities, or taxes; these are recorded in other parts
of the income statement.4 Our analysis focuses on noninterest
4
BHC net income in Schedule HI is calculated as follows: net income = net
interest income + noninterest income − noninterest expense − provision
for loan and lease losses + realized securities gains (losses) − taxes +
extraordinary items and other adjustments − net income attributable to
noncontrolling interests. See Copeland (2012) for descriptive information on
how the main components of BHC income have evolved over time.
expense because it is the most likely area in which firms would
realize operating cost advantages from size.
We compute several normalized measures of noninterest
expense. The first measure, widely used by practitioners and
industry analysts, is the “efficiency ratio,” defined as the ratio
of noninterest expense to “net operating revenue,” the sum of
net interest income and noninterest income:
noninterest expense
Efficiency ratio = ________________________________
net interest income + noninterest income
A higher efficiency ratio indicates higher expenses, or equivalently, lower efficiency. Effectively, this ratio measures the
operating cost incurred to earn each dollar of revenue. Efficiency ratios vary widely across BHCs, as we document below,
but typical values range from 50 to 80 percent. Efficiency ratios
are sometimes computed excluding certain noncash items from
noninterest expense, such as amortization of intangible assets.
We refer to such measures as “cash” efficiency ratios.
One limitation of the efficiency ratio is that it is sensitive to
quarter-to-quarter movements in net operating revenue. For
example, ratios spiked for many BHCs during the financial
crisis, because of trading losses and other noninterest losses.
(In rare cases, the efficiency ratio even flips sign, because the
sum of net interest and noninterest income is negative.) To
provide an alternative normalization that is less sensitive to
these concerns, we also present results based on scaling noninterest expense by total assets or risk-weighted assets (RWA),
rather than net operating revenue:
noninterest expense
Expense asset ratio = ____________________________
total assets (or risk-weighted assets)
These normalizations can be computed for total noninterest
expense, or for NIE subcomponents such as compensation.
3.1 Descriptive Statistics
Table 1 presents descriptive statistics for noninterest expense
over the period from first-quarter 2001 to fourth-quarter
2012. We selected this period to take advantage of additional
detail on noninterest income expense that was added to the
Y-9C in 2001, thereby allowing us to separate noninterest
income (which we use as a control) into components such as
investment banking fees, income from insurance fees, deposit
fees, and servicing fees. Note that the sample period for our
regression analysis in Section 4 begins in first-quarter 2002
because we incorporate lagged income variables from the
previous four quarters. A total of 2,810 BHCs are present in
the sample for at least one quarter.
Panel A of the table reports summary statistics for four
normalized measures of noninterest expense: the efficiency
ratio, the cash efficiency ratio (which excludes goodwill
impairment and amortization from noninterest expense),
noninterest expense scaled by total assets, and noninterest
expense scaled by RWA. The industry efficiency ratio averages
66.3 percent over 2001-12, although it is somewhat higher
(71.7 percent) in 2012. The standard efficiency ratio and the
cash efficiency ratio differ little on average, reflecting the fact
that goodwill impairment and amortization expense generally
represent a small total of total noninterest expense.
The distribution of the expense ratios is skewed to the
right. For example, the difference between the 5th percentile
of the efficiency ratio and its median is 19.5 percent, significantly smaller than the difference of 28.0 percent between the
median and the 95th percentile value. Furthermore, the right
tail includes some extremely high values (for example, the
99.5th percentile is 198.4 percent), likely driven by one-time
spikes in revenue. To reduce the influence of outliers, our
regression analysis winsorizes the top and bottom 0.5 percent
of observations for each noninterest expense ratio (all data
below and above the bottom and top 0.5th percentiles, respectively, are set equal to the 0.5th and 99.5th percentiles).
We examine the components of noninterest expense in
Panel B of the table, based on the five noninterest expense
categories reported on Schedule HI.5
• Compensation (49.4 percent of industry total over the
sample time period, reported on FR Y-9C as “salaries and
employee benefits”). This category includes wages and salaries, bonus compensation, contributions to social security,
retirement plans, health insurance, employee dining rooms,
and other components of employee compensation.
• Premises and fixed assets (11.6 percent of total, reported on
Y-9C as “expenses of premises and fixed assets net of rental
income”) includes depreciation, lease payments, repairs,
insurance and taxes on premises, equipment, furniture,
and fixtures. It excludes mortgage interest on corporate real
estate.
• Goodwill impairment (1.8 percent of total, reported on
Y-9C as “goodwill impairment losses”) represents losses
incurred when goodwill exceeds implied fair value and is
revalued downwards. This item is reported separately from
“other noninterest expense” from 2002 onwards.
• Amortization expense (1.9 percent of total, reported on
Y-9C as “amortization expense and impairment losses for
other intangible assets”) includes amortization of goodwill
5
A detailed definition of these five variables can be found in the Federal
Reserve Microdata Reference Manual data dictionary, available at http://www
.federalreserve.gov/apps/mdrm/data-dictionary.
FRBNY Economic Policy Review / December 2014
5
Table 1
Noninterest Expense Summary Statistics
Industry
Full Sample
Individual Observations
2012
p0.5
p5
p25
p50
p75
p95
p99.5
Mean
Standard
Deviation
Panel A: Efficiency Measures, in Percent: 2001-12
Efficiency ratio
66.32
71.68
29.07
46.31
58.26
65.77
74.44
93.71
198.40
68.10
18.69
Cash efficiency ratio
63.29
70.39
28.69
45.81
57.72
65.17
73.72
92.07
168.11
67.05
16.64
Expense-to-asset ratio
0.82
0.82
0.25
0.45
0.63
0.75
0.88
1.25
3.95
0.80
0.37
Expense-to-RWA ratio
1.22
1.35
0.35
0.61
0.87
1.05
1.28
1.89
6.02
1.15
0.58
Panel B: Components of Noninterest Expense, as a Percentage of Total: 2001-12
Compensation
49.36
48.68
18.08
40.45
50.31
54.67
58.58
64.59
74.30
53.96
13.54
Premises and fixed assets
11.63
9.64
2.79
7.78
11.47
13.67
16.01
20.16
26.53
13.84
5.45
1.75
0.02
0.00
0.00
0.00
0.00
0.00
0.00
16.28
0.29
5.03
Goodwill impairment
Amortization expense
Other
1.93
1.78
-0.03
0.00
0.00
0.00
0.97
3.57
9.03
0.76
1.72
34.95
39.88
10.02
20.93
26.22
30.04
34.71
45.82
69.29
31.11
16.15
Source: Board of Governors of the Federal Reserve System, Consolidated Financial Statements of Bank Holding Companies (FR Y-9C data).
Notes: The table reports summary statistics for 2,810 unique bank holding companies from 2001:Q1 to 2012:Q4, a total of 58,217 firm-quarter observations.
The column labeled “industry” reports the average industry efficiency ratio, calculated by summing across all bank holding companies each quarter, taking
the ratio, and then taking the time-series mean, either over the 2001:Q1 – 2012:Q4 sample period or over calendar year 2012. The denotation “p” refers to
percentiles of individual observations (for example, “p50” is the median). Variables are defined in Appendix A. RWA is risk-weighted assets.
and other intangible assets owned by the BHC, as well as
impairment losses for intangible assets other than goodwill.
This item is also available from 2002 onwards.
• Other (35.0 percent of total) includes a broad range of
other operating costs, such as telecommunication and
information technology costs, legal fees, deposit insurance,
advertising, printing, postage, and so on. Additional information on these expenses is provided in the memoranda to
Schedule HI, as we explain in detail below.
Chart 1 plots the time series evolution of the four normalized measures of total industry NIE. Each expense measure
declined between 2001 and mid-2007, a period when the
revenues and assets of the banking system grew rapidly. For
example, the industry efficiency ratio fell from 65.4 percent in
quarter-one 2001 to 58.8 percent in quarter-two 2007, while
the expense asset ratio declined from 0.96 percent to 0.72 percent over the same period. This downward trend was reversed
during the 2007-09 financial crisis. Since the efficiency ratio
is mechanically inversely related to net operating revenue,
the reversal for that NIE measure is perhaps unsurprising.
However, the expense asset ratio also increased, whether normalized by total assets or risk-weighted assets. In recent years
6
Do Big Banks Have Lower Operating Costs?
noninterest expense ratios have stabilized at levels higher than
those prevailing prior to the onset of the crisis. The rise in the
efficiency ratio in part simply reflects the decline in net operating revenue and measures of profitability for the banking
industry, owing to compression of net interest margins and
lower noninterest income.
Appendix B also plots the evolution of the relative shares of
the five noninterest expense subcategories.6 Goodwill impairment expenses are almost entirely concentrated in 2008, with
negligible levels for this expense category before and after
2008. Other noninterest expense makes up a progressively
larger fraction of total NIE over the past five years. (In 2012,
this category represented 39.9 percent of total NIE, a share
similar to that held by compensation expenses).
As a first look at the relationship between firm size and
normalized noninterest expense, the main focus of this paper,
we present scatter plots of BHC size and the efficiency ratio
(Chart 2). The plots are based on year-to-date 2012 expense
data and assets as of the end of 2012. A striking feature of the chart
6
Appendix B is available at http://www.newyorkfed.org/research
/epr/2014/1403kovn_appendixB.pdf.
Chart 1
Chart 2
Noninterest Expense Ratios over Time
Scatter Plots of Operating Cost Ratios and BHC Size
Percent
110 Efficiency Ratios
Percent
300 Efficiency Ratio
Efficiency ratio
100
250
90
200
80
150
70
100
$1 billion
$10 billion
$100 billion
$1 billion
$10 billion
$100 billion
50
60
Cash efficiency ratio
50
0
1.6 Expense Asset Ratios
4 Expense Asset Ratio
NIE/RWA Ratio
1.4
3
1.2
2
1.0
NIE/Assets Ratio
1
0.8
0.6
0
2001 02
04
06
08
10
12
10
Source: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data).
3.2 Classifying Other Noninterest Expense
14
16
18
Size (log of assets in $000s)
20
22
Source: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data).
Notes: Income data are quarterly and are not annualized. Ratios are
reported in percentages. NIE is noninterest expense; RWA is
risk-weighted assets.
is the variability in noninterest expense across firms, particularly
among smaller BHCs. This finding is also borne out in our
multivariate analysis in Section 4. The variability points to the
importance of adding controls for those observable differences
in BHCs’ activities that are associated with different types of
expenses. These controls are described in Section 3.3.
12
Notes: Scatter plots are based on average quarterly noninterest
expenses over 2012 and total BHC assets as of the end of 2012. BHC
is bank holding company.
classify other NIE into eleven standardized subcategories;7
in addition, space is provided for BHCs to report additional
“write-in” expense items that were not captured by the standardized fields. For the eleven standardized subcategories,
BHCs are instructed to record items for amounts greater than
$25,000 that also exceed 3 percent of total other noninterest
expense. Write-in items bear the additional requirement that
the expense item exceed 10 percent of total other noninterest
7
The category “other NIE” represents more than one-third
of industry noninterest expenses since 2001. To shed light
on these costs, we examine data from the memoranda to
Schedule HI. Since 2008, Schedule HI has allowed BHCs to
The eleven standardized memoranda categories are (a) data processing
expenses, (b) advertising and marketing expenses, (c) directors' fees, (d)
printing, stationery, and supplies, (e) postage, (f) legal fees and expenses,
(g) FDIC insurance assessments, (h) accounting and auditing expenses,
(i) consulting and advisory expenses, (j) automated teller machine (ATM)
and interchange expenses and (k) telecommunications expenses. See FR Y-9C
Schedule HI Memorandum Item 7.
FRBNY Economic Policy Review / December 2014
7
expense. Since 2008, amounts in the eleven standardized
categories have made up 38 percent of total other noninterest
expense, while the write-in fields have constituted another
28 percent of other NIE. The remaining 34 percent of other
noninterest expense is not reported in the Schedule HI memoranda, presumably because it does not meet the reporting
thresholds described above.
It is particularly challenging to classify and analyze items
recorded in the write-in expense fields, because these amounts
are reported using nonstandardized language by each BHC.
For example, noninterest expenses related to foreclosures and
to properties that are “other real estate owned”8 are variously
written in as “reo,” “ore,” “R.E.O,” “oreo,” “foreclose,” and so on,
as well as various misspelled text strings such as “oero” and
“forclosuer” (sic). Overall, more than 30,000 text strings are
written in by the BHCs in our sample between 2008 and 2012.
Approximately 5,500 of these strings are unique. Individual
BHCs often tend to use the same text field from one quarter
to the next when referring to a given data item, a practice that
reduces the total number of fields to be classified.
We classify each unique text string into broad categories,
proceeding in two steps. First, we classify each string into one
of ninety subcategories, such as “card rewards,” “custodian
fees,” “affordable/low-income housing,” “servicing,” “dues/
memberships/subscriptions,” and “lockbox fee.” We chose
these subcategories by grouping together apparently similar
items, employing our institutional knowledge where possible,
as well as internet searches and our best judgment. A list of
these subcategories, along with the percentage of nonmissing values, is presented in Appendix B to this paper. This
classification was in part done by hand, and in part via Stata
code that conducted Boolean searches for keywords within
each text string. The subcategories include four separate
“miscellaneous/other” categories, one for text strings that
are well-defined but do not fit into any obvious category (for
example, “cattle feed,” “livestock,” and “image processing”),
one for items that we did not understand (for example, “tops
expense”), one for items that are vague or otherwise unclassifiable (for example, “sundry loss”), and one for text strings that
combine multiple items with values listed.
Since most of the subcategories are fairly sparsely populated,
as documented in Appendix B, we then aggregate them into
nine categories that are better suited to statistical analysis. We
also assign each of the eleven standardized memoranda items to
one of the same nine author-defined categories. By doing this,
we are able to classify 66.2 percent of other noninterest expense
into the nine high-level categories, which are listed below:
8
“Other real estate owned” refers to real estate owned by a bank as a result of
the foreclosure of a mortgage loan.
8
Do Big Banks Have Lower Operating Costs?
• Corporate overhead (18.6 percent of other NIE). This category,
which is intended to measure general corporate expenses,
includes four standardized Y-9C items: “accounting and
auditing,” “printing, stationery, and supplies,” “postage,”
and “advertising and marketing.” It also includes write-in
expenses related to corporate overhead costs, such as travel,
business development, recruitment, professional memberships and subscriptions, and charitable contributions.
• Information technology and data processing (12.6 percent
of other NIE). This category covers the standardized
Y-9C item “data processing expenses,” as well as write-in
expenses related to information technology, software, and
internet banking.
• Consulting and advisory (11.1 percent of other NIE).
This category is the standardized Y-9C item “consulting
and advisory expenses.” It does not include any write-in
expenses.
• Legal (6.7 percent of other NIE). This category includes the
standardized Y-9C item “legal fees and expenses,” as well
as write-in line items related to “litigation,” “settlement,”
“records retention,” “legal reserve,” and similar items.9
• Retail banking (6.4 percent of other NIE). This category
is intended to reflect operating costs related to lending and
deposit-taking. It includes the standardized NIE category
“ATM and interchange expenses,” as well as write-in items
related to loans, retail banking, or credit cards (for example,
costs related to real estate owned properties, credit reports,
credit card rewards, branch closing costs, lockbox fees,
check fraud, and so on).
• Federal Deposit Insurance Corporation (FDIC) assessments
and other government-related expenses (5.8 percent of
other NIE). This category includes the standardized Y-9C
item “FDIC deposit insurance assessments” and write-in
expenses related to the Community Reinvestment Act,
compliance with regulation, and other items. In practice,
deposit insurance fees make up the bulk of these expenses.
• Other financial services (3.0 percent of other NIE). This
category embraces written-in expense items for financial
activities other than traditional lending and depository
services—in particular, asset management, insurance, and
miscellaneous derivatives- and trading-related expenses.
• Directors’ fees and other compensation (0.3 percent of
other NIE). This category includes the standardized Y-9C
category “directors’ fees,” as well as write-in fields related to
director compensation or other compensation costs.
9
The standardized “legal fees and expenses” other NIE category includes fees
and retainers paid for legal services obtained, but excludes legal settlements
and legal expenses associated with owned real estate. Legal settlements and
legal reserves established against expected future settlements are recorded in
the write-in text fields, if separately reported.
Table 2
Components of Other Noninterest Expense
Panel A: FR Y-9C Classification of Other Noninterest Expense: 2008-12
Category
Percentage of Total
Other Noninterest
Expense, Industry
In Y-9C
37.99
Text classified
28.21
Unclassified
Total
33.80
100.00
Panel B: Components of Other Noninterest Expense, as a Percentage of Total Other Noninterest Expense: 2008-12
Individual Observations
Component (Author-Defined)
Industry
p0.5
p5
p25
p50
p75
p95
p99.5
Mean
Standard
Deviation
Corporate overhead
18.63
0.00
2.43
10.29
16.26
22.70
34.58
50.95
17.07
10.07
Information technology and data processing
12.63
0.00
0.64
8.21
13.84
19.81
29.91
45.01
14.54
8.69
Consulting and advisory
5.23
11.07
0.00
0.00
0.00
2.31
5.78
12.73
29.97
3.74
Legal
6.68
0.00
0.00
0.00
3.53
6.19
12.43
24.71
4.16
4.71
Retail banking
6.35
0.00
0.00
0.00
6.41
13.48
29.64
55.24
9.24
10.55
FDIC assessments and other government
5.81
0.00
0.00
6.80
11.53
16.95
25.54
37.34
12.26
7.58
Other financial services
3.01
0.00
0.00
0.00
0.00
0.00
4.00
15.85
0.56
2.72
Directors’ fees and other compensation
0.25
0.00
0.00
0.00
0.00
3.45
6.99
14.60
1.91
2.85
Miscellaneous
1.76
0.00
0.00
0.00
0.00
0.00
5.75
24.91
0.84
3.98
Total classified
66.20
4.02
35.11
55.83
66.87
75.05
85.72
95.35
64.32
15.73
Unclassified
33.80
Source: Board of Governors of the Federal Reserve System, Consolidated Financial Statements of Bank Holding Companies (FR Y-9C data).
Notes: The table reports summary statistics for 2,810 unique bank holding companies from 2008 to 2012. Annual data are as of year-end, for a total of 4,999
firm-year observations. Panel A summarizes information on the following types of noninterest expense: (i) FR Y-9C line items: eleven standardized other
noninterest expense items reported in FR Y-9C Schedule HI: Memoranda, (ii) text classified: other noninterest expense items reported in Schedule HI: Memoranda as text fields, and (iii) unclassified: other noninterest expense items not classified in Schedule HI (for example, because the amounts do not exceed the
reporting threshold). Panel B includes summary statistics for the nine author-defined other noninterest expense categories, which are constructed from the
FR Y-9C line items and the text fields. These data are described in Section 3.2. FDIC is Federal Deposit Insurance Corporation.
• Miscellaneous (1.8 percent of other NIE). The final category
reflects the four types of miscellaneous categories described
above—that is, items that cannot be easily classified or are not
understood by us based on the content of the write-in field.
In a small minority of cases, the write-in field content suggests an expense item that may have been classified as other
NIE by mistake (for example, costs related to employee compensation). We did not attempt to reclassify these expenses,
given the limited context and detail in the write-in fields.
Descriptive statistics for these nine author-defined categories of other NIE are presented in Panel B of Table 2. Note that
the individual percentiles and standard deviations reported in
the table are based on annual expenses, rather than quarterly
values. We adopt this approach because of the significant
number of zero values reported for even these nine aggregated
categories. Our analysis of the other NIE subcategories is
based on these year-end cumulative expenses.
The variation across BHCs in the relative size of different components of other NIE is striking. For example, the
category “other financial services,” which includes noninterest
expense related to insurance and other nonbanking financial
services, has a median value of zero, but at the 99.5th percentile, it is 15.9 percent of total other noninterest expense. This
varied distribution of expenses is consistent with the dispersion in products and services offered by BHCs.
FRBNY Economic Policy Review / December 2014
9
3.3 Controls
Operating costs are likely to vary significantly across BHCs
engaged in different business activities. While the decision to
enter different businesses is endogenous, and may be related
to size, we are primarily interested in understanding how size
is related to operating expenses on an apples-to-apples basis.
For this reason, our regression analysis controls for a variety
of BHC characteristics reported in the FR Y-9C. Summary statistics for these controls are presented in Table 3. In order to
show how these controls are related to bank size, we also present industry averages for the following size cohorts: largest
1 percent, 95 to 99 percent, 75 to 95 percent, 50 to 75 percent,
and smallest 50 percent.10 Differences in BHC characteristics
by size are clear from differences in sample means within the
cohorts. However, there is substantial variation in business
models apparent within size cohorts as well.
The controls in Table 3 are grouped into six categories, as
follows:
• Asset shares. Our asset composition control variables
measure the fraction of balance sheet assets held in various
types of loans and other assets (for example, trading assets,
securities, cash, and fixed assets). As shown in Table 3,
small firms hold a higher fraction of total assets in the form
of loans, while trading assets are a significantly higher
share of total assets for the largest BHCs than for any
other group.
• Risk. We control for two additional measures of asset
risk: risk-weighted assets as a percentage of total assets,
and nonperforming loans (NPLs) as a percentage of
total loans. The relationship between firm size and risk is
non-monotonic for both risk measures, although we note that
the largest firms have significantly higher nonperforming loan
ratios than other BHCs.
• Revenue composition. Revenue composition refers to the percentage of net operating revenue (the sum of interest and
noninterest income) that is earned from different sources:
(i) interest income, (ii) trading income, and (iii) five different
components of noninterest nontrading income. Since these
components can be volatile, in the regressions we include
these variables in the form of a four-quarter rolling average
lagged value. (The standard deviation reported in the table is
10
To compute the industry average for the asset and income ratios, we sum
the numerator and denominator of the ratio across all firms in the size cohort,
and then take the ratio of the two sums. In contrast, the mean and standard
deviation reported in the first two columns represent the unweighted mean
and standard deviation of the individual observations in the sample. Of
course, the mean of the individual observations may differ substantially from
the industry mean if the ratio in question is correlated with firm size.
10
Do Big Banks Have Lower Operating Costs?
based on this four-quarter rolling average.) It is notable that
large BHCs earn a significantly higher percentage of revenue
from noninterest income.
• Funding structure. In some specifications, we include two
controls for funding structure, the ratio of deposits to
assets, and a dummy for whether the BHC is a publicly
traded company (firms with foreign parents are coded
as private, regardless of whether their ultimate parent is
public). Large firms fund less of their assets with deposits,
on average.
• Business concentration. Research in organizational economics has found that diversified firms tend to be less
efficient and less profitable than focused firms. In studies
that are most relevant to our analysis, Goetz, Laeven, and
Levine (2013) find that geographically diversified commercial
banks have lower valuations, while Laeven and Levine (2007)
find a diversification discount (based on the firm’s activity
mix) in an international cross-section of banks. In the spirit
of these studies, we include Herfindahl-Hirschman Index
(HHI)–style measures of asset and income concentration,
computed as the sum of squared asset weights and income
weights, respectively, based on the categories presented in
Table 3. Higher values of these measures indicate greater
concentration. As the table shows, large firms have more
diversified assets and activities (lower HHI), reflecting
their greater reliance on financial activities outside of
traditional lending and deposit taking.
• Organizational complexity. Organizationally complex firms
may also have higher operating costs, because of various
internal inefficiencies (for example, duplication of efforts
across different subsidiaries or divisions within the same
firm). It is important to attempt to disentangle the effects
of size and structure, given that large firms are likely to be
organizationally complex. Our analysis includes three measures of organizational structure, the log number of subsidiaries (following Avraham, Selvaggi, and Vickery [2012]),
the percentage of subsidiaries domiciled overseas, and
a dummy for whether the BHC has a foreign parent. As
shown by the sample means across size cohorts, large
firms have more complex organizational structures than
small firms on each of these dimensions. The differences
are striking: the largest BHCs (those in the top 1 percent
of the size distribution) have 962 subsidiaries on average,
22.7 percent of which are domiciled overseas. BHCs below
the sample median in size, however, have only 4 subsidiaries on average, and only 4.8 percent of these subsidiaries
are domiciled outside the United States.
Table 3
Summary Statistics for Control Variables
Individual
Observations
Industry, by Size Cohort
Top 1%
95-99%
75-95%
50-75%
Bottom
50%
Industry
Mean
Standard
Deviation
42.08
59.58
64.65
67.84
67.57
48.39
66.44
13.36
Residential real estate loans
14.94
16.63
16.55
17.32
18.08
15.53
17.78
10.62
Commercial real estate loans
4.26
15.65
28.12
31.47
29.77
9.48
28.27
15.02
Commercial and industrial loans
8.64
12.54
11.20
10.25
9.94
9.65
10.42
6.84
Credit card loans
3.53
2.33
0.59
0.26
0.17
2.93
0.32
2.93
Other consumer loans
4.68
6.11
4.19
3.72
3.87
4.89
4.25
5.14
All other loans
6.03
6.32
4.00
4.83
5.73
5.91
5.40
7.83
Trading assets
15.52
1.45
0.24
0.04
0.04
10.89
0.20
1.75
Federal funds and repurchase agreements
13.67
2.20
1.24
1.61
2.07
9.95
2.14
3.93
Asset shares (percentage of total assets)
Total loans
Cash
5.49
5.76
4.41
4.65
4.91
5.43
4.64
4.01
12.65
20.60
22.94
20.56
20.46
15.34
21.35
12.38
Other real estate owned
0.11
0.12
0.31
0.42
0.49
0.14
0.36
0.89
Fixed assets
0.70
1.24
1.62
1.92
2.02
0.93
1.90
1.05
Investments in unconsolidated subsidiaries
0.33
0.18
0.09
0.12
0.07
0.27
0.09
1.38
Investments in real estate ventures
0.08
0.05
0.02
0.03
0.02
0.07
0.02
0.94
Intangible and other assets
8.02
6.77
3.89
3.19
2.97
7.24
3.19
2.11
Risk-weighted assets (percentage of total assets)
63.85
75.08
71.72
72.95
71.82
67.04
71.68
11.89
Nonperforming loans (percentage of total loans)
2.94
1.85
2.05
1.83
1.95
2.51
1.65
2.65
Interest income
50.61
51.56
65.08
73.25
77.26
53.01
77.62
12.54
Trading income
7.38
1.58
0.28
0.08
0.09
5.44
0.19
1.14
45.38
46.85
34.65
26.68
22.66
43.90
22.26
12.30
Investment securities
Risk
Revenue composition (percentage of net operating revenue)
Noninterest nontrading income
Fiduciary income
7.86
9.63
4.54
3.96
2.64
7.83
2.84
4.97
12.96
7.32
8.60
1.38
0.83
10.73
0.99
2.83
Service charges on deposits
5.43
6.53
7.40
7.84
7.79
5.93
7.87
4.56
Net servicing fees
3.48
1.52
0.65
0.47
0.52
2.69
0.60
1.58
15.55
21.85
13.45
13.03
10.88
16.66
9.77
9.32
Investment banking fees
Other income
Funding structure
Deposits/assets (percent)
43.67
62.76
74.85
79.58
81.17
51.49
79.21
10.42
Publicly traded (percentage of sample)
76.85
79.16
60.18
30.81
12.69
30.02
27.75
44.78
HHI assets
0.25
0.41
0.48
0.51
0.51
0.29
0.52
0.13
HHI income
0.53
0.56
0.59
0.64
0.67
0.53
0.69
0.17
Business Concentration
Organizational complexity
Number of subsidiaries
962.25
68.78
10.76
6.22
4.07
18.29
15.75
139.99
Percentage of subsidiaries foreign
22.71
14.46
3.88
4.54
4.83
16.15
0.75
5.18
BHC is foreign-owned (percentage of sample)
23.15
18.06
3.28
0.39
0.62
2.02
1.78
13.24
FRBNY Economic Policy Review / December 2014
11
Table 3 (continued)
Summary Statistics for Control Variables
Industry, by Size Cohort
Top 1%
95-99%
75-95%
50-75%
Bottom 50%
Industry
604
2,405
12,197
15,181
27,830
58,217
14
56
282
352
705
1,410
599,180
42,761
3,153
838
424
9,065
Sample statistics: Regression sample (2002-12)
N
Average number of firms
Average asset size (millions of dollars)
Source: Board of Governors of the Federal Reserve System, Consolidated Financial Statements of Bank Holding Companies (FR Y-9C data).
Notes: The table reports summary statistics for 2,810 unique bank holding companies from 2001:Q1 to 2012:Q4, a total of 58,217 firm-quarter observations.
The first six columns are industry ratios (computed by first summing numerator and denominator across all firms in the relevant size class), or are statistics
weighted by firm size, except for the two indicator variables “publicly traded” and “BHC is foreign-owned.” Size cohorts are recalculated in each quarter. The
last two columns are unweighted statistics across all firms. Note that the sample period for the regression analysis begins in 2002:Q1, not 2001:Q1, because
specifications include lagged income variables from the previous four quarters. See Appendix A for variable definitions. HHI is Herfindahl-Hirschman
Index; BHC is bank holding company.
4. Analysis
In this section, we study the relationship between BHC size
and measures of total noninterest expense scaled by revenue or assets, examining how this relationship is affected by
controlling for differences in firms’ business models and by
the normalization of noninterest expense used. Our analysis
progressively adds controls for a wide range of measures of
the composition of BHC assets and sources of income, on the
presumption that some types of assets or activities are likely to
be more complex and time-consuming to manage than others.
For example, a BHC with a large portfolio of other real estate
owned assets will likely incur significant property maintenance and management expenses associated with these assets,
compared with an otherwise similar banking firm that has liquidated such properties in return for cash, government securities, or other simple assets. Similarly, a portfolio of consumer
loans is likely to have different screening and monitoring costs
than a portfolio of commercial loans. Including these controls
seems particularly important given that asset composition
varies significantly by firm size, as documented in Section 3.
4.1 Total Noninterest Expense
Table 4 presents ordinary least squares estimates of the relationship between the efficiency ratio and BHC size measured
by the log of total assets. We find a statistically and eco-
12
Do Big Banks Have Lower Operating Costs?
nomically significant inverse relationship between size and
the efficiency ratio in each regression specification. That is,
noninterest expenses per dollar of net operating revenue are
lower for large BHCs.
The first column of results controls only for time-series
variation in the efficiency ratio, through the inclusion of
quarter fixed effects. Each subsequent regression specification
successively adds more explanatory variables associated with
differences in BHCs’ business activities. We begin with simple
controls for the composition of BHC assets and add more
detailed measures of the risk of those assets, the composition
of revenue, funding structure, business concentration, organizational complexity, and geography.
Looking across the models, we see that the inclusion of
additional controls tends to steepen the inverse relationship
between BHC size and the efficiency ratio. Including controls
for BHC asset composition (for example, the percentage of
assets in fixed assets, residential real estate loans, trading
assets, and so on) increases the magnitude of the coefficient
on bank size by 54 percent (from -1.32 in specification 1 to
-1.96 in specification 3), and increases the explanatory power
of the model by 13 percentage points. Controlling for the percentage of income generated by different activities (for example, trading, investment banking, and deposit service charges)
shifts the coefficient to -2.63 (specification 6). The inclusion
of controls for organizational complexity further steepens the
association between BHC size and the efficiency ratio; the
coefficient increases in magnitude from -2.98 (specification 8)
to -4.13 (specification 9).
Table 4
BHC Size and the Efficiency Ratio
Specification
(1)
Log assets
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
-1.320***
-1.892*** -1.962***
-2.044***
-2.509***
-2.631***
-2.886***
-2.983***
-4.131***
-4.151***
-2.471*
(0.235)
(0.228)
(0.246)
(0.239)
(0.240)
(0.271)
(0.273)
(0.334)
(0.326)
(1.156)
-30.446*** -31.136*** -23.170*
-21.549*
-22.379*
-31.408**
(8.859)
(8.910)
(10.472)
(0.226)
Asset shares
(percentage of total assets)
Total loans
-50.105***
(7.446)
Residential real
-41.250*** -42.777*** -28.889**
estate loans
(7.850)
Commercial real estate
(8.211)
(8.877)
(8.367)
(8.579)
(9.415)
-55.329*** -63.050*** -46.223*** -46.866*** -47.723*** -38.003*** -36.868*** -31.123**
loans
(7.452)
Commercial and
(9.352)
(10.172)
-41.365*** -43.923*** -30.428**
industrial loans
(8.235)
(10.014)
(10.676)
(9.729)
(9.922)
(9.990)
-32.324** -32.581**
-24.657*
-25.291*
-15.721
-43.188***
(10.189)
(10.748)
(10.249)
(10.201)
(10.512)
(10.276)
(9.894)
-45.328***
(10.596)
(10.340)
Credit card loans
-70.539*** -84.648*** -79.998*** -81.301*** -80.567*** -69.742*** -66.710*** -59.817*** -36.635
Other consumer loans
-63.106*** -67.709*** -54.509*** -53.905*** -54.258*** -45.243*** -45.078*** -34.291**
(11.455)
(8.749)
All other loans
Trading assets
Federal funds and
repurchase agreements
Investment securities
(10.068)
(9.973)
(11.430)
(10.805)
(10.945)
(10.353)
(10.950)
(10.466)
(12.164)
(11.060)
(11.620)
(10.654)
(10.812)
(10.619)
(19.167)
-37.861***
(11.343)
-69.382*** -74.193*** -59.828*** -61.058*** -60.776*** -52.092*** -51.257*** -41.791*** -60.073***
(8.442)
(9.793)
(10.711)
(10.216)
(10.442)
(10.901)
(10.321)
(10.233)
-2.154
-2.418
-1.657
-3.909
-12.428
-10.508
-5.105
-3.128
-1.641
-9.133
(18.177)
(18.105)
(17.966)
(17.525)
(16.434)
(16.871)
(18.359)
(17.552)
(18.084)
(33.833)
-20.466*
-18.125
-22.468*
-17.305
-19.636*
-18.727*
-18.063
-16.537
-15.062
-16.323*
(9.194)
(9.378)
(9.526)
(9.598)
(9.278)
(9.253)
(9.220)
(8.875)
(8.654)
(13.084)
(7.514)
-44.233*** -46.246*** -47.976*** -35.704*** -36.532*** -36.918*** -35.623*** -32.975*** -29.990*** -28.246***
(7.538)
(7.420)
(7.135)
(7.792)
(7.487)
(7.660)
(7.625)
(7.248)
(7.193)
(6.448)
Other real estate owned
511.223*** 516.118*** 218.441*** 224.027*** 227.645*** 228.260*** 224.115*** 223.890*** 248.885*** 264.291***
Fixed assets
195.591*** 195.896*** 213.179*** 182.093*** 190.166*** 197.031*** 187.538*** 189.759*** 223.443*** 289.553***
(59.960)
(31.754)
Investments in
unconsolidated subsidiaries
Investments in real
estate ventures
Intangible and other assets
(58.233)
(31.448)
(50.156)
(30.379)
(52.325)
(29.035)
(51.683)
(29.664)
(51.743)
(29.974)
(52.201)
(29.496)
(51.959)
(28.939)
(51.125)
(29.775)
-74.519*** -64.972*** -56.758*** -69.983*** -75.613*** -74.270*** -75.580*** -86.452*** -81.657***
(13.295)
-72.295*** -64.503*** -54.043*** -66.178*** -71.900*** -70.348*** -29.115
-42.377*
-36.690*
58.462
(15.963)
(19.251)
(17.849)
(50.434)
92.308***
90.825***
(17.868)
(15.216)
55.478**
(21.111)
For the model including all controls but excluding firm
fixed effects (specification 10), the coefficient on size of -4.151
implies that a 10 percent increase in size is associated with a
42 basis point decrease in the efficiency ratio, equivalent to
0.6 percent of the sample average efficiency ratio. In dollar
(12.469)
(14.355)
(11.868)
(13.837)
(12.733)
(14.470)
(13.733)
7.582
(13.386)
(16.499)
(13.768)
(36.789)
(13.632)
(18.720)
(16.201)
(54.925)
(19.204)
(42.429)
34.231
31.273
26.103
23.238
19.702
16.813
0.999
(20.543)
(19.928)
(20.804)
(20.893)
(20.411)
(20.255)
(21.117)
terms, the coefficient implies that for a BHC at the mean of
the data ($9.1 billion in assets), an increase in size of $1 billion is associated with a reduction in operating expenses of
$437,000 per quarter, relative to a counterfactual in which the
efficiency ratio is not associated with size. The corresponding
FRBNY Economic Policy Review / December 2014
13
Table 4 (continued)
BHC Size and the Efficiency Ratio
Specification
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Revenue composition
(percentage of net operating
revenue)
Trading income
Noninterest nontrading
49.008
45.614
47.794
44.346
30.746
46.602
35.616
(26.304)
(25.079)
(25.203)
(26.351)
(25.803)
(25.765)
(43.903)
19.746***
income
(3.151)
Fiduciary income
Investment banking fees
Service charges
on deposits
Net servicing fees
Other noninterest income
30.172*** 29.695***
25.165*** 27.327*** 24.057***
(4.570)
(4.580)
(4.793)
(4.822)
(4.718)
(8.471)
37.832**
37.510**
33.487**
29.794**
35.915***
46.586**
(12.036)
(12.140)
13.020*
13.072*
(6.356)
(6.284)
(11.527)
(11.075)
(9.925)
3.950
5.965
14.567*
(6.448)
(6.294)
(7.094)
34.635***
(14.453)
49.324***
(10.446)
-1.060
1.707
-5.177
-1.426
14.615
-9.113
(16.367)
(16.477)
(16.582)
(16.699)
(14.275)
(15.153)
21.814*** 21.688***
20.629*** 20.181*** 21.462***
(3.837)
(3.919)
(3.751)
(3.716)
(3.730)
0.801
-0.497
-2.194
-0.643
-1.061
4.577
(3.119)
(3.075)
(2.980)
(2.903)
(3.770)
1.418*
-0.704
(0.626)
(1.705)
(3.656)
Funding structure
Deposits/assets (percent)
Public [1=yes]
1.474*
1.314*
(0.606)
(0.608)
(0.621)
1.787**
-10.565*
-9.828*
(4.220)
(4.093)
(3.907)
(5.091)
-8.101**
-7.205*
-8.681**
-8.903**
(3.023)
(2.934)
(2.902)
(3.447)
Log number of subsidiaries
1.883***
1.771***
(0.395)
(0.396)
Percentage of subsidiaries
-3.813
-5.668
2.694
(5.341)
(5.139)
(8.515)
Business concentration
HHI assets
HHI income
-10.531** -10.581*
Organizational complexity
that are foreign
Foreign-owned [1=yes]
(2.436)
(5.529)
101.061*** 143.904*** 146.053*** 144.782*** 136.250*** 138.941*** 142.911*** 152.872*** 161.137*** 157.186*** 122.139***
(3.377)
(8.397)
(8.432)
(8.075)
calculation for the smaller coefficient from column 2 implies a
reduction in operating expenses of $199,000 per quarter.
14
(0.534)
14.895*** 13.512*** 15.046**
(2.481)
Constant
1.404**
Do Big Banks Have Lower Operating Costs?
(8.276)
(8.036)
(9.438)
(9.380)
(9.324)
(9.372)
(19.637)
The final specification in Table 4 includes BHC fixed
effects, and thus examines only changes in size within bank
holding companies. This within-firm analysis includes both
Table 4 (continued)
BHC Size and the Efficiency Ratio
Specification
Time fixed effects
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
State fixed effects
Yes
Firm fixed effects
Yes
R2
0.080
0.195
0.207
0.247
0.258
0.261
0.262
0.264
0.271
0.296
0.549
N
58,217
58,217
58,217
58,217
58,217
58,217
58,217
58,217
58,217
58,217
58,217
Source: Authors' calculations.
Notes: The table presents an analysis of the relationship between size, measured by log of total assets, and efficiency ratio, defined as total noninterest expense
normalized by net operating revenue. All explanatory variables are lagged by one quarter. Revenue composition variables are the rolling average for the absolute value of the income share over net operating revenue. HHI (Herfindahl-Hirschman Index) assets is the sum of squared asset shares, by asset type, and
HHI income is the sum of squared four-quarter rolling average income shares, by income type. See Appendix A for further detail on controls included in the
models. Models are estimated with robust standard errors and two-way clustering by firm and quarter. Standard errors are in parentheses.
*** p<0.01
** p<0.05
* p<0.1
size changes from organic growth and size changes from
mergers. While still statistically significant, this coefficient is
somewhat smaller in magnitude than that of specification 10
(-2.47 compared with -4.15). There is some evidence that
noninterest expenses after mergers are inflated by one-time
merger related costs (Kwan and Wilcox 2002), which may
account for this difference. The standard error of the size
coefficient estimate from specification 11 is much larger
than in the other specifications; in other words, the coefficients are estimated with lower power, owing to the smaller
residual variation in the efficiency ratio not absorbed or
accounted for by the fixed effects and other controls.
As expected, observable differences among BHCs explain
a significant fraction of the variation in noninterest expenses.
Simple asset controls alone more than double the adjusted
R2 of the initial specification. However, even the fixed effects
specification in column 11 has an R2 of only 54.9 percent,
implying a large amount of residual variation in operating
costs. Furthermore, the inclusion of BHC fixed effects nearly
doubles the R2 relative to specification 10, a result suggestive
of large persistent differences in operating costs across observably similar firms. This finding seems consistent with prior
literature on X-inefficiency, which shows that many banking
firms operate significantly inside the efficient production frontier (see, for example, Berger, Hunter, and Timme [1993]). It
is worth noting that BHC size alone explains only a very small
fraction (less than 1 percent) of the total variation in noninterest expense in the data, as illustrated graphically in Chart 2.
In sum, Table 4 provides consistent evidence that large
BHCs have lower operating costs as measured by the efficiency ratio, although the strength of the relationship is
sensitive to the set of controls used. Instead of taking a strong
stance on the “appropriate” set of controls, throughout the
paper we present results for specifications using controls
from columns 1, 2, and 10 from Table 4. A comparison of
the results across these specifications enables the reader to
observe how the relationship between noninterest expenses
and size is influenced by the inclusion or exclusion of controls
for the mix of BHC assets and business activities.
Although our main focus is on the relationship between
operating costs and firm size, estimates for several of the
controls included in Table 4 are also of independent interest. In
particular, BHC organizational complexity, measured by the log
number of subsidiaries, is associated with higher noninterest
expense ratios. BHCs with a foreign parent also have higher
expenses. Proxies for greater organizational focus are associated
with lower noninterest expense: BHCs that have more concentrated asset portfolios and more concentrated sources of noninterest income have lower expenses, all else equal, although
the marginal explanatory power of additional concentration is
relatively low. Each of these relationships is robust to the inclusion of BHC fixed effects (column 11). Although not shown
in Table 4, these relationships are also robust to specification
changes that allow for a more flexible linkage between size and
the efficiency ratio. This finding suggests that our results are not
likely to be driven only by the largest BHCs.
FRBNY Economic Policy Review / December 2014
15
Caution should be exercised in applying a causal interpretation to these associations, given that we do not have
a convincing econometric instrument for organizational
complexity or focus. But taken at face value, each of these
estimates implies that complex, diversified firms have higher
operating expenses than focused or organizationally simple
firms, consistent with the conclusions of prior literature on
the diversification discount in banking (Goetz, Laeven, and
Levine 2013; Laeven and Levine 2007).
Chart 3
Efficiency Ratio and BHC Size, Flexible
Functional Forms
30
Normalized efficiency ratio
$1 billion
20
$10 billion
$100 billion
10
0
-10
4.2 Other Functional Forms
The specifications so far assume a log-linear relationship
between BHC size and the efficiency ratio. Next we allow
for a more flexible functional form by estimating fractional
polynomial specifications that permit the data to determine
the shape of the relationship between size and the NIE ratio.
An alternative to regular polynomials, fractional polynomials
provide flexible parameterization for continuous variables. We
use the Stata function fracpoly to determine an optimal polynomial specification (optimal polynomial) and also estimate a
specification with exponents ranging from -2 to 2—that is, log
assets raised to the -2, -1, 0, 1, and 2 power (flex polynomial).
These best-fit polynomials are shown in Chart 3 along with
the ordinary least squares line of best fit.
Overall, the log-linear functional form assumed in Table 4
appears to be a good approximation, although we note that,
based on point estimates, the point-estimated relationship
between log assets and the efficiency ratio is somewhat concave at the tails. Specifically, the relationship between BHC
size and the NIE ratio is relatively flat among small BHCs
(those with assets below $150 million), while the relationship
is steeper among the largest BHCs (those with assets above
$750 billion). For the vast range of asset sizes, the relationship
between log size and efficiency ratio is close to linear, and the
95 percent confidence interval of the alternative forms is very
similar. Thus, we use a log-linear specification for the remainder of the analysis.
In addition to investigating flexible polynomial specifications, we separate the sample into different size cohorts,
re-sorted in each quarter, and estimate separate specifications
for each cohort. This approach allows the relationship between
NIE and control variables, as well as size, to vary by BHC size
class. (In the fractional polynomial approach, the coefficients
on explanatory variables other than size are unrelated to size.)
Each column of Table 5 represents specifications 1, 2, and 10
of Table 4 estimated on a subset of the BHCs sorted by size in
16
Do Big Banks Have Lower Operating Costs?
-20
-30
-40
10
Ordinary least squares (OLS)
Optimal polynomial
Flex polynomial (flex)
95% confidence interval flex
95% confidence interval OLS
12
14
16
18
Size (log of assets in $000s)
20
22
Source: Authors’ calculations, based on statistical analysis of
FR Y-9C data.
Note: Functional forms are partial predictions based on varying log of
assets ($000s), holding other covariates fixed at their sample means.
The efficiency ratio is normalized to be equal to zero for a bank
holding company with $10 billion in assets.
each year. The first column replicates the results on the entire
sample, for comparison. Without including controls for BHC
asset mix, it appears that much of the coefficient on size is
driven by BHCs below the median asset size (column 6). As
additional controls are included, economies of scale become
apparent in many of the size cohorts. In the specification
including all controls, the estimated coefficient on size is negative in all cohorts and statistically significant. As suggested
by the flexible polynomial specifications, the point estimate
coefficient on size is largest in the top 1 percent of the sample.
What do these findings imply for the policy debate around
size limits for the largest BHCs? We find no evidence that the
inverse relationship between size and operating costs disappears above any particular size threshold; indeed, our point
estimates suggest that, if anything, the relationship is steeper
among the largest firms. This result is consistent with scale
economies from sources other than bargaining power to the
extent that we believe that differences in bargaining power
may be small within the top 1 percent of BHCs. The statistical
precision of our estimates is limited, however, given the small
number of observations for the largest BHCs.
Table 5
Coefficient on Log Assets, by Size Cohort
Table 4, Specification (1)
Table 4, Specification (2)
Table 4, Specification (10)
N
(1)
(2)
(3)
(4)
(5)
(6)
All
Top 1%
95-99%
75-95%
50-75%
Bottom 50%
Controls
Time fixed effects
-1.320***
1.860
1.273
-1.790**
-0.768
-6.140***
(0.235)
(1.647)
(1.164)
(0.687)
(1.509)
(1.633)
-1.892***
-2.864
-0.379
-1.888**
-1.914
-3.195*
(0.228)
(2.020)
(1.278)
(0.674)
(1.352)
(1.334)
-4.151***
-8.018*
-5.138***
-4.132***
-4.238***
-5.055***
(0.326)
(3.931)
(1.442)
(0.696)
(1.204)
(1.311)
58,217
604
2,405
12,197
15,181
27,830
Asset shares
All controls
Source: Authors’ calculations.
Notes: The table presents an analysis of the relationship between size, measured by the log of total assets (lagged by one quarter), and efficiency
ratio, defined as total noninterest expense as a percentage of net operating revenue. Each row represents the coefficient on size for specifications
(1), (2), and (10) of Table 4, estimated on a subset of bank holding companies sorted by size in each quarter. Specification (1) includes time fixed effects.
Specification (2) includes time fixed effects as well as controls for the percentage of assets in each broad category (asset shares). Specification (10) includes the
controls from specification (2) as well as controls for types of loans, revenue composition, funding structure, business concentration, organizational complexity, and headquarters state fixed effects. Robust standard errors reported in parentheses are clustered by bank holding company and quarter.
*** p<0.01
** p<0.05
* p<0.1
4.3 Alternative Measures of Operating Costs
The efficiency ratio may be distorted in periods when net
operating income is temporarily low.11 Next, we test the sensitivity of our results to other normalizations of noninterest
expense: the expense asset ratio discussed in Section 3 (NIE /
total assets), NIE / risk-weighted assets, and a “cash” efficiency
ratio, which excludes noncash expenses such as goodwill
amortization in the numerator. We do this because noncash
expenses are often associated with one-time costs relating to
mergers and acquisitions that are not likely to persist, and may
be associated with size. We also estimate a specification using
the log of noninterest expense as an alternative measure of
operating costs.
As before, for each normalization of NIE, we re-estimate
specifications with the set of right-hand-side variables from
columns 1, 2, and 10 of Table 4 and present the coefficient on
asset size. Results are presented in Table 6. Regardless of the
normalization used, the coefficient on size is negative and
statistically significant once BHC controls are included. In
the specification including all controls, the estimated coeffi11
During the 2007-08 financial crisis, trading losses and other losses brought
net operating income close to zero for several large BHCs.
cient on size is approximately 7 to 10 percent of the average
expense ratio.
For the specifications using the log of noninterest expense
as the dependent variable, the coefficient on log assets can be
directly interpreted as the elasticity of operating costs with
respect to size. In line with our other results, this elasticity is
less than unity—in other words, a 10 percent change in BHC
size is associated with a less than 10 percent change in NIE
operating costs, a finding consistent with the presence of scale
economies in operating costs. For the specification including
all controls, the operating cost elasticity is 0.899, much smaller
than one, although it is significantly closer to one for the
specification just including asset controls (0.979). Both these
estimates are statistically significantly smaller than unity.
5. Decomposition of Noninterest
Expense
This section examines the relationship between BHC size and
components of noninterest expense. First, we consider the
five major components of noninterest expense reported in the
Y-9C income statement. Probing more deeply, we then analyze
FRBNY Economic Policy Review / December 2014
17
Table 6
Alternative Measures of Operating Costs
Noninterest Expense/
Risk-Weighted Assets
Table 4, Specification:
Log assets
(1)
0.007
(2)
-0.044*** -0.115***
(0.010) (0.011)
Asset share controls
(10)
Yes
All controls
(0.013)
Noninterest Expense/
Assets
(1)
(2)
(10)
0.003 -0.018** -0.083***
-1.686***
-2.239***
-4.339***
0.993***
(0.006) (0.007)
(0.231)
(0.217)
(0.303)
(0.007)
Yes
Yes
R2
0.016
0.231
N
58,217
58,217
0.487
58,217
(10)
Log Noninterest Expense
(1)
Yes
(2)
Cash Noninterest Expense/
Net Revenue (Cash Efficiency Ratio)
(0.009)
Yes
Yes
Yes
0.007
0.171
58,217 58,217
(1)
Yes
(2)
0.979***
(0.007)
0.078
0.208
0.325
58,217
58,217
58,217
58,217
0.899***
(0.008)
Yes
Yes
0.949
0.968
58,192
58,192
Yes
0.430
(10)
Yes
0.935
58,192
Source: Authors’ calculations.
Note: The table presents an analysis of the relationship between size, measured by the log of total assets (lagged by one quarter), and different measures of
efficiency. The dependent variables in the first three specifications are cash efficiency ratio, defined as total noninterest expense less goodwill impairment
and amortization expense over net operating revenue; in the next three specifications, NIE/assets ratio, defined as total noninterest expense (NIE) over total
assets; and in the final three specifications, NIE/RWA ratio, defined as total noninterest expense over total risk-weighted assets (RWA). For each alternative
measure of efficiency ratio, specifications (1), (2) and (10) of Table 4 are presented. Specification (1) includes controls for quarter fixed effects. Specification (2) includes the controls from specification (1) as well as controls for the percentage of assets in each broad category. Specification (10) includes the controls from specification (2) as well as controls for types of loans, revenue composition, funding structure, business concentration, organizational complexity,
and headquarters state fixed effects. Models are estimated with robust standard errors and two-way clustering by firm and quarter.
*** p<0.01
** p<0.05
* p<0.1
the nine subcomponents of “other noninterest expense,” using
our manual classification of these expenses as described in
Section 3.
One goal of this disaggregated analysis is to shed additional
light on the sources of the lower operating costs enjoyed by
large BHCs. Although these lower costs could be due to scale
economies or other efficiency benefits of size, they could also
reflect implicit government guarantees for large BHCs, or the
greater bargaining power of these firms. For example, large
banks may endogenously select riskier activities, but invest
less in risk management because of implicit insurance associated with being “too big to fail.” Alternatively, large banks may
simply take advantage of greater bargaining power to reduce
expenses. These different explanations have very different normative welfare implications. Efficiency benefits of size imply
that limiting size would impose deadweight economic costs,
while explanations relating to bargaining power and TBTF
primarily relate to the allocation of economic rents. Although
the breakdown of expenses in the Y-9C does not allow us to
fully disentangle these different explanations, we are able to
draw some suggestive conclusions.
18
Do Big Banks Have Lower Operating Costs?
5.1 Major Components of Noninterest
Expense
We begin by studying the five expense categories reported
on Schedule HI: compensation (49.4 percent of noninterest
expense), premises and fixed assets expense (11.6 percent),
goodwill impairment (1.8 percent), amortization (1.9 percent), and other (35.0 percent). Results are presented in
Table 7. As before, we normalize each expense by net operating revenue, and for parsimony, focus on the coefficient on log
assets for specifications 1, 2, and 10 from Table 4.
Each of the three largest categories of noninterest expense
declines as a percentage of net revenue as size increases, all
else equal, with or without the inclusion of controls for BHC
characteristics. The final column of the table presents the
estimated coefficient scaled by the mean of the dependent
variable in question (that is, an elasticity of the component
efficiency ratio with respect to firm size). Focusing on the specifications including these controls (either for asset composition
alone, or for all controls), we find that the inverse relationship
between BHC size and scaled noninterest expense is steepest
for compensation, followed by other noninterest expense, based
on this calculated elasticity. For the specifications including
Table 7
Bank Holding Company Size and the Efficiency Ratio, by Component of Noninterest Expense
Total noninterest expense
Table 4
Specification
Log
Assets
Standard Significance
Error
Level
Adjusted
Mean
R2
(Percent)
1
-1.320
(0.235)
***
0.080
2
-1.892
(0.228)
***
0.195
10
-4.151
(0.326)
***
0.296
66.32
Controls
Coefficient/ Mean
(Percent)
Time FE
-1.99
Asset shares
-2.85
All
-6.26
Components of noninterest expense
Compensation
Premises and fixed assets
Other
Amortization expense
Goodwill impairment
1
-1.135
(0.126)
***
0.048
2
-1.472
(0.133)
***
0.103
10
-2.385
(0.175)
***
0.242
1
-0.265
(0.045)
***
0.025
2
-0.103
(0.048)
*
0.135
10
-0.365
(0.073)
***
0.257
1
-0.283
(0.127)
*
0.111
2
-0.658
(0.125)
***
0.256
10
-1.585
(0.167)
***
0.354
1
0.181
(0.016)
***
0.077
2
0.164
(0.018)
***
0.106
10
0.159
(0.024)
***
0.163
1
0.044
(0.015)
**
0.031
2
0.042
(0.014)
**
0.032
10
0.017
(0.011)
32.44
7.64
23.20
1.29
1.46
0.039
Time FE
-3.50
Asset shares
-4.54
All
-7.35
Time FE
-3.47
Asset shares
-1.35
All
-4.78
Time FE
-1.22
Asset shares
-2.84
All
-6.83
Time FE
14.00
Asset shares
12.68
All
12.29
Time FE
3.01
Asset shares
2.88
All
1.16
Components of other noninterest expense
Corporate overhead
Information technology and data processing
Consulting and advisory
Legal
Retail banking
FDIC assessments and other government
Other financial services
1
-0.002
(0.073)
2
-0.212
(0.063)
***
0.074
0.018
Time FE
-0.04
Asset shares
-4.45
10
-0.334
(0.074)
***
0.212
All
-7.00
1
-0.106
(0.044)
*
0.006
Time FE
-3.28
Asset shares
-4.64
All
-6.59
Time FE
9.92
2
-0.150
(0.054)
**
0.023
10
-0.213
(0.068)
**
0.139
1
0.285
(0.047)
***
0.069
2
0.208
(0.053)
***
0.097
10
0.053
(0.054)
1
0.006
(0.035)
7.24
0.210
All
1.84
0.008
Time FE
0.33
2
-0.022
(0.034)
-0.118
(0.045)
**
0.263
0.141
***
0.017
1
-0.225
(0.058)
2
-0.068
(0.087)
0.108
10
-0.205
(0.118)
1
-0.249
(0.048)
***
2
-0.103
(0.042)
*
10
-0.036
(0.068)
1
0.038
(0.019)
-0.022
(0.011)
10
-0.058
(0.017)
3.23
Asset shares
10
2
4.77
2.87
1.79
Time FE
-13.59
-4.11
All
-12.38
0.242
Time FE
-16.51
Asset shares
-6.83
All
-2.39
1.51
0.009
0.146
***
-6.57
0.208
0.536
*
-1.23
All
Asset shares
0.393
1.66
Asset shares
0.211
0.78
Time FE
4.86
Asset shares
-2.81
All
-7.42
FRBNY Economic Policy Review / December 2014
19
Table 7 (continued)
Bank Holding Company Size and the Efficiency Ratio, by Component of Noninterest Expense
Directors’ fees and other compensation
Miscellaneous
Unclassified other noninterest expenses
Table 4
Specification
Log
Assets
Standard
Error
Significance
Level
Adjusted
R2
1
-0.142
(0.012)
***
0.095
2
-0.182
(0.015)
***
0.139
10
-0.190
(0.019)
***
0.259
1
0.026
(0.014)
0.002
2
0.017
(0.017)
0.010
10
-0.004
(0.022)
1
-0.129
(0.115)
2
-0.063
(0.102)
10
-0.289
(0.134)
0.06
Controls
Coefficient/ Mean
(Percent)
Time FE
-221.31
Asset shares
-283.65
All
-296.12
Time FE
3.68
0.042
All
-0.87
0.004
Time FE
-1.48
Asset shares
-0.72
All
-3.32
0.229
0.46
5.62
Asset shares
0.147
*
Mean
(Percent)
8.72
Source: Authors' calculations.
Notes: The table presents an analysis of the relationship between size, measured by the log of total assets (lagged by one quarter), and the components of
noninterest expense normalized by net operating revenue. The first nineteen rows present the specifications for NIE and its large components: compensation, premises and fixed assets, other, amortization expense, and goodwill impairment. The remaining rows present three specifications each for the nine
subcomponents of other, as well as for unclassified expense, the total other noninterest expense less the nine constructed components of other noninterest
expense. All noninterest expense components are normalized by net operating revenue. Each row presents specifications (1), (2), and (10) of Table 4 for each
main component of noninterest expense. Specification (1) includes time fixed effects. Specification (2) includes time fixed effects as well as controls for the
percentage of assets in each broad category (asset shares). Specification (10) includes the controls from specification (2) as well as controls for types of loans,
revenue composition, funding structure, business concentration, organizational complexity, and headquarters state fixed effects. See Appendix A for further
detail. The sample mean for each component is presented, and the final column is the estimated coefficient on size normalized by the sample mean for the
NIE component. Robust standard errors reported in parentheses are clustered by BHC and quarter. FE is fixed effects; FDIC is Federal Deposit Insurance
Corporation.
*** p<0.01
** p<0.05
* p<0.1
all controls, a 10 percent increase in size is associated with a
0.735 percent decline in compensation scaled by net operating revenue and a 0.683 percent decline in the corresponding
ratio for other noninterest expense. The result for employee
compensation is perhaps surprising, given that large BHCs
have more employees in highly compensated roles such as
investment banking and trading. However, the higher productivity and additional revenue earned by these employees
(the denominator of the efficiency ratio) appears to offset this
higher compensation.
Expenses related to premises and fixed assets may represent a category of operating costs for which scale efficiencies are lower (for example, building lease costs are roughly
proportionate to the size of the leased space, at least within a
specific geographic area). Given this, it is perhaps unsurprising that estimated economies of scale are smaller for premises
and fixed assets expense: for this category, our point estimate
20
Do Big Banks Have Lower Operating Costs?
implies that a 10 percent increase in size is associated with a
0.478 percent decline in expenses scaled by operating revenue.
Significantly, expenses related to the impairment and
amortization of goodwill and other intangible assets are
actually proportionately higher for large firms—a fact that
distinguishes these expenses from the other categories. We
estimate a positive, statistically significant (in most specifications) coefficient on these expenses. The likely key reason
for this finding is that large BHCs often have grown by way of
acquisitions, which will sometimes result in goodwill when
the acquisition purchase price exceeds the tangible book value
of assets purchased. Consequently, these firms report higher
expenses related to the amortization or impairment of these
assets. Although the positive slope for these two expense categories is economically significant, the two categories together
make up only a relatively small proportion (3.7 percent) of
total industry NIE.
5.2 Subcomponents of Other Noninterest
Expense
In this section, we examine the nine subcomponents of “other
NIE” identified in section 3.2. (Recall that these categories
reflect both standardized memoranda items reported on the
Y-9C since 2008 and “write-in” text strings classified by us.)
Previous work estimating scale curves for these disaggregated
categories has been based on case studies or has had limited
sample size (for example, Clearing House Association [2012]).
Overall, we find evidence that scaled expense falls with size
for most, but not all, components of other noninterest expense,
especially after including controls for BHC asset and income
composition. When controls for the composition of assets and
income sources are included in the specification, large BHCs
exhibit lower expenses in categories in which a fixed cost can be
spread across an expanded scale of operations, such as corporate overhead, information technology, and data processing.
The lower part of Table 7 presents results for the other
NIE components, listed in descending order of size. Corporate overhead is the largest component of other noninterest
expense, and a component for which we estimate significant
scale efficiencies (a high estimated coefficient on size relative to mean level of expense). Corporate overhead includes
expenses such as accounting and auditing, advertising and
marketing, treasury expenses, travel and business development, charitable donations, insurance, and utilities. These
expenses appear to have significant operational leverage; the
estimated coefficient on size is -0.33, approximately 7 percent
of the mean level of corporate overhead expenses.
Similar scale economies are observed for expenses associated with information technology and data processing, with
an estimated coefficient on size that is -6.6 percent of mean IT
expense. This finding is consistent with the view that spreading overhead expenses associated with technology may be one
source of cost advantage for large banking firms.
In contrast to these two categories, we find that expenses
associated with consulting and advisory services are proportionately higher for large BHCs. Prior to adding controls for
BHC characteristics, our estimates show that the coefficient
on size and consulting expenses is positive and statistically
significant. This coefficient remains significant when asset
composition controls are included, although once all controls
are included, the coefficient is positive but no longer statistically
significant. This suggests that consulting and advisory services
may be related to noninterest income, rather than to the composition of BHC assets. Despite recent publicity surrounding
large BHCs’ legal issues and large-dollar-value settlements,
over the 2008-12 period, legal expenses also increase less than
proportionately with BHC size, particularly in the specification
including the full set of controls (specification 10 from Table
4). This expense category includes both legal fees and retainers
paid for legal services performed, as well as expenses associated
with legal settlements and reserves, to the extent we can identify
these expenses from the write-in text fields. Some part of this
finding may reflect the fact that small banks may lack internal
legal departments, for which expenses would be recorded as
part of compensation, and thus have higher external legal fees.
The assignment of write-in fields to retail banking requires
perhaps the most judgment on our part. This category
includes collection expenses, credit reports, mortgage-related
expenses such as appraisal and title fees, branch expenses,
checks, lockboxes, and robbery, among many others. After
including asset composition controls, the estimated coefficient remains negative although not statistically significant.
This result may reflect the wide variation in the types of retail
banking businesses that are not well captured by our BHC
characteristics. Alternatively, economies of scale may be
limited or not present for branch banking (at least among the
set of expenses classified into this category), since many costs
only scale until the next branch is opened.
Similarly, we find a negative but statistically insignificant
relationship between size and normalized FDIC assessments
and other government-related expenses after including the
full complement of BHC characteristics. The majority of the
expenses in this line item are due to deposit insurance, and
thus it would be surprising to uncover economies of scale
once we control for the amount of deposit financing. This
coefficient would likely shrink further if our regression specification included a control for the fraction of insured deposits,
rather than total deposits.
The category “other financial services” represents the sum
of expenses associated with BHCs’ non-banking businesses,
such as asset management, trust and custody services, and
insurance. Given likely differences in the noninterest expenses
of these businesses, it is not surprising that the estimated
coefficient changes sign from positive to negative once we
control for the composition of BHCs’ assets and noninterest
income. Banking firms that earn a high percentage of income
from fee income should naturally have higher expenses. But
holding all else equal and controlling for income composition,
we find that larger BHCs have lower scaled expenses in this
category: we estimate a coefficient of 7.4 percent of the mean
value. This result is consistent with cost economies of scale in
noncompensation expenses associated with businesses such as
insurance and asset management.
FRBNY Economic Policy Review / December 2014
21
The component of other noninterest expense for which
scale economies are largest in percentage terms is directors’
fees and other compensation. For this category, the coefficient
on size is almost three times as large as the sample mean.
This makes intuitive sense; even though directors of large
BHCs have higher compensation, board size does not increase
dramatically with firm size. This coefficient is negative and
significant regardless of the set of controls used.
Miscellaneous expenses include items as varied as expenditures for cattle feed and reducing gold to market. It also
includes nonspecific write-in text fields such as “miscellaneous expense,” “miscellaneous fee,” and “other expense.”
Regardless of the controls for bank businesses used, we
do not see economies of scale in these varied expenses,
although some economies may exist in the residual category
“other expenses,” which includes all noninterest expenses not
otherwise classified.
6. Conclusion
We find a robust inverse relationship between the size of bank
holding companies and scaled measures of operating costs.
Quantitatively, a 10 percent increase in assets is associated
with a 0.3 to 0.6 percent decline in noninterest expense scaled
by income or assets, depending on the specification. In dollar
terms, our estimates imply that for a BHC of mean size in our
sample, an additional $1 billion in assets reduces noninterest
expense by $1 million to $2 million per year, relative to a base
case where operating cost ratios are unrelated to size. This
inverse relationship is robust to various changes in model
specification, although the magnitude of the relationship is
sensitive to the set of controls used.
22
Do Big Banks Have Lower Operating Costs?
Unpacking our results, we find that while size is associated
with lower scaled operating costs for most components of
noninterest expense, the largest contributions in dollar terms
come from employee compensation, premises and fixed assets,
corporate overhead, and information technology and data
processing. While not a large component of total noninterest
expense, directors’ fees and other compensation account for
the largest proportionate savings, presumably a reflection of
the fact that corporate boards do not expand with firm size,
even if their members are better paid on average.
Our results likely reflect a combination of three factors: First,
large BHCs benefit from “operational leverage” or economies
of scale, whereby they effectively spread costs over a higher
revenue or asset base. Second, “X-efficiency”—a factor closely
related to operational leverage—may be higher for large BHCs;
that is, these firms may operate closer to the production frontier
on average. Third, large BHCs may have greater bargaining
power than smaller firms with suppliers or employees. We are
not able to pin down with confidence the relative contribution
of these three factors. We emphasize, however, that the inverse
relationship between BHC size and scaled measures of NIE is
not limited to particular components of expense or particular
segments of the BHC size distribution.
Consistent with recent research that identifies the presence of scale economies in banking, our results suggest that
imposing size limits on banking firms would be likely to
involve real economic costs. Although the limitations of our
econometric methodology must be borne in mind, a backof-the-envelope calculation applied to our estimates implies
that limiting BHC size to be no larger than 4 percent of GDP
would increase total noninterest expense by $2 billion to
$4 billion per quarter. These costs should be weighed against
the potential benefits of size limits as policymakers address
the “too-big-to-fail” problem.
Appendix A: Variable Definitions
Income Statement Variables
Variable
Definition
Y-9C Mnemonic Construction/Variable Source
Net interest income
bhck4074 [1981:Q2 - present]
Noninterest income
bhck4079 [1981:Q2 - present]
Trading revenue
Includes the net gain or loss from trading cash
instruments and off-balance-sheet derivative contracts
(including commodity contracts) that has been recognized during the calendar year-to-date
bhcka220 [1996:Q1 - present]
Fiduciary income
Includes income from fiduciary activities, fees and
commissions from annuity sales, underwriting income
from insurance and reinsurance activities, and income
from other insurance activities
bhck4070 + bhckb494 [2001:Q1 - 2002:Q4],
bhck4070 + bhckc386 + bhckc387 [2003:Q1 2006:Q4], bhck4070 + bhckc887 + bhckc385 +
bhckc387 [2007:Q1 - present]
Investment banking income
Includes venture capital revenue, fees and commissions
from securities brokerage, and investment banking, advisory, and underwriting fees and commissions
bhck b491 + bhckb490 [2001:Q1 - 2006:Q4],
bhckb491 + bhckc886 + bhckc888 [2007:Q1 present]
Service charges on deposits
Service charges on deposit accounts in domestic offices
bhck4884 [1981:Q2 - present]
Net servicing fees
Includes income from servicing real estate mortgages,
credit cards, and other financial assets held by others
bhckb492 [2001:Q1 - present]
Other income
Total noninterest income not accounted for in the five
categories listed above
Derived
Net interest income plus noninterest income
bhck4074 + bhck4079 [1981:Q2 - present]
Net operating revenue
Noninterest expense
Compensation
bhck4093 [1981:Q2 - present]
Salaries and employee benefits
bhck4135 [1981:Q2 - present]
Premises and fixed assets
bhck4217 [1981:Q2 - present]
Amortization expense
Amortization expense and impairment losses for other
intangible assets
bhckc232 [2002:Q1 - present]
Goodwill impairment
Goodwill impairment losses
bhckc216 [2002:Q1 - present]
Other
Total noninterest expense not accounted for in the four
categories listed above
Derived
Eleven standardized other noninterest expense items
reported in Schedule HI: Memoranda of the FR Y-9C beginning either in 2002 or in 2008. BHC filers only report
amounts greater than $25,000 that exceed 3 percent of
total other noninterest expense
bhckc017 [2002:Q1 - present]
Data processing expenses
Advertising and marketing expenses
Directors’ fees
Printing, stationery, and supplies
bhck0497 [2002:Q1 - present]
bhck4136 [2002:Q1 - present]
bhckc018 [2002:Q1 - present]
Postage
bhck8403 [2002:Q1 - present]
Legal fees and expenses
bhck4141 [2002:Q1 - present]
FDIC deposit insurance assessment
bhck4146 [2002:Q1 - present]
Accounting and auditing expenses
bhckf556 [2008:Q1 - present]
Consulting and advisory expenses
bhckf557 [2008:Q1 - present]
ATM and interchange expenses
bhckf558 [2008:Q1 - present]
Telecommunications expenses
bhckf559 [2008:Q1 - present]
TEXT8565
TEXT8566
TEXT8567
Description of the “write-in” components of other noninterest expense. BHCs only report amounts that exceed 10
percent of total other noninterest expense
bhck8565 [1994:Q1 - present]
bhck8566 [1994:Q1 - present]
bhck8567 [1994:Q1 - present]
FRBNY Economic Policy Review / December 2014
23
Appendix A:
(Continued)
Variable Definitions (continued)
Consolidated Balance Sheet Variables
Variable
Definition
Total assets
bhck2170 [1991:Q1 - present]
Total loans
bhck2122 [1991:Q1 - present]
Residential real estate loans
The sum of 1) all other loans secured by one-to-four-family residential properties: secured by first liens; 2) all other loans secured by
one-to-four-family residential properties: secured by junior liens;
3) revolving, open-end loans secured by one-to-four-family residential properties and extended under lines of credit
bhdm1797 + bhdm5367 + bhdm5368
[1991:Q1 - present]
Commercial real estate loans
The sum of 1) one-to-four-family residential construction loans;
2) other construction loans and all land development and other
land loans; 3) real estate loans secured by multifamily (five or more)
residential properties; 4) loans secured by owner-occupied nonfarm
nonresidential properties; 5) loans secured by other nonfarm nonresidential properties
bhdm1415 + bhdm1460 + bhdm1480
[1990:Q3 - 2006:Q4], bhckf158 + bhckf159 +
bhdm1460 + bhckf160 + bhckf161 [2007:Q1 present]
Credit card loans
Loans to individuals for household, family, and other personal
expenditures (that is, consumer loans). Includes purchased paper:
credit cards
bhck2008 [1991:Q1-2000:Q4], bhckb538
[2001:Q1 - present]
Other consumer loans
The sum of 1) loans to individuals for household, family, and other
personal expenditures—that is, consumer loans (includes purchased
paper): other revolving credit plans; 2) automobile loans to individuals for household, family, and other personal expenditures—that is,
consumer loans (includes purchased paper); 3) other consumer loans
to individuals, for household, family, and other personal expenditures
(includes single payment, installment, and all student loans)
bhck2011 [1991:Q1 - 2000:Q4], bhck2011 +
bhckb539 [2001:Q1 - 2010:Q4], bhckb539 +
bhckk137 + bhckk207 [2011:Q1 - present]
All other loans
Total loans minus the sum of residential real estate loans, commercial derived
real estate loans, credit card loans, and other consumer loans
Cash and balances due
from depository institutions
The sum of 1) non-interest-bearing balances and currency and coin;
2) interest-bearing balances in U.S. offices; 3) interest-bearing balances
in foreign offices, edge and agreement subsidiaries, and international
banking facilities
bhck0081 + bhck0395 + bhck0397 [1991:Q1 present]
Trading assets
Assets held in trading accounts include but are not limited to U.S.
Treasury securities; U.S. government agency and corporation
obligations; securities issued by states and political subdivisions in
the United States; other bonds, notes, and debentures; certificates of
deposit; commercial paper; and bankers acceptances. Assets held in
trading accounts also include the amount of revaluation gains from
the “marking to market” of interest rate, foreign exchange rate, and
other off-balance-sheet commodity and equity contracts held for
trading purposes
bhck2146 [1981:Q2 - 1994:Q4], bhck3545
[1995:Q1 - present]
Federal funds and repurchase
agreements
The sum of 1) outstanding amount of federal funds sold—that is, immediately available funds lent (in domestic offices) under agreements
or contracts that have an original maturity of one business day or roll
over under a continuing contract, excluding such funds lent in the
form of securities purchased under agreements to resell and overnight lending for commercial and industrial purposes; 2) securities
resale agreements, regardless of maturity, if the agreement requires
the bank to resell the identical security purchased or a security that
meets the definition of substantially the same in the case of a dollar
roll, and purchases of participations in pools of securities, regardless
of maturity
bhck1350 [1981:Q2 - 1988:Q1][1997:Q1 2001:Q4], bhck0276 + bhck0277
[1988:Q2 - 1996:Q4], bhdmb987 + bhckb989
[2002:Q1 - present]
Investment securities
Held-to-maturity securities (at amortized cost) plus available for sale
securities (at fair value)
bhck0390 [1981:Q2 - 1993:Q4], bhck1754 +
bhck1773 [1994:Q1 - present]
Other real estate owned
The book value (not to exceed fair value), less accumulated depreciation, if any, of all real estate other than bank premises actually owned
by the bank and its consolidated subsidiaries.
bhck2150[1981:Q2-1990:Q2][2001:Q1 - present],
bhck2744 + bhck2745 [1990:Q3 - 2000:Q4]
Premises and fixed assets
24
Y-9C Mnemonic Construction/Variable Source
Do Big Banks Have Lower Operating Costs?
bhck2145
Appendix A:
(Continued)
Variable Definitions (continued)
Consolidated Balance Sheet Variables
Variable
Definition
Y-9C Mnemonic Construction/Variable Source
Investments in unconsolidated
subsidiaries and associated
companies
Includes the amount of the bank holding company’s investments in
subsidiaries that have not been consolidated; associated companies;
and corporate joint ventures, unincorporated joint ventures, general
partnerships, and limited partnerships over which the bank exercises
significant influence (collectively referred to as “investees”). Also
includes loans and advances to investees and holdings of their bonds,
notes, and debentures
bhck2130 - bhck3656 [1981:Q2 - 2009:Q1],
bhck2130 [2009:Q2 - present]
Investments in real estate
ventures
The book value of direct and indirect investments in real estate
ventures
bhck3656 [1981:Q2 - present]
Intangible and other assets
Other identifiable intangible assets plus other assets
bhck3165 + bhck2160 + bhck2155
[1985:Q2 - 1991:Q4], bhck3164 + bhck5506 +
bhck5507 + bhck2160 + bhck2155 [1992:Q1 1998:Q4], bhck0426 + bhck2160 + bhck2155
[2001:Q1 - 2005:Q4], bhck0426 + bhck2160
[2006:Q1 - present]
Nonperforming loans
The sum of 1) total loans and leasing financing receivables that are
ninety days or more past due and still accruing; 2) total loans and
leasing financing receivables in nonaccrual status.
bhck5525 - bhck3506 + bhck5526 - bhck3507
[1990:Q3 - present]
Risk-weighted assets
BHC risk-weighted assets net of all deductions
bhcka223 [1996:Q1 - present]
Total deposits
1) Non-interest-bearing deposits 2) total interest-bearing deposits in
foreign and domestic offices
bhdm6631 + bhdm6636 + bhfn6631 + bhfn6636
[1981:Q2 - present]
Other Characteristics and Organizational Structure Variables
Variable
Definition
Y-9C Mnemonic Construction/Variable Source
Public
Dummy=1 if firm has PERMCO, Dummy=0 otherwise
Federal Reserve Bank of New York. 2013. CRSP-FRB Link
Number of subsidiaries
Total number of offspring entities whose relationship to the bank
holding company is regulated, that is, governed by applicable
banking statutes, which are either federal or state banking laws
NIC Top Holder Table: top holder variable rssd9003
Foreign subsidiaries
Total number of offspring entities that are not domiciled in the
United States
NIC Country Name Directory: domestic indicator rssd9101
Foreign parent
Dummy=1 if the highest entity in the organization is not
domiciled in the United States, Dummy=0 otherwise
NIC Board Derived Items Table: foreign family ID rssd9360
Source: Board of Governors of the Federal Reserve System, Microdata Reference Manual.
Note: BHC is bank holding company; FDIC is Federal Deposit Insurance Corporation; CRSP is Center for Research in Securities Prices; NIC is National
Information Center.
Note to Readers:
Appendix B, “Additional Materials,” is available as a separate file at http://www.newyorkfed.org/research/
epr/2014/1403kovn_appendixB.pdf.
FRBNY Economic Policy Review / December 2014
25
References
Avraham, D., P. Selvaggi, and J. Vickery. 2012. “A Structural View
of U.S. Bank Holding Companies.” Federal Reserve Bank of
New York Economic Policy Review 18, no. 2: 65-81.
DeYoung, R. 1998. “Management Quality and X-Inefficiency in
National Banks.” Journal of Financial Services Research 13,
no. 1 (February): 5-22.
Benston, G. 1972. “Economies of Scale of Financial Institutions.”
Journal of Money, Credit, and Banking 4 (May): 312-41.
Feng, G., and A. Serletis. 2010. “Efficiency, Technical Change, and
Returns to Scale in Large U.S. Banks: Panel Data Evidence from
an Output Distance Function Satisfying Theoretical Regularity.”
Journal of Banking and Finance 34, no. 1 (January): 127-38.
Benston, G., G. A. Hanweck, and D. B. Humphrey. 1982. “Scale Economies
in Banking: A Restructuring and Reassessment.” Journal of Money,
Credit, and Banking 14 (November): 435-56.
Berger, A., and D. Humphrey. 1991. “The Dominance of Inefficiencies
over Scale and Product Mix Economies in Banking.” Journal of
Monetary Economics 28, no. 1 (August): 117-48.
———. 1994. “Bank Scale Economies, Mergers, Concentration, and
Efficiency: The U.S. Experience.” Wharton Financial Institutions
Center Working Paper no. 94-25.
Berger, A., W. Hunter, and S. Timme. 1993. “The Efficiency of
Financial Institutions: A Review and Preview of Research Past,
Present, and Future.” Journal of Banking and Finance 17,
no. 2-3 (April): 221-49.
Bernanke, B. 2013. Semiannual Monetary Policy Report to the
Congress. Testimony before the Committee on Financial
Services, U.S. House of Representatives, Washington, D.C.,
July 17, and before the Committee on Banking, Housing, and
Urban Affairs, U.S. Senate, Washington, D.C., July 18.
Clearing House Association, L.L.C. 2011. Understanding the
Economics of Large Banks. TCH Research Study, November.
Copeland, A. 2012. “Evolution and Heterogeneity among Larger
Bank Holding Companies: 1994- 2010.” Federal Reserve Bank of
New York Economic Policy Review 18, no. 2: 83-93.
Cornett, M., J. McNutt, and H. Tehranian. 2006. “Performance
Changes around Bank Mergers: Revenue Enhancements versus
Cost Reductions.” Journal of Money, Credit, and Banking 38,
no. 4 (June): 1013-50.
Davies, R., and B. Tracey. 2014. “Too Big to Be Efficient? The Impact
of Implicit Subsidies on Estimates of Scale Economies for Banks.”
Journal of Money, Credit, and Banking 46, no. s1 (February):
219-53.
26
Do Big Banks Have Lower Operating Costs?
Goetz, M., L. Laeven, and R. Levine. 2013. “Identifying the Valuation
Effects and Agency Costs of Corporate Diversification: Evidence
from the Geographic Diversification of U.S. Banks.” Review of
Financial Studies. 26, no. 7: 1787-1823.
Hannan, T., and S. Pilloff. 2006. “Acquisition Targets and Motives in
the Banking Industry.” Board of Governors of the Federal Reserve
System Finance and Economics Discussion Series, no. 2006-40.
Hughes J., and L. Mester. 2013. “Who Said Large Banks Don’t
Experience Scale Economies? Evidence from a Risk-ReturnDriven Cost Function.” Journal of Financial Intermediation
22, no. 4 (October): 559-85.
Hughes, J. P., W. Lang, L. J. Mester, and C. G. Moon. 2001. “Are Scale
Economies in Banking Elusive or Illusive? Evidence Obtained by
Incorporating Capital Structure and Risk-Taking into Models of
Bank Production.” Journal of Banking and Finance 25, no. 12
(December): 2169-208.
Johnson, S., and J. Kwak. 2010. Thirteen Bankers: The Wall
Street Takeover and the Next Financial Meltdown.
New York: Pantheon.
Kwan, S., and J. Wilcox. 2002. “Hidden Cost Reductions in Bank
Mergers. Accounting for More Productive Banks.” In J. Kensigner,
ed., Research in Finance 19, 109-24. Bingley, U.K.: Emerald
Group Publishing Limited.
Laeven, L., and R. Levine. 2007. “Is There a Diversification
Discount in Financial Conglomerates?” Journal of Financial
Economics 85, no. 2 (August): 331-67.
Niepmann, F. 2013. “Banking across Borders with Heterogeneous
Banks.” Federal Reserve Bank of New York Staff Reports,
no. 609, April.
References
Peristiani, S. 1997. "Do Mergers Improve the X-Efficiency and Scale
Efficiency of U.S. Banks? Evidence from the 1980s." Journal of
Money, Credit, and Banking 29, no. 3 (August): 326-37.
Wheelock, D. C., and P. W. Wilson. 2012. “Do Large Banks Have
Lower Costs? New Estimates of Returns to Scale for U.S. Banks.”
Journal of Money, Credit, and Banking 44, no. 1 (February):
171-99.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2014
27
Do Big Banks Have Lower Operating Costs?
Anna Kovner, James Vickery, and Lily Zhou
Appendix B: Additional Materials
Contents
1. Chart: Time Series Plot: Components of Noninterest Expense
2. Table B1: Estimated Effect of Changing Bank Holding Company Size on
Operating Costs
3. Table B2: Methodology for Classifying Memorandum “Write-In” Items
Time Series Plot: Components of Noninterest Expense
60
Percent of total noninterest expense
Compensation expense
50
40
Other noninterest expense
30
Goodwill expense
20
Fixed assets expense
10
Amortization expense
0
2001
02
03
04
05
06
07
08
09
10
11
12
Source: Board of Governors of the Federal Reserve System, Consolidated Financial Statements of Bank Holding Companies (FR Y-9C data).
FRBNY Economic Policy Review / December 2014
1
Appendix B: Additional Materials
Table B1
Estimated Effect of Changing Bank Holding Company Size on Operating Costs
Panel A. Limiting Asset Size to 4 Percent of GDP: Estimated Impact on Bank Holding Company Operating Costs, as of 2012:Q4
Name
(1)
(2)
(3)
(4)
(5)
(6)
Quarterly Net
Operating
Revenue ($bn)
Quarterly
Noninterest
Expense ($bn)
Quarterly
Efficiency
Ratio (%)
Assets
($bn)
4 percent
of GDP Size
Limit ($bn)
Excess over
GDP Size
Limit ($bn)
(7)
(8)
Log of Total Log of 4 percent
Assets, $000s
of GDP
(9)
(10)
(11)
(12)
Change
in Logs
Coefficient
from Table 4,
Column 10a
Change to
Efficiency
Ratio
Reduction in Operating
Cost Efficiency
($bn, quarterly)
JP Morgan Chase
23.60
15.98
67.74
2,359.14
656.81
1,702.33
21.58
20.30
-1.28
-4.15
5.31
1.25
Bank of America
18.62
18.48
99.22
2,212.00
656.81
1,555.19
21.52
20.30
-1.21
-4.15
5.04
0.94
Wells Fargo
22.06
12.98
58.85
1,422.97
656.81
766.16
21.08
20.30
-0.77
-4.15
3.21
0.71
Citigroup
18.15
14.48
79.80
1,864.66
656.81
1,207.85
21.35
20.30
-1.04
-4.15
4.33
0.79
Morgan Stanley
6.90
6.07
87.94
780.96
656.81
124.15
20.48
20.30
-0.17
-4.15
0.72
0.05
Goldman Sachs
9.59
5.37
56.01
938.77
656.81
281.96
20.66
20.30
-0.36
-4.15
1.48
0.14
Sum
3.88
Source: Authors’ calculations.
Notes: This table computes an estimate of the reduction in operating cost efficiencies that would occur if each of the six BHCs with assets greater than 4 percent of nominal GDP were reduced in size to 4 percent of
GDP. Nominal GDP in 2012:Q4 was $16,420.30 billion. Columns 1 and 2 present the net operating revenue and total interest expense for each firm in 2012:Q4. Column 3 presents the 2012:Q4 efficiency ratio (percent) for each firm, defined as total interest expense over net operating revenue. Column 4 presents the 2012:Q4 total assets for each firm, column 5 shows the asset size threshold of 4 percent of nominal GDP, and
column 6 shows the dollar difference between each bank’s asset size and the asset size threshold (column 4 less column 5). Column 7 shows the natural log of each firm’s total assets, column 8 shows the natural log
of the asset size threshold, and column 9 shows the difference between the natural log of total assets and the natural log of the asset threshold (column 8 less column 7). Column 10 is the coefficient on the natural
log of assets from Table 4, column 10. The change to the efficiency ratio (percent), column 11, is calculated as the change in logs (column 9) multiplied by the coefficient on the log assets (column 10). Column 12, the
quarterly change in noninterest expense, is calculated as the change to the efficiency ratio, column 11, multiplied by net operating revenue, column 1.
If we instead apply the smaller coefficient from Table 4, column 2, of -1.892, we obtain a total estimate of $1.77 billion per quarter instead of $3.88 billion per quarter.
a
2
Do Big Banks Have Lower Operating Costs?
Appendix B: Additional Materials
Table B1
Estimated Effect of Changing Bank Holding Company Size on Operating Costs
Panel B: Estimated Impact of Increasing Total Assets by $1 Billion, for a Bank Holding Company of Mean Size
(1)
(2)
Mean Industry Net
Operating Revenue
($m)
Mean Industry
Noninterest Expense
($m)
100.60
66.83
(3)
(4)
(5)
(6)
(7)
(8)
Log Mean
Log Mean Industry
Mean Industry
Mean Industry
Mean Industry
Industry Assets,
Assets + $1bn,
Change
Efficiency Ratio (%)
Assets ($m)
Assets + $1bn ($m)
$000s
$000s
in Logs
66.43
9,064.66
10,064.66
16.02
16.12
(9)
(10)
(11)
Coefficient
from Table 4,
Column 10a
Change to
Efficiency
Ratio
Increase in Operating
Cost Efficiency
($m, quarterly)
-4.15
-0.43
-0.44
0.10
Source: Authors’ calculations.
Notes: This table computes an estimate of the operating cost efficiencies that would occur if a BHC of mean size increased its assets by $1 billion. Column 1 presents the mean industry net operating revenue, defined
as the sum of net interest income and noninterest income, in millions of dollars. Column 2 presents the mean industry total noninterest expense, and column 3 shows the mean industry efficiency ratio, defined
as net operating revenue over total interest expense. Columns 4 and 5 present the mean industry assets and the mean industry assets increased by $1 billion, respectively. Columns 6 and 7 show the log of mean
industry assets and the log of the mean industry assets increased by $1 billion, respectively. Column 8 shows the change in the log of mean industry assets and the log of mean industry assets + $1 billion, that is,
the difference between columns 6 and 7. Column 9 presents the coefficient on the natural log of assets from Table 4, column 10 in the first row, and the coefficient on the natural log of assets from Table 4, column
2, in the second row. The change to the efficiency ratio (percent), column 10, is calculated as the change in logs (column 8) multiplied by the coefficient on the log assets (column 9). Column 11, the quarterly change in
noninterest expense is calculated as the change to the efficiency ratio, column 10, multiplied by the mean industry net operating revenue, column 1.
a
If we instead apply the smaller coefficient from Table 4, column 2, of -1.892, we obtain a total estimate of $-0.20 million per quarter instead of $-0.43 million per quarter.
FRBNY Economic Policy Review / December 2014
3
Appendix B: Additional Materials (Continued)
Table B2
Methodology for Classifying Memorandum “Write-In” Items
Panel A: Classification of Standardized Subcategories into Nine Author-Defined Categories
Standardized Other Noninterest Expense Categories
Mnemonic
Data processing expenses
BHCKC017
Information technology and data processing
Advertising and marketing expenses
BHCK0497
Corporate overhead
Directors’ fees
BHCK4136
Directors’ fees and other compensation
Printing, stationery, and supplies
BHCKC018
Corporate overhead
Postage
BHCK8403
Corporate overhead
Legal fees and expenses
BHCK4141
Legal
FDIC deposit insurance assessments
BHCK4146
FDIC assessments and other government-related expenses
Accounting and auditing expenses
BHCKF556
Corporate overhead
Consulting and advisory expenses
BHCKF557
Consulting and advisory
Automated teller machine (ATM) and interchange expenses
BHCKF558
Retail banking
Telecommunications expenses
BHCKF559
Information technology and data processing
Note: FDIC is Federal Deposit Insurance Corporation.
4
Do Big Banks Have Lower Operating Costs?
Author-Defined Noninterest Expense Categories
Appendix B: Additional Materials (Continued)
Table B2
Methodology for Classifying Memorandum “Write-In” Items
Panel B: Classification of “Write-In” Items into Subcategories
Midlevel Subcategory
Code
Author-Defined Other Noninterest Expense Category
Percentage Nonmissing
Brokerage / clearing
A01
Other financial services
0.58
Custodian fees
A02
Other financial services
0.20
Trust-related
A03
Other financial services
0.92
Advisor commissions / management fees
A04
Other financial services
0.72
Deferred compensation
B01
Directors’ fees and other compensation
0.68
Agent expenses / insurance commissions
B02
Directors’ fees and other compensation
0.08
Benefits
B03
Directors’ fees and other compensation
0.24
Options / incentives
B04
Directors’ fees and other compensation
0.06
Severance
B05
Directors’ fees and other compensation
0.02
Unclassified commissions
B06
Directors’ fees and other compensation
0.14
Other / unknown compensation
B07
Directors’ fees and other compensation
0.42
Computer software / IT / internet banking
C01
Information technology and data processing
13.10
Bank card unknown
D01
Retail banking
1.46
Card interchange / card exchange
D02
Retail banking
0.78
Card processing
D03
Retail banking
0.62
Card rewards
D04
Retail banking
0.40
Card service
D05
Retail banking
0.32
Other / unknown credit card
D06
Retail banking
3.14
Affordable / low-income housing
E01
FDIC assessments and other government-related expenses
1.24
Regulation / assessment / compliance
E02
FDIC assessments and other government-related expenses
4.50
New markets tax credit investments
E03
FDIC assessments and other government-related expenses
0.06
Fees for small business loans
E04
FDIC assessments and other government-related expenses
0.02
Community Reinvestment Act
E05
FDIC assessments and other government-related expenses
0.30
Other / unknown government-related
E06
FDIC assessments and other government-related expenses
0.04
Insurance losses / insurance provision
F01
Other financial services
0.26
Premiums
F02
Other financial services
0.26
Other / unknown insurance-related
F03
Other financial services
5.74
Servicing
G01
Retail banking
0.52
Reps and warranties
G02
Retail banking
0.26
Owned real estate
G03
Retail banking
27.05
Collection / repossession
G04
Retail banking
7.28
Bad debt
G05
Retail banking
1.82
Credit reports
G06
Retail banking
0.42
Mortgage-related
G07
Retail banking
2.60
Fraud
G08
Retail banking
0.44
Write-in items
Other / unknown loan
G09
Retail banking
8.20
Corporate debt and equity issuance / repayment
H01
Corporate overhead
3.70
Hedging
H02
Corporate overhead
0.56
Foreign Exchange
H03
Corporate overhead
0.00
Other / unknown treasury-related
H04
Corporate overhead
0.48
Taxes
I01
Corporate overhead
12.32
FRBNY Economic Policy Review / December 2014
5
Appendix B: Additional Materials (Continued)
Table B2
Methodology for Classifying Memorandum “Write-In” Items
Panel B: Classification of “Write-In” Items into Subcategories
Midlevel Subcategory
6
Code
Author-Defined Other Noninterest Expense Category
Percentage Nonmissing
Investment in unconsolidated subsidiaries
I02
Corporate overhead
0.46
Unclassified depreciation
I03
Corporate overhead
0.28
M&A / restructuring / non-FDIC indemnification costs
I04
Corporate overhead
1.88
Travel / business development / recruitment /
transportation / staff relations
I05
Corporate overhead
9.20
Leases / equipment
I06
Corporate overhead
0.68
Charity
I07
Corporate overhead
3.72
General / mixed corporate
I08
Corporate overhead
2.30
Operating leases
I09
Corporate overhead
0.62
Directly-owned electric / utility / janitorial / security
/ rent
I10
Corporate overhead
0.98
Directors and officers’ insurance
I11
Corporate overhead
0.36
Account analysis otherwise uncategorized
I12
Corporate overhead
0.04
Dues / memberships / subscriptions
I13
Corporate overhead
0.90
Other / unknown corporate / overhead
I14
Corporate overhead
0.10
Branch closing
J01
Retail banking
0.12
Brokered deposits
J02
Retail banking
0.04
Deposit costs / unknown banking
J03
Retail banking
0.76
Due from account / bank charge
J04
Retail banking
0.10
Home banking
J05
Retail banking
0.04
Foreign offices
J06
Retail banking
0.10
Lockbox fee
J07
Retail banking
0.04
Checks
J08
Retail banking
0.50
NOW accounts
J09
Retail banking
0.06
Robbery / bad checks / forgery / overdrawn accounts
J10
Retail banking
0.12
Correspondent banking / affiliate
J11
Retail banking
2.30
Armored car
J12
Retail banking
0.56
Other / unknown retail banking
J13
Retail banking
0.12
Off-balance-sheet
K01
Miscellaneous
0.48
Writedowns and writeoffs otherwise uncategorized
L01
Miscellaneous
1.22
Amortization otherwise uncategorized
L02
Miscellaneous
0.18
Miscellaneous - well defined, but without category
M01
Miscellaneous
0.20
Miscellaneous - item not understood
M02
Miscellaneous
0.32
Miscellaneous - unclassifiable / vague
M03
Miscellaneous
5.64
Miscellaneous - multiple items with values listed
M04
Miscellaneous
0.48
Trading expenses
N01
Other financial services
0.02
Litigation, settlements, and other legal
OF1
Legal
1.04
Provision of legal reserves
OF2
Legal
0.04
Realized gains / losses
ZA1
Miscellaneous
0.14
Preferred dividends
ZA2
Miscellaneous
0.04
Do Big Banks Have Lower Operating Costs?
Appendix B: Additional Materials (Continued)
Table B2
Methodology for Classifying Memorandum “Write-In” Items
Panel B: Classification of “Write-In” Items into Subcategories
Midlevel Subcategory
Code
Author-Defined Other Noninterest Expense Category
Percentage Nonmissing
Write-in expenses that appear to fit in the standardized categories
Data processing expenses
XA1
Information technology and data processing
0.24
Advertising and marketing expenses
XB1
Corporate overhead
1.42
Directors’ Fees (I)
XC1
Directors’ fees and other compensation
0.42
Directors’ Fees (II)
XC2
Directors’ fees and other compensation
0.14
Printing, stationery, and supplies
XD1
Corporate overhead
0.06
Postage
XE1
Corporate overhead
2.46
FDIC deposit insurance assessments
XG1
FDIC assessments and other government-related expenses
0.76
Accounting and auditing expenses
XH1
Corporate overhead
0.02
Consulting and advisory expenses
XI1
Consulting and advisory
9.40
Automated teller machine (ATM) and interchange expenses
XJ1
Retail banking
0.36
Telecommunications expenses
XK1
Information technology and data processing
0.18
Notes: This table presents our methodology for classifying “other” noninterest expense items reported in the Memoranda of Schedule HI in the FR Y-9C.
Panel A shows the eleven standardized subcategories of other noninterest expense (NIE) that are included in the Memoranda, and the author-defined
category that each standardized subcategory is aggregated into. BHCs are only required to report amounts in these standardized subcategories if the amount
is greater than $25,000 and exceeds 3 percent of total other NIE. The Memoranda also provides space for BHCs to report additional “write-in” expense items
that are not captured by the eleven standardized fields. These write-in items are only reported if the item exceeds 10 percent of total other NIE. Overall,
30,457 text strings are written in by the BHCs in our sample since 2008, and 5,418 of these are unique. Panel B shows the classification of all write-in items
from 2008 to 2012 into seventy-seven new subcategories created by the authors and the reallocation of the write-in items that should have been classified
into the eleven standardized subcategories. The write-in items are not reported using standardized language, and as a result, classification was in part done
by hand, and in part via Stata code that conducted Boolean searches for keywords within each text string. For each of the eighty-eight total subcategories,
Panel B presents the percentage of nonmissing values (column “percentage nonmissing”), and the author-defined category into which the subcategory is
aggregated (column “author-defined other NIE category”). M&A is mergers and acquisitions; FDIC is Federal Deposit Insurance Corporation.
FRBNY Economic Policy Review / December 2014
7
João A. C. Santos
Evidence from the Bond
Market on Banks’
“Too-Big-to-Fail” Subsidy
Expectations that the government will step
in to save the largest banks from failure
could create a “subsidy” for these banks by
encouraging investors to discount risk when
they provide funding.
A look at bond data over the 1985-2009
period suggests that investors accept lower
credit spreads on bonds issued by the largest
banks than on bonds issued by small banks.
The funding advantage enjoyed by the largest
banks appears to be significantly larger than
that of the largest nonbanks and nonfinancial
corporations.
This evidence is consistent with the idea that
“too-big-to-fail” status gives the largest banks
a competitive edge.
João A. C. Santos is a vice president at the Federal Reserve Bank of New York.
joao.santos@ny.frb.org
1. Introduction
T
he idea that some firms may be too big to fail appears
to go back as far as 1975 in connection with Lockheed
Corporation and the financial difficulties experienced by that
firm at the time.1 It was, however, the demise of Continental
Illinois Bank in 1984 that provided solid supporting evidence
for this idea.
Continental Illinois, which was the seventh-largest U.S.
bank by deposits, experienced runs by large depositors following news it had incurred significant losses in its loan portfolio.
Concerns that a failure of Continental Illinois would have
significant adverse effects on the banks that had deposits with
it led regulators to take the unprecedented action of assuring
all of Continental’s depositors—large and small—that their
1
In 2008, in his New York Times column on language, William Safire explored
the origins of the phrase, citing a 1975 Business Week article about Lockheed
Corporation that carried the headline “When Companies Get Too Big to Fail”
(“Too Big to Fail or to Bail Out?” New York Times, April 6, 2008).
The author thanks Mark Flannery and seminar participants at the Federal
Reserve Bank of New York for valuable comments on an earlier draft of
this article, and Vitaly Bord for outstanding research assistance. The views
expressed in the article are those of the author and do not necessarily
reflect the position of the Federal Reserve Bank of New York or the Federal
Reserve System.
FRBNY Economic Policy Review / December 2014
29
money was fully protected.2 Subsequently, during Congressional hearings on Continental Illinois, the Comptroller of the
Currency indicated that the eleven largest banks in the United
States were too big to fail and would not be allowed to fail.3
The perception that some banks will be rescued because
they are too big to fail is important because it can have
far-reaching implications. If investors, creditors in particular,
believe that certain banks are too big to fail, they will discount
risk when providing those banks with funding. This insensitivity of financing costs to risk will encourage too-big-to-fail
banks to take on greater risk. The largest banks’ risk taking, in
turn, will drive the smaller banks that compete with them to
take on additional risk as well.4
That perception has triggered a large body of research
attempting to determine whether bank investors, including
depositors, believe that the largest banks are too big to fail,
and whether those banks behave differently because they
expect to be rescued if they get into financial difficulties.
A number of studies have tried to test the too-big-to-fail
hypothesis by investigating spreads on bank bonds. Flannery
and Sorescu (1996), for example, find that yield spreads on
bank bonds were not risk sensitive after the Continental
Illinois bailout, suggesting that bond investors believed large
banks were too big to fail. However, the authors find that bond
spreads came to reflect the specific risks of individual issuing
banks starting around 1988 when conjectural guarantees no
longer covered (many) bank debentures. Balasubramnian
and Cyree (2011) document that the relationship between
spread and risk for the largest banks flattened after the rescue
of Long-Term Capital Management in 1998. Anginer and
Warburton (2014) find a positive relationship between risk
and bond spreads in the secondary market but only for midsize and small institutions. Acharya, Anginer, and Warburton
(2013) document that bond credit spreads continued to be less
sensitive to risk for the largest financial institutions even after
the passage of the Dodd-Frank act.5 Penas and Unal (2004),
2
Simultaneously, the Federal Reserve Board, the Federal Deposit Insurance
Corporation (FDIC), and the Comptroller of the Currency, together with
twenty-four U.S. banks, announced a $7.3 billion bailout for Continental
Illinois. The rescue package comprised a $2 billion capital injection by the
FDIC and the group of twenty-four banks and an unsecured line of credit by
the banks of $5.3 billion.
3
See O’Hara and Shaw (2000) for further details on the Comptroller of the
Currency’s announcement.
in turn, focus on bank mergers. They find that bondholders of
medium-sized banks that may push the merging bank into the
too-big-to-fail category realize the highest returns around the
merger and only these banks benefit from some savings when
they issue in the bond market after they merge.
Some studies have considered instead credit default swap
(CDS) spreads. Demirguc-Kunt and Huizinga (2010) report
that, in countries with weak finances, too-big-to-fail banks
could increase their value by downsizing (they are too big to
save) while, in stronger regimes, CDS spreads tend to decrease
with bank size.6
Other studies have focused on support ratings, which
attempt to capture the likelihood that the bank will receive
government support if it runs into financial difficulties. Rime
(2005) shows that proxies for the too-big-to-fail status of a
bank, such as size and market share, have a positive effect
on a large bank’s support rating relative to its stand-alone
rating. Haldane (2010) documents that the stand-alone versus
support ratings differential was between 1.5 and 4 notches for
a sample of U.K. banks, building societies, and global banks
between 2007 and 2009. Ueda and Weder di Mauro (2011) in
turn report that, for the top forty-five U.S. banks, the mean
support rating differential increased from 3.2 in 2007 to 4.1 in
2009, suggesting an increase in the importance of the too-bigto-fail status over that period.
Still other studies have considered the cost of deposits and
bank merger premiums. Baker and McArthur (2009), for
example, report that the average cost of deposits is lower for
large banks. They also report that the difference in the cost of
deposits for banks with more than $100 billion in assets and
those with less increased in the period from the fourth quarter
of 2008 to the fourth quarter of 2009. Jacewitz and Pogach
(2013) report that the risk premium on uninsured deposits
paid by the largest banks was 15 to 40 basis points lower than
at other banks, based on deposit rates offered at the branch
level over the 2005-08 time period.
Brewer and Jagtiani (2007), meanwhile, study the purchase
premium that acquirers are willing to pay for becoming too
big to fail and gaining the presumed benefits of that status.
The authors estimate that, over the 1991-2004 period, acquirers in nine mergers were willing to pay about $14 billion in
additional premiums in order to become too big to fail.7
Lastly, a set of studies has unveiled evidence that banks
believed to be too big to fail take on additional risk. Gropp,
4
As Hakenes and Schnabel (2010) show, lower financing costs induce large
banks to behave more aggressively, increasing competition and decreasing
margins and hence charter values for competing banks—developments
that push these banks toward higher risk taking. See Gropp, Hakenes, and
Schnabel (2011) for evidence of this effect on smaller competing banks.
5
See Sironi (2003) and Morgan and Stiroh (2005) for further studies of bank
bond spreads in Europe and the United States, respectively.
30
Evidence from the Bond Market
6
Li, Qu, and Zhang (2011) also consider CDS spreads to investigate whether
investors believe the largest U.S. banks are too big to fail.
7
Molyneux, Schaeck, and Zhou (2010) also investigate the merger premiums,
but their analysis is based on a sample of bank mergers and acquisitions in
nine European Union economies.
Hakenes, and Schnabel (2011), for example, find support for
this conclusion by looking at bank balance sheet data, and
Gadanecz, Tsatsaronis, and Altunbas (2012), by looking at
bank lending in the syndicated loan market. Brandao Marques
et al. (2013) and Afonso, Santos, and Traina (2014), in turn,
uncover evidence of bank risk taking by studying various
measures of bank risk. These studies are important because
they show that too-big-to-fail status does have an effect on
banks’ policies.
Although this article, like other studies reviewed here,
focuses on the primary bond market, our approach differs
from that of other researchers who look for evidence of a toobig-to-fail subsidy in bond spreads. Specifically, we ascertain
whether investors perceive the largest banks to be too big to
fail by investigating whether these banks benefit from a cost
advantage when they raise funding in the bond market. We
start by examining how the bonds issued by the largest banks
over the 1985-2009 period compare with those issued by
smaller banks in terms of their credit spreads over Treasury
securities of the same maturity, controlling for bond risk and
other factors that may affect bond spreads.
The results of this part of our investigation show that the
top-five banks by asset size pay significantly lower spreads
than their smaller peers. In particular, the spreads of bonds
issued by the largest banks are, on average, 41 basis points
below the smaller banks’ bond spreads, after controlling for
bond characteristics, including the credit rating, maturity, and
amount of the issue, as well as conditions in the bond market
at the time of issue. However, this cost difference does not
necessarily imply that investors believe that the largest banks
are too big to fail. For example, if the largest banks are better
positioned to diversify risk because they offer more products
and operate across more businesses (something not fully captured in their credit rating), this advantage could explain part
of that difference in the cost of bond financing.
To address this concern, we extend the analysis and
compare the largest banks’ cost advantage over smaller banks
in the bond market with the cost advantages that the large
nonbank financial institutions (nonbanks) and the largest
nonfinancial corporations enjoy relative to their smaller peers.
If what drives the difference in the cost of bond issuance for
the largest and smaller banks is a size-specific factor or a
perception by investors that the largest firms in general are
all too big to fail, then the cost advantage of the largest banks
should be similar to the cost advantages possessed by the
largest nonbanks and the largest nonfinancial corporations in
the bond market. If, however, investors believe that the largest
banks are more likely to be considered too big to fail, then the
cost advantage of these banks will exceed that of the largest
nonbanks and nonfinancial corporations.
The results of this part of our investigation show that the
largest nonbanks and the largest nonfinancial corporations
pay less than their smaller peers to raise funding in the bond
market. However, in contrast to our findings on banks, that
discount is generally not statistically different from zero.
Given these findings, it is not surprising that our results show
that the largest banks enjoy a significantly larger discount
than both the largest nonbanks and the largest nonfinancial corporations. The largest banks that issue bonds rated
double A and single A—the two main rating categories for
these banks’ bonds—benefit from a discount (relative to their
smaller peers) that is larger by 92 and 16 basis points, respectively, than the discount enjoyed by the largest nonbanks that
issue bonds with those same ratings (relative to their smaller
peers), though the difference is statistically significant only in
the former case. When compared with the largest nonfinancial corporations, the largest banks that issue bonds rated
double A and single A benefit from an additional discount of
53 and 50 basis points, respectively, although only the latter
difference is statistically significant.
Our finding that the largest banks, the largest nonbanks,
and the largest nonfinancial corporations all benefit from a
discount relative to their smaller peers in the bond market
can be interpreted as some support for the view that the toobig-to-fail status does not apply solely to banks. However,
our evidence that the largest banks benefit from a bigger discount than the largest nonbanks and the largest nonfinancial
corporations suggests that investors believe that the largest
banks are more likely to be rescued if they get into financial
difficulties.
The rest of the paper is organized as follows. Section 2
describes the methodology and data sources used and
characterizes the sample. Section 3 compares the spreads
that the largest banks pay to raise funding in the bond
market with those paid by smaller banks. Section 4 conducts
a similar exercise for nonbanks and nonfinancial corporations, respectively. Section 5 compares the discount that the
largest banks enjoy (relative to their smaller peers) with the
discount available to the largest nonbanks and the largest
nonfinancial corporations in the bond market. Section 6
summarizes our findings.
FRBNY Economic Policy Review / December 2014
31
2. Methodology, Data, and Sample
Characterization
the largest nonbank issuers relative to their smaller peers. To
that end, we estimate the following model of bond spreads:
SPREADi= c + θTOP5i + ϑBKi+ αBK × TOP5i + δBONDi
+ βBKi × BONDi + γTIMEi + εi .
2.1 Methodology
To ascertain whether too-big-to-fail banks benefit from a
discount in the bond market, we begin by estimating the
following model of bond spreads on the sample of bonds
issued by U.S. banks:
SPREADi= c + αTOP5i + βBONDi+ γTIMEi+ εi,
where SPREAD is the bond yield over the Treasury security
(with the same maturity as the bond) at the time of the bond
origination. TOP5, the key variable of interest, is a dummy
variable equal to 1 for bonds issued by the top-five banks (by
asset size) in the year. If large banks benefit from a discount
in the bond market relative to their smaller peers, then we
should find that TOP5 is negative and statistically significant.
We attempt to identify that effect while controlling for a
set of bond characteristics, BOND, which includes a dummy
variable for the rating of the bond (AAA, AA, A . . .), the log
of the size of the bond issue (LAMOUNT), and the maturity
of the bond (MATURITY). Everything else equal, we should
expect bonds with higher ratings to carry lower spreads. With
regard to the size of the bond issue, banks that are more creditworthy usually find it easier to make larger issues, but they
may have to offer higher yields to create a sufficiently large
demand for their bond issues. So the effect of the size of the
bond issue on the spread is ambiguous. Similarly, banks that
are more creditworthy may find it easier to issue longer-term
bonds, but these bonds tend to carry a higher risk. Finally, we
include a set of year-quarter dummy variables to control for
any effects that economic conditions at the time of the issue
may have on the bond spread.
The large-bank discount identified by the model of bond
spreads we presented above may not be solely attributable to
a too-big-to-fail subsidy. For example, if bonds of the largest
banks are safer in a way that is not captured in their credit
ratings, this will lower the coefficient on TOP5; yet it is not
the result of investors “offering” a discount to the largest banks
because they believe these banks will be protected in the event
of financial difficulties. In an attempt to disentangle these
effects, we expand the sample to include bonds issued by nonbanks and nonfinancial firms. We then investigate whether the
largest banks benefit from a discount relative to their smaller
peers and consider how that discount compares with that of
32
Evidence from the Bond Market
This is an extension of the previous model. TOP5 is a
dummy variable equal to 1 if the bond issuer is a top-five firm
by assets in its group (banks, nonbanks, and nonfinancial corporations). BK is a dummy variable equal to 1 if the bond was
issued by a bank. As in the previous model, the key variable
of interest is the dummy variable BK × TOP5. This variable
will indicate whether the largest banks benefit from a bigger
discount in the bond market than the largest nonbank issuers.
We attempt to identify that difference in the cost paid by
the largest firms while using the same set of controls we use in
our base model of bond spreads. To allow for the possibility
that bank bonds are priced differently from the bonds of the
remaining firms, we include not only the set of bond controls, BOND, but also its interactions with our bank dummy
variable, BK. As in the base model, we include year-quarter
dummy variables to control for the potential effects of economic conditions at the time of the bond issue.
Since there are important differences between the two
control groups considered, we estimate that model separately
on the sample of bonds issued by banks and by nonbanks, and
on the sample of bonds issued by banks and by nonfinancial
corporations. Finally, since the pool of bonds issued by the
largest firms may carry a different level of risk than the set of
bonds issued by the remaining firms, we estimate our bond
spread model separately for bonds with the same credit rating.
In this case, we restrict the sample to bonds most commonly
issued by the largest banks, that is, bonds rated single A and
those rated double A.
2.2 Data
The data for this analysis come from the Securities Data
Corporation’s Domestic New Bond Issuances (SDC) database
and from Compustat. We use the SDC database to obtain
information on all bonds issued in the United States, including their maturity and yield at origination, and whether they
are callable or convertible or have a floating rate. We also use
the SDC database to get information about the identity of the
bond issuer.
We complement these data with information on issuers’
assets from Compustat and from banks’ Consolidated Reports
of Condition and Income (call reports), which are used to
identify the largest firms among banks, nonbanks, and nonfinancial corporations.
Table 1
Ratings Distribution of Bonds in the Sample
Financials
Banks
2.3 Sample Characterization
To select our sample of bonds, we start out with all the bonds
issued in the U.S. bond market by banks, nonbanks, and
nonfinancial corporations between 1985 and 2009. We begin
in 1985 since the claim that some banks were too big to fail
was first made in connection with the demise of Continental
Illinois in 1984. Next, we drop the bonds that do not have the
information we need to estimate the bond spread model (ex
ante yield to maturity, issue date, maturity date, and Standard
& Poor’s rating). Finally, we drop bonds with “unique” features
that affect their pricing (such as floating-rate bonds, as well
as callable bonds and convertible bonds). These criteria leave
us with a sample of 8,399 bonds, of which 436 were issued by
banks, 1,696 were issued by nonbanks, and 6,267 were issued
by nonfinancial corporations.
We identify the top-five firms by asset size in each group
and isolate their bonds. Of the 436 bonds issued by banks, 243
were issued by the top-five banks. Of the 1,696 bonds issued
by nonbanks, 241 were issued by the top-five firms. Lastly, of
the 6,267 bonds issued by nonfinancial corporations, 139 were
issued by the top-five firms. Table 1 reports the rating distribution of the bonds issued by each of these groups.
Significant differences emerge in the risk profile of the sample
of bonds issued by each of the three groups in the sample. For
example, only about 16 percent of the bonds issued by banks
are rated below investment grade. In the case of bonds issued
by nonbanks, that percentage goes up to 20 percent, and it rises
further to 33 percent in the case of nonfinancial corporations.
These differences are even more striking when we consider
the bonds issued by the top-five firms within each group. For
example, none of the bonds in the sample issued by the top-five
banks are rated below investment grade. It is for this reason that,
when comparing the difference in credit spreads at origination
across the three groups of firms, we focus on single-A- and
double-A-rated bonds, which are the two most populated rating
categories among bonds issued by the largest banks.
3. Do the Largest Banks Issue Bonds
at a Discount?
To ascertain whether the largest banks benefit from a discount
in the bond market, we use our model of bond spreads to
TOP5
243
Nonbanks
Nonfinancials
All
Others
TOP5
All
Others
TOP5
All
Others
193
241
1,455
139
6,128
Percentage of Bonds by Bond Rating
AAA
0.058
0.010
0.095
0.014
0.007
0.006
AA
0.152
0.150
0.320
0.086
0.266
0.035
A
0.790
0.446
0.581
0.333
0.410
0.253
BBB
0.238
0.004
0.353
0.108
0.382
BB
0.119
0.058
0.007
0.130
B
0.031
0.054
0.007
0.116
CCC
0.006
0.037
0.122
0.053
CC
0.003
0.004
C
0.002
0.001
D
0.060
0.073
0.020
Source: Author’s calculations.
Notes: Our sample includes 8,399 bonds issued by banks (436), nonbank
financial institutions (1,696), and nonfinancial corporations (6,267) over
the 1985-2009 time period. TOP5 is a dummy variable for the top-five
issuers by asset size. AAA, AA . . . are dummy variables for the S&P rating
of the bond.
compare the credit spreads (over Treasuries with the same
maturity) on their bonds in the primary market with the
spreads on the bonds of the remaining banks. Table 2 reports
the results. Model 1 distinguishes the bonds issued by the
top-five banks (as measured by asset size) from those issued
by the remaining banks, controlling only for the year-quarter
when the bond was issued in order to account for the overall
macroeconomic effects on the cost to issue in the bond market. According to our results, the largest banks benefit from a
discount of 44 basis points relative to the spread paid by the
remaining banks to issue in the bond market.
Model 2 shows that when we control for the risk of the
bond as determined by its Standard & Poor’s rating and for the
maturity and size of the bond issue, the discount enjoyed by
the largest banks drops to 41 basis points, although it continues to be statistically different from zero. As one would expect,
safer bonds carry lower credit spreads, and bonds with longer
maturity carry higher credit spreads, probably to compensate investors for the higher risk associated with these bonds.
Lastly, our controls show that larger bond issues carry larger
FRBNY Economic Policy Review / December 2014
33
Table 2
Spreads on Bonds of Banks
TOP5
Model 1:
All Bonds
Model 2:
All Bonds
Model 3:
AA Bonds
Model 4:
A Bonds
-0.440***
-0.406***
-1.208**
-0.308*
(3.48)
(3.01)
(2.13)
(1.84)
AAA
-4.151***
AA
-1.433***
A
-1.064***
BBB
-0.45
BB
-0.39
B
-0.773***
(7.55)
(5.25)
(3.92)
(1.51)
(1.40)
(3.60)
MATURITY
0.036***
(3.44)
LAMOUNT
Constant
0.250***
1.620***
(9.43)
Observations
R2
0.081**
(2.65)
0.319
0.031***
(2.66)
(1.13)
(4.03)
0.255
-3.275*
-1.169*
(0.58)
(1.79)
(1.93)
436
436
0.539
66
278
0.799
0.579
Source: Author’s calculations.
Notes: The dependent variable in these models is the bond spread in the
primary market (computed over the Treasury security with the same
maturity as the bond). TOP5 is a dummy variable for the top-five issuers
by assets size. AAA, AA . . . are dummy variables for the S&P rating of the
bond. Maturity is the maturity of the bond. LAMOUNT is the log of the
amount of the issue. Included in all of the models are also year-quarter
dummy variables. Models estimated with robust standard errors clustered
at the bond issuer. The t-statistics are reported in parentheses.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
yields, suggesting that economies of scale are not prevalent in
the bond underwriting business.
As we saw in Table 1, the largest banks issue, on average,
safer bonds than their smaller peers—an observation that
helps explain part of the discount that these banks enjoy in
the bond market, as captured in model 2. To account for this
risk difference in the pool of bonds issued by the two groups,
we reestimate the bond spread model on bonds with the same
34
4. Do Large Firms Enjoy a Discount
in the Bond Market?
0.329***
(4.24)
0.375
credit rating. We limit this exercise to bonds rated double A
and single A because they are the ones most commonly issued
by the largest banks. Models 3 and 4 of Table 2 report the
results of this exercise. The negative coefficient on the dummy
variable that isolates the bonds issued by the largest banks,
TOP5, in the new models indicates that the largest banks
enjoy a discount in the bond market relative to their smaller
peers that issue bonds with the same credit rating.
These last findings suggest that the status of too big to fail
may give the largest banks a competitive edge by virtue of
their ability to raise funding in the bond market at a discount
relative to their smaller peers. However, it is possible that
the discount enjoyed by the largest banks reflects only their
unique ability to diversify risk because of their presence in a
larger number of markets—a distinction that is not fully captured in their credit rating. We investigate this possibility next
by comparing banks with nonbank financial institutions and
with nonfinancial corporations, respectively.
Evidence from the Bond Market
To investigate whether the largest firms outside the banking
sector also benefit from a discount when they raise funding in
the bond market, we repeat the same exercise we conducted
for banks, but now for the bonds issued by nonbanks and
nonfinancial corporations. The results of this investigation are
reported in Tables 3 and 4, respectively.
We find that the largest nonbanks also appear to benefit
from a discount relative to their smaller peers when they issue
bonds (Table 3). The top-five nonbanks are able to issue bonds
with spreads about 79 basis points lower than those issued by
their smaller peers (model 1). When we control for the rating
of the bond, its maturity, and the size of the issue, that discount comes down to 22 basis points (model 2). These results
suggest that the largest nonbanks, like the largest banks, benefit from a discount in the bond market. As we will show, this
similarity disappears when we investigate how that discount
varies with the credit rating of the issuer.
For bonds rated triple A, double A, and single A (models
3-5), TOP5 is negative in all of the models, but not statistically
significant.8 Thus it appears that the largest nonbanks also
benefit from a discount when they issue in the bond market;
however, in contrast to banks, that discount is generally not
statistically different from zero within risk categories.
8
We omit from this exercise bonds rated triple B because the sample contains
only one such bond that is issued by the largest nonbanks.
Table 3
Spreads on Bonds of Nonbank Financial Institutions
Model 1: Model 2:
Model 3:
Model 4: Model 5:
All Bonds All Bonds AAA Bonds AA Bonds A Bonds
TOP5
-0.788***
-0.220**
-0.156
-0.007
-0.177
(7.92)
(2.29)
(0.90)
(0.04)
(1.53)
AAA
-1.761***
AA
-0.448**
A
-0.229
(4.83)
(2.42)
(1.39)
BBB
0.451***
(2.71)
BB
0.553***
(2.60)
B
1.756***
(6.34)
CCC
1.190***
(4.23)
CC
-0.071
(0.14)
C
4.771***
(4.12)
MATURITY
0.051***
(12.71)
LAMOUNT
Constant
Observations
R2
0.043**
0.152***
0.077***
(7.87)
(6.40)
0.025
0.025
0.053***
(6.93)
0.064**
(2.24)
(0.41)
(0.57)
(2.13)
1.092***
-0.275
-0.291
-2.613***
-0.940***
(6.07)
(1.06)
(1.19)
(4.21)
(10.48)
1,696
0.249
1,696
0.472
44
202
625
0.978
0.633
0.574
Turning to nonfinancial corporations (Table 4), we see that
the results are very similar to those for nonbanks. The largest
nonfinancial corporations enjoy a discount of about 76 basis
points relative to their smaller peers when we do not account
for any bond characteristics (model 1). This discount drops to
47 basis points when we account for the characteristics of the
bonds (model 2). Once again, we see that this discount does
not continue to hold when we estimate our model separately
for the ratings of the bonds issued by the largest nonfinancial
corporations (models 3-6).9
Overall, these results suggest that the cost advantage that
the largest banks enjoy in the bond market relative to their
smaller peers is unique to banks. When we do not restrict the
comparison to bonds with the same credit rating, it appears
as if both the largest nonbanks and the largest nonfinancial
corporations benefit from a discount relative to their smaller
peers, as happens with banks. This similarity is not present,
however, when we restrict the comparison to bonds with the
same rating. Looking at bonds rated double A or single A, we
continue to find that the largest banks benefit from a statistically significant discount relative to their smaller peers. The
largest nonbanks benefit from a discount, but it is not statistically different from zero, and the results show mixed effects
for the largest nonfinancial corporations. The largest nonfinancials rated double A benefit from a discount, while those
rated single A pay a premium, but in either case the difference
relative to their smaller peers is not statistically significant.
It is unclear from these findings, however, whether the
discount that the largest banks enjoy relative to their smaller
peers is statistically different from the discount for the largest
nonbanks or even that for the largest double-A-rated nonfinancial corporations. We investigate this issue next.
Source: Author’s calculations.
Notes: The dependent variable in these models is the bond spread in the
primary market (computed over the Treasury security with the same
maturity as the bond). TOP5 is a dummy variable for the top-five issuers
by asset size. AAA, AA . . . are dummy variables for the S&P rating of the
bond. Maturity is the maturity of the bond. LAMOUNT is the log of the
amount of the issue. Included in all of the models are also year-quarter
dummy variables. Models estimated with robust standard errors clustered
at the bond issuer. The t-statistics are reported in parentheses.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
5. Do the Largest Banks Benefit
from a Unique Discount?
To determine whether the discount that the largest banks
enjoy in the bond market (relative to their smaller peers) is
unique to banks, we estimate our expanded model of bond
spreads separately on the set of bonds issued by banks and
nonbanks, and on the set of bonds issued by banks and
nonfinancial corporations. The results of these investigations,
reported in Tables 5 and 6, reveal whether the discount for the
largest banks is significantly larger than the discounts for the
largest nonbanks and nonfinancial corporations.
9
We omit from this exercise bonds rated triple A, single B, and D because of
their reduced number in the sample.
FRBNY Economic Policy Review / December 2014
35
Table 4
Spreads on Bonds of Nonfinancial Corporations
TOP5
Model 1:
All Bonds
Model 2:
All Bonds
-0.76***
-0.47***
(6.52)
(4.30)
AAA
Model 3:
AA Bonds
Model 4:
A Bonds
Model 5:
BBB Bonds
Model 6:
CCC Bonds
-0.17
0.14
-0.17
0.52
(1.18)
(1.34)
(0.82)
(1.21)
-3.85***
(15.36)
AA
-3.64***
(21.08)
A
-3.28***
(20.02)
BBB
-2.73***
(16.03)
BB
-1.44***
B
-0.36**
CCC
-0.3
(8.61)
(2.06)
(1.57)
CC
0.54
(1.18)
C
-0.73
(1.06)
MATURITY
0.02***
(9.7)
LAMOUNT
1.04***
(10.17)
Observations
6,267
R2
0.175
4.33***
(15.11)
0.03***
0.02***
(10.89)
(7.05)
0.01
-0.02
0.03
(0.09)
(1.30)
(1.35)
-0.07***
(4.24)
Constant
0.05***
(7.94)
-0.45
0.46***
0.06
(1.15)
(3.21)
(0.4)
250
1,609
2,355
0.717
0.478
6,267
0.423
0.227
-0.02
(1.38)
-0.59***
(4.36)
5.71***
(4.74)
339
0.636
Source: Author’s calculations.
Notes: The dependent variable in these models is the bond spread in the primary market (computed over the Treasury security with the same maturity as the
bond). TOP5 is a dummy variable for the top-five issuers by asset size. AAA, AA . . . are dummy variables for the S&P rating of the bond. MATURITY is the
maturity of the bond. LAMOUNT is the log of the amount of the issue. Included in all of the models are also year-quarter dummy variables. Models estimated with robust standard errors clustered at the bond issuer. The t-statistics are reported in parentheses.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
Looking at Table 5 and the variable BK × TOP5, which
tells us whether the discount for the largest banks is different
from the discount for the largest nonbanks (relative to their
36
Evidence from the Bond Market
smaller peers), we see that there is no statistically significant
difference between these discounts when we consider all of
the bonds of these issuers together (models 1 and 2). However,
Table 5
Table 6
Spreads on Bonds of Banks and Nonbanks
TOP5
BK
BK × TOP5
Constant
Model 3:
AA Bonds
Spreads on Bonds of Banks and
Nonfinancial Corporations
Model 1:
All Bonds
Model 2:
All Bonds
Model 4:
A Bonds
-0.74***
-0.22**
0.1
(7.68)
(2.36)
(0.59)
(1.82)
-0.45***
-2.53***
-1.24
-1.32**
(5.48)
(0.85)
(2.52)
0.24
-0.18
-0.92**
-0.16
(1.61)
(1.18)
(2.15)
(0.92)
0.19
-0.54***
(0.58)
(4.33)
(0.29)
Observations
2,132
2,132
268
903
R2
0.252
0.614
Model 3:
AA Bonds
Model 4:
A Bonds
-0.77***
-0.49***
-0.21
0.12
(6.71)
(4.43)
(1.49)
(1.16)
-1.11***
-4.64***
-1.47
-2.17***
(12.59)
(11.55)
(1.11)
(4.33)
0.19
0.16
-0.53
-0.50***
(1.09)
(0.94)
(1.38)
(2.99)
1.50***
4.27***
-0.56
0.61***
(5.1)
(16.75)
(1.47)
(3.24)
Observations
6,703
6,703
316
1,887
R2
0.189
0.439
0.695
0.479
TOP5
BK
BK × TOP5
0.09
(15.07)
0.476
Model 2:
All Bonds
-0.20*
(5.00)
2.13***
Model 1:
All Bonds
0.543
Source: Author’s calculations.
Notes: The dependent variable in these models is the bond spread in the
primary market (computed over the Treasury security with the same
maturity as the bond). TOP5 is a dummy variable for the top-five issuers
by asset size. BK is a dummy variable for bonds issued by banks. All of
the models include year-quarter dummy variables. Additionally, models
2 through 4 include dummy variables for the S&P rating of the bond,
MATURITY, LAMOUNT, and the interaction of these variables with BK.
Models estimated with robust standard errors clustered at the bond issuer.
The t-statistics are reported in parentheses.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
when we estimate the model separately on the bonds rated
double A and single A, the most common ratings of the bonds
issued by the largest institutions in the two groups, we see that
largest banks benefit from a bigger discount than the largest
nonbanks, which is statistically significant in the case of bonds
rated double A.
We get a similar picture when we compare banks with
nonfinancial corporations (Table 6). Again, the largest banks
do not appear to benefit from a bigger discount when we
consider all of the bonds together (models 1 and 2). However,
when we estimate the model separately on the bonds of each
rating category, we see that the largest banks do benefit from
a bigger discount than the largest nonfinancial corporations,
and the difference is statistically significant in the case of
bonds rated single A.
Constant
Source: Author’s calculations.
Notes: The dependent variable in these models is the bond spread in the
primary market (computed over the Treasury security with the same
maturity as the bond). TOP5 is a dummy variable for the top-five issuers
by asset size. BK is a dummy variable for bonds issued by banks. All of
the models include year-quarter dummy variables. Additionally, models
2 through 4 include dummy variables for the S&P rating of the bond,
MATURITY, LAMOUNT, and the interaction of these variables with BK.
Models estimated with robust standard errors clustered at the bond issuer.
The t-statistics are reported in parentheses.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
5.1 Robustness Tests
In this exercise, we considered bonds issued since 1985
because the claim that some banks were too big to fail was first
made in connection with the demise of Continental Illinois
in 1984. However, our use of a long sample period may give
rise to certain concerns. For example, several bank regulations
were introduced in the post-1984 period. One in particular,
the depositor preference rule, introduced in 1993, could be
important because it likely increased the compensation that
bondholders demand to invest in banks. However, we have
year-quarter fixed effects in all of our models. Further, limiting the sample period to the years after 1994 does not affect
our key findings in any meaningful way.
Another potential concern with the length of the sample
period is that it allows for several changes in the top-five firms
in each sector of activity, either because of firms’ different
FRBNY Economic Policy Review / December 2014
37
organic growth rates or because of mergers and acquisitions.
Recall that we rank firms in each sector of activity according
to their size each year. Again, shortening the sample period
and restricting it to, for example, the last decade does not
affect our key findings.
Yet another potential concern derives from our focus on
the top-five firms in each sector of activity. The number of
firms investors perceive to be too big to fail is likely to vary
over time and across sectors of activity. We experimented with
other cutoffs, including using the top-ten firms in each sector
of activity, and obtained similar results.
5.2 Is the Too-Big-to-Fail Discount
Economically Relevant?
The evidence presented thus far indicates that the largest
banks do benefit from a discount in the bond market that is
statistically different from zero. A related question is whether
this discount is economically meaningful. A possible way to
investigate this question is to compute the savings that the
largest banks enjoy per bond issue relative to their smaller
counterparts.
Looking at Table 2, we see that the largest banks that issue
bonds rated double A benefit from a reduction in their cost
of bond financing of about 121 basis points compared with
smaller banks that also issue double-A-rated bonds. The
largest banks that issue bonds rated single A benefit from a
reduction of about 31 basis points in the cost of bond financing. Taking into account the average bond issue by the largest
banks in each group, this reduction in spreads translates into
savings of about $80 million and $3 million for an average
issue, respectively.
As noted above, these calculations will likely overestimate
the too-big-to-fail subsidy that the largest banks enjoy in the
bond market. A more conservative way of estimating that subsidy is to determine the additional cost savings of the largest
banks (relative to their smaller peers) as opposed to the cost
savings that the largest nonbanks enjoy (also relative to their
smaller peers). Table 5 shows that the discount (relative to
their smaller peers) of the largest banks that issue bonds rated
38
Evidence from the Bond Market
double A is about 91 basis points bigger than the discount for
the largest nonbanks relative to their smaller peers. This translates into cost savings for the largest banks of about $60 million for an average bond issue. Doing the same exercise for the
largest banks that issue bonds rated single A reveals that they
enjoy cost savings of about $1.5 million.
In sum, the findings reported in this section confirm the
results from models 1 and 2 that the largest banks benefit
from a bigger discount (relative to smaller banks) when they
raise funding in the bond market than do either the largest
nonbank financial institutions or the largest nonfinancial
corporations. The results reported in this section further
show that the discount the largest banks enjoy is statistically
different from that of the largest nonbanks or the largest nonfinancial corporations. This difference suggests that investors
believe that the largest banks are likelier to be classified as too
big to fail, and thus to be rescued if they run into financial
trouble, than either the largest nonbanks or the largest nonfinancial corporations.
6. Conclusion
The evidence presented in this article—demonstrating the
additional discount that bond investors offer the largest banks
compared with the return they demand from the largest nonbanks and nonfinancial corporations—is novel and consistent
with the idea that investors perceive the largest U.S. banks to
be too big to fail.
Since the sample ends in 2009, these findings do not reflect
any changes in bond investors’ expectations resulting from the
regulatory interventions that occurred during the financial
crisis. Similarly, our findings do not account for any effects
that the regulatory changes introduced following the financial crisis may have had, in particular those changes aimed at
addressing the too-big-to-fail problem. However, our findings
are pertinent to the ongoing debate on requiring bank holding
companies to raise part of their funding with long-term bonds,
particularly if the post-crisis regulatory changes are unable to
fully address the too-big-to-fail status of the largest banks.
References
Acharya, V. V., D. Anginer, and J. Warburton. 2013. “The End of
Market Discipline? Investor Expectations of Implicit State
Guarantees.” Unpublished paper, New York University, Stern
School of Business.
Afonso, G, J. A. C. Santos, and J. Traina. 2014. “Do ‘Too Big To Fail’
Banks Take on More Risk?” Federal Reserve Bank of New York
Economic Policy Review 20, no. 2 (December): 41-58.
Anginer, D., and A. J. Warburton. 2014. “The Chrysler Effect: The
Impact of Government Intervention on Borrowing Costs.”
Journal of Banking and Finance 40: 62-79.
Balasubramnian, B., and K. B. Cyree. 2011. “Market Discipline of
Banks: Why Are Yield Spreads on Bank-Issued Subordinated
Notes and Debentures Not Sensitive to Bank Risks?” Journal of
Banking and Finance 35, no. 1: 21-35.
Baker, D., and T. McArthur. 2009. “The Value of the ‘Too Big to Fail’
Big Bank Subsidy.” Center for Economic and Policy Research
Issue Brief, September.
Brandao Marques, L., R. Correa, and H. Sapriza. 2013. “International
Evidence on Government Support and Risk Taking in the Banking
Sector.” International Monetary Fund Working Paper no. 13/94.
Brewer, E., and J. Jagtiani. 2007. “How Much Would Banks Be
Willing to Pay to Become ‘Too-Big-To-Fail’ and to Capture
Other Benefits?” Federal Reserve Bank of Kansas City Research
Working Paper no. 07-05.
Demirguc-Kunt, A., and H. Huizinga. 2010. “Are Banks Too Big to
Fail or Too Big to Save? International Evidence from Equity Prices
and CDS Spreads.” World Bank Policy Research Working Paper
no. 5360.
Hakenes, H., and I. Schnabel. 2010. “Banks without Parachutes:
Competitive Effects of Government Bail-Out Policies.” Journal
of Financial Stability 6: 156–68.
Flannery, M. J., and S. M. Sorescu. 1996. “Evidence of Bank Market
Discipline in Subordinated Debenture Yields: 1983-1991.”
Journal of Finance 51, no. 4: 1347-77.
Haldane, A. G. 2010. “The $100 Billion Question.” BIS Review,
no. 40/2010. Available at http://www.bis.org/review/r100406d.pdf.
Jacewitz, S., and J. Pogach. 2013. “Deposit Rate Advantages at the
Largest Banks.” Unpublished paper, Federal Deposit Insurance
Corporation.
Li, Z., S. Qu, and J. Zhang. 2011. “Quantifying the Value of Implicit
Government Guarantees for Large Financial Institutions.”
Moody’s Analytics Quantitative Research Group, January.
Molyneux, P., K. Schaeck, and T. Zhou. 2010. “‘Too-Big-to-Fail’ and its
Impact on Safety Net Subsidies and Systemic Risk.” Unpublished
paper, Bangor Business School.
Morgan, D. P., and K. J. Stiroh. 2005. “Too Big to Fail after All These
Years.” Federal Reserve Bank of New York Staff Reports,
no. 220, September.
O’Hara, M., and W. Shaw. 1990. “Deposit Insurance and Wealth
Effects: The Value of Being ‘Too Big to Fail.’” Journal of
Finance 45, no. 5: 1587-1600.
Penas, M. F., and H. Unal. 2004. “Gains in Bank Mergers: Evidence from
the Bond Markets.” Journal of Financial Economics 74: 149-79.
Rime, B. 2005. “Do ‘Too Big To Fail’ Expectations Boost Large Banks’
Issuer Ratings?” Unpublished paper, Swiss National Bank.
Gadanecz, B., K. Tsatsaronis, and Y. Altunbas. 2012. “Spoilt and
Lazy: The Impact of State Support on Bank Behavior in the
International Loan Market.” International Journal of
Central Banking 8, no. 4: 121-73.
Sironi, A. 2003. “Testing for Market Discipline in the European
Banking Industry: Evidence from Subordinated Debt Issues.”
Journal of Money, Credit, and Banking 35, no. 1: 443-72.
Gropp, R., H. Hakenes, and I. Schnabel. 2011. “Competition, RiskShifting, and Public Bail-Out Policies.” Review of Financial
Studies 24, no. 6: 2084-120.
Ueda, K., and B. Weder di Mauro. 2011. “Quantifying the Value of
the Subsidy for Systemically Important Financial Institutions.”
International Monetary Fund Working Paper no. 12/128.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2014
39
Gara Afonso, João A. C. Santos, and James Traina
Do “Too-Big-to-Fail” Banks
Take On More Risk?
• Large or complex banks might have a
greater appetite for risk if they expect
future rescues.
• Using data for more than 200 banks in
45 countries, the authors find higher
levels of impaired loans after an increase
in government support, as measured by
Fitch Ratings’ support rating floors (SRFs).
• A one-notch rise in the SRF increases an
average bank’s impaired loan ratio by roughly
8 percent; the authors show similar effects
on net charge-offs and for U.S. banks only.
• The authors also show that riskier banks
are more likely to take advantage of
potential government support.
1. Introduction
I
n 1984, U.S. regulators made the unprecedented move
of insuring all of Continental Illinois’s liabilities. The
Comptroller of the Currency indicated during the hearings
after Continental’s resolution that regulators would not allow
the eleven largest banks in the Unites States to fail. Ever
since, there have been many concerns with banks deemed
“too big to fail.”1
These concerns derive from the belief that the too-big-tofail status gives large banks a competitive edge and incentives
to take on additional risk. If investors believe the largest
banks are too big to fail, they will be willing to offer them
funding at a discount. Together with expectations of rescues,
this discount gives the too-big-to-fail banks incentives to
engage in riskier activities. This, in turn, could drive the
smaller banks that compete with them to take on further risks,
1
• The findings suggest that banks
classified by rating agencies as more
likely to receive government support
engage in more risk taking.
Gara Afonso is an economist, João A. C. Santos a vice president, and
James Traina a former senior research analyst at the Federal Reserve
Bank of New York.
Correspondence: gara.afonso@ny.frb.org
Continental Illinois, which was the seventh-largest bank by deposits,
experienced runs by large depositors following news that it had incurred
significant losses in its loan portfolio. Concerns that a failure of Continental
would have significant adverse effects on other banks that had deposits with
it led the Federal Reserve Board, the Federal Deposit Insurance Corporation
(FDIC), and the Comptroller of the Currency, together with twenty-four
U.S. banks, to announce a $7.3 billion bailout. The rescue package comprised
a $2 billion capital injection by the FDIC and the group of twenty-four banks
and a $5.3 billion unsecured line of credit from the banks.
The authors thank Christian Cabanilla, Nicola Cetorelli, Mark Flannery, David
Marqués-Ibañez, Stavros Peristiani, William Riordan, and Tony Rodrigues for
valuable comments. They are grateful to Alex Entz for research assistance. The views
expressed in this article are those of the authors and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
41
exacerbating the negative effects of having too-big-to-fail
banks in the financial system.
The debate around too-big-to-fail banks has given
rise to a large literature. Part of this literature attempts to
determine whether bank investors, including depositors,
believe the largest banks are too big to fail. Some studies
seek to answer this question by investigating spreads on
bank bonds (Flannery and Sorescu 1996; Sironi 2003;
Morgan and Stiroh 2005; Anginer and Warburton 2010;
Balasubramnian and Cyree 2011; Santos, forthcoming). Other
studies consider spreads on bank credit default swap contracts
(Demirgüç-Kunt and Huizinga 2013; Li, Qu, and Zhang
2011), bank stock returns (Correa et al. 2012), and deposit
costs (Baker and McArthur 2009). Yet others focus on the
premiums that banks pay in mergers and acquisitions (Brewer
and Jagtiani 2007; Molyneux, Schaeck, and Zhou 2011).
Another part of that literature investigates whether
too-big-to-fail banks behave differently by looking at
balance-sheet data (Gropp, Hakenes, and Schnabel 2011),
syndicated loans (Gadanecz, Tsatsaronis, and Altunbas
2012), and bank z-scores (Brandão Marques, Correa, and
Sapriza 2013), among other measures.
Our paper is closer to the latter studies in that we are also
interested in finding out whether the too-big-to-fail status
affects bank behavior. Specifically, we study whether banks
that rating agencies classify as likely to receive government
support increase their risk-taking.
An important novelty of our paper is the way we measure
the likelihood of a bank receiving government support.
Previous studies, including Haldane (2010), Lindh and
Schich (2012), and Hau, Langfield, and Marqués-Ibañez
(2013), attempt to infer support from the difference between
Moody’s all-in credit ratings (long-term bank deposit
ratings, which capture a bank’s ability to repay its deposit
obligations and include external support) and Moody’s
stand-alone ratings (bank financial strength ratings, which
exclude external support). The difference between Moody’s
all-in credit and stand-alone ratings is commonly known as a
ratings “uplift.” Using uplifts, however, presents two potential
issues. First, a change in uplift may arise from movement in
either of the two underlying ratings, with completely different
implications. Second, uplift incorporates any type of external
support, including from governments, parent companies,
and other institutions.
To avoid the first concern, some studies rely on support
ratings issued by Fitch Ratings (Gropp, Hakenes, and
Schnabel [2011] and Molyneux, Schaeck, and Zhou
[2010], among others). As with uplift, support ratings also
include institutional, cooperative, local government, and
regional government support. We sidestep both problems
42
Do “Too-Big-to-Fail” Banks Take On More Risk?
by considering a new Fitch rating. Starting in March 2007,
Fitch began to issue support rating floors (SRFs), which
reflect its opinion of potential sovereign support only
(including a government’s ability to support a bank). The
main advantage of using this rating is that, in contrast with
earlier approaches used in the literature, the support rating
floor explicitly captures government support. That is, it
does not incorporate other forms of external support, such
as the institutional support of a high-holder in a banking
organization to a bank within its own hierarchy.2
The results of our investigation show that a greater
likelihood of government support leads to a rise in bank risktaking. Following an increase in government support, we see
a larger volume of bank lending becoming impaired. Further,
and in line with this finding, our results show that stronger
government support translates into an increase in net chargeoffs. Additionally, we find that the effect of government
support on impaired loans is stronger for riskier banks than
safer ones, as measured by their issuer default ratings.
Our findings offer novel evidence that government support
does play a role in bank risk-taking incentives. The results are
also important because they already include the effects of the
government interventions undertaken throughout the latest
financial crisis. At the same time, however, not enough time
has elapsed since the crisis for our results to reflect the impact
of the regulatory changes enacted in its wake.
The rest of our paper is organized as follows. The next section
introduces our measure of government support. Section 3
describes the data sources and characterizes our sample.
Section 4 introduces our methodology. Section 5 discusses
our results. Section 6 presents robustness analysis. Section 7
concludes with some final remarks.
2
Fitch Ratings (2013a) explicitly defines support rating floors as based on
potential sovereign support (not on the intrinsic credit quality of the bank).
In the case of the landesbanks, Fitch assumes that Germany’s and the German
states’ creditworthiness are linked. For example, in August 2013, Landesbank
Baden-Wuerttemberg (LBBW) had a support rating floor of A+ even though
Fitch does not rate the State of Baden-Wuerttemberg. The assessment
implicitly assumes that the creditworthiness of the support “is underpinned
by the strength of the German solidarity system, which links the state’s
creditworthiness to that of the Federal Republic of Germany (AAA/Stable)”
(Fitch Ratings 2013b).
2. Measuring the Likelihood of
Government Support
There are a number of different methods for measuring sovereign
support based on rating agency assessments. Previous work
uses two ratings published by Moody’s to derive a measure of
government support (Haldane [2010], Lindh and Schich [2012],
and Hau, Langfield, and Marqués-Ibañez [2013], among others).
Moody’s issues bank deposit ratings based on its opinion of a
bank’s ability to repay punctually its deposit obligations. These
ratings are all-in credit ratings that reflect intrinsic financial
strength, sovereign transfer risk (for foreign currency deposits),
and both implicit and explicit external support elements. Moody’s
also issues bank financial strength ratings, which exclude
sovereign risk and external support. Uplifts—calculated as the
difference between these two ratings—provide an estimate of
the implicit guarantees. This measure incorporates any type
of external support (not just sovereign support), including
institutional backing from parent companies. To control for this
support, some recent studies exclude all bank subsidiaries from
their samples and focus their analysis on high-holders of banking
organizations only (Brandão Marques, Correa, and Sapriza
[2013], among others). Uplifts also capture cooperative, local
government, and regional government support.
Although intuitive, this methodology assumes a linear
functional form for the difference between these two ratings,
but the relationship between external support and stand-alone
ratings may be more complex. It also makes it difficult to
identify the source of variation in uplifts. For example, suppose
there is a one-notch increase in the stand-alone rating, but
no change in the all-in credit rating. Uplift would decrease,
indicating weaker external support when, in practice, there
has been no change. Moreover, even if both ratings were to
change, differences in Moody’s publication timing would lead
to spurious variation in external support.
An alternative approach relies on ratings issued by Fitch
that explicitly measure external support, independent of the
intrinsic credit quality of the bank. Support ratings (SRs) rely
on Fitch’s assessment of a supporter’s propensity and ability
to support a bank. Supporters can be of two types: sovereign
states and institutional owners. Studies that use SRs include
Gadanecz, Tsatsaronis, and Altunbas (2012) and Gropp,
Hakenes, and Schnabel (2011).
In addition to support ratings, Fitch issues support rating
floors based on its opinion of potential sovereign support
only (including a government’s ability to support a bank).3
3
According to Fitch Ratings (2013a), support typically extends to the
following obligations: senior debt (secured and unsecured), including
insured and uninsured deposits (retail, wholesale, and interbank); obligations
Comparison of Ratings Issued by
Moody’s and Fitch Ratings
Moody’s
Longterm
Bank
bank
deposit financial
rating strength
Fitch Ratings
Longterm
issuer
default
rating
Support
rating
Support
rating
floor
Intrinsic
credit quality
✓
✓
✓
✗
✗
Institutional
support
✓
✗
✓
✓
✗
Sovereign
support
✓
✗
✓
✓
✓
Sources: Moody’s and Fitch Ratings.
Notes: Comparison of several ratings issued by Moody’s and Fitch Ratings
that are typically used in the calculation of government support. A check
mark denotes that the definition of a given rating includes one of three
characteristics listed in the table above. An “x” indicates that a characteristic is not included in the definition of the rating. For example, bank
financial strength measures intrinsic credit quality, but not institutional or
sovereign support.
The main difference with respect to SRs is that SRFs do not
incorporate external support other than sovereign support,
such as the institutional support of a high-holder in a banking
organization to a bank within its own hierarchy. Isolating the
support coming from the government is crucial to addressing
the question of whether too-big-to-fail banks increase their
risk-taking, because, in contrast to other sources of external
support, sovereign support is typically unpriced and not
risk-sensitive. The exhibit shows a comparison of these ratingsbased approaches to measuring sovereign support.
To stress the difference between these two ratings, let
us consider the case of Bank of America. Table 1 shows the
history of changes in support ratings and support rating floors
for Bank of America Corporation (the parent company) and
Bank of America National Association (the largest national
bank within the organization). Fitch expresses SRs on a fivenotch, 1-to-5 scale, where a rating of 1 denotes a bank with
extremely high probability of external support. SRFs use
the AAA long-term scale, where AAA ratings indicate an
extremely high probability of government support. SRFs
include one additional point on the scale, “no floor” (NF),
arising from derivatives transactions and from legally enforceable guarantees
and indemnities, letters of credit, and acceptances; trade receivables; and
obligations arising from court judgments.
FRBNY Economic Policy Review / December 2014
43
Table 1
Example of Fitch Ratings
Bank of America
Corporation
Date
06/01/88
02/01/89
02/15/89
06/01/90
02/01/91
05/27/94
10/03/95
04/11/96
04/26/96
05/20/96
10/01/98
10/15/99
07/22/03
09/29/03
04/01/04
02/15/07
03/16/07
07/16/08
01/16/09
12/15/11
Bank of America
National Association
IDR
SR
SRF
IDR
SR
SRF
BBB
BBB+
A
A
A+
A+
A+
A
AAA+
AAAAAAAA
AAAA
AA
A+
A+
A
•
•
•
•
•
•
5
5
5
5
5
5
5
5
5
5
5
5
1
1
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
NF
NF
A+
A
•
•
•
•
•
AAAA
AA
AA
AA
AA
AA
AA
AA
AAAA
AA
AAA+
A
•
•
•
1
1
1
1
1
1
1
1
2
2
2
1
1
1
1
1
1
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
AAA+
A
Source: Fitch Ratings.
Notes: History of long-term issuer default ratings (IDRs), support ratings
(SRs), and support rating floors (SRFs) of Bank of America Corporation
and Bank of America National Association. NF is “no floor.”
bringing the total number of notches to twenty. According to
Fitch, NF designates no reasonable presumption of potential
support and translates to a probability of support of less than
40 percent (Fitch Ratings 2013a).
From March 16, 2007, to January 16, 2009, Bank of
America Corporation (the parent) had the lowest level of
external support (SR = 5), while Bank of America National
Association enjoyed the highest level of external support
(SR = 1). By looking at support ratings only, we cannot
disentangle if the strong support of Bank of America
National Association comes from the government or from
the parent company. To answer this question, we turn to its
support rating floor. The SRF of Bank of America National
Association was A- over this period, indicative of strong
government support.
The evolution of Bank of America National Association’s
support rating floors also shows how sovereign support to the
national bank heightened two notches in January 2009 and
44
Do “Too-Big-to-Fail” Banks Take On More Risk?
lessened one notch in December 2011, while external support
(measured by SRs) remained constant. The difference in granularity between these two ratings is yet another advantage to
using SRFs over SRs since they allow for higher precision and
more variability in support.
A similar measure based on S&P ratings is currently not
available since S&P does not issue ratings that allow measurement of sovereign support.
3. Data and Sample
Characterization
3.1 Data
The data for this paper come from several sources. We
use Bureau van Dijk’s Bankscope to gather balance-sheet
data on banks in our sample, including our key measures
of bank risk-taking—impaired loans and net charge-offs.
In addition, we use two data sets from Fitch Ratings: one
containing information on government support ratings
(described in detail in section 2 above) and the other
containing information on bank strength ratings (long-term
issuer default ratings [IDRs]). IDRs reflect Fitch’s opinion
on an entity’s relative vulnerability to default on its financial
obligations. IDRs are Fitch’s primary issuer rating for financial
institutions and are expressed on a AAA long-term scale,
where AAA ratings denote the lowest expectation of default.
IDRs incorporate not only intrinsic strength, but also external
support. Even though stand-alone ratings are a cleaner
measure of a bank’s intrinsic strength than IDRs, we cannot
rely on these ratings in our analysis because of the lack of a
consistent time series during our sample period.4
4
Historically, Fitch issued individual ratings on an A-E scale to assess a bank’s
creditworthiness on a stand-alone basis. Similar to Moody’s bank financial
strength ratings, these ratings aimed to capture the strength of a bank if it
was unable to rely on external support. On March 7, 2011, Fitch announced a
revision to the methodology used to calculate the stand-alone ratings, as well
as a change from a nine-point scale (using letter ratings such as A and A/B) to
a lowercase variation of the traditional nineteen-point long-term rating scale
(using letter ratings such as aaa and aa+). On July 20, 2011, Fitch introduced new
stand-alone ratings called viability ratings, designed to reflect the same core risks
as individual ratings but with renewed definitions and greater granularity.
Chart 1
Government Support by Origin
Average rating
Average rating (no NF)
BBB-
A
BB+
Foreign
A-
All
BBB+
BB
BBB+
U.S.
All
BBB
B
Foreign
BBB-
BU.S.
CCC
BB+
CC
13
1/
1/
20
12
1/
1/
20
11
1/
1/
20
10
20
1/
1/
1/
1/
20
08
20
1/
1/
1/
1/
20
13
12
1/
1/
20
11
1/
1/
20
10
1/
1/
20
09
1/
1/
20
08
20
1/
1/
09
BB
C
Source: Authors’ calculations, based on data from Fitch Ratings.
Notes: The left panel displays the average government support (measured by the support rating floor [SRF]) from March 16, 2007, to August 15, 2013,
including “no floor” (NF) ratings. The right panel shows the average SRF excluding NF ratings. Trend lines capture daily ratings.
3.2 Sample Characterization
To construct our data set, we start with the universe of banks
that have support rating floors, which Fitch began issuing on
March 16, 2007. Though the most recent ratings are easily
accessible online, historical ratings need manual collection.
Our sample includes daily SRF observations for 612 banks
(bank holding companies, commercial banks, and savings
banks) from March 16, 2007, to August 15, 2013. The data
span 92 countries, with 182 banks from the United States.
Our sample of changes in support rating floors
comprises increases and decreases in ratings. The first
change in our sample occurs on July 2, 2007, and the last
one on August 14, 2013. There are 446 changes in SRFs
(234 increases and 212 decreases) across 234 unique banks
and 177 unique event dates. On average, each change shifts
the rating about two notches.
The left panel of Chart 1 seems to support the commonly
understood idea that foreign countries tend to provide
stronger support to their banks than the United States does.
We see the average support rating floor of a foreign bank is
about four times larger than that of a U.S. bank.5 Interestingly,
this pattern changes dramatically when we zoom in on the
set of banks with an SRF different from an NF rating: the
5
As standard in the ratings literature, we assign numeric values to the notches
on the rating scale, where a value of nineteen denotes a AAA rating and zero
a “no floor” rating.
“supported” banks. As the right panel of Chart 1 shows,
average sovereign support remains slightly humped in foreign
countries (according to Fitch’s ratings), but the pattern
changes significantly for the United States, where, over the last
six years, average government support has increased markedly.
Since 2010, average sovereign support for U.S. banks has been
stronger than that for foreign banks.
This difference in patterns seems to be driven by the
larger proportion of U.S. banks that have a probability of
government support lower than 40 percent. The data show
that 80 percent of banks in the United States have “no floor”
ratings compared with 21 percent in foreign countries. The
larger the number of banks in a country with “no floor”
ratings, the starker the difference between the left and right
panels of Chart 1. Whether or not government support to
banks is more prevalent in the United States than abroad
depends on whether we take “no floor” ratings into account.
Making this distinction matters because it portrays a
different picture of how government support has evolved
in the United States.6
6
The heat map in Chart 4 highlights the unique character of the “no floor” (NF)
rating. At first glance, since SRFs act as a floor for IDRs, one might think the NF
rating is located one notch below D on the SRF scale. However, the distribution of
IDRs for banks with NF SRFs is significantly different from IDRs for banks with
SRFs expressed on the AAA scale. While banks with SRFs ranging from CCC to
AA- typically have an IDR between zero to two notches higher, a bank with an NF
SRF is more likely to have a BBB or A- IDR rating. This suggests a definition of
average government support that excludes banks with NF ratings.
FRBNY Economic Policy Review / December 2014
45
Chart 2
Government Support by Country
Average rating
A
Support rating floor
Support rating floor (no NF)
ABBB+
BBB
BBBBB+
BB
BBB+
0
0
0
40
0
3
0
7
0
AT
BE
DE
SA
36
18
0
0
0
73
0
22
16
VE OM JP
PA
0
80
0
0
5
22
0
TH
IN
ES
SI
IT
B
BCCC
CC
C
FR KW AE
CH QA
GB CA
KR
SG BH
CN US
Source: Authors’ calculations, based on data from Fitch Ratings.
Notes: Average government support (measured by the support rating floor) by country from March 16, 2007, to August 15, 2013. Dark green bars
represent average SRFs including “no floor” ratings; light green bars exclude NF ratings. The numbers in the middle of the bars indicate the percentage
of “no floor” ratings in each country: France (FR), Kuwait (KW), United Arab Emirates (AE), Switzerland (CH), Qatar (QA), Austria (AT), Belgium
(BE), Germany (DE), Saudi Arabia (SA), United Kingdom (GB), Canada (CA), Republic of Korea (KR), Singapore (SG), Bahrain (BH), Venezuela (VE),
Oman (OM), Japan (JP), Panama (PA), China (CN), United States (US), Thailand (TH), India (IN), Spain (ES), Slovenia (SI), and Italy (IT).
Chart 2 captures this idea. It presents, for the top twenty-five
countries with the strongest government support, average
support rating floors including “no floor” ratings (dark green)
and excluding “no floor” ratings (light green).
The cases of the United States and Venezuela stand
out in that overall average sovereign support is weak but
average support to banks that have a rating other than “no
floor” (the “supported” banks) is very strong. Consistent
with the findings of Ueda and Weder di Mauro (2012),
banks headquartered in Switzerland, France, and Germany
enjoy high probability of sovereign support. We also find
that Arabic countries, including Kuwait, the United Arab
Emirates, and Qatar, provide strong support to their banks.
Table 2 shows the average level of sovereign support for the
top twenty-five countries with the strongest government
support as well as the number of banks per country rated
by Fitch. There is significant heterogeneity in the number
of rated banks per country, perhaps reflective of differences
in size of each country’s financial system and in the level of
concentration of their banking sectors.
For information on credit quality and exposure to default,
we use long-term issuer default ratings issued by Fitch. For
46
Do “Too-Big-to-Fail” Banks Take On More Risk?
each bank in our sample, we obtain the history of changes in
IDRs from January 1, 1988, to August 15, 2013. To present
summary statistics on a comparable sample, we restrict our
attention to IDR observations for which we also see an SRF.
Chart 3 shows the distribution of SRFs (left) and IDRs (right)
for the sample of 612 banks.
Recall from sections 2 and 3 that support rating floors
reflect government support while long-term issuer default
ratings incorporate both intrinsic and external support. As
such, a bank’s SRF acts as a floor for its IDR. Chart 4 highlights
this relationship by presenting the distribution of IDRs by
SRFs. The intensity of each symbol denotes the frequency (that
is, a darker square indicates a more frequent relationship).
As expected, many bank ratings lie on the diagonal,
indicating that Fitch’s assessment of a bank’s relative
vulnerability to default and of a government's propensity
to support a bank are identical. The rest of the observations
are on the upper diagonals of the heat map, which denote
that the overall strength of a bank exceeds its sovereign
support. It is also interesting to note that banks rated
with a probability of sovereign support of less than
40 percent (SRF = NF) are rated with IDRs ranging
Table 2
Average Government Support
Name
SRF (no NF)
SRF
Percent NF
Banks
Days
Observations
1
France
14.3
14.3
0
5
2,345
9,303
2
Kuwait
14.3
14.3
0
5
2,283
11,415
3
United Arab Emirates
14.1
14.1
0
8
2,283
15,866
4
Switzerland
14.0
8.4
40
5
2,345
11,725
5
Qatar
13.9
13.9
0
5
2,283
9,283
6
Austria
13.8
13.4
3
5
2,345
7,295
7
Belgium
13.8
13.8
0
5
2,345
9,687
8
Germany
13.6
12.6
7
7
2,345
12,215
9
Saudi Arabia
13.5
13.5
0
9
2,283
20,547
10
United Kingdom
13.4
8.7
36
20
2,345
39,843
11
Canada
13.2
10.8
18
6
2,345
13,100
12
Republic of Korea
12.8
12.8
0
5
2,283
11,415
13
Singapore
12.7
12.7
0
5
2,345
11,725
14
Bahrain
12.6
12.6
0
5
2,283
8,589
15
Venezuela
12.3
3.4
73
8
2,345
16,901
16
Oman
12.3
12.3
0
5
2,283
8,330
17
Japan
12.2
9.5
22
10
2,345
21,480
18
Panama
11.9
10.0
16
7
2,283
12,222
19
China
11.8
11.8
0
13
2,283
14,233
20
United States
11.1
2.3
80
186
2,345
342,905
21
Thailand
10.5
10.5
0
8
2,345
14,836
22
India
10.5
10.5
0
8
2,345
13,949
23
Spain
10.4
9.9
5
17
2,345
22,677
24
Slovenia
10.2
8.0
22
5
2,283
11,240
25
Italy
10.0
10.0
0
8
2,345
16,365
Source: Authors’ calculations, based on data from Fitch Ratings.
Notes: The table reports each country’s mean support rating floor (SRF) for countries with at least five rated banks (top twenty-five only). Ratings were
issued from March 16, 2007, to August 15, 2013. NF is “no floor.”
from D to AA+. Having risky banks among those with a
probability of sovereign support of less than 40 percent
suggests that risk alone does not drive the probability of
government support. This would be the case, for example,
for small banks that may not receive government support
regardless of their overall financial strength.
Finally, we use the Bankscope database to augment
the ratings data with quarterly information on bank
characteristics spanning 2007:Q1 to 2013:Q3. Fitch issues
support rating floors at the entity level, so we keep in our
sample parent banks and their subsidiaries when there are
multiple entities for a consolidated bank in Bankscope.
The matched sample consists of 11,929 bank-quarter
observations for 601 banks.
Because of the global nature of our data, we are missing
balance-sheet information for approximately 59 percent
of our bank-quarter observations for which we have SRFs.
To alleviate this problem, we linearly interpolate adjacent
data if they are missing for less than one year in duration.
Interpolation recovers approximately 15 percent of our
potential data, reducing the proportion missing to 44 percent.7
After matching and interpolation, we further limit our sample
7
Results are qualitatively similar in the analysis without interpolation.
FRBNY Economic Policy Review / December 2014
47
Chart 3
Distribution of Fitch Ratings
A
A+
AA
AA
AA
+
A-
B
B+
BB
BB
BB
+
BB
BBB
B
BB
B+
AA
A
A+
A-
B-
BB
BB
BB
CC
BB
B
BB
B+
0
+
0
BB
5
B
5
B+
10
B-
10
C
15
C
15
Issuer default rating (no NF)
C
Support rating floor (no NF)
Percent
B-
20
C
Percent
CC
20
Source: Authors’ calculations, based on data from Fitch Ratings.
Notes: Histograms include observations for banks with support rating floors and issuer default ratings from March 16, 2007, to August 15, 2013.
Both panels exclude observations where the banks have a support rating floor of “no floor” (NF).
Chart 4
Chart 5
Fitch Ratings Heat Map
Distribution of Bank Size by Government Support
Issuer default rating
Assets in billions of dollars
800
AA+
AA
AA−
A+
A
A−
BBB+
BBB
BBB−
BB+
BB
BB−
B+
B
B−
CCC
CC
C
D
700
600
500
400
300
200
100
0
NF
A A+ AA-
NF
C
CC
CCC
B−
B
B+
BB−
BB
BB+
BBB−
BBB
BBB+
A−
A
A+
AA−
AA
AA+
AAA
Support rating floor
Support rating floor
Source: Authors’ calculations, based on data from
Fitch Ratings.
Notes: The chart shows the distribution of issuer default
ratings by support rating floor. The intensity of each
symbol indicates the frequency; darker squares denote a
more frequent relationship.
Source: Authors’ calculations, based on data from
Fitch Ratings.
48
C B- B B+ BB- BB BB+ BB- BB BB+ AB B B
Notes: Chart 4 shows the distribution of issuer default ratings by
support rating floors. The intensity of each symbol indicates the
frequency;Do
darker
squares denote a Banks
more frequent
relationship.
“Too-Big-to-Fail”
Take On
More Risk?
Source: Authors’ calculations, based on data from Fitch Ratings
and Bureau van Dijk’s Bankscope.
Note: The chart shows total assets of banks with support rating
floors and issuer default ratings from March 16, 2007, to
August 15, 2013, by category of government support.
Table 3
Summary Statistics
Support Rating Floors
Total assets
Impaired loans
Net charge-offs
Return on assets
Tier 1 capital
Trading assets
NF
C-CCC
B
BB
BBB
A
AA-AAA
Total
Mean
110
4.2
53
92
150
600
370
200
Median
16
4.2
33
46
51
190
180
22
Standard deviation
380
•
45
110
180
780
690
500
Mean
2.50
1.81
3.23
2.48
2.78
2.24
1.82
2.48
Median
1.97
1.81
2.96
1.80
0.95
1.38
1.77
1.85
Standard deviation
2.46
•
1.99
2.44
4.56
2.77
0.45
2.61
Mean
0.66
0.44
0.66
0.34
0.17
0.50
0.07
0.59
Median
0.29
0.44
0.51
0.07
0.05
0.10
0.06
0.22
Standard deviation
1.02
•
0.66
0.56
0.27
1.22
0.11
1.01
Mean
0.17
1.09
0.25
0.64
0.55
0.40
0.41
0.27
Median
0.21
1.09
0.14
0.56
0.63
0.27
0.33
0.24
Standard deviation
0.59
•
0.57
0.50
0.85
0.45
0.30
0.59
Mean
11.34
6.44
8.45
8.99
7.78
11.24
6.03
10.89
Median
9.38
6.44
8.60
8.54
7.38
7.40
4.99
8.86
Standard deviation
11.16
•
1.79
3.00
2.99
14.23
2.45
11.08
Mean
1.16
0.10
2.22
2.07
3.21
3.72
3.14
1.83
Median
0.04
0.10
1.10
0.67
0.73
0.50
3.29
0.13
4.27
•
3.45
3.47
4.20
5.35
2.49
4.53
1,153
1
52
131
65
327
10
1,739
Standard deviation
Observations
Source: Authors’ calculations, based on data from Fitch Ratings and Bureau van Dijk’s Bankscope.
Notes: The table presents summary statistics on total assets and our risk variable ratios by bins of government support. We rely on the following
variables from Bankscope (series in parentheses): total assets (DATA2025), impaired loans (DATA2170), net charge-offs (DATA2150), net income
(DATA2115), tier 1 capital (DATA2140), and trading assets (DATA29190). We normalize each risk measure by total assets, converted to 2012
U.S. dollars and presented in millions. NF is “no floor.”
to banks with information on total assets, impaired loans, net
charge-offs, tier 1 capital, and trading assets. This step leads
to a final data set with 1,739 bank-quarter observations.
Most banks in the sample (75 percent) have investmentgrade ratings. Many (38 percent) also have government
support of BBB- or above. The median bank has total
assets of $22 billion, while the average bank has assets
of $200 billion. Size, however, changes significantly by level
of government support, with highly supported banks being
typically larger. The bank with a C-CCC rating (the lowest
SRF in our sample) has close to $4 billion in total assets while
those with an AA-AAA rating are almost 100 times larger on
average. Chart 5 shows this pattern, which is consistent with
the literature that documents a positive relationship between
size and government support.
Banks with a higher probability of government support
also have more trading assets on average. However, as shown
in Table 3, we do not find a similar pattern with return on
assets (ROA), impaired loans, net charge-offs, or tier 1 capital.
In our sample, the average bank has an ROA of 0.27 percent,
an impaired loan ratio of 2.48 percent, a net charge-off ratio of
0.59 percent, and a tier 1 capital ratio of 10.89 percent. Table 3
tabulates descriptive statistics for our sample.
FRBNY Economic Policy Review / December 2014
49
4. Methodology And Empirical
Strategy
The goal of our analysis is to investigate whether banks with
higher government support engage in riskier activities. To
test this hypothesis, we use a panel of bank-level data. After
matching and interpolating, we further limit our sample to
banks with information on total assets, impaired loans, net
charge-offs, tier 1 capital, and trading assets. This restriction
leads to a final panel data set with 1,739 bank-quarter
observations. Although 85 percent of our bank-quarter
observations correspond to domestic banks, our sample retains
a global nature, spanning 224 banks in 45 countries.
We first measure the riskiness of a bank’s activities by
the ratio of impaired loans to total assets. We also present
results for alternative measures of risk, including ratios of net
charge-offs, net income, tier 1 capital, and trading assets to
total assets.8 Specifically, we investigate whether the ratio of
impaired loans to total assets relates to government support
of banks. Since we expect that a bank’s response to sovereign
support might take time to show up on its balance sheet, we
estimate specifications of our model with progressively higher
lags for all right-hand-side variables. To that end, we estimate
the following model:
1)
Riskb,t = β * SRFb,t-i + δ * IDRb,t-i
+ η * Assetsb,t-i + μ * OtherRiskb,t-i
+ γ * Zb + τ * Xt + εb,t ,
where b indexes banks, t denotes time in quarters, and
i = {1,...,11} indicates the number of lags. The availability of
data determines the maximum number of lags (eleven). The
dependent variable Riskb,t is a measure of bank riskiness. In
our baseline specification, we measure riskiness as the ratio
of impaired loans to total assets. SRFb,t denotes the support
rating floor of bank b at the end of quarter t; IDRb,t indicates
the long-term issuer default rating of bank b at the end of
quarter t; and Assetsb,t is the natural logarithm of total assets
in U.S. dollars, normalized using the consumer price index.9
OtherRiskb,t is a vector of our remaining risk measures as bank
controls. In the baseline specification, this vector includes
net charge-offs/total assets, return on assets (net income/total
assets), tier 1 capital/total assets, and trading assets/total assets.
8
Data on these risk measures are from Bankscope. In particular, we
use the following series: DATA2170 (impaired loans), DATA2025 (total
assets), DATA2115 (net income), DATA2140 (tier 1 capital), DATA2150
(net charge-offs), and DATA29190 (total trading assets).
9
We use 2012 dollars as the baseline. We pull the “All Urban Consumers, All
Items, Not Seasonally Adjusted” series from Federal Reserve Economic Data.
50
Do “Too-Big-to-Fail” Banks Take On More Risk?
εb,t is the error term. All specifications control for country fixed
effects Zb and quarter-year fixed effects Xt. We also consider
specifications in which we control for bank-fixed effects
instead of country-fixed effects. We refer to this alternative
specification as Model 2. The standard errors are robust and
adjusted to control for clustering at the bank level.
Finally, since a bank’s creditworthiness will likely play a
role in the effect of government support on its risk-taking
activities, we also consider a version of our model that
includes the interaction between the support rating floor and
the long-term issuer default rating, φ * SRFb,t-i * IDRb,t-i.
5. Results
5.1 Impaired Loans
Impaired loans are those that are either in default or close
to default. These loans are typically behind in payments or
restructured from a previous loan. They constitute a good
measure of the amount of bad debt currently in the loan
portfolio of a bank. Regulatory agencies require banks to write
down loans as impaired under specific delinquency criteria,
which may vary by country. Typically, regulators classify loans
that are delinquent for ninety days (one quarter) as impaired.
In our analysis, we use impaired loans (from Bankscope) as
our baseline measure of a bank’s riskiness. The main hypothesis that we intend to test is that banks with higher government
support engage in riskier (lending) activities. Specifically, if
the level of government support affects bank preferences for
risk, we would expect that banks with stronger SRFs would
engage in riskier lending activity. This, in turn, implies that
more loans would become delinquent, resulting in an increase
in impaired loans in the following quarters.
Table 4 summarizes our results. It presents the value of the
coefficient β on the SRF in our models of risk for different
lags (one to eleven quarters) of sovereign support. The top
rows of panel A show the effect of government support on
the level of impaired loans. The main finding is that stronger
sovereign support is associated with an increase in the ratio
of impaired loans to total assets. In the model that includes
country-fixed effects but no bank-fixed effects (Model 1),
this result is statistically significant at the 1 percent level and
the effect is economically meaningful; each notch increase in
the SRF increases the impaired loan ratio by just under 0.2,
which is an approximately 8 percent increase for the average
Table 4
Bank Risk Response to Government Support
Panel A: Risk Measures
Variable
Model
Lags
1
Impaired loans
Net charge-offs
2
3
4
5
6
7
8
9
10
11
1
0.17***
0.18***
0.18***
0.19***
0.18***
0.19***
0.20***
0.20***
0.20***
0.20***
0.21***
2
0.01
0.01
0.01
0.02
0.02
0.03
0.24*
0.26**
0.24*
0.20**
0.12***
1
-0.00
0.00
0.00
0.00
0.01
0.01
0.01
0.01
0.01
0.01
0.01
0.02***
0.02**
0.02*
0.03***
0.02
0.08***
0.08***
0.05***
0.03***
0.06***
1,313
1,149
1,003
888
697
613
528
443
363
6
7
8
9
10
11
2
Observations
0.02***
1,491
790
Panel B: Other Measures
Variable
Model
Lags
1
Return on assets
Tier 1 capital
Trading assets
Observations
2
3
4
5
1
0.00
0.01
0.01
0.01
0.01
0.01
0.00
0.00
0.01
0.01
0.01
2
-0.00
-0.00
-0.00
-0.00
-0.00
-0.01
0.02
0.02
0.03
0.02
-0.02
1
0.38*
0.38*
0.39*
2
-0.04*
-0.04**
-0.05*
0.40*
-0.02
0.42*
0.42**
0.41**
0.40**
0.41**
0.36**
0.25*
0.04
0.17
1.36
1.50
1.76
1.32
0.85
1
0.06
0.07
0.07
0.06
0.06
0.05
0.05
0.05
0.05
0.06
0.06
2
-0.06
-0.08*
-0.08
-0.10
-0.02
-0.05
-0.09
-0.05
-0.03
0.02
0.06
1,313
1,149
1,003
528
443
363
1,491
888
790
697
613
Source: Authors’ calculations, based on data from Fitch Ratings and Bureau van Dijk’s Bankscope.
Notes: The table presents results on the relationship between government support and bank risk-taking. For each measure of bank risk, we report the
value of the estimated coefficient on the support rating floor for different lags (one to eleven quarters). Model 1 corresponds to the analysis with country-fixed effects and without bank-fixed effects. Model 2 includes bank-fixed effects, but no country-fixed effects. Each specification uses robust standard
errors clustered by bank.
*** Statistically significant at the 1 percent level.
** Statistically significant at the 5 percent level.
* Statistically significant at the 10 percent level.
bank. The effect is persistent and roughly constant through
the following ten quarters.
Results are similar but weaker in the analysis that includes
bank- instead of country-fixed effects (Model 2). In particular, we find a statistically and economically significant effect
of sovereign support on the proportion of a bank’s impaired
loans approximately seven quarters ahead. The lack of
within-bank variation in government support may drive this
weakness, as suggested by the lower t-statistics.
Chart 6 presents the relevant coefficients for both models. The circles and closed circles correspond, respectively,
to the values and significance at the 10 percent level of the
support-rating floor coefficient through time. This graphing
of our results illustrates the importance of timing after a
change in the SRF. The black line of Chart 6 shows that an
increase in sovereign support leads to a rise in the ratio of impaired loans as early as a quarter after the change in support
in the model with country-fixed effects. We also see that this
result is persistent and statistically significant through the
following ten quarters. The green line presents the results of
the specifications that control for bank-fixed effects (but no
country-fixed effects). An increase in government support to
a bank also leads to a higher impaired loan ratio, but the effect
is only significant seven quarters after the change.
FRBNY Economic Policy Review / December 2014
51
Chart 6
Chart 7
Effect of Government Support
on Impaired Loans
0.30
Effect of Government Support
on Net Charge-Offs
Coefficient
0.10
0.25
Country-fixed
Bank-fixed effects
0.06
0.15
0.04
0.10
0.02
Bank-fixed effects
0.05
1
2
3
4
5
6
7
Quarter
8
9
10
Country-fixed effects
0
11
-0.02
1
2
3
4
5
6
7
Quarter
8
9
10
11
Source: Authors’ calculations, based on data from Fitch Ratings
and Bureau van Dijk’s Bankscope.
Source: Authors’ calculations, based on data from Fitch Ratings
and Bureau van Dijk’s Bankscope.
Notes: The chart presents results on the relationship between
government support and impaired loans. The circles illustrate
the value of the estimated coefficient on the support rating floor
through time (one- to eleven-quarter lags). The closed circles
denote significance at the 10 percent level. The black and green
lines correspond to Models 1 and 2, respectively. Each
specification uses robust standard errors clustered by bank.
Notes: The table presents results on the relationship between
government support and net charge-offs. The circles illustrate
the value of the estimated coefficient on the support rating floor
through time (one- to eleven-quarter lags). The closed circles
denote significance at the 10 percent level. The black and green
lines correspond to Models 1 and 2, respectively. Each
specification uses robust standard errors clustered by bank.
5.2 Net Charge-Offs
For robustness, we also look at alternative measures of a
bank’s riskiness. Net charge-offs are often used as a proxy
for bank risk because they tend to increase with riskier
lending activities. They are defined as the difference
between charge-offs and recoveries, where charge-offs are
debts that a bank declares likely uncollectible and recoveries
are collections on debts that a bank had previously written
down as charge-offs. As with impaired loans, we scale
net charge-offs by the total assets of the bank. Similar to
our test based on impaired loans, if changes in sovereign
support affect bank preferences for risk, then we expect
that increases in support rating floors would lead to riskier
lending activity, resulting in an increase in net charge-offs.
The second set of rows in panel A of Table 4 presents the
results of the analysis where the dependent variable is net
charge-offs, with country-fixed (Model 1) and bank-fixed
(Model 2) effects. Our findings support and complement our
52
Bank-fixed effects
0.08
0.20
0
Coefficient
Do “Too-Big-to-Fail” Banks Take On More Risk?
previous result that stronger sovereign support is associated
with an increase in riskier lending activity. When we control
for bank-fixed effects (Model 2), we find that the effect is
statistically and economically meaningful, comprising a
change in net charge-offs of approximately 0.04 per SRF
notch, or 7 percent of an average bank’s net-charge-off level.
Chart 7 shows these results. The coefficients on sovereign
support are positive but not statistically significant in the
model with country-fixed effects.
The dynamics and timing of debt charge-offs are complex. On the one hand, there is guidance from governments
and regulators to encourage early charge-offs through tax
exemptions and regulatory enforcement. On the other hand,
banks still retain some discretion and may prefer to delay
charging off debt within the timing established by the regulatory guidelines. Consistent with this pattern in the timing of
charge-offs, we find that the effect is strongly significant for
the two quarters following a change in support; it becomes
weaker for the third to sixth quarters and then strongly
significant after seven quarters.
Table 5
Impaired Loan Response, Interaction
Panel A: Model 1
Coefficient
Lags
1
SRF
0.75**
(2.23)
SRF * IDR
IDR
Observations
2
0.81**
(2.24)
3
0.86**
(2.21)
4
0.93**
(2.22)
5
6
1.04**
1.05**
(2.39)
(2.38)
-0.04*
-0.04*
-0.04*
-0.05*
-0.06*
-0.06*
(-1.78)
(-1.80)
(-1.76)
(-1.79)
(-1.97)
(-1.94)
-0.46***
-0.45***
-0.44***
-0.41**
-0.39**
-0.38**
7
1.19**
(2.41)
-0.06**
(-2.00)
-0.37**
8
1.29**
(2.43)
-0.07**
(-2.04)
9
1.34**
(2.52)
11
1.35**
1.30**
(2.55)
-0.07**
(-2.14)
10
-0.07**
(-2.17)
(2.54)
-0.07**
(-2.14)
-0.35*
-0.33*
-0.33*
(-3.38)
(-3.18)
(-2.91)
(-2.58)
(-2.27)
(-2.17)
(-2.04)
(-1.97)
(-1.87)
(-1.90)
(-2.00)
-0.34**
1,491
1,313
1,149
1,003
888
790
697
613
528
443
363
7
8
9
10
11
Panel B: Model 2
Coefficient
Lags
1
SRF
SRF * IDR
IDR
Observations
2
3
4
0.41**
5
6
0.28
0.29
0.34
0.47*
0.35*
(1.35)
(1.31)
(1.61)
(2.01)
(3.88)
0.60***
(1.85)
(2.42)
(2.16)
(2.37)
(1.89)
(1.93)
-0.02
-0.02
-0.02
-0.03*
-0.04***
-0.04*
-0.04*
-0.02
-0.02
-0.02
-0.01
(-1.29)
(-1.26)
(-1.56)
(-1.93)
(-1.79)
(-1.88)
(-1.39)
(-1.66)
(-1.13)
(-1.33)
(-3.72)
0.63*
0.21**
0.80**
0.22***
0.63**
0.16
0.60**
-0.24
-0.12
0.03
0.19
0.25*
(-1.65)
(-0.88)
(0.20)
(1.47)
(1.89)
(2.26)
(2.88)
(1.44)
(2.83)
0.22***
(2.35)
0.18**
(3.72)
0.24***
1,491
1,313
1,149
1,003
888
790
697
613
528
443
363
Source: Authors’ calculations, based on data from Fitch Ratings and Bureau van Dijk’s Bankscope.
Notes: The table presents results on the relationship between government support, credit quality, and impaired loans. We report the value of the estimated
coefficient on the support rating floor (SRF), issuer default rating (IDR), and their interaction for different lags (one to eleven quarters). Model 1 in panel A
corresponds to the analysis with country-fixed effects and without bank-fixed effects. Model 2 in Panel B includes bank-fixed effects, but no country-fixed
effects. Each specification uses robust standard errors clustered by bank.
*** Statistically significant at the 1 percent level.
** Statistically significant at the 5 percent level.
* Statistically significant at the 10 percent level.
5.3 Does Government Support Have a Bigger
Effect on Riskier Banks?
The results that we have reported thus far suggest that
government support influences bank preference for risk.
Given that finding, a natural question to ask is whether the
link between government support and bank risk-taking
varies with a bank’s creditworthiness. We are particularly
interested in finding out whether the link is stronger for
riskier banks because, all else equal, we would expect these
banks to be more prone to taking on additional risks. To test
this hypothesis, we extend our impaired-loans regression
analysis and include a term for the interaction of the support
rating floor and the issuer default rating. The size of the
interaction captures the marginal effect of government
support for safe banks relative to risky banks. As before,
we estimate two models: one with country-fixed effects,
Model 1, and the other with bank-fixed effects, Model 2. We
include the same controls for bank size and risk, that is, (the
natural logarithm of) assets and our remaining risk ratios
(net charge-offs/total assets, ROA [net income/total assets],
tier 1 capital/total assets, and trading assets/total assets). In
each model, we estimate the different specifications for onethrough eleven-quarter lags.
FRBNY Economic Policy Review / December 2014
53
Chart 8
Effects on Impaired Loans, Interaction
1.4
Coefficient
0
Government support
1.2
Coefficient
Interaction of government support and risk
Bank-fixed effects
Country-fixed effects
-0.02
1.0
Bank-fixed effects
0.8
-0.04
0.6
Country-fixed effects
0.4
-0.06
0.2
0
1
2
3
4
5
6
7
Quarter
8
9
10
11
-0.08
1
2
3
4
5
6
7
Quarter
8
9
10
11
Source: Authors’ calculations, based on data from Fitch Ratings and Bureau van Dijk’s Bankscope.
Notes: The chart presents results on the relationship between government support, credit quality, and impaired loans in our interaction regressions.
The left panel represents the support rating floor coefficient; the right panel represents the support rating floor interacted with the issuer default rating
coefficient. The circles illustrate the respective values of the estimated coefficients through time (one- to eleven-quarter lags). The closed circles denote
significance at the 10 percent level. The black and green lines correspond to Models 1 and 2, respectively. Each specification uses robust
standard errors clustered by bank.
Table 5 summarizes our results. Our main variables of
interest are SRF and SRF * IDR. For completeness, we also
present the coefficient on the IDR. Panel A shows Model 1,
which includes country-fixed effects, while panel B presents
Model 2, which includes bank-fixed effects. Each column
indicates a different quarter-lag specification. Chart 8
illustrates the timing of the SRF and SRF * IDR coefficients
in the left and right panels, respectively.
Looking across the eleven specifications in Model 1, each
with a different lag, we find a persistent, statistically significant
relationship for all three coefficients. As before, the level of
impaired loans in a bank loan portfolio increases directly
with the level of government support. Reflecting the timing of
impairment, this effect increases with higher lags. Similarly,
the interaction of the SRF and the IDR grows increasingly
negative and significant, indicating that riskier banks are
more likely to take advantage of potential sovereign support.
In other words, though all banks increase impaired loans
proportionately to their SRF, riskier banks do so even more.
For each one-notch level of the IDR, a one-notch change in
the SRF increases the impaired loan ratio by approximately
2 percent for the average bank. When we control for bankfixed effects in Model 2, the interaction effect is still present,
but it is significant only if we examine lags four through seven.
54
Do “Too-Big-to-Fail” Banks Take On More Risk?
6. Robustness
6.1 Other Measures of Risk
For completeness of our analysis, we consider three additional
measures of bank risk: the tier 1 capital ratio (tier 1 capital/
total assets), return on assets (net income/total assets), and
trading assets (trading assets/total assets). The traditional
rationale behind capital requirements is that capital acts as a
buffer for protection against unexpected losses. In that sense,
a higher capital ratio implies a safer bank. However, capital
can also act as a measure of bank risk: The amount of capital
a bank needs for protection against losses is closely related
to the risk profile of the bank that will ultimately lead to
those losses. From this perspective, a higher capital ratio is
indicative of a riskier bank because of the requirement of a
higher buffer against losses. ROA captures the profitability of
a bank’s assets. Banks with higher ROA typically have riskier
asset portfolios and, as such, ROA can be considered a proxy
for the risk preference of a bank. In a related spirit, trading
assets can also act as an indirect measure of bank risk. Trading
Table 6
Bank Risk Response to Government Support, Domestic Subsample
Panel A: Baseline
Variable
Impaired loans
Net charge-offs
Coefficient
SRF
SRF
Model
Lags
1
2
3
4
5
6
7
8
9
10
11
1
0.18***
0.19***
0.19***
0.19***
0.19***
0.19***
0.20***
0.20***
0.20***
0.20***
0.21***
2
0.01
0.01
0.01
0.02
0.02
0.04
0.24*
0.26**
0.24*
0.20**
0.12***
1
0.00
0.00
0.01
0.00
0.01
0.01
0.01
0.01
0.01
0.01
0.01
2
0.02***
0.02***
0.02**
0.02*
0.03***
0.02
0.09***
0.08***
0.05***
0.04***
0.06***
1,267
1,155
1,047
943
854
768
684
604
522
440
361
Observations
Panel B: Interactions
Variable
Impaired loans
Impaired loans
Coefficient
SRF
SRF * IDR
Model
Lags
1
2
3
4
1
1.30**
1.32**
1.30**
2
0.30
0.29
0.34
1
-0.07**
-0.07**
-0.07**
2
-0.02
-0.02
-0.02
1,267
1,155
1,047
Observations
5
6
7
8
9
10
11
1.25**
1.29**
1.22**
1.28**
1.29**
1.34**
1.35**
1.30**
0.41**
0.60***
0.65*
0.81**
0.63**
0.60**
0.47*
0.35*
-0.07**
-0.07**
-0.07**
-0.07**
-0.07**
-0.07**
-0.08**
-0.07**
-0.03*
-0.04***
-0.04*
-0.04*
-0.02
-0.02*
-0.02
-0.01
943
854
768
684
604
522
440
361
Source: Authors’ calculations, based on data from Fitch Ratings and Bureau van Dijk’s Bankscope.
Notes: The table presents results on the relationship between government support and bank risk-taking for U.S. banks only. Panel A corresponds to the
baseline specification. Panel B corresponds to the interactions specification. We report the value of the relevant estimated coefficient for different lags (one
to eleven quarters). Model 1 corresponds to the analysis with country-fixed effects and without bank-fixed effects. Model 2 includes bank-fixed effects,
but no country-fixed effects. Each specification uses robust standard errors clustered by bank. SRF is the support rating floor. IDR is the long-term issuer
default rating.
*** Statistically significant at the 1 percent level.
** Statistically significant at the 5 percent level.
* Statistically significant at the 10 percent level.
assets are securities that banks hold for reselling at a profit
(as opposed to investment purposes). As a result, we could
expect that banks with a higher ratio of trading assets to total
assets would engage in riskier activities. We do not discuss
composite measures of bank risk, such as z-scores, because of
data-availability limitations.
As shown in panel B of Table 4, banks with stronger government support have a higher tier 1 capital ratio, ROA, and
trading asset ratio in the specifications with country-fixed
effects. The effect is statistically significant only for the tier 1
capital ratio. As an additional robustness test to this interesting result, we consider an alternative definition of the capital
ratio, calculated as the ratio of tier 1 capital to risk-weighted
assets. This analysis takes into account the riskiness of bank
asset portfolios. Results are similar (statistically significant at
the 5 percent level in the model with country-fixed effects)
and consistent with the second interpretation of bank capital,
in which riskier banks hold higher capital.10
6.2 Domestic Banks
In our analysis, we derive all of our results with country-fixed
effects (Model 1) or bank-fixed effects (Model 2). Nonetheless,
one may still worry about the large diversity of countries
included in our sample. To address this concern, we limit our sample to include only banks headquartered in the
10
Analysis not included, available upon request.
FRBNY Economic Policy Review / December 2014
55
United States, which is the country with the largest number
of banks in the sample. We are interested in understanding if
the relationship between sovereign support and risk-taking
documented in sections 5.1-5.3 is also present in the United
States. Table 6 summarizes our main results.
We see in panel A of Table 6, consistent with our previous
findings, that an increase in government support leads to a
higher ratio of impaired loans and to higher net charge-offs.
Similar to our results for the global sample, the effect on
impaired loans is stronger for riskier banks, reflecting the fact
that they are more likely to exploit potential sovereign support
by engaging in even riskier activities than their safer counterparts do (panel B of Table 6).
As an additional robustness test, we also consider a
variation of our sample in which we exclude banks that
experience a simultaneous (within-quarter) change in
both sovereign support and credit quality. The idea behind
this analysis is to consider a sample without potential
contamination of the identification. After dropping such
banks from our sample (23 percent of SRF changes), we find
qualitatively similar results. Overall, all these findings support
our initial hypothesis that banks with stronger government
support take on more risk.
7. Final Remarks
6.3 Alternative Hypothesis
In this paper, we find evidence that suggests that banks
with stronger sovereign support engage in riskier lending
activities, which translates into a higher ratio of impaired
loans. One alternative hypothesis could be that financial
conditions were already deteriorating, which would lead
to a higher ratio of impaired loans. Although we cannot
completely rule out this premise, all of our specifications
control for bank credit quality. Specifically, as shown in
section 4, we control for the long-term issuer default rating
of each bank at the end of each quarter to take into account
variation in bank financial strength.
In addition, our results in Table 4 and Chart 6 show that the
effect becomes stronger, rather than weaker, over time (that is, the
value of the coefficient on government support is increasing with
the number of lags). This finding is inconsistent with a story in
which the deterioration was already taking place and the change
in sovereign support is a response to worsening conditions.
Also inconsistent with the alternative hypothesis are our
findings on the tier 1 capital ratio. If stronger government support was the response to a bank’s weaker conditions, we would
expect the tier 1 capital ratio to decrease rather than increase
(panel B of Table 4).
56
Do “Too-Big-to-Fail” Banks Take On More Risk?
This study offers new and relevant evidence on a long-debated
question: Does the too-big-to-fail status increase bank
risk-taking incentives? Our evidence is novel because it
focuses on Fitch’s new support rating floors, which aim at
isolating the likelihood of governmental support from other
sources of external support. Of course, SRFs only reflect
Fitch’s opinion of potential government support and of the
government’s ability to support a bank. As is the case in all
studies based on ratings, our results hinge on this assessment’s
reliability. The key advantage of our approach is that support
rating floors only include (Fitch’s views on) sovereign support,
and exclude parent corporations’ support.
Our findings are also innovative in that we focus on
impaired loans to measure bank risk-taking incentives. This
analysis is important because impaired loans, in contrast
to other, more general measures of risk, are more directly
under bank control. Our results account for the governmental
interventions during the financial crisis, but do not reflect
the long-term effects that may arise from the regulatory
changes introduced in its aftermath. An interesting area for
future research would be to investigate to what extent the new
regulations, in particular those dealing with the too-big-to-fail
banks, affect the link we unveiled between the likelihood of
governmental support and bank risk-taking policies.
References
Anginer, D., and A. J. Warburton. 2010. “The Chrysler Effect:
The Impact of the Chrysler Bailout on Borrowing Costs.”
Unpublished paper.
Flannery, M. J., and S. M. Sorescu. 1996. “Evidence of Bank Market
Discipline in Subordinated Debenture Yields: 1983-1991.”
Journal of Finance 51, no. 4 (September): 1347-77.
Baker, D., and T. McArthur. 2009. “The Value of the ‘Too Big to Fail’
Big Bank Subsidy.” CEPR Reports and Issue Briefs 36.
Gadanecz, B., K. Tsatsaronis, and Y. Altunbas. 2012. “Spoilt and
Lazy: The Impact of State Support on Bank Behavior in the
International Loan Market.” International Journal of
Central Banking 8, no. 4 (December): 121-73.
Balasubramnian, B., and K. B. Cyree. 2011. “Market Discipline of
Banks: Why Are Yield Spreads on Bank-Issued Subordinated
Notes and Debentures Not Sensitive to Bank Risks?” Journal of
Banking and Finance 35, no. 1 (January): 21-35.
Gropp, R., H. Hakenes, and I. Schnabel. 2011. “Competition, RiskShifting, and Public Bail-out Policies.” Review of Financial
Studies 24, no. 6 (June): 2084-120.
Brandão Marques, L., R. Correa, and H. Sapriza. 2013. “International
Evidence on Government Support and Risk Taking in the
Banking Sector.” Unpublished paper.
Haldane, A. G. 2010. “The $100 Billion Question.” Bis Review 40.
Available at http://www.bis.org/ review/r100406d.pdf.
Brewer, E., and J. Jagtiani. 2007. “How Much Would Banks Be
Willing to Pay to Become ‘Too-Big-to-Fail’ and to Capture Other
Benefits?” Unpublished paper.
Hau, H., S. Langfield, and D. Marqués-Ibañez. 2013. “Bank Ratings:
What Determines Their Quality?” Economic Policy 28, no. 74
(April): 289-333.
Correa, R., K. Lee, H. Sapriza, and G. Suarez. 2012. “Sovereign Credit
Risk, Banks’ Government Support, and Bank Stock Returns
around the World.” International Finance Discussion
Papers, no. 2012-1069.
Li, Z., S. Qu, and J. Zhang. 2011. “Quantifying the Value of Implicit
Government Guarantees for Large Financial Institutions.”
Moody’s Analytics Quantitative Research Group, January.
Demirgüç-Kunt, A., and H. Huizinga. 2013. “Are Banks Too Big to
Fail or Too Big to Save? International Evidence from Equity Prices
and CDS Spreads.” Journal of Banking and Finance 37, no. 3
(March): 875-94.
Fitch Ratings. 2013a. “Definitions of Ratings and Other Forms of
Opinion.” Available at http://www.fitchratings.com/web_content/
ratings/fitch_ratings_definitions_and_scales.pdf.
---. 2013b. “Landesbank Baden-Wuerttemberg: Full Rating
Report.” Available at http://www.lbbw.de/imperia/md/content/
lbbwde/investorrelations/en/2013/Fitch_Full_Rating_Report
_final_20130823.pdf.
Lindh, S., and S. Schich. 2012. “Implicit Guarantees for Bank Debt:
Where Do We Stand?” OECD Journal of Financial Market
Trends 2012, no. 1 (June): 45-63.
Molyneux, P., K. Schaeck, and T. Zhou. 2010. “‘Too-Big-to-Fail’ and
Its Impact on Safety Net Subsidies and Systemic Risk.” CAREFIN
Research Paper no. 09/2010, June.
---. 2011. “Too Systemically Important to Fail in Banking.”
Unpublished paper.
Morgan, D., and K. Stiroh. 2005. “Too Big to Fail after All These
Years.” Federal Reserve Bank of New York Staff Reports,
no. 220, September.
FRBNY Economic Policy Review / December 2014
57
References (Continued)
Santos, J. 2014. “Evidence from the Bond Market on Banks’ ‘Too-BigTo-Fail’ Subsidy.” Federal Reserve Bank of New York Economic
Policy Review 20. no. 3 (December): 29-39.
Ueda, K., and B. Weder di Mauro. 2012. “Quantifying Structural
Subsidy Values for Systemically Important Financial Institutions.”
Unpublished paper.
Sironi, A. 2003. “Testing for Market Discipline in the European Banking
Industry: Evidence from Subordinated Debt Issues.” Journal of
Money, Credit, and Banking 35, no. 3 (June): 443-72.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
58
Do “Too-Big-to-Fail” Banks Take On More Risk?
Samuel Antill, David Hou, and Asani Sarkar
Components of
U.S. Financial-Sector
Growth, 1950-2013
• The U.S. financial sector grew steadily relative
to the entire business sector from 1975
until its growth was interrupted in the recent
financial crisis. Recovery has been tepid since.
• Large financial firms have had moderately
higher average growth rates than small
financial firms, especially since the 1990s.
The shift followed regulatory changes that
facilitated bank consolidation.
• Shadow banking grew rapidly at the expense
of traditional banks, becoming a significant
portion of the financial sector in the mid-1990s
and peaking just before the crisis.
• The study’s results show that growth in finance
has mainly occurred in opaque, complex, and
less-regulated subsectors of finance.
Samuel Antill is a senior research analyst, David Hou a risk analytics
associate, and Asani Sarkar an assistant vice president at the Federal
Reserve Bank of New York.
1. Introduction
T
here has been a resurgence of interest in the issue of whether
financial-sector growth is necessarily good for the economy.1
Earlier literature generally supported the idea that financial
and economic development go together (King and Levine
1993; Rajan and Zingales 1998) or even that financial growth
is a precondition for economic development (Wright 2002).
More recently, the “dark” side of finance has been emphasized,
with commentators questioning the social value of certain
financial activities.2 This change is an outcome, in part, of the
experience of the recent financial crisis. For example, Turner
(2010) argues that the financial sector extracts rent from the
nonfinancial sector. Other studies (Philippon 2012; Greenwood
1
See, for example, the symposium issue on “The Growth of the Financial
Sector” in the Journal of Economic Perspectives, Spring 2013.
2
Wouter den Haan, “Why Do We Need a Financial Sector?”
Vox, October 24, 2011, http://www.voxeu.org/debates/
why-do-we-need-financial-sector.
The views expressed in this article are those of the authors and do not
necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System.
Correspondence: asani.sarkar@ny.frb.org
FRBNY Economic Policy Review / December 2014
59
and Scharfstein 2013; Philippon and Reshef 2013) find that,
globally, the size of finance relative to gross domestic product
(GDP) has been increasing and reached a historical maximum
before the recent financial crisis. It is difficult to reconcile this
fact with standard models of growth (Philippon 2012).
It is important to understand the evolution of finance and
its subsectors since it weighs on many questions of policy
interest. First, to what extent is credit being intermediated by
shadow banks rather than by commercial banks, which have
traditionally been the main conduits of funds to households
and businesses?3 The relative growth of shadow banks has
implications for regulatory policies geared toward enhancing
the safety and soundness of commercial banks (such as those
governing deposit insurance, central bank liquidity, and
capital requirements). Second, what was the relative growth of
large financial firms that pose risk for the rest of the economy?
Third, what was the role of leverage in the growth of firms,
especially of large firms, given that leverage constraints are
now an important tool in bank regulation? Finally, to what
extent has growth occurred within privately held firms that
are more opaque than publicly listed companies?
To investigate these questions, we must first measure
the size of the financial sector. So far, the literature has
produced measures based on value-added and on liabilities
of broad sectors using data from the Bureau of Economic
Analysis (BEA) and the Federal Reserve Board’s Flow of Funds
Accounts (FOF).4 Others rely on aggregate wages and
income. These measures, however, cannot be used to estimate
accurately the growth of shadow banks or publicly listed firms.
Accordingly, this paper provides new descriptive measures
of financial-sector size using firm-level balance sheet data
from the Center for Research in Security Prices (CRSP) and
Compustat from the first quarter of 1975 to the first quarter of
2013. Our disaggregated approach allows us to examine how
financial-sector growth relates to firm size, financing choice
(whether equity or debt), and industry type.
The balance sheet data have the disadvantage of excluding
private firms which are an important source of economic
growth.5 To address this concern, we also measure the size
of finance based on the FOF data reporting total liabilities
for private and publicly listed firms at the sectoral level. In
addition, we examine data for individual commercial banks
3
Shadow banks are entities such as structured investment vehicles that
(like traditional banks) perform credit intermediation services, but
(unlike banks) lack central bank liquidity or public sector credit guarantees
(Pozsar, Adrian, Ashcraft, and Boesky 2013).
4
The FOF data release is now titled Financial Accounts of the United States.
5
The growth potential of private firms is indicated by evidence that
these firms invest more than publicly listed firms of similar sizes
(Asker, Farre-Mensa, and Ljungqvist 2013).
60
Components of U.S. Financial-Sector Growth, 1950-2013
from the Federal Financial Institutions Examination Council’s
Consolidated Reports of Condition and Income (Call Reports)
that include both private and publicly listed banks. These data
provide a second source for examining the relative growth of
the commercial banking sector.
Our measures are of the form ______
F +SNF , where S, F, and
NF are sizes of a particular financial subsector S, the entire
financial sector F, and the entire nonfinancial sector NF,
respectively. When S = F, we are estimating the size of the
finance sector relative to the total business (that is, financial
plus nonfinancial) sector. By normalizing by the size of the
business sector, we control for economy-wide factors that
impact all firms. For firm-level or bank-level data, size is
the value of firm or bank assets (comprising either debt plus
equity or equity only).6 For FOF data, size equals the total
liabilities of a sector.7
Using these measures, we find that the U.S. financial
sector grew steadily relative to the entire business sector from
1975 until the recent financial crisis. Further, publicly listed
financial firms had lower average size relative to the total
business sector than private financial firms. For example,
while publicly listed financial firms were about 50 percent
of the business sector based on total asset values (representing
debt plus equity) on average, financial-sector liabilities
inclusive of private financial firms were almost 70 percent
of total liabilities based on the FOF data.
We also measure the size of the credit intermediation
subsectors, starting with shadow banks. Following an
approach described in Financial Stability Board (2011)
and Financial Stability Board (2012), we consider all
nonbank credit intermediation activities and use FOF
sector categories to identify the corresponding liabilities. For
our CRSP-Compustat measures, we identify shadow banks
by using the Standard Industrial Classification (SIC) and the
North American Industrial Classification System (NAICS)
codes that map to the FOF sectors. This broad measure of
shadow banking is consistent with Financial Stability Board
(2011, 3), which argues that “authorities should cast the net
wide, looking at all nonbank credit intermediation to ensure
that data gathering and surveillance cover all areas where
shadow banking–related risks to the financial system might
potentially arise.” For comparison, we also report a narrower
measure of shadow banking developed by Adrian and
Ashcraft (2012) using specific types of FOF liabilities.
6
Many small publicly listed firms do not file with the Securities and Exchange
Commission and, thus, we do not have debt data available. To account for
these firms, we also report the equity-only measure of size.
7
Sectoral assets and liabilities need not be equal in the FOF data since these
are not aggregated from firm-level balance sheets. However, our results are
qualitatively the same whether we use assets or liabilities.
In keeping with the previous literature, we find that the
share of shadow banking in the total business sector has
grown from less than 4 percent in 1975 to a high of between
9 percent and 37 percent in recent years (depending on
the measure). Growth in shadow banking has been fueled
by rapid expansion in credit intermediation services
performed by asset management and securities firms
(including open-end investment funds and securities and
commodities brokerages). We also see that housing-related
credit intermediation (provided by real estate credit firms and
real estate investment trusts [REITs]) is a substantial part of
shadow banking, but its share has been declining since the
1980s. The average share of shadow banking in the business
sector was at least twice as large when calculated with private
liabilities (about 16 percent) than without (about 4 percent to
8 percent). Our results, which are qualitatively similar using
broad and narrow definitions of shadow banking, emphasize
the predominantly private nature of shadow bank liabilities
and thereby heighten concerns about the opacity of the sector.
Shadow banks are a potential source of systemic risk
(Adrian and Ashcraft 2012) in part because their activities are
intertwined with those of traditional banks and depository
credit institutions (DCIs) (Cetorelli and Peristiani 2012;
Cetorelli, McAndrews, and Traina 2013). Boyd and Gertler
(1995) find that between 1976 and 1993 the share of
commercial banks in financial intermediation was stable.
Recent evidence shows that shadow banks have grown
relative to DCIs (Pozsar et al. 2013; Adrian and Ashcraft
2012). Greenwood and Scharfstein (2013) suggest that
incremental growth in household credit origination was due
to securitization, implying a growth in shadow banks at the
expense of traditional banks. We measure the share of shadow
banks in total credit intermediation (TCI) as _______
SB +SBDCI , where
DCI (SB) is the size of the DCI (shadow banking) sector.
We find that the share of shadow banking in TCI
grew from less than 9 percent in 1975 to a high of at least
33 percent in the period from 2004 to 2013. The estimate
may understate the share of shadow banking in TCI because
DCIs have increasingly acquired shadow banks, with this
type of acquisition occurring at a greater rate than the reverse
(Cetorelli et al. 2013). After being acquired, these shadow
banks may be counted as part of the DCI sector, provided that
DCI activity is considered the main business of the merged
firm by SIC and NAICS. In this case, the shadow banking
activity becomes part of the DCI sector.8
8
In some cases, SIC codes may be reclassified and changed retroactively.
We were unable to verify how frequently this occurs, but it appears that
at least in some cases a firm will have different SIC and NAICS codes in
different periods due to corporate structure changes, as discussed here.
Large financial firms (those in the top 10 percent of
firms by value) were a substantially greater share of all
large firms than small financial firms (those in the bottom
90 percent of firms by value) were of all small firms. Further,
large financial firms had moderately higher average growth
rates than small financial firms, especially since the 1990s.
Size-related differences were most pronounced in the
DCI sector, with large DCIs outgrowing small DCIs by an
average of at least 3 percent over the sample period. Some
of this shift followed regulatory changes that facilitated bank
consolidation (such as the Riegle-Neal Act of 1994 and the
Graham-Leach-Bliley Act of 1999).
The recent financial crisis adversely affected the size of
the financial sector, but its impact differed by subsector and
type of firm. For example, the shadow banking subsector did
poorly relative to other sectors by most measures, with its
size shrinking from the peak pre-crisis quarter to the trough
crisis quarter more than that of the financial sector as a whole.
These effects were even more pronounced when we excluded
government-sponsored enterprises (GSEs) from our sample of
financial firms. While small financial firms generally suffered
the most of all firms during the crisis, larger shadow banks
did worse than small shadow banks. Large DCIs actually
grew in size during the crisis based on book values, especially
during 2008 and 2009 when the Federal Deposit Insurance
Corporation (FDIC) guaranteed the debt issuances by these
firms. We estimate that the size of DCIs issuing guaranteed
debt between October 2008 and October 2009 under this
program increased by an average of roughly 11 percent
compared with all other firms, an economically but not
statistically significant number.
To understand the effect of balance sheet leverage on
the size of financial firms, we look at total assets versus
equity-only measures for publicly listed firms. We find that,
on average, financial firms are three times smaller, shadow
banks are one-and-a-half times smaller, and DCIs are five
times smaller under equity-only measures than they are by
total asset value, attesting to the importance of leverage in the
capital structure of financial firms and of DCIs in particular.
This article contributes to the literature by proposing new
firm-based and sector-based measures of financial-sector size
in line with an approach by the Financial Stability Board
(2012, 5), which recommends the use of more granular data
and market prices “to adequately capture the magnitude and
nature of risks in the shadow banking system.” While our
metrics do not speak to risk exposures directly, they may
be used as starting points for determining the location of
vulnerabilities. Our findings also have policy implications,
such as for the regulation of shadow banks, that we discuss
more fully in the conclusion.
FRBNY Economic Policy Review / December 2014
61
The remainder of our article is organized as follows. We
review the literature in Section 2 and explain our measures
of financial-sector size in Section 3. In Section 4, we describe
our results on the size and growth of the financial sector.
Section 5 discusses the effects of leverage, firm size, and
financial regulation on financial-sector growth. Section 6
summarizes our findings. Results discussed in the article
but not reported in the tables and charts can be found in the
online appendixes.9
2. Literature Survey
One of the earliest papers on trends in financial-sector size
is Boyd and Gertler (1995), who use value-added data
from the BEA and other measures to examine whether the
commercial banking sector was declining or not. Already in
the mid-1990s, there was concern over the growth of nonbank
credit intermediaries—shadow banks, in today’s terminology—
and its effect on traditional banks. The authors conclude
that the share of banking in total financial intermediation
was generally stable, with small losses in the 1980s and
1990s, and that financial intermediation had grown relative
to GDP. They suggest that the apparent decline in banking
reflects the movement of activities from on-balance-sheet to
off-balance-sheet as well as the significant increase in the share
of foreign-owned banks in U.S. banking activity.
More recently, a number of papers that were part of a
Journal of Economic Perspectives Spring 2013 symposium
examined the evolution of finance. Specifically, Greenwood and
Scharfstein (2013) and Philippon and Reshef (2013) propose
metrics of financial-sector size and evaluate hypotheses on the
sources of growth, while Cochrane (2013) argues that the focus
should be on the functions of financial firms and not on their
sizes. Separately, Philippon (2012) and Philippon and Reshef
(2012) have also contributed to this literature.
Greenwood and Scharfstein examine financial-sector size
using several measures, including value-added and liabilities
data for broad sectors from the BEA and the FOF, as
well as industry output, fees, and traded value for more
specific sectors (such as asset management). They find
that financial-sector growth has accelerated since 1980,
fueled by the securities and credit intermediation sectors,
and accounted for a quarter of the growth in the services
sector as a whole. Considering the source of financial growth,
the authors emphasize the role of asset management, which
9
The online appendixes for this article are available at http://
www.newyorkfed.org/research/epr/2014/1403anti_appendixA-D.pdf.
62
Components of U.S. Financial-Sector Growth, 1950-2013
grew as a class largely because of the increase in stock market
valuations, and the provision of household credit, especially
residential mortgages, which increased through fees derived
from loan origination, underwriting, and trading activities.
They also question the social value of this growth, given the
high cost and persistent underperformance of professional
asset management and the fallout from an excess of
credit-financed consumption.
Cochrane (2013) argues that the growth in finance was
an outcome of increased demand for financial services
and higher wages for finance employees with scarce skills
(although both these arguments appear to be inconsistent
with the results of Philippon [2012], whose work is
described below). Cochrane proposes a supply-and-demand
model, based on Berk and Green (2004), to explain the
underperformance of actively managed funds. He points
to the persistence of proportional fees across different
professions and over time to suggest that asset management
fees may not represent suboptimal contracts. More generally,
he makes an important distinction between form and function
of firms that we discuss further in the conclusion.
Philippon (2012) shows that, while the income of financial
intermediaries as a share of GDP has generally varied
over time, it increased rapidly from 1980 to 2010. Using
value-added data from the BEA, liabilities data from the FOF,
and financial flow variables (such as for corporate issuance
and mergers-and-acquisitions), he constructs a measure of
financial-intermediation output as the weighted average of
various types of credit, equity issuances, and liquid assets, with
the relative weights based on theory. He finds that the annual
unit cost of financial intermediation (defined as income over
output) is around 2 percent and relatively stable over time.
Philippon and Reshef (2012) examine wages, complexity of
jobs, and skill levels in finance, relative to the economy, and
find that they all follow a U-shaped pattern, peaking before
World War II and then again from 1980 on. They point out,
however, that growth in the financial industry and growth
in skills and wages of finance employees did not always go
together. Philippon and Reshef (2013) investigate the income
share of finance in international data, using the ratio of bank
loans to GDP as a proxy for financial-sector output.
These papers indicate the difficulty of measuring
financial-sector output consistently over time and across
countries in the context of financial innovation and other
structural changes and given differences in accounting
methodologies. Our balance-sheet-based measures are also
open to the same critique, as they are affected by changes in
accounting systems over time, assets moving off balance sheet,
and changes in industry structure (in particular, mergers and
acquisitions) that make industry classification ambiguous.
We discuss these issues further in the conclusion. The benefits
of our measures, relative to the earlier literature, are the level of
disaggregation (that is, firm-level observation) and the ability to
use the same data sources (CRSP and Compustat) consistently
for measuring sizes of all sectors. Previous papers use different
data sets depending on which sector is being measured.
Unconventional measures have also been suggested,
sizing the sector, for example, by the percentage of firms
on the Forbes 400 list whose wealth is derived primarily
from financial activities (Kaplan and Rauh 2013) or by
the percentage of graduates from top colleges entering into
financial services employment (Goldin and Katz 2008).
3. Methodology
We propose seven measures of financial-sector size, which
are summarized in Table 1 along with their respective
sample periods, data sources, and definitions. Appendix A
describes the data used.
The first set of metrics is based on firm-level balance
sheet data which are aggregated to the sectoral level to
derive measures of sectoral size:
(1)
ϵ Value
iS
i,t
Size St = ___________
,
jϵF,NFValuej,t
where Valuei,t is the value of firm i in day or quarter t, S is
either the entire financial sector or a financial subsector,
F denotes the entire financial sector, and NF denotes the
entire nonfinancial sector. In other words, we define the size
of a sector S as the value of all firms in sector S relative to the
value of all firms in the nonfinancial and financial sectors.
Financial sectors S are classified using the SIC and NAICS
systems, as described in Appendix B. When S = F, our metric
is a measure of the size of finance relative to the total business
sector. This methodology of calculating the size of a financial
sector relative to the nonfinancial sector, similar to Philippon
and Reshef (2012), controls for a spurious increase in the size
of finance due to a general increase in the number of publicly
listed firms over time.
The first four rows of Table 1 list the metrics derived from
firm-level data, which correspond to whether we use equity
value, total value (debt plus equity), the market value of equity
(MVE), or the book value of equity (BVE).10 The measures
10
For asset management firms, we capture only liabilities of the firm, not
funds held by a firm for other companies. So long as these funds belong
to publicly listed companies in the same sector, the sectoral aggregates
will remain unaffected.
Table 1
Relative Size Measures
Name of
Measure
Sample
Tsize - bv 1975:Q1-2013:Q1
Formula
iϵS (BVEi,t + BVDi,t )
Tsize - bv St = _________________
jϵF,NF (BVEj,t + BVDj,t
iϵS (MVEi,t + BVDi,t )
_______________
Tsize - qmv 1975:Q1-2013:Q1 Tsize - qmv St = ___
jϵF,NF (MVEj,t + BVDj,t)
Esize - bv
BVE
iϵS
i,t
1975:Q1-2013:Q1 Esize - bv St = ________
jϵF,NF BVEj,t
iϵS MVEi,t
__
Esize - mv 1950:Q1-2013:Q1 Esize - mv St = _________
jϵF,NF MVEj,t
Fsize
1952:Q1-2013:Q1
sϵS Liabilitiess,t
___
Fsize St = ___________
jϵF,NF Liabilitiesj,t
AA
1990:Q2-2013:Q1
kϵSB Liabilitiesk,t
AA = ______________
jϵF,NF Liabilitiesj,t
Csize
1975:Q1-2013:Q1
iϵB(BVEi,t + BVDi,t )
Csizet = _______________
jϵF,NF Liabilitiesj,t
Notes: This table defines the relative size measures used in this article
and their sample periods.
The size of a sector is defined as the value of assets in the sector
relative to the asset values of the financial and nonfinancial sectors.
MVE is market value of equity. The data for MVE are from the
Center for Research in Security Prices.
BVE is book value of equity. BVD is book value of liabilities.
The data for BVD and BVE are from Compustat.
For the Tsize − qmv and Esize − mv measures, an observation
is the asset value of a firm i in day t belonging to a sector S.
Since BVD and BVE are only observed at the quarterly level, the
quarterly value is repeated each day of the quarter for Tsize − qmv.
For the Tsize − bv and Esize − bv measures, an observation
is the asset value of a firm i in day t belonging to a sector S.
When S = F or S = NF, the F and the NF indicate the
financial and nonfinancial sectors, respectively.
For Fsize, an observation is the liability of a sector S in quarter t.
For AA, the numerator is the liability of a financial instrument k aggregated
over all shadow banking (SB) instruments. The SB instruments
are commercial paper, repo, debt, pools of mortgages backed by
government-sponsored enterprises, asset-backed securities, and money
market mutual funds. The denominator of AA is the aggregate liabilities
of the financial and nonfinancial sectors (which is the same as the
denominator of Fsize). The data source for Fsize and AA is the Federal
Reserve Board’s Flow of Funds Accounts.
For Csize, the numerator is the book value of assets of a commercial bank j,
obtained from Call Reports data aggregated over the banking sector B.
The denominator of Csize is the aggregate liabilities of the financial and
nonfinancial sectors from flow-of-funds data (which is the same as the
denominator of Fsize and AA).
FRBNY Economic Policy Review / December 2014
63
using equity value are Esize − mv (based on MVE) and
Esize − bv (based on BVE). The measures using total value
are Tsize − qmv (based on MVE plus the book value of debt
[BVD])11 and Tsize − bv (based on BVE plus BVD).
For firm i and day t, the MVE-based measures are:
(2)
(3)
MVE
iϵS
i,t
Esize - mv St = __________
,
jϵF,NF MVEj,t
(MVE + BVD )
jϵF,NF (MVEj,t + BVDi,t )
BVE
(4)
iϵS
i,t
Esize - bv St = _________
,
(5)
iϵS
i,t
i,t
Tsize - bv St = ________________
.
jϵF,NF BVEj,t
(BVE + BVD )
jϵF,NF (BVEj,t + BVDj,t )
A downside of our firm-level measures is that, because
they comprise only publicly listed firms, the estimated sizes
are affected by the relative shares of private firms in the
financial and nonfinancial sectors. Over time, these effects
are magnified if financial and nonfinancial firms go public
at different rates. To address these concerns, we consider
an additional measure based on FOF data, which captures
most assets and liabilities in the economy, although it is
available only at the sectoral level:
(BVE + BVD )
iϵB
i,t
i,t
Csizet = _______________
,
jϵF,NF Liabilitiesj,t
where B is the commercial banking sector, equity and debt
values of commercial banks are from Call Reports, and the
denominator represents the total liabilities of the financial and
nonfinancial sectors from the FOF data (which is identical to
the denominator of equation 6).
As a check, we compare the total book value (BVE plus
BVD) of banks for all our metrics and find that the mean is
smaller using Call Reports data (about $4.36 trillion) than
with the FOF data ($5.9 trillion) or the CRSP-Compustat
data (about $7.9 trillion). This discrepancy may be due to the
fact that Call Reports provide data for individual commercial
banks while the other data sets report information for bank
holding companies.12
Finally, we also calculate an alternative measure of shadow
banking using an approach developed by Adrian and Ashcraft
(2012) based on specific types of FOF liabilities.13 The
measure sums all liabilities recorded in the flow of funds that
relate to securitization activity including mortgage-backed
securities (MBS), asset-backed securities (ABS), and other
GSE liabilities, as well as all short-term money market
liabilities that are not backstopped by deposit insurance
(such as repos, commercial paper, and other money market
mutual fund liabilities). We adjust the aggregate to mitigate
double-counting. So, we have:
(8)
Liabilities
kϵSB
k,t
AAtSB = _____________
.
jϵF,NF Liabilitiesj,t
Liabilities
sϵS
s,t
Fsize St = _____________
,
jϵF,NF Liablitiesj,t
where Liabilitiess,t is the total liability of sector s in
quarter t. Thus, for the shadow banking sector, for
example, we sum the liabilities of the subsectors s making
up the shadow banking industry and express that figure as
We only calculate Tsize − qmv when both CRSP MVE data and
Compustat BVD data are available.
11
64
(7)
iϵS
i,t
i,t
Tsize - qmv St = _________________
,
Since MVE is reported daily and BVD quarterly, the latter is
carried over for each day of the quarter in order to obtain a
daily measure of Tsize − qmv. For comparability, all measures
are reported at the quarterly frequency and so Esize − mv
and Tsize − qmv are averaged quarterly. We focus on the
MVE-based measures 2 and 3 for most of the paper.
For firm i and day t, the book-value based measures,
discussed in Section 5.2, are:
(6)
a ratio to the total liabilities of the business sector. When
S = F, we are measuring total financial liabilities relative to
total business sector liabilities.
To further address the potential biases of using only
publicly listed firms, we provide an alternative measure of the
size of the DCI sector using Call Reports data that include
public and private banks:
Components of U.S. Financial-Sector Growth, 1950-2013
12
For more consistency across databases, we could have used Federal Reserve
Y-9C forms that are filed by all bank holding companies of a certain size,
which report consolidated data that include both commercial banking activity
as well as other activity (such as investment banking) unrelated to commercial
banking. Since we only want to focus on commercial banking activity, we
prefer to use Call Reports.
13
We thank the authors for providing the data. The liabilities are
described in Table 1.
where Liabilitiesk,t is a particular liability k (such as MBS)
used by the shadow banking sector SB in quarter t. The
denominator represents the total liabilities of the financial
and nonfinancial sectors from the FOF data (which is
identical to the denominator of equation 6).
Chart 1
The Relative Size of Finance
Percent
80
Fsize
70
60
Tsize – qmv
50
4. The Size and Growth of Finance
and Its Subsectors
40
30
Esize – mv
20
In this section, we describe evolution of the aggregate
financial sector, the DCI and shadow banking credit
intermediation subsectors, and the remaining subsectors, in
particular, asset management, securities, and insurance. While
the subsets of asset management and securities firms involved
in credit intermediation are included in the shadow banking
sector, this analysis encompasses the asset management and
securities industries as a whole.
4.1 Growth of Finance
We find that, for all measures, the relative size of finance
was growing consistently, particularly in the 1980s and from
2000 until just before the crisis in the third quarter of 2007.
Chart 1 plots the values of Tsize − qmv, Esize – mv, and Fsize,
while Table 2 reports summary statistics of these metrics
for the full sample, pre-crisis (1980 to the third quarter of
2007), and crisis (the fourth quarter of 2007 to first quarter
of 2013) periods. Chart 2 shows the median percent changes
of the size measures for the pre-1980 period and subsequent
decades. We report median instead of mean growth rates
because the distribution of quarterly growth rates is skewed
right, especially in the earlier part of the sample when
some of our measures started from a low value (resulting
in unusually high growth rates).
The financial sector was smaller but grew faster when
measured using publicly listed firm assets instead of total
(private and publicly listed) sectoral liabilities. Thus, the
sample means for Tsize − qmv of about 50 percent and for
Esize – mv of 17 percent were smaller than the mean for
Fsize of almost 70 percent (see Table 2). The average growth
over the full sample in the size of finance using Fsize was
half that using public firm-based measures (0.40 percent
for Fsize versus more than 0.80 percent for Tsize − qmv
and Esize − mv) (Chart 2). Moreover, the growth in the
relative size of publicly listed financial firms occurred even
before 1980, whereas the Fsize measure had negative median
10
0
1950
Q1
1960
Q1
1970
Q1
1980
Q1
1990
Q1
2000
Q1
2010
Q1
Source: Authors’ calculations.
Notes: This chart shows measures of the size of finance relative to the
financial and nonfinancial sectors. See Table 1 for variable definitions.
growth during this period. This result is consistent with that
obtained by using BEA data (which also include private firm
liabilities).14 Finally, the relative size of finance was larger
using Tsize − qmv instead of Esize − mv (Chart 1), and the
gap was increasing over time, which is indicative of rising
leverage ratios for finance relative to the nonfinancial sector,
as discussed further in Section 5.2.
As expected, the financial crisis had a deleterious effect on
the size of the financial sector. From the peak pre-crisis quarter
(the third quarter of 2007 for Tsize − qmv and Fsize and the
first quarter of 2007 for Esize − mv) to the trough crisis quarter
(the first quarter of 2009 for Fsize and Esize − mv and the
first quarter of 2013 for Tsize − qmv), its total value shrank
between 2 and 6 percentage points relative to the nonfinancial
sector (Table 2).15 The Esize − mv and Fsize measures indicate
that finance barely recovered to pre-crisis levels by the first
quarter of 2013 whereas Tsize − qmv shows that the relative
14
For example, Greenwood and Scharfstein (2013) and Philippon (2012)
find that finance became prominent in the 1980s.
Unlike the other measures, Tsize − qmv does not reach its minimum
during the depth of the crisis (first quarter of 2009). However, the crisis
period declines in finance remain similar across measures even if we use
a common measurement period (such as the third quarter of 2007 to the
first quarter of 2009). For the interested reader, the value for Esize − mv
in the third quarter of 2007 was 23.22 percent and the value for Tsize −
qmv in the first quarter of 2009 was 68.26 percent.
15
FRBNY Economic Policy Review / December 2014
65
Table 2
The Relative Size of Finance
Full Sample
Tsize - qmv
Esize - mv
Fsize
8,017,993
17,099,300
245
Mean
49.73
16.84
69.43
Median
46.5
17.36
68.18
Observations
Min / min quarter
38.04 / 1976:Q3
9.34 / 1980:Q4
61.47 / 1981:Q1
Max / max quarter
68.40 / 2008:Q4
25.03 / 2007:Q1
77.67 / 2013:Q1
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv
Esize - mv
Fsize
6,242,561
11,433,458
111
Mean
48.51
17.02
70.51
Median
46.41
16.26
71.54
Observations
Min / min quarter
39.8 / 1981:Q3
9.34 / 1980:Q4
61.47 / 1981:Q1
Max / max quarter
64.00 / 2007:Q3
25.03 / 2007:Q1
77.17 / 2007:Q3
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Median
Tsize - qmv
Esize - mv
Fsize
1,293,724
3,487,929
22
64.34
21.4
76.68
64.43
21.55
76.66
Min / min quarter
60.41 / 2013:Q1
19.05 / 2009:Q1
75.74 / 2009:Q1
Max / max quarter
68.40 / 2008:Q4
22.96 / 2013:Q1
77.67 / 2013:Q1
Source: Authors’ calculations.
Notes: This table reports summary statistics of measures of the size of finance relative to the financial and nonfinancial sectors. Observation units are firm days
for Tsize − qmv and Esize − mv and quarters for Fsize. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then
average across days for each quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions. Min (max) quarter refers to
the quarter in which the measure achieves its minimum (maximum) value in the sample.
size of finance remains almost 2 percent lower than its peak in
the pre-crisis quarter (Chart 2).
While the post-crisis recovery in finance has been tepid by
any measure, it would have been even worse if we excluded
GSEs from our sample. As discussed in Appendix B, we
consider GSEs to be financial firms (in keeping with Financial
Stability Board [2011] and Financial Stability Board [2012]).
To examine the effect of GSEs on size measures, we recalculate
our metrics excluding GSEs and agency- and GSE-backed
mortgages from our definitions of both finance and
nonfinance. Although they typically account for a small share
66
Components of U.S. Financial Sector Growth, 1950-2013
of finance, GSEs expanded greatly during the recent crisis
in response to the credit crunch. For example, if we subtract
GSEs, the peak in finance shifts from the first quarter of 2013
to third quarter of 2007 using Fsize.
While our results show that finance grew relative to the
nonfinancial sector in the sample period, that increase may
have been part of a general growth in services. Using SIC and
NAICS codes to classify the services industry, we find that
finance grew even relative to the nonfinancial services sector,
consistent with Greenwood and Scharfstein (2013) and
Philippon (2012).
Chart 2
Median Percentage Change in the Relative Size of Finance, by Period
Tsize – qmv
Esize – mv
Fsize
Percent
4
3
2
1
0
-1
-2
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows median annualized quarter-to-quarter percentage changes in the relative size of finance for each measure for specific
periods. Size is relative to the financial andnonfinancial sectors. For Tsize − qmv and Esize − mv, we first aggregate from the firm level to the
sector level and then calculate quarterly changes. See Table 1 for variable definitions.
4.2 The Size and Growth of
Credit Intermediation
While credit intermediation has always been an essential
component of finance, its nature has changed over time.
Traditional credit intermediation is carried out by DCIs or banks
that take insured deposits and give loans, and are regulated by
and receive liquidity support from the central bank. Increasingly,
though, shadow banks outside the purview of regulatory
authorities intermediate credit. In this section, we discuss
the growth of shadow banking and its sources, the size of the
traditional banking and DCI sectors, and the relative share of
shadow banking in total credit intermediation.
Shadow Banking
Shadow banking is, in essence, any form of nondepository
credit intermediation. Pozsar et al. (2013) explain that
shadow banking credit is intermediated by a variety of
nonbank financial specialists such as asset managers,
broker-dealers, and finance companies. For the Fsize
measure, we follow Financial Stability Board (2011)
and Financial Stability Board (2012) and use FOF
sector categories to define the shadow banking sector.
For our CRSP-Compustat measures, we define equivalent
sectors by using SIC and NAICS industry codes that map
to the FOF sectors.16 We also report an AA measure using
the approach of Adrian and Ashcraft (2012), who size the
shadow banking sector based on specific types of FOF
liabilities (see Section 3 for more discussion).
All measures show shadow banking growing relative to
the rest of the economy from at least the 1980s until the
recent financial crisis. The sector was small and growing
unevenly in the 1970s. Then its growth accelerated in the
1980s and 1990s before slowing down in the 2000s and
finally plummeting in the crisis (Charts 3 and 4). This result
is consistent with Pozsar et al. (2013), Adrian and Ashcraft
(2012), and Financial Stability Board (2012). The relative size
of the shadow banking sector was less than 4 percent of the
business sector in 1975 but reached a high of between 9 and
37 percent (depending on the measure) in the recent decade
(Table 3). As for finance in general, the relative size of
shadow banking is smaller when public firm-based measures
are used. For example, the sample mean relative size of
shadow banking is 8 percent based on Tsize − qmv and
about 16 percent per the Fsize and AA measures.
The share of shadow banking in the business sector
decreased during the recent financial crisis based on
all measures, with pre-crisis peak quarter to crisis
trough quarter declines of at least 6 percentage points
by all measures except Esize − mv (Table 3). The average
16
Details of this mapping are in Appendix B.
FRBNY Economic Policy Review / December 2014
67
Chart 3
The Relative Size of Shadow Banking
Percent
40
Fsize
35
30
25
AA
20
Tsize – qmv
15
Esize – mv
10
5
0
1950:Q1
1960:Q1
1970:Q1
1980:Q1
1990:Q1
2000:Q1
2010:Q1
Source: Authors’ calculations.
Notes: This chart shows measures of the size of shadow banking relative to the financial and nonfinancial sectors. See Table 1 for variable definitions.
Chart 4
Median Percentage Change in the Relative Size of Shadow Banking, by Period
Tsize – qmv
Esize – mv
Fsize
AA
Percent
12
10
8
6
4
2
0
-2
-4
-6
-8
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows median annualized quarter-to-quarter percentage changes in the relative size of shadow banking for each measure for specific
periods. Size is relative to the financial and nonfinancial sectors. For Tsize − qmv and Esize − mv, we first aggregate from the firm level to the sector
level and then calculate quarterly changes. See Table 1 for variable definitions.
decline in the share of shadow banking during the crisis
was particularly sharp using the AA measure, which is
based on financial liabilities such as commercial paper and
asset-backed securities that suffered the most during the
crisis (Chart 4). In contrast, the Esize − mv measure shows
only a modest decline from pre-crisis peak to crisis trough
quarters and, in fact, indicates a positive median growth
rate in shadow banking since the crisis.
68
Components of U.S. Financial-Sector Growth, 1950-2013
As for finance overall, the crisis effect was harsher for
shadow banks when GSEs and agency- and GSE-backed
mortgages are excluded. Indeed, when Fannie Mae and
Freddie Mac are not counted, the crisis-period spike in
Tsize − qmv (Chart 3) disappears entirely. However, the
general trend of a growing shadow banking sector in the
pre-crisis period is robust to whether GSEs are included or
excluded from the sample.
Table 3
The Relative Size of Shadow Banking
Full Sample
Tsize - qmv
Esize - mv
Fsize
AA
1,524,082
7,186,652
245
91
Mean
7.85
4.49
16.36
16.71
Median
7.75
4.52
11.32
16.48
Min / min quarter
2.83 / 1975:Q2
1.27 / 1982:Q2
3.51 / 1952:Q1
13.12 / 1990:Q2
Max / max quarter
15.34 / 2010:Q2
9.45 / 2013:Q1
36.74 / 2007:Q4
20.82 / 2007:Q2
Observations
Pre-crisis Period (1980:Q1-2007:Q3)
Observations
Tsize - qmv
Esize - mv
Fsize
AA
1,243,879
4,199,269
111
69
Mean
8.25
4.3
22.74
16.56
Median
7.61
4.51
22.93
16.36
Min / min quarter
3.17 / 1981:Q3
1.27 / 1982:Q2
9.25 / 1980:Q1
13.12 / 1990:Q2
Max / max quarter
14.46 / 2007:Q3
8.02 / 2007:Q3
36.72 / 2007:Q3
20.82 / 2007:Q2
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Tsize - qmv
Esize - mv
Fsize
AA
165,822
2,311,902
22
22
10.2
7.99
32.71
17.19
9.55
8.25
31.78
17.01
Min / min quarter
7.88 / 2012:Q4
6.57 / 2008:Q3
30.53 / 2011:Q3
13.82 / 2013:Q1
Max / max quarter
15.34 / 2010:Q2
9.45 / 2013:Q1
36.74 / 2007:Q4
20.72 / 2008:Q1
Median
Source: Authors’ calculations.
Notes: This table reports summary statistics of measures of the size of finance relative to the financial and nonfinancial sectors. Observation units are firm days
for Tsize − qmv and Esize − mv and quarters for Fsize. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then
average across days for each quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions. Min (max) quarter refers to
the quarter in which the measure achieves its minimum (maximum) value in the sample.
To understand the source of growth of shadow banking,
we examine the types of credit intermediation that make
up shadow banking: securities credit intermediation
(SCI) (such as securities and commodities brokerages
and investment banking), asset management credit
intermediation (AMCI) (including mutual funds, closed-end
funds, exchange-traded funds, and other financial vehicles),
and real estate credit intermediation (RECI) (like mortgage
credit, mortgage brokerages, agency GSEs, agency- and
GSE-backed mortgages, and REITs). We define these
sectors consistently in all our data sets (although, due
to differences in data construction, it is unlikely that the
industry composition of sectors is identical in the different
data sets). Table 4 reports the relative shares of various types
of credit intermediation in shadow banking for the full
sample and subsamples of interest.
We see that AMCI and RECI liabilities make up the
bulk of total shadow banking liabilities, per the Fsize
measure. For example, the share of AMCI liabilities in
shadow banking (based on Fsize) grew from 28 percent
in the 1980s to almost 43 percent in the crisis period of
2007-13. In contrast, and perhaps surprisingly, the share
of RECI declined steadily from 42 percent in the 1980s to
32 percent in the crisis, based on Fsize. Of publicly listed
FRBNY Economic Policy Review / December 2014
69
Table 4
Share of Shadow Banking, by Types of Credit Intermediation
Full Sample
Tsize - qmv
Esize - mv
SCI
29.04
15.02
8.64
AMCI
0.00
22.99
29.63
Fsize
RECI
2.06
5.21
30.50
Other
68.89
56.78
31.23
Pre-crisis Period (Start-2007:Q3)
Tsize - qmv
Esize - mv
Fsize
SCI
27.01
17.62
8.70
AMCI
0.00
27.97
28.36
RECI
1.96
6.08
30.35
Other
71.02
48.32
32.58
1980:Q1-1989:Q4
Tsize - qmv
Esize - mv
Fsize
SCI
53.88
37.35
5.97
AMCI
0.00
28.55
28.22
RECI
2.76
12.73
42.04
Other
43.35
21.36
23.76
1990:Q1-1999:Q4
Tsize - qmv
Esize - mv
Fsize
SCI
51.27
19.45
7.32
AMCI
0.00
36.04
36.72
RECI
3.69
9.72
36.14
Other
45.03
34.79
19.82
2000:Q1-2007:Q3
Tsize - qmv
Esize - mv
SCI
47.48
23.61
8.86
AMCI
0.00
29.29
38.24
RECI
2.78
6.05
31.95
Other
49.74
41.05
20.95
Fsize
Crisis Period (2007:Q4-2013:Q1)
Tsize - qmv
Esize - mv
Fsize
SCI
50.39
16.15
8.05
AMCI
0.00
69.62
42.97
RECI
3.12
2.3
31.97
Other
46.49
11.94
17.00
Source: Authors’ calculations.
Notes: This table shows the sample averages for the share of total shadow banking for which each type of credit intermediation accounts. All statistics are
percentages of the total size of shadow banking. See Table 1 for variable definitions. AMCI stands for asset management credit intermediation, which we
define as the component of asset management which occurs in the shadow banking sector. SCI and RECI are securities credit intermediation and real estate
credit intermediation, respectively. Exact definitions of these types of credit intermediation can be found in Appendix B.
70
Components of U.S. Financial-Sector Growth, 1950-2013
shadow banks, SCI firms accounted for the largest shares
by Tsize − qmv and Esize − mv. Greenwood and Scharfstein
(2013) note that the rise of asset management firms is
closely correlated with asset prices, which rose strongly
in the 1990s. Consistent with this finding, publicly listed
AMCI firms grew from about 29 percent in the 1980s to
36 percent in the 1990s, according to the Esize − mv metric.
Interestingly, the market capitalization of AMCI firms grew
strongly even during the recent crisis, with their share
using Esize − mv jumping from about 29 percent of all
shadow banking in the early 2000s to about 70 percent in
the crisis. Since many of these AMCI firms are funds which
do not file quarterly or annual reports with the Securities
and Exchange Commission, we do not have balance sheet
data for them and thus exclude them from all Tsize − qmv
calculations. The share of publicly listed “other” shadow
banks was also large and growing in the pre-crisis period,
mostly due to increases in assets of secondary market
financing and general finance companies.
To determine how shadow banking has evolved relative
to the traditional banking sector, we measure the share
of shadow banks in total credit intermediation (TCI). In
particular, we look at ratios of the form _______
SB +SBDCI , where
DCI (SB) is either the asset or equity value of DCI (shadow
banking) firms in CRSP-Compustat or the total liability of
the DCI (shadow banking) sector in the FOF data.
The share of shadow banking in TCI has grown steadily
since 1980 (Chart 5). While shadow banking has always
made up a nontrivial portion of TCI (at least 9 percent),
it grew to a peak of between 33 percent and 69 percent
post-2004, depending on the measure) as Table 5 shows. By
all measures except Esize, the share of shadow banking in
TCI grew consistently until the period between 2000 and
the third quarter of 2007, when the growth rate decelerated
while remaining positive, and then turned negative in the
recent crisis (Chart 6).
Depository Credit Intermediation
Did the DCI sector shrink over time or did it simply
expand at a slower pace than shadow banking? To
examine its size and evolution, we also consider the
metric Csize, based on commercial bank assets reported
in Call Reports (as described in Section 3).
The Fsize and Csize measures show a striking pattern
of persistent decline for the DCI sector (Chart 7).
These measures attain their peak early in the sample (the
fourth quarter of 1954 for Fsize and the fourth quarter of
1975 for Csize; see Table 6) and have negative average growth
Chart 5
The Share of Shadow Banking in Total
Credit Intermediation
Percent
80
70
Fsize
60
50
40
AA
Esize – mv
30
20
10
0
1950
Q1
Tsize – qmv
1960
Q1
1970
Q1
1980
Q1
1990
Q1
2000
Q1
2010
Q1
Source: Authors’ calculations.
Notes: This chart shows measures of the share of shadow banking in
total credit intermediation (TCI). The TCI sector is the sum of credit
intermediation by the shadow banking sector and the depository
credit institutions sector. See Table 1 for variable definitions.
rates over the entire sample period (-0.95 percent for Fsize
and -2.20 percent for Csize; see Chart 8). The median growth
rates based on these measures became particularly negative
in the 1980s and 1990s and were only mildly positive in
the 2000s (Chart 8). Growth rates for publicly listed DCIs,
per Tsize − qmv, were also negative on average. Only the
Esize − mv measure shows positive average growth for
DCIs over the whole sample period.
4.3 Asset Management, Securities,
Real Estate, and Insurance
We next focus on the size of the entire asset management,
securities and real estate sectors, rather than specifically
examining their credit intermediation components. We also
examine the insurance sector. The results are not reported
here but are available in Appendix C online.
For the asset management sector, Table C1 indicates
almost 5.7 million firm-day observations for the Esize − mv
sample, but only about 600,000 firm-day observations in the
Tsize − qmv sample. This difference is because most open-end
funds do not report balance sheet data, and so we exclude
them from our Tsize − qmv calculations (see Appendix A).
FRBNY Economic Policy Review / December 2014
71
Table 5
Share of Shadow Banking in Total Credit Intermediation
Full Sample
Tsize - qmv
Esize - mv
Fsize
AA
1,524,082
7,186,652
245
91
Mean
19.94
39.58
35.53
28.58
Median
17.39
39.84
27.86
29.1
Min / min quarter
8.61 / 1975:Q2
24.07 / 1982:Q1
8.86 / 1952:Q1
22.04 / 1990:Q2
Max / max quarter
33.08 / 2004:Q2
61.26 / 2011:Q4
69.42 / 2007:Q2
34.81 / 2008:Q1
Observations
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv
Esize - mv
Fsize
AA
1,222,813
4,199,269
111
69
Mean
21.73
38.36
49.44
28.35
Median
20.67
39.62
52.33
29.1
Min / min quarter
9.73 / 1981:Q3
24.07 / 1982:Q2
22.53 / 1980:Q1
22.04 / 1990:Q2
Max / max quarter
33.08 / 2004:Q2
51.53 / 2000:Q3
69.42 / 2007:Q2
34.57 / 2007:Q2
Observations
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Tsize - qmv
Esize - mv
Fsize
AA
165,346
2,311,902
22
22
20.69
55.9
63.15
29.31
19.08
55.96
62.5
29.12
Min / min quarter
16.60 / 2011:Q4
49.01 / 2008:Q4
60.1 / 2011:Q3
24.56 / 2013:Q1
Max / max quarter
29.38 / 2010:Q2
61.26 / 2011:Q4
69.2 / 2007:Q4
34.81 / 2008:Q1
Median
Source: Authors’ calculations.
Notes: This table shows the summary statistics of measures of the share of shadow banking in total credit intermediation (TCI). TCI is the sum of credit
intermediation by the shadow banking sector and the depository credit institutions sector. Observation units are firm days for Tsize − qmv and Esize − mv
and quarters for Fsize and AA. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then average across days
for each quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions.
Accordingly, we place more emphasis on the results based
on Esize − mv when evaluating the performance of the asset
management sector.
Asset management had a relatively small average share
of the business sector ranging from about 2 percent to
3 percent using the MVE-based measures to 6 percent by
Fsize (see Table C1). However, the sector has been growing
rapidly by all measures except Tsize − qmv. The Fsize
and Esize − mv measures show average growth rates of
about 8 percent and 4 percent in the sample, respectively,
including during the recent crisis (Chart C1). While the
Fsize measure marks consistent growth in all decades, the
72
Components of U.S. Financial-Sector Growth, 1950-2013
MVE-based measures suggest more intermittent growth
that has surged since 2000, consistent with Greenwood and
Scharfstein (2013), who find a similar pattern of rapid recent
growth based on industry revenues.
The securities sector has been about 1 percent to 4 percent
of the business sector on average, peaking at 2 percent to
8 percent just before the recent crisis (Table C2). We find
a consistent pattern of growth in most decades, with an
acceleration since 2000, in contrast with Greenwood and
Scharfstein (2013), who find that securities growth peaked
in 2001 (see Chart C2). Our measures unanimously show
securities firms shrinking during the crisis.
Chart 6
Median Percentage Change in the Share of Shadow Banking in Total Credit Intermediation, by Period
Tsize – qmv
Esize – mv
Fsize
AA
Percent
8
6
4
2
0
-2
-4
-6
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows median annualized quarter-to-quarter percentage changes in the share of shadow banking in total credit intermediation (TCI)
for each measure over several periods. The TCI sector is the sum of credit intermediation by the shadow banking sector and the depository credit
institutions sector. For Tsize − qmv and Esize − mv, we first aggregate from the firm level to the sector level and then calculate quarterly changes.
See Table 1 for variable definitions.
The size and evolution of the real estate sector present
sharply contrasting pictures depending on whether we use
the MVE-based measures or the Fsize measure. Real estate
firms were small relative to the universe of publicly listed
firms over the whole period, based on Tsize − qmv and
Esize − mv, with sample averages under 0.40 percent of total
publicly listed firm assets (Table C3), but they have grown
since the 1980s and especially during the crisis (Chart C3).
In contrast, the Fsize metric shows that real estate accounted
for more than 5 percent of total liabilities on average, with
its share peaking at 12 percent in first quarter of 2003
(constituting almost a third of all shadow banking liabilities)
before shrinking during the crisis (Table C3).
Finally, the insurance sector is the largest of the noncredit
intermediation sectors, with an average relative size of
more than 21 percent over the sample period (peaking at
27 percent in first quarter of 1998) per Fsize and about
9 percent (peaking at 15 percent in the third quarter of 2004)
per Tsize − qmv (Table C4). The sector grew steadily but
moderately over most of the sample period and then crashed
in the recent crisis (Chart C4).
Chart 7
The Relative Size of Depository Credit Institutions
45
Percent
C
40
Tsize – qmv
35
F
30
E
25
T
Fsize
20
15
Csize
10
Esize – mv
5
1965
Q1
1975
Q1
1985
Q1
1995
Q1
2005
Q1
2013
Q1
Source: Authors’ calculations.
Notes: This chart shows measures of the size of depository credit
institutions relative to the financial and nonfinancial sectors. See
Table 1 for variable definitions.
FRBNY Economic Policy Review / December 2014
73
Table 6
The Relative Size of Depository Credit Institutions
Full Sample
Tsize - qmv
Esize - mv
Fsize
Csize
4,237,897
5,236,753
245
1,816,776
Mean
30.49
6.09
27.69
20.21
Median
29.88
5.81
29.7
17.28
Observations
Min / min quarter
22.43 / 2000:Q3
4.71 / 1976:Q3
15.44 / 2000:Q1
14.14 / 2000:Q3
Max / max quarter
41.66 / 2009:Q1
16.17 / 2013:Q1
36.59 / 1954:Q4
32.04 / 1975:Q4
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv
Esize - mv
Fsize
Csize
3,225,438
4,063,827
111
1,349,320
Mean
28.87
6.38
22.23
19.04
Median
29.37
6.04
20.87
16.79
Min / min quarter
22.43 / 2000:Q3
3.21 / 1980:Q4
15.44 / 2000:Q1
14.14 / 2000:Q3
Max / max quarter
31.94 / 1991:Q1
9.42 / 2002:Q3
31.81 / 1980:Q1
30.78 / 1980:Q4
Observations
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Tsize - qmv
Esize - mv
Fsize
Csize
745,412
766,541
22
166,487
38.89
6.28
19.05
17.12
39.59
6.18
19.3
17.14
Min / min quarter
34.56 / 2007:Q4
5.23 / 2009:Q1
16.35 / 2007:Q4
16.05 / 2007:Q4
Max / max quarter
41.66 / 2009:Q1
7.18 / 2010:Q2
20.27 / 2011:Q3
18.08 / 2012:Q2
Median
Source: Authors’ calculations.
Notes: This table reports summary statistics of measures of the size of depository credit institutions relative to the financial and nonfinancial sectors.
Observation units are firm days for Tsize − qmv, Esize − mv, and Csize and quarters for Fsize. Units for all other statistics are percentages.
For Tsize − qmv and Esize − mv, we first sum over firms, then average across days for each quarter, and finally take means and medians of quarterly averages.
See Table 1 for variable definitions. Min (max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
5. Firm Size and Heterogeneity,
Leverage, and Regulation
In this section, we examine the effects of firm size and
heterogeneity, leverage, and regulation on financialsector growth. Philippon and Reshef (2013) suggest
that increased concentration in the banking sector may
be responsible for the increasing income share of finance.
Policy initiatives have sought to mitigate negative externalities
generated by too-big-to-fail firms.17 Motivated by these
concerns, we estimate our size measures for large and small
financial firms separately. Regarding firm heterogeneity,
17
See http://www.federalreserve.gov/newsevents/reform_milestones.htm for
examples of policy proposals for regulation of large and complex institutions.
74
Components of U.S. Financial-Sector Growth, 1950-2013
Philippon (2012) notes that the mixture of new and old
firms changes significantly over time, reflecting waves of
technological change (Jovanovic and Rousseau 2005) and
affecting measures of financial-sector size. We adjust for
firm-level heterogeneity through firm fixed effects in a panel
regression. Heightened awareness of the risks of leverage led
to a minimum leverage ratio of 3 percent for banks under
the Basel III regulatory framework as well as a proposal
for additional capital requirements for large bank holding
companies by U.S. regulators.18 Calomiris and Nissim (2012)
18
See https://www.bis.org/publ/bcbs270.htm for Basel III leverage
ratio requirements and http://www.federalreserve.gov/newsevents/
press/bcreg/20130709a.htm for the proposal to strengthen leverage
ratio standards by the Federal Reserve, the FDIC, and the Office of
the Comptroller of the Currency.
Chart 8
Median Percentage Change in the Relative Size of Depository Credit Institutions, by Period
Tsize – qmv
10
Esize – mv
Fsize
Csize
Percent
8
6
4
2
0
-2
-4
-6
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows median annualized quarter-to-quarter percentage changes in the relative size of depository credit institutions for each measure for
specific periods. Size is relative to the financial and nonfinancial sectors. For Tsize − qmv and Esize − mv, we first aggregate from the firm level to the sector
level and then calculate quarterly changes. See Table 1 for variable definitions.
find that leverage is an important determinant of the market
value of commercial banks. Thus, to investigate the effect
of leverage on our size measures, we compare equityonly with total asset-based measures. Finally, we consider
the effect of select financial regulations on changes in
financial-sector size. Philippon and Reshef (2012) suggest
that regulation discourages skilled workers and conclude that
it is the main determinant of the demand for skill and wages
in the U.S. financial sector. Philippon and Reshef (2013)
find that, with some exceptions, countries that deregulate
more also experience larger increases in the relative skill
intensity in finance.
5.1 Firm Size and Heterogeneity
Our disaggregated data allow us to evaluate whether the
growth of finance is mainly due to the growth of large
financial firms or whether it is more broadly based. We
first take a look at trends in the Herfindahl-Hirschman
index (HHI) of market concentration for the financial
and nonfinancial sectors. Both sectors show low levels of
concentration that have changed little over time. Given
the low and stable concentration in both the financial and
nonfinancial sectors, we estimate the relative size of small
and large financial firms separately. For each metric and each
year, we partition firms at the beginning of the year into two
subsets. Large firms are those in the top 10 percent of firms,
while small firms are defined as the remaining 90 percent
of firms, based on Tsize − qmv or Esize − mv.19 We then
estimate the share of value of large (small) firms in sector S
as a percentage of the total value of large (small) firms in the
financial and nonfinancial sectors.20 Thus, for large firms i on
day t, the size measure for sector S is:
(9)
Large
Value
jϵS
i,t
Size SLarge,t = _______________
.
Large Large
jϵF
,NF
Valuej,t
Similarly, for small firms i on day t, the size metric
for sector S is:
(10)
jϵSSmall Valuei,t
Size SSmall,t = _______________
.
Small Small
jϵF
,NF
Valuej,t
19
We also tried a lower cutoff for small firms (such the bottom 50 percent of
firms) and obtained similar mean shares but substantially larger volatility in
the shares from year to year.
20
The share of finance in small firms may increase because large financial
firms have decreased in size and become small, or vice versa. Likewise,
an increase in the share of finance in large firms could be due to small
financial firms growing and joining the large sample. Thus, growth in the
share of finance in the large (small) firm sample need not be the same
as the relative growth of large (small) finance firms. We use SLarge (SSmall)
to denote the intersection of sector S with the top ten percent (bottom
90 percent) of all firms.
FRBNY Economic Policy Review / December 2014
75
Table 7
The Relative Size of Large and Small Financial Firms
Full Sample
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
6,396,748
1,892,004
6,396,748
1,892,004
Mean
18.54
55.56
8.21
16.81
Median
19.25
51.44
8.34
16.43
Min / min quarter
8.76 / 1984:Q1
46.60 / 1976:Q3
3.41 / 1975:Q4
9.53 / 1981:Q3
Max / max quarter
27.86 / 1994:Q2
73.15 / 2008:Q4
12.94 / 2003:Q1
26.12 / 2006:Q3
Observations
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
5,010,409
1,468,739
5,010,409
1,468,739
Mean
18.26
53.73
8.77
17.35
Median
19.14
50.96
9.53
16.43
Min / min quarter
8.76 / 1984:Q1
47.17 / 1983:Q1
3.81 / 1981:Q1
9.53 / 1981:Q3
Max / max quarter
27.86 / 1994:Q2
69.68 / 2007:Q3
12.94 / 2003:Q1
26.12 / 2006:Q3
Observations
Crisis Period (2007:Q4-2013:Q1)
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
1,094,084
217,514
1,094,084
217,514
Mean
20.81
70.10
8.98
19.28
Median
20.76
70.09
8.67
19.52
Min / min quarter
19.20 / 2012:Q1
67.12 / 2013:Q1
8.20 / 2011:Q3
16.07 / 2009:Q1
Max / max quarter
24.03 / 2008:Q4
73.15 / 2008:Q4
11.05 / 2008:Q4
22.11 / 2007:Q4
Observations
Source: Authors’ calculations.
Notes: This table reports, for each size measure, summary statistics of the relative size of large and small financial firms in the sample. For each year and each
size measure, we rank all publicly listed firms by Tsize − qmv. The top 10 percent of firms are included in the large firm sample, while the remaining firms are
included in the small firm sample. We estimate our size measures separately for the large and small firm samples. Observation units are firm days for Tsize − qmv
and Esize − mv. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then average across days for each
quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions. Min (max) quarter refers to the quarter in which the
measure achieves its minimum (maximum) value in the sample.
We find that financial firms are far more prevalent in the
sample of large firms than they are in the sample of small
firms. Within any period and for any measure, the relative
size of finance is two to three times bigger in the large firm
sample than in the small firm sample (Table 7). For example,
by Tsize − qmv, large financial firms account for 56 percent
of all large firms on average whereas small financial firms are
19 percent of all small firms on average for the full sample.
Median annualized growth rates show the relative size
of large financial firms growing moderately more than the
small financial firms (Chart 9). According to Tsize − qmv,
76
Components of U.S. Financial-Sector Growth, 1950-2013
small financial firms grew more until the 1990s, but large
financial firms have grown more (or declined less) since then.
Esize − mv shows large financial firms growing more in every
decade since the 1980s. Both metrics show that small financial
firms did worse than large financial firms during the crisis.
Large shadow banks also make up a larger proportion of
all large firms than do small shadow banks of all small firms,
although the difference is not as pronounced as for financial firms
in general. Thus, the sample mean of the relative size of large
shadow banks is over 8 percent whereas it is less than 3 percent
for smaller shadow banks, according to Tsize − qmv (Table 8).
Chart 9
Median Percentage Change in the Relative Size of Large and Small Financial Firms
Tsize – qmv_small
Tsize – qmv_large
Esize – mv_small
Esize – mv_large
Percent
8
6
4
2
0
-2
-4
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows, for each size measure, the median annualized quarter-to-quarter percentage changes in the relative size of large and small
financial firms in the sample. We estimate our measures separately for the large and small firm samples. We first aggregate from the firm level to the
sector level and then calculate quarterly changes. See Table 1 for variable definitions.
The corresponding means for Esize − mv are 3 percent of the
large firm sample and 2 percent of the small firm sample. The
share of large shadow banks in the large firm sample has grown
more than the share of smaller shadow banks, although the
difference is moderate according to Esize − mv (Chart 10).
We do see that the recent crisis had a harsher effect on large
shadow banks whose share in the large firm sample declined
by more than 4 percent during the crisis, while the share of
smaller shadow banks grew in the same period.
Large DCIs are a bigger share of all large firms than
are small DCIs of all small firms, and the difference is
substantial. For example, the sample mean of the relative size
of large DCIs is about 34 percent by Tsize − qmv, more than
three times the sample mean of 11 percent for small DCIs
(Table 9). In addition, the gap between the relative shares of
small and large DCIs has been increasing. We see in Chart 11
that the relative size of small DCIs has been declining over
time, whereas the reverse is true for large DCIs. Moreover,
large DCIs have consistently outgrown small DCIs in most
decades since the 1980s. In the recent crisis period, small
DCIs shrank more than large DCIs by Tsize − qmv while the
opposite was true based on Esize − mv.
Firm-size effects illustrate the impact of firm heterogeneity
generally. Since our measures are aggregated up to sectors
from firm-level data, the sectoral means are potentially
affected by firm-level heterogeneity. To account for this,
we estimate a firm-level panel regression using firm size
(relative to the total size of the business sector, as in the
denominator of equation 1) as the dependent variable. We
include all financial firms in the sample and regress the
relative firm-size variable upon period and firm fixed effects.
Chart 12 shows estimates of these period fixed effects,
divided by the estimate of the regression intercept, using
Tsize − qmv as the size measure. We find that, when firmlevel heterogeneity is accounted for, financial-sector growth
becomes more consistent. In particular, the dips in size
around 2000, and during the crisis, are considerably
muted, suggesting that these may have been largely
firm-level effects.
To quantify the effect of firm heterogeneity on the size
of different credit intermediation subsectors, we regress
estimates of the period fixed effects, in a second stage, on
sector-level dummy variables, omitting the nonfinancial
sector. The results confirm the descriptive statistics.
Specifically, the coefficient on the shadow banking sector is
positive and significant for all measures, while the coefficient
on the DCI sector is negative and significant for Fsize and
positive and significant for Tsize − qmv, indicating the
relative expansion of the shadow banking sector and the
relative decline of the DCI sector per the Fsize measure.
FRBNY Economic Policy Review / December 2014
77
Table 8
The Relative Size of Large and Small Shadow Banking Firms
Full Sample
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
1,372,634
195,870
1,372,634
195,870
Mean
2.81
8.63
1.81
2.68
Median
2.68
8.42
1.56
2.3
Observations
Min / min quarter
1.46 / 1981:Q4
2.98 / 1976:Q2
0.91 / 1975:Q4
0.64 / 1975:Q4
Max / max quarter
4.64 / 1998:Q3
17.02 / 2010:Q2
3.56 / 1997:Q4
6.28 / 2004:Q1
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
1,129,896
148,134
1,129,896
148,134
Mean
3.00
9.07
2.09
3.01
Median
3.02
8.34
1.91
2.54
Observations
Min / min quarter
1.46 / 1981:Q4
3.65 / 1981:Q3
0.95 / 1981:Q4
0.74 / 1982:Q1
Max / max quarter
4.64 / 1998:Q3
16.09 / 2007:Q3
3.56 / 1997:Q4
6.28 / 2004:Q1
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
135,935
33,416
135,935
33,416
2.29
11.22
1.12
2.73
2.19
10.51
1.10
2.78
Min / min quarter
1.9 / 2008:Q4
8.64 / 2012:Q4
0.91 / 2009:Q1
1.77 / 2009:Q1
Max / max quarter
2.9 / 2012:Q4
17.02 / 2010:Q2
1.33 / 2012:Q4
4.31 / 2007:Q4
Median
Source: Authors’ calculations.
Notes: This table reports, for each size measure, summary statistics of the relative size of large and small shadow banking firms in the sample. For each year
and each size measure, we rank all publicly listed firms by Tsize − qmv. The top 10 percent of firms are included in the large firm sample, while the remaining
firms are included in the small firm sample. We estimate our size measures separately for the large and small firm samples. Observation units are firm days for
Tsize − qmv and Esize − mv. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then average across days for
each quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions. Min (max) quarter refers to the quarter in which the
measure achieves its minimum (maximum) value in the sample.
5.2 Leverage
To examine the effect of leverage on the growth pattern
of financial firms, we consider the equity-only metrics
Esize − mv and Esize − bv and compare them with the total
value measures Tsize − qmv and Tsize − bv, respectively.
The Esize − bv and Tsize − bv measures use the BVE rather
than the MVE of firms, as shown in Table 1. The BVE-based
results are reported in Appendix D online.
78
Components of U.S. Financial-Sector Growth, 1950-2013
The equity-only measures show finance to be smaller
than the total liabilities measures, but growing at a faster
rate. Thus, Esize measures had sample means of 19 percent
or less (Tables 2 and D1) compared with at least 50 percent
using the Tsize measures. The difference increased during
the crisis, with the Esize measures being 40 percentage
points lower than the respective Tsize measures. This result
indicates that balance sheet leverage has become relatively
more prevalent in the capital structure of financial firms
Chart 10
Median Percentage Change in the Relative Size of Large and Small Shadow Banking Firms
Tsize – qmv_small
8
Tsize – qmv_large
Esize – mv_large
Esize – mv_small
Percent
6
4
2
0
-2
-4
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows, for each size measure, the median annualized quarter-to-quarter percentage changes in the relative size of large and small
shadow banking firms in the sample. We estimate our measures separately for the large and small firm samples. We first aggregate from the firm level
to the sector level and then calculate quarterly changes. See Table 1 for variable definitions.
Chart 11
Median Percentage Change in the Relative Size of Large and Small Depository Credit Institutions
Tsize – qmv_small
8
Tsize – qmv_large
Esize – mv_small
Esize – mv_large
Percent
6
4
2
0
-2
-4
Full sample
Pre-1980
1980s
1990s
2000-07:Q3
Crisis
Source: Authors’ calculations.
Notes: This chart shows, for each size measure, the median annualized quarter-to-quarter percentage changes in the relative size of large and small
depository credit institutions in the sample. We estimate our measures separately for the large and small firm samples. We first aggregate from the
firm level to the sector level and then calculate quarterly changes. See Table 1 for variable definitions.
than in that of nonfinancial firms. The median annualized
growth rate for finance was higher using the equity-only
metrics, being 2.6 percent to 3.6 percent for the whole
sample according to the Esize measures compared
with 0.75 percent to 1.9 percent for the Tsize measures
(Charts 2 and D1).
Our measures also highlight the importance of
balance sheet leverage for the DCI subsector, more
so than for shadow banks. For example, the mean
relative size of DCI over the sample period is between
6 percent and 9 percent for the equity-only measures
(Tables 6 and D3) and between 30 percent and 33 percent
FRBNY Economic Policy Review / December 2014
79
Table 9
The Relative Size of Large and Small Depository Credit Institutions
Full Sample
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
3,199,397
1,140,087
3,199,397
1,140,087
Mean
10.75
34.36
2.30
7.60
Median
12.35
34.35
2.20
7.40
Min / min quarter
2.99 / 1984:Q2
23.99 / 2000:Q3
0.35 / 1984:Q3
4.66 / 1980:Q4
Max / max quarter
17.37 / 1994:Q2
44.75 / 2011:Q3
4.97 / 2003:Q1
11.12 / 2003:Q4
Observations
Pre-crisis Period (1980:Q1-2007:Q3)
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
2,413,211
903,242
2,413,211
903,242
Mean
10.07
32.23
2.35
7.76
Median
12.15
33.21
2.25
7.52
Min / min quarter
2.99 / 1984:Q2
23.99 / 2000:Q3
0.35 / 1984:Q3
4.66 / 1980:Q4
Max / max quarter
17.37 / 1994:Q2
37.73 / 1980:Q2
4.97 / 2003:Q1
11.12 / 2003:Q4
Observations
Crisis Period (2007:Q4-2013:Q1)
Observations
Mean
Tsize - qmv_small
Tsize - qmv_large
Esize - mv_small
Esize - mv_large
663,933
91,295
663,933
91,295
12.12
42.47
2.62
8.15
12.21
43.64
2.45
8.30
Min / min quarter
10.31 / 2013:Q1
37.37 / 2007:Q4
2.19 / 2011:Q2
6.33 / 2009:Q1
Max / max quarter
15.62 / 2008:Q4
44.75 / 2011:Q3
4.19 / 2008:Q4
9.49 / 2010:Q2
Median
Source: Authors’ calculations.
Notes: This table reports, for each size measure, summary statistics of the relative size of large and small depository credit institutions in the sample. For
each year and each size measure, we rank all publicly listed firms by Tsize − qmv. The top 10 percent of firms are included in the large firm sample, while the
remaining firms are included in the small firm sample. We estimate our size measures separately for the large and small firm samples. Observation units are firm
days for Tsize − qmv and Esize − mv. Units for all other statistics are percentages. For Tsize − qmv and Esize − mv, we first sum over firms, then average across
days for each quarter, and finally take means and medians of quarterly averages. See Table 1 for variable definitions. Min (max) quarter refers to the quarter in
which the measure achieves its minimum (maximum) value in the sample.
for the total value-based measures. While the shadow
banking subsector also had a larger measured size based on
leverage, its dependence on balance sheet leverage was not as
stark.21 However, given the importance of off-balance-sheet
leverage for shadow banks, this result need not indicate a
lower overall dependence on leverage of shadow banks.
21
For example, the sample means of the relative size of shadow banks using the
equity-only measures were about 3 to 6 percentage points (Tables 3 and D2)
smaller than those using total value measures. The two DCI measures differed
by more than 20 percentage points.
80
Components of U.S. Financial-Sector Growth, 1950-2013
5.3 Regulation
We examine the effects of three important pieces of banking
regulation on financial-sector size: the Riegle-Neal Act, the
Gramm-Leach-Bliley Act, and the FDIC program of debt
guarantees. The Riegle-Neal Act repealed interstate bank
branching restrictions and allowed interstate bank mergers, while
the Gramm-Leach-Bliley Act rolled back additional restrictions
on bank consolidations.22 By facilitating bank mergers and
22
See http://en.wikipedia.org/wiki/Bank_Holding_Company_Act.
book value of banks issuing guaranteed debt compared with all
other firms), but statistically insignificant.24
Chart 12
The Relative Size of Finance, Accounting
for Firm Heterogeneity
Percent
80
6. Conclusion
60
40
Tsize – qmv
20
0
-20
-40
1975
Q2
1980
Q2
1985
Q2
1990
Q2
1995
Q2
2000
Q2
2005
Q2
2010
Q2
Source: Authors’ calculations.
Notes: This chart shows estimates of period fixed effects as a
percentage of the estimated intercept from these regressions. Using
only finance firms, we create a quarterly firm-level panel of relative
size, as measured by Tsize − qmv. We estimate a panel regression of
Tsize − qmv on firm level and period fixed effects. See Table 1 for
Tsize − qmv definition.
consolidations, these acts may have led to an increase in the
relative share of large banks in all large firms, as compared with
the relative share of small banks in all small firms. We find
evidence consistent with this hypothesis. For example, before
the fourth quarter of 1999, the relative share of large DCIs in all
large firms compared with small DCIs in small firms was about
1.4 percentage points higher on average (by Tsize − qmv). But
after that time, the relative share of large DCIs in all large firms
was 6.6 percentage points higher on average than that of small
DCIs in small firms. This difference of five percentage points is
statistically significant. We see a similar increase in the relative
size of large DCIs after the passage of the Riegle-Neal Act.
The shrinkage of finance during the crisis may have been
mitigated, at least temporarily, by the FDIC’s Temporary
Liquidity Guarantee Program (TLGP) program, which backed
in full the senior unsecured debt issued by participating
entities between October 14, 2008, and October 31, 2009.23 We
investigate the effect of the TLGP program by comparing banks
that issued guaranteed debt under the program with the rest of
the firms in our sample. We find a positive treatment effect that
is economically meaningful (that is, an 11 percent increase in the
23
See http://www.fdic.gov/regulations/resources/TLGP/index.html.
In this article, we provide a comprehensive picture of the
historical growth of finance and its subsectors using a
variety of firm- and sector-level size measures. We define
financial-sector size relative to the business sector (financial
plus nonfinancial). We find that, with one exception, finance
grew relative to the nonfinancial sector, especially from the
late 1980s, whether one considers publicly listed firm liabilities
or total sectoral liabilities (inclusive of private firms), equity
or total asset values, large or small firms, or book or market
values. The only exception is that, based on total value (market
value of equity plus book value of debt), small financial firms
did not increase their relative size on average, mainly due to
the effects of the recent financial crisis. Indeed, the finance
sector shrank relative to the nonfinancial sector during the
recent crisis, and its recovery has been tepid.
Our analysis further shows that shadow banking grew
rapidly at the expense of traditional banks, becoming a
significant portion of the financial sector in the mid-1990s,
and peaking just before the crisis, consistent with the
literature. The growth in shadow banking was driven by the
securities and asset management subsectors, and we find that
small and large shadow banks grew similarly. The traditional
banking sector, in contrast, declined by some measures, with
growth in this sector being mostly explained by large banks.
Finance was smaller but grew faster when measured based
on the liabilities of publicly listed firms than when measured
based on the liabilities of all firms. That financial liabilities
make up a substantially larger portion of total liabilities
when private firms are included may be of importance to
policymakers. Private firms not only face less regulation
than publicly listed firms, but also operate with far less
transparency. Indeed, comprehensive and reliable data on
private firms are not available and most private firms are
not required to submit quarterly financial statements to
regulators. Similar concerns have been raised about shadow
banks, leading to an internationally coordinated effort to
collect data on shadow banks as well as proposals to regulate
24
We used a two-period difference-in-difference specification, where the
dependent variable was the change in average relative size from the year of the
TLGP (fourth quarter of 2008 to fourth quarter of 2009) to the year preceding
the program (third quarter of 2007 to third quarter of 2008). It was regressed on a
dummy for the program year, a dummy for the issuing banks, and an interaction
term between the two. Results are available upon request.
FRBNY Economic Policy Review / December 2014
81
the sector (Financial Stability Board 2012). However, no such
initiative exists generally for private firms. While many private
firms are small and may not pose significant systemic risk
presently, opacity can hide the buildup of vulnerabilities.
Financial firms are relatively larger based on their total
asset values (equity plus debt) than on their equity values only.
Large traditional banks are particularly dependent on balance
sheet leverage, which indicates that the leverage restrictions
on banks, as proposed under the Basel III agreement, can
be effective policy tools for restricting the size of banks.
By contrast, shadow banks are less dependent on leverage,
suggesting that policymakers might need a different toolkit to
monitor and regulate them.
A concern with our approach (and of the literature) is
the inability to distinguish sufficiently between form and
function (for example, when considering how to categorize
a traditional bank that carries on shadow banking activities).
We use NAICS and SIC codes to classify firms into industries.
These classifications are based on the primary business of a
company, which may lead to classification errors in some cases.
For example, though many financial holding companies may
be bank holding companies, if NAICS has determined that
banking is not their primary business, we do not categorize
them as banks or DCIs but rather as “finance, other.”
Fortunately, given the small number of firms in this category,
these potential misclassifications have little effect on our
results. Moreover, we can mitigate these errors to some extent.
In some cases, we use Call Reports to identify banks directly.
In particular, if our mapping indicates that a publicly listed
company is a Call Report–filing commercial bank, then we call
it a DCI regardless of what NAICS calls it. Further, to the extent
that market prices accurately incorporate information about
a firm’s activities, our use of market values may mitigate this
concern. Nevertheless, more research is needed on this issue.25
25
Cetorelli and Peristiani (2012) and Cetorelli, McAndrews, and Traina
(2014) are important steps in this direction. Cetorelli and Peristiani (2012)
find that regulated banks played a dominant role in all aspects (issuance,
underwriting, trustee, and servicing) of the securitization of asset-backed
securities between 1978 and 2008. Cetorelli, McAndrews, and Traina (2014)
find that banks expanded horizontally by acquiring shadow banking firms.
82
Components of U.S. Financial-Sector Growth, 1950-2013
References
Financial Stability Board. 2011. “Shadow Banking: Strengthening
Oversight and Regulation.” October.
Greenwood, R., and D. Scharfstein. 2013. “The Growth of Finance.”
Journal of Economic Perspectives 27, no. 2 (Spring): 3-28.
_______. 2012. “Global Shadow Banking Monitoring Report
2012.” November.
Jovanovic, B., and P. L. Rousseau. 2005. “General Purpose
Technologies.” In P. Aghion and S. Durlauf, eds., Handbook of
Economic Growth. Vol. 1, part B. Amsterdam: Elsevier.
Adrian, T., and A. Ashcraft. 2012. “Shadow Banking: A Review of the
Literature.” Federal Reserve Bank of New York Staff Reports,
no. 580, October.
Asker, J., J. Farre-Mensa, and A. Ljungqvist. 2014. “Corporate
Investment and Stock Market Listing: A Puzzle?” ECGI-Finance
working paper, January. Available at http://ssrn.com/
abstract=1603484 and http://dx.doi.org/10.2139/ssrn.1603484.
Berk, J. B., and R. C. Green. 2004. “Mutual Fund Flows and
Performance in Rational Markets.” Journal of Politcal
Economy 112, no. 6 (December): 1269-95.
Boyd, J. H., and M. Gertler. 1995. “Are Banks Dead? Or Are
the Reports Greatly Exaggerated?” NBER Working Paper
no. 5045, February.
Calomiris, C. W., and D. Nissim. 2012. “Crisis-Related Shifts in the
Market Valuation of Banking Activities.” NBER Working Paper
no. 17868, February.
Cetorelli, N., and S. Peristiani. 2012. “The Role of Banks in Asset
Securitization.” Federal Reserve Bank of New York Economic
Policy Review 18, no. 2 (July): 1-12.
Cetorelli, N., J. McAndrews, and J. Traina. 2014. “Evolution in Bank
Complexity.” Federal Reserve Bank of New York Economic
Policy Review 20, no. 2 (December): 85-106.
Cochrane, J. H. “Finance: Function Matters, Not Size.” Journal of
Economic Perspectives 27, no. 2 (Spring): 29-50.
Goldin, C., and L. Katz. 2008. “Transitions: Career and Family Life
Cycles of the Educational Elite.” American Economic Review:
Papers and Proceedings 98, no. 2 (May): 363-69, Papers and
Proceedings of the 120th Annual Meeting of the American
Economic Association.
Kaplan, S., and J. Rauh. 2013. “Family, Education, and Sources of
Wealth among the Richest Americans, 1982-2012.” American
Economic Review: Papers and Proceedings 103, no. 3 (May):
158-62, Papers and Proceedings of the 125th Annual Meeting of
the American Economic Association.
King, R. G., and R. Levine. 1993. “Finance and Growth: Schumpeter
Might Be Right.” Quarterly Journal of Economics 108, no. 3
(August): 717-37.
Philippon,T. 2012. “Has the U.S. Finance Industry Become Less
Efficient? On the Theory and Measurement of Financial
Intermediation.” NBER Working Paper no. 18077, May.
Philippon, T., and A. Reshef. 2012. “Wages and Human Capital in the
U.S. Financial Industry: 1909-2006.” Quarterly Journal of
Economics 127, no. 4 (November): 1551-1609.
______. 2013. “An International Look at the Growth of Modern
Finance.” Journal of Economic Perspectives 27, no. 2
(Spring): 73-96.
Pozsar, Z., T. Adrian, A. Ashcraft, and H. Boesky. 2013. “Shadow
Banking.” Federal Reserve Bank of New York Economic Policy
Review 19, no. 2 (December): 1-16.
Rajan, R., and L. Zingales. 1998. “Financial Dependence and
Growth.” American Economic Review 88, no. 3 (June): 559-86.
Turner, A. 2010. “What Do Banks Do? Why Do Credit Booms and
Busts Occur and What Can Public Policy Do about It?” In The
Future of Finance: The LSE Report. London: London School
of Economics and Political Science.
Wright, R. E. 2002. The Wealth of Nations Rediscovered.
Cambridge: Cambridge University Press.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2014
83
Components of U.S. Financial-Sector Growth,
1950-2013
Samuel Antill, David Hou, and Asani Sarkar
Appendixes A–D: Additional Materials
Contents
Appendix A: Data
Appendix B: Classifications
Appendix C: Additional Subsector Tables
Appendix D: Equity Tables and Charts
FRBNY Economic Policy Review / December 2014
Appendix A: Data
Appendix A: Data
CRSP and Compustat
CRSP and Compustat
The market value of equity is computed as the market price of shares times the shares outstanding.
The data on public equity prices and shares outstanding comes from the Center for Research in
Security Prices (CRSP). We pull the universe of public equity data from January 1, 1950, to March
31, 2013. Data on prices and shares outstanding are adjusted for dividends, stock splits, and other
capital actions using the cumulative factors to adjust prices and shares, respectively. We also
correct for the convention of entering the negative bid ask spread for price on days with no trading
activity. We drop observations for which there are neither SIC nor NAICS data, since we can not
reliably classify such a firm-day as finance or nonfinance. Finally, we exclude from our sample dates
that are well-known outliers (Black Monday 1987; September 11 2001; Hurricane Sandy, October
29 2012).
The data on book value of equity and debt comes from the Compustat database. All data are
collected at the consolidated balance-sheet level and not at the level of firm subsidiaries. We
similarly drop observations for which there are neither SIC nor NAICS data. Our measure of total
book value of debt is calculated as the total book value of assets (ATQ) minus the total book value
of stockholder equity or book equity (SEQQ). The quarterly balance-sheet data from Compustat
begin in 1961, but coverage of firms in Compustat for the earlier years (prior to 1975) is incomplete.
As an example, our liabilities variable is missing for over 85 percent of observations prior to 1975.
Since the data are incomplete and the NAICS and SIC coverage is far more sparse, we begin our
Compustat series from 1975 rather than 1961. Fortunately, coverage is fairly complete for recent
data. To the best of our knowledge, CRSP coverage does not suffer substantially from gaps for our
sample period.1
Identification of sectors is performed using two systems: the Standard Industrial Classification (SIC)
and the North American Industry Classification System (NAICS). SIC served as the standard from
its origination in 1937 to 1997, when NAICS was developed as its eventual successor. However,
NAICS codes are not available in the CRSP database prior to August 24, 2001. Therefore, we
rely on a combination of SIC and NAICS codes to group firms. Where discrepancies arise, we take
NAICS as the standard. Both classification systems break out institutions at a level of granularity
sufficient to allow us to distinguish between financial and nonfinancial firms and to define our
specific sector categories such as shadow banking, DCIs, securities, and insurance firms. We have
discovered no discrepancies in the classification systems of SIC and NAICS that would materially
affect our results. The classification of sectors in our paper is described in appendix B.
1
As a precautionary measure, we also used supplementary data sources to verify our metrics. The SEC
EDGAR database indicates that the total number of firms filing 10-K forms is roughly in line with our CRSP
calculations.
A.1
Flow-of-Funds
Flow-of-Funds
We use liability data from the Flow-of-Funds (FOF) Accounts of the United States (Z.1) database
released by the Federal Reserve Board of Governors which reports the book value of liabilities of
various sectors. To classify subsectors within FOF, we use the levels dataset within the FOF data.
We match the categories in this dataset as closely as possible with the same finance subsectors that
we defined using SIC and NAICS data. The FOF sector classifications are described in appendix
B.
Flow-of-Funds
Call Reports
We use liability data from the Flow-of-Funds (FOF) Accounts of the United States (Z.1) database
For banks, we supplement our previously described measures with additional data from the Consolreleased by the Federal Reserve Board of Governors which reports the book value of liabilities of
idated Reports of Condition and Income (call reports). The advantage of these data is that it covers
various sectors. To classify subsectors within FOF, we use the levels dataset within the FOF data.
private and publicly listed commercial banks and is available at the bank level. The banks required
Call Reports
We match the categories in this dataset as closely as possible with the same finance subsectors that
to file this report include all national banks, state member banks, and nonmember insured banks.
we defined using SIC and NAICS data. The FOF sector classifications are described in appendix
In particular, we use the book value of liabilities and the book value of equity for all commercial
B.
banks that are required to file the call report regulatory form.
Call Reports
For banks, we supplement our previously described measures with additional data from the Consolidated Reports of Condition and Income (call reports). The advantage of these data is that it covers
private and publicly listed commercial banks and is available at the bank level. The banks required
to file this report include all national banks, state member banks, and nonmember insured banks.
In particular, we use the book value of liabilities and the book value of equity for all commercial
banks that are required to file the call report regulatory form.
B.1
Appendix B: Classifications
Appendix B: Classifications
Financial firms according to the SIC system are taken as those in Division H: Finance, Insurance,
and Real Estate but outside of Major Group 65: Real Estate. Financial firms according to NAICS
are those in Sector 52: Finance and Insurance. Tables B.1 and B.2 show the NAICS and SIC codes
corresponding to each financial subsector.
Within financial firms, depository credit institutions are firms involved in “the activities typically associated with traditional banking — lending to consumers and corporations, deposit taking,
and processing financial transactions ”Greenwood and Scharfstein (2013, 6). Shadow banks, also
referred to as nondepository credit institutions (NDCI), are of particular interest as a subsector.
They can be broadly defined as firms that conduct “credit intermediation involving entities and
activities outside the regular banking system ”(FSB 2011, 1). In classifying shadow banking, FSB
(2011, 3) has the philosophy that “authorities should cast the net wide, looking at all nonbank
credit intermediation to ensure that data gathering and surveillance cover all areas where shadow
banking-related risks to the financial system might potentially arise.”
In its definition of shadow banking, FSB (2012) includes hedge funds, finance companies, structured
finance vehicles, financial holding companies, broker dealers, MMMFs, U.S. funding corporations,
and other investment funds. In a similar vein, we consider several subsectors of shadow banking.
Securities firms are those involved in “securities trading and market making, securities underwriting ”Greenwood and Scharfstein (2013, 6). Asset management firms are firms involved in
portfolio management and investment management. Poszar et al. (2013, 14) describe the shadow
banking sector as containing an “internal shadow banking subsystem ... the credit intermediation
process of a global network of banks, finance companies, broker-dealers, and asset managers.”In
light of this, we include a subset of asset management and securities firms in our definition of
shadow banking (see tables B.1 and B.2). We also include a real estate section in our definition
of shadow banking, which contains real estate credit firms and mortgage loan brokers.
Outside of shadow banking, insurance firms are firms that provide or facilitate the provision of
a type of insurance or pension. All other financial firms not classified as asset management, securities firms, depository credit, shadow banks, or insurance are aggregated in the others category.
Nonfinancial firms are all firms not classified as financials.
One area of discrepancy between SIC and NAICS is in the treatment of financial holding companies.
Bank and other financial holding companies are lumped in Division H in SIC, but are broken out
into Sector 55: Management of Companies and Enterprises in NAICS. We treat NAICS code
551111 Offices of Bank Holding Companies as a direct indicator of a financial firm. Firms with
codes 551112 Offices of Other Holding Companies and 551114 Corporate, Subsidiary, and Regional
Managing Offices are treated as finance only if their SIC code also meet the criteria listed above.
B.2
It is worth noting that both SIC and NAICS classify firms based on their primary type of economic
activity. Publicly owned conglomerates with significant concentration in multiple lines of business
will necessarily be funneled into one category. The same phenomenon occurs with financial firms;
JPMorgan Chase, for example, is treated as depository credit (6021) despite having a nonnegligible
securities arm.
In cases where there were partial NAICS codes available that were nonambiguous (e.g., 524 is
always an insurance prefix), we classify them as the relevant subsector. In particular, for old SIC
codes that we were not able to find the names of, we classified them according to the two digit
prefix: 60 for DCI, 61 for NDCI, 62 for Securities, and 63 and 64 for Insurance. All SIC codes
beginning in 67 that we were not able to identify were called Other. SIC code 6711 is in some cases
a Bank Holding Company while in other cases it is a nonbank holding company. To remedy this,
we consider 6711 to be Other, except for any firms that correspond to banks that appear in the
call reports data, we call them DCI.
For FOF data, table B.3 shows how the FOF categories correspond to each financial subsector.
In the FOF data, we consider depository credit institutions to be those companies listed in
L.109, representing private depository institutions. This category encompasses L.110 through L.113,
which are U.S. Chartered Depository Institutions, Foreign Banking Offices in the U.S., Banks in
U.S.-Affiliated Areas, and Credit Unions, respectively, as well as L.128, Holding companies. We
consider asset management to be those companies listed in L.119 through L.121, which are
Money Market Mutual Funds, Mutual Funds, and Closed-End and Exchange Traded Funds (ETFs),
respectively. We consider securities firms to be those companies listed in L.127, Security Brokers
and Dealers. Real estate firms are those companies listed in L.122,L.123, and L.126, GovernmentSponsored Enterprises (GSE), Agency and GSE backed Mortgage, and Real Estate Investment
Trusts, respectively. Since we were unable to replicate the shadow banking classification used by
FSB (2011), we follow our CRSP/Compustat classification and consider shadow banks to include
all those firms considered asset management, real estate, or securities firms, as well as L.124,
Issuers of Asset Backed Securities, and L.125, Finance Companies. We consider insurance to
be those companies listed in L.115 through L.118, which are Life Insurance Companies, Private
Pension Funds, State and Local Government Employee Retirement Funds, and Federal Government
Retirement Funds, respectively. Finally, we categorize other as those firms listed in L.129, Funding
Corporations, and L.108, Monetary Authority.2
2
Because Monetary Authority, GSE, and Agency and GSE backed Mortgage are government sectors,
including them in the Finance category might distort our inference on the effect of the crisis. We explain
the effect of excluding GSEs in the main text. Excluding Monetary Authority had little to no effect on our
results– it is virtually nonexistent in the publicly-listed measures, and it is a small portion of finance by
F size.
B.3
Table B.1: Financial and Shadow Banking Categorization, by NAICS code
NAICS
521110
522110
522120
522130
522190
522210
522220
522291
522292
522293
522294
522298
522310
522320
522390
523110
523120
523130
523140
523210
523910
523920
523930
523991
523999
524113
524114
524126
524127
524128
524130
524210
524291
524292
524298
525110
525120
525190
525910
525920
525930
525990
551111
Name
Monetary Authorities-Central Bank
Commercial Banking
Savings Institutions
Credit Unions
Other Depository Credit Intermediation
Credit Card Issuing
Sales Financing
Consumer Lending
Real Estate Credit
International Trade Financing
Secondary Market Financing
All Other Nondepository Credit Intermediation
Mortgage and Nonmortgage Loan Brokers
Financial Transactions Processing, Reserve, and Clearinghouse Activities
Other Activities Related to Credit Intermediation
Investment Banking and Securities Dealing
Securities Brokerage
Commodity Contracts Dealing
Commodity Contracts Brokerage
Securities and Commodity Exchanges
Miscellaneous Intermediation
Portfolio Management
Investment Advice
Trust, Fiduciary, and Custody Activities
Miscellaneous Financial Investment Activities
Direct Life Insurance Carriers
Direct Health and Medical Insurance Carriers
Direct Property and Casualty Insurance Carriers
Direct Title Insurance Carriers
Other Direct Insurance (except Life, Health, and Medical) Carriers
Reinsurance Carriers
Insurance Agencies and Brokerages
Claims Adjusting
Third Party Administration of Insurance and Pension Funds
All Other Insurance Related Activities
Pension Funds
Health and Welfare Funds
Other Insurance Funds
Open-End Investment Funds
Trusts, Estates, and Agency Accounts
Real Estate Trusts
Other Financial Vehicles
Offices of Bank Holding Companies
Financial Subsector
Other
DCI
DCI
DCI
DCI
NDCI
NDCI
NDCI
NDCI
NDCI
NDCI
NDCI
NDCI
Other
Other
Sec
Sec
Sec
Sec
Other
AM
AM
AM
AM
AM
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
AM
AM
NDCI
AM
DCI
SB Subsector?
No
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
Yes
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
No
Yes
Yes
No
Type of SB CI
OCI
OCI
OCI
RECI
OCI
OCI
OCI
OCI
SCI
SCI
SCI
SCI
AMCI
RECI
AMCI
-
This table reports classifications by NAICS code. We abbreviate Shadow Banking as SB and Credit Intermediation as CI.
AM is Asset Management, AMCI is Asset Management Credit Intermediation, Sec is Securities, and SCI is Securities Credit Intermediation.
DCI is Depository Credit Intermediation, NDCI is Nondepository Credit Intermediation, and OCI is Other Nondepository Credit Intermediation.
RECI is Real Estate Credit Intermediation, Ins is Insurance, and Other is our catchall for financial firms that do not fit into the above categories.
B.4
Table B.2: Financial and Shadow Banking Categorization, by SIC code
SIC
Name
Financial Sector
SB Subsector?
Type of SB CI
5932
6011
6019
6021
6022
6029
6035
6036
6061
6062
6081
6082
6091
6099
6111
6141
6153
6159
6162
6221
6231
6282
6289
6311
6321
6324
6331
6351
6361
6371
6399
6411
6711
6712
6719
6722
6726
6733
6792
6798
6799
7389
Used Merchandise Stores
Federal Reserve Banks
Central Reserve Depository, Nec
National Commercial Banks
State Commercial Banks
Commercial Banks, Nec
Federal Savings Institutions
Savings Institutions, Except Federal
Federal Credit Unions
State Credit Unions
Foreign Bank and Branches and Agencies
Foreign Trade and International Banks
Nondeposit Trust Facilities
Functions Related To Depository Banking
Federal and Federally Sponsored Credit
Personal Credit Institutions
Short-term Business Credit
Miscellaneous Business Credit
Mortgage Bankers and Correspondents
Security Brokers and Dealers
Security and Commodity Exchanges
Investment Advice
Security and Commodity Service
Life Insurance
Accident and Health Insurance
Hospital and Medical Service Plans
Fire, Marine, and Casualty Insurance
Surety Insurance
Title Insurance
Pension, Health, and Welfare Funds
Insurance Carriers, Nec
Insurance Agents, Brokers, and Service
Financial Holding Companies
Bank Holding Companies
Holding Companies, Nec
Management Investment, Open-ended
Investment Offices, Nec
Trusts, Nec
Oil Royalty Traders
Real Estate Investment Trusts
Investors, Nec
Business Services, Nec
NDCI
Other
NDCI
DCI
DCI
DCI
DCI
DCI
DCI
DCI
NDCI
NDCI
AM
DCI
NDCI
NDCI
NDCI
NDCI
NDCI
Sec
Other
AM
AM
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Ins
Other
DCI
Other
AM
AM
AM
AM
NDCI
Sec
Other
Yes
No
Yes
No
No
No
No
No
No
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
Yes
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Yes
Yes
No
No
Yes
Yes
No
OCI
OCI
OCI
OCI
AMCI
OCI
OCI
OCI
OCI
RECI
SCI
AMCI
AMCI
RECI
SCI
-
This table reports classifications by SIC code. We abbreviate Shadow Banking as SB and Credit Intermediation as CI.
AM is Asset Management, AMCI is Asset Management Credit Intermediation, Sec is Securities, and SCI is Securities Credit Intermediation.
DCI is Depository Credit Intermediation, NDCI is Nondepository Credit Intermediation, and OCI is Other Nondepository Credit Intermediation.
RECI is Real Estate Credit Intermediation, Ins is Insurance, and Other is our catchall for financial firms that do not fit into the above categories.
B.5
Table B.3: Financial and Shadow Banking Subsector Categorization, by Flow-of-Funds Category
Flow-of-Funds Category
Financial Sector
SB Subsector?
Type of SB CI
L.108 Monetary Authority
L.110 US Chartered Depository Institutions, ex. Credit Unions
L.111 Foreign Banking Offices in U.S.
L.112 Banks in U.S.-Affiliated Areas
L.113 Credit Unions
L.114 Property-Casualty Insurance Companies
L.115 Life Insurance Companies
L.116 Private Pension Funds
L.117 State and Local Gobernment Employee Retirement Funds
L.118 Federal Government Retirment Funds
L.119 MMMF
L.120 Mutual Funds
L.121 Closed-End and ETF
L.122 GSE
L.123 Agency and GSE backed Mortgage
L.124 Issuers of ABS
L.125 Finance Companies
L.126 REITs
L.127 Security Brokers and Dealers
L.128 Holding Companies
L.129 Funding Corporations
Other
DCI
DCI
DCI
DCI
Ins
Ins
Ins
Ins
Ins
AM
AM
AM
NDCI
NDCI
NDCI
NDCI
NDCI
Sec
DCI
Other
No
No
No
No
No
No
No
No
No
No
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
No
No
AMCI
AMCI
AMCI
RECI
RECI
OCI
OCI
RECI
SCI
-
This table reports classifications by flow-of-funds category. We abbreviate Shadow Banking as SB and Credit Intermediation as CI.
AM is Asset Management, AMCI is Asset Management Credit Intermediation, Sec is Securities, and SCI is Securities Credit Intermediation.
DCI is Depository Credit Intermediation, NDCI is Nondepository Credit Intermediation, and OCI is Other Nondepository Credit Intermediation.
RECI is Real Estate Credit Intermediation, Ins is Insurance, and Other is our catchall for financial firms that do not fit into the above categories.
B.6
Appendix C: Additional Subsector Tables
Table C.1: The Relative Size of Asset Management
Full Sample
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
612,792
1.96
1.89
0.51/2000q1
3.59 / 2013q1
Esize-mv
5,696,913
2.67
2.08
0.74 / 1981q3
8.23 / 2013q1
Fsize
245
5.73
3.05
0.65 / 1953q3
15.73 / 2013q1
Precrisis: 1980Q1-2007Q3
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
458,254
1.78
1.72
0.51 / 2000q1
3.59 / 2007q1
Esize-mv
3,159,196
2.13
2.05
0.74 / 1981q3
4.96 / 2007q1
Fsize
111
8.17
8.09
1.66 / 1980q1
14.45 / 2007q3
Crisis: 2007Q4-2013Q1
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
124,399
2.88
2.85
2.47 / 2010q1
3.34 / 2008q1
Esize-mv
2,268,862
6.9
7.02
5.19 / 2007q4
8.23 / 2013q1
Fsize
22
13.99
14.04
11.97 / 2009q1
15.73 / 2013q1
This table reports summary statistics of measures of the size of asset management, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-days for T size − qmv and Esize − mv, and quarters for F size. Units for all other statistics are percentages.
For T size − qmv and Esize − mv we first sum over firms, then average across days for each quarter, and finally take means and medians of quarterly averages.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
C.4
Table C.2: The Relative Size of Securities
Full Sample
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
359,077
3.86
4.12
0.59 / 1975q1
7.91 / 2007q3
Esize-mv
670,854
0.99
0.96
0.32 / 1978q1
1.96 / 2007q2
Fsize
245
1.34
0.81
0.36 / 1975q3
4.07 / 2007q2
Precrisis: 1980Q1-2007Q3
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
274,723
4.17
4.12
1.06 / 1981q3
7.91 / 2007q3
Esize-mv
424,656
1.04
0.95
0.39 / 1980q2
1.96 / 2007q2
Fsize
111
1.75
1.73
0.50 / 1980q1
4.07 / 2007q2
Crisis: 2007Q4-2013Q1
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
55,456
5.01
4.58
4.25 / 2012q4
7.47 / 2008q1
Esize-mv
64,997
1.27
1.27
0.96 / 2009q1
1.62 / 2007q4
Fsize
22
2.65
2.43
2.2 / 2013q1
4.00 / 2008q1
This table reports summary statistics of measures of the size of securities, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-days for T size − qmv and Esize − mv, and quarters for F size. Units for all other statistics are percentages.
For T size − qmv and Esize − mv we first sum over firms, then average across days for each quarter, and finally take means and medians of quarterly averages.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
C.5
Table C.3: The Relative Size of Real Estate
Full Sample
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
682,763
0.25
0.19
0.10 / 1982q2
0.54 / 2013q1
Esize-mv
739,939
0.31
0.32
0.09 / 2008q2
0.60 / 1997q4
Fsize
245
5.43
4.71
0.63 / 1952q4
11.93 / 2003q1
Precrisis: 1980Q1-2007Q3
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
589,998
0.26
0.22
0.10 / 1982q2
0.48 / 1998q2
Esize-mv
613,617
0.37
0.39
0.14 / 1981q3
0.60 / 1997q4
Fsize
111
8.04
8.39
4.21 / 1980q1
11.93 2003q1
Crisis: 2007Q4-2013Q1
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
41,905
0.28
0.20
0.15 / 2008q4
0.54 / 2013q1
Esize-mv
49,711
0.19
0.16
0.09 / 2008q2
0.32 / 2013q1
Fsize
22
10.44
10.32
9.53 / 2013q1
11.46 / 2009q1
This table reports summary statistics of measures of the size of real estate, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-days for T size − qmv and Esize − mv, and quarters for F size. Units for all other statistics are percentages.
For T size − qmv and Esize − mv we first sum over firms, then average across days for each quarter, and finally take means and medians of quarterly averages.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
C.6
Table C.4: The Relative Size of Insurance
Full Sample
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
1,480,709
8.70
7.84
3.77 / 1975q1
14.81 / 2004q3
Esize-mv
2,055,054
4.50
4.48
2.43 / 1976q2
6.67 / 2005q4
Fsize
245
21.14
20.99
16.45 / 2009q1
26.78 / 1998q1
Precrisis: 1980Q1-2007Q3
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
1,190,676
8.81
7.74
4.63 / 1980q4
14.81 / 2004q3
Esize-mv
1,475,086
4.70
4.71
2.82 / 1980q4
6.67 / 2005q4
Fsize
111
22.45
21.91
18.29 / 1981q3
26.78 / 1998q1
Crisis: 2007Q4-2013Q1
Observations
Mean
Median
Min/Min quarter
Max/Max quarter
Tsize-qmv
225,982
11.86
11.30
9.94 / 2013q1
13.6 / 2008q2
Esize-mv
233,403
4.74
4.59
4.34 / 2011q3
5.87 / 2007q4
Fsize
22
19.41
19.72
16.45 / 2009q1
20.58 / 2012q3
This table reports summary statistics of measures of the size of insurance, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-days for T size − qmv and Esize − mv, and quarters for F size. Units for all other statistics are percentages.
For T size − qmv and Esize − mv we first sum over firms, then average across days for each quarter, and finally take means and medians of quarterly averages.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
C.7
Figure C.1: Median Percent Change in the Relative Size of Asset
Management, By Period
Tsize-qmv
Esize-mv
Fsize
20
15
10
5
0
Full
Pre 1980
1980s
1990s
2000-2007:Q3
Crisis
-5
-10
-15
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of asset management firms
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-qmv and Esize-mv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Figure C.2: Median Percent Change in the Relative Size of
Securities, By period
Tsize-qmv
Esize-mv
Fsize
20
15
10
5
0
Full
Pre-1980
1980s
1990s
2000-2007:Q3
Crisis
-5
-10
-15
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of securities firms
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-qmv and Esize-mv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Figure C.3: Median Percent Change in the Relative Size of Real
Estate, By Period
Tsize-qmv
Esize-mv
Fsize
20
15
10
5
s
isi
Cr
7:
00
-2
00
20
Pr
-5
e1
98
0
Q3
0
-10
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of real estate firms
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-qmv and Esize-mv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Figure C.4: Median Percent Change in the Relative Size of
Tsize-qmv
Insurance, By Period
Esize-mv
Fsize
10
8
6
4
2
0
Full
Pre-1980
1980s
1990s
2000-2007:Q3
Crisis
-2
-4
-6
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of insurance firms
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-qmv and Esize-mv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Appendix D: Equity Tables and Charts
Online Appendix D: Book Value Tables and Charts
Table D.1: Finance, % of business sector (F+NF)
Full Sample
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
127,707
53.15
38.74/1976q1
72.08/2008q1
Esize-bv
127,707
19.23
9.04/1975q4
32.02/2013q1
Precrisis: 1980Q1-2007Q3
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
99,663
52.39
39.92/1981q1
70.54/2007q3
Esize-bv
99,663
18.94
9.39/1980q1
28.77/2007q3
Crisis: 2007Q4-2013Q1
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
20,271
69.02
66.74/2012q4
72.08/2008q1
Esize-bv
20,271
29.66
27.13/2008q3
32.02/2013q1
This table reports summary statistics of measures of the size of finance, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-quarters for T size − bv and Esize − bv. Units for all other statistics are percentages.
For T size − bv and Esize − bv we first sum over firms, then take means and medians of quarterly values.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
D.1
Table D.2: Shadow Banking, % of business sector (F+NF)
Full Sample
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
24,304
8.39
2.90/1975q2
16.19/2010q2
Esize-bv
24,304
2.66
0.95/1982q2
5.14/2005q2
Precrisis: 1980Q1-2007Q3
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
19,873
8.86
3.17/1981q3
15.76/2007q3
Esize-bv
19,873
2.70
0.95/1982q2
5.14/2005q2
Crisis: 2007Q4-2013Q1
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
2,599
10.91
8.59/2012q4
16.19/2010q2
Esize-bv
2,599
3.97
3.68/2010q1
4.44/2007q4
This table reports summary statistics of measures of the size of shadow banking, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-quarters for T size − bv and Esize − bv. Units for all other statistics are percentages.
For T size − bv and Esize − bv we first sum over firms, then take means and medians of quarterly values.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
D.2
Table D.3: Depository Credit Institutions, % of business sector (F+NF)
Full Sample
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
67,700
32.71
26.13/2001q1
43.82/2011q3
Esize-bv
67,700
9.31
4.71/1976q3
16.17/2013q1
Precrisis: 1980Q1-2007Q3
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
51,619
31.40
26.13/2001q1
35.07/2007q3
Esize-bv
51,619
8.98
4.77/1980q1
12.25/2007q3
Crisis: 2007Q4-2013Q1
Observations
Mean
Min/Min quarter
Max/Max quarter
Tsize-bv
11,739
42.03
38.62/2007q4
43.82 / 2011q3
Esize-bv
11,739
15.02
13.32/2008q2
16.17/2013q1
This table reports summary statistics of measures of the size of depository credit institutions, relative to the financial and nonfinancial sectors (in percent).
Observation units are firm-quarters for T size − bv and Esize − bv. Units for all other statistics are percentages.
For T size − bv and Esize − bv we first sum over firms, then take means and medians of quarterly values.
See table 1 for variable definitions.
Min (Max) quarter refers to the quarter in which the measure achieves its minimum (maximum) value in the sample.
D.3
Figure D.1: Median Percent Change in the Relative Size of
Tsize-bv
Finance, By Period
Esize-bv
8
7
6
5
4
3
2
1
0
-1
Full
Pre-1980
1980s
1990s
2000-2007:Q3
Crisis
T
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of finance
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-bv and Esize-bv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Figure D.2: Median Percent Change in the Relative Size of
Shadow Banking, By Period
Tsize-bv
Esize-bv
14
12
10
8
6
4
2
0
-2
Full
Pre-1980
1980s
1990s
2000-2007:Q3
Crisis
-4
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of shadow banking
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-bv and Esize-bv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Figure D.3: Median Percent Change in the Relative Size of Depository
Tsize-bv
Credit Intermediation, By Period
Esize-bv
6
5
4
3
2
1
0
-1
Full
Pre-1980
1980s
1990s
2000-2007:Q3
Crisis
-2
-3
-4
The figure shows median quarter to quarter annualized changes (in percent) in the relative size of depository credit institutions
for each measure for specific periods.
Size is relative to the financial and nonfinancial sectors.
For Tsize-bv and Esize-bv, we first aggregate from the firm level to the sector level, then calculate quarterly changes.
See table 1 for variable definitions.
Nicola Cetorelli, James McAndrews, and James Traina
Evolution in Bank
Complexity
• In the 1980s, the top ten bank holding
companies accounted for about 20 percent
of total bank assets; that percentage is now
above 50 percent.
1. Introduction
T
• The findings suggest that greater complexity
is a natural adaptation on the part of banks
to a new model of finance oriented to
securitization.
he financial intermediation industry has experienced
significant structural transformations over the past twenty
to thirty years. Some of these changes are well known. Since
the 1980s, for instance, the number of commercial banks
operating in the United States fell from about 14,000 to 6,000.
Most of this reduction was the result of a well-documented
process of consolidation, encouraged in large part by geographic deregulation. Along the way, both the average size
of bank holding companies (BHCs) and their market shares
increased remarkably. In the 1980s, the top ten BHCs accounted for about 20 percent of total bank assets; that percentage is
now above 50 percent. Not only did they grow in size, but the
remaining entities also grew substantially in organizational
complexity, incorporating a large and growing number of subsidiaries spanning the entire spectrum of business activities
within the financial sector.
In particular, the transformation of the financial intermediation industry has generated a few banking behemoths,
and public debate has focused on ways to regulate such
supersized institutions. There are a number of proposed
Nicola Cetorelli is an assistant vice president, James McAndrews an executive
vice president and the director of research, and James Traina a former senior
research analyst at the Federal Reserve Bank of New York.
The views expressed in this article are those of the authors and do not
necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System.
• Bank holding companies have not only
grown in size, but they have also become
substantially more complex, incorporating a
large number of subsidiaries that span the
entire spectrum of business activities within
the financial sector.
• The authors document and analyze banks’
organizational evolution, posing questions
about the forces driving the industry and
firm structures evident today.
Correspondence: nicola.cetorelli@ny.frb.org
FRBNY Economic Policy Review / December 2014
85
approaches to such regulation, including breakups, size caps,
or business activity limits. Other suggestions include enhanced regulations in the form of capital and long-term debt
requirements, capital surcharges, stress tests, and improved
resolution planning.
Although the discussion around the largest entities is certainly important, we suggest that their emergence is part of a
larger process that has transformed the financial intermediation industry more broadly. In this paper, we document and
analyze how the industry evolved and pose questions about
what might have been the forces that drove the industry and
firm structures we see today.
Despite the intense debate on bank complexity, very little
documentation or analysis exists on the dynamics leading
to the current industry configuration. In fact, even the
meaning and metrics of complexity are debatable; in both
comparative and absolute terms, we lack a clear consensus
on how to assess an entity’s complexity. This problem is
important not only from a positive angle, as we strive to understand the economics behind the phenomenon, but also
from a normative angle, as we decide on policy measures
exclusively for complex institutions. How do we establish
how complex entities are? Where do we draw the line across
institutions?
In this paper, we focus on organizational complexity.1 We
look at organizational structure as gauged by the number
and types of subsidiaries organized under common ownership and control. A focus on organizational complexity
has multiple implications for policy analysis. It seems, for
instance, a natural way to look at issues of resolvability and
systemic importance. An institution with more legally organized affiliates, perhaps engaged in diverse business activities or located across geographic borders, presents greater
challenges for orchestrating an orderly resolution. Similarly,
entities with complex organizational structures may experience systemic events of broader scope: shocks can spread to
multiple industries within the financial sector as they propagate across the many affiliates of the organization, perhaps
accelerated by “cross-default” clauses in debt and derivative
contracts. Finally, a complex organizational structure is a
direct gauge of how complex regulation itself might be, or
need to be, and thus of the challenges to effective oversight
of complex organizations.2
This paper is the first to offer a rich documentation of
the evolution in organizational structure of U.S. financial
intermediation firms. Using comprehensive data on the
universe of U.S. financial mergers and acquisitions over
the past thirty years, we track the process of consolidation
and cross-industry acquisitions and show a significant
expansion in the complexity of banking institutions. Our
study indicates that banks have transformed into increasingly expanding holding companies, extending their
organizational footprint into nontraditional bank business
lines through acquisitions of already formed specialized
subsidiaries. This process of organizational transformation
is substantial and far-reaching and is not confined simply to
the largest entities of today. The massive sequence of transactions was also surprisingly gradual and “hidden in plain
sight”: given the regulated nature of bank holding companies,
this process occurred with the explicit authorization of the
regulator.
Multiple factors likely drove the rise in organizational
complexity of banking institutions in the early 1990s. The
process of geographic deregulation that has taken place in the
past thirty years or so, which allowed banks to consolidate and
expand both within and across state lines, may be one such
factor; it allowed banks to reach sufficient scale to expand into
nonbank sectors. The passage of the Gramm-Leach-Bliley Act
(GLB), also known as the Financial Modernization Act of 1999,
sanctioned and reinforced this process, even though, as we
show in the data, a great deal of nonbank acquisition activity
had already taken place.3
Banks became complex bank holding companies
with control over many subsidiaries and across multiple
sectors of the financial sector. However, we posit that this
intense transformation was the result of a natural process
of adaptation to a changing financial intermediation
“technology.” The traditional bank-centered model, familiar
from textbooks on banking, puts banks as the central
brokers between funding supply and demand. With this
2
1
Alternative metrics focus instead on what an entity does. For instance, the
methodology for the designation of global systemically important banks
proposes as metrics of complexity the notional value of over-the-counter
derivatives, the balance-sheet presence of "Level 3" assets (assets for which
prices cannot be inferred by either markets or models), and the size of
the trading and available-for-sale books. This is a narrower definition of
complexity, likely captured adequately by metrics of scope and diversity in
business lines of the subsidiaries of an organization.
86
Evolution in Bank Complexity
This implication applies directly to the regulation of U.S. bank holding
companies. The Federal Reserve is the regulator of BHCs. However, other
agencies are the principal regulators of specific types of subsidiaries.
3
There is ample literature on the dynamic evolution of the original GlassSteagall Act restrictions on banks’ activities. See, for example, Carpenter
and Murphy (2010) and Omarova and Tahyar (2011). Also see Fein (2004):
“Although the Gramm-Leach-Bliley Act was expected to trigger a cascade of
new consolidation proposals, the onslaught had not materialized . . . perhaps
because much of the consolidation had already occurred prior to the Act.”
model, general-purpose deposit-taking and loan-making
operations define an intermediary and its organizational
boundaries. However, asset securitization changed the
technology of intermediation. Loans no longer have to
reside on the balance sheet of an intermediary. Alternatives to bank deposits can fulfill the liquidity needs of
fund suppliers. Hence, general-purpose banks—in their
traditional form—are less necessary for all intermediation
services. Instead, highly specialized entities have emerged,
each able to offer specific services that taken together fulfill
the functions traditionally provided by banks. This is the
model of intermediation that we are now accustomed to
describing as shadow banking (see, for example, Pozsar et
al. [2010], and Adrian, Ashcraft, and Cetorelli [2013]). This
transformation in the technology of intermediation can
also explain the observed evolution in bank organizational
form: as modern intermediation increasingly relies on
nonbank entities to provide specialized services, banking
organizations can adapt and survive by incorporating
these specialists as subsidiaries under common ownership
and control. Hence, as shadow banking has grown and
become a prevalent model of intermediation, we should
expect banks to enlarge their organizational footprint. In
other words, and in truly Coasian terms (Coase 1937), the
boundaries of the banking firm have expanded progressively to include the activities of nonbank intermediaries,
and this evolution should be reflected in the forms of
increasingly complex bank holding companies.
The debate around the repeal of the Glass-Steagall Act,
which brewed for decades before the passage of GLB, actually
reflected the argument that the technology of intermediation
was changing. For instance, already in 1988, Isaac and Fein
wrote, “Congress [should not] ignore the technological, economic, and competitive forces shifting the financial markets
away from traditional banking channels toward increased use
of the securities markets for financial intermediation. . . .The
securitization of assets has reduced the need for bank loans
even further.” (Isaac and Fein 1988). And two years earlier,
the president of the Federal Reserve Bank of New York stated
that “if securitization were to continue to spread rapidly to
other types of credit, the historic role of the deposit-based
credit intermediation process could be seriously jeopardized”
(Federal Reserve Bank of New York 1986).4
In the next section, we develop this rationalization of
the observed organizational evolution of banking firms in
further detail. In our discussion, in line with the observation above, we purposefully use the terms banks and bank
holding companies interchangeably, in recognition of the
dynamic evolution in the organizational structure of entities
involved in i ntermediation activities. We are, of course, aware
of specific regulatory meanings attached to these terms and to
the existence of other types of entities that are authorized to
conduct banking activity without a BHC organizational form,5
but in practice it turns out that the BHC “model” is the one
that dominates over this time period. Chart 1, from Avraham,
Selvaggi, and Vickery (2012), clearly shows that dominance. In
terms of dollars of assets, BHCs have consistently represented
almost the totality of all bank assets. Section 3 presents the
data, a comprehensive panel of merger-and-acquisition
transactions that have occurred in the U.S. financial sector
over the past thirty years. Section 4 illustrates our method of
using transaction data to construct metrics of complexity for
bank “families,” matched to regulated bank holding companies. Section 5 describes our findings and our interpretation
of the observable evolution of the complexity of bank holding
companies. Section 6 draws concluding remarks.
4
5
Pavel (1986) documents the growing importance of asset securitization and
its implication for traditional banking.
Chart 1
Trends in Number and Total Size
of U.S. Bank Holding Companies (BHCs)
20
Trillions of U.S. dollars
Growth in commercial banking
industry assets over time
15
10
Banks not part of BHC
Small BHCs
Large BHCs: outside of
commercial bank subsidiaries
Large BHCs: commercial
bank subsidiaries
5
Gramm-Leach-Bliley Act
0
1991
1995
2000
2005
2010
Source: Avraham, Selvaggi, and Vickery 2012.
Note: This chart presents financial data up to fourth-quarter 2011. A large
bank holding company is defined as a top-tier firm that files a Federal
Reserve Y-9C report. Commercial bank assets of large BHCs are measured as
the sum of consolidated assets reported by each banking subsidiary in its
Federal Financial Institutions Examination Council Call Report filing.
Nonbank assets of large BHCs are the difference between total assets as
reported in the Y-9C and commercial bank assets as defined above. Assets of
small BHCs reflect only their commercial bank subsidiaries.
Likewise, we are aware of the regulatory evolution even in the meaning of
the word bank (see, again, Omarova and Tahyar [2011]).
FRBNY Economic Policy Review / December 2014
87
2. A Rationale for Increasing Bank
Complexity
Our approach places the evolution of financial intermediaries
within the broader context of the evolution of the financial
intermediation industry. In a recent special issue of the
Economic Policy Review, Cetorelli, Mandel, and Mollineaux
(2012) expounded the main thesis that, with the rising
importance of asset securitization, banks adapted and remained
central players in the process of financial intermediation.6 They
did so by embracing the new activities related to securitization
(Cetorelli and Peristiani 2012) and expanding the footprint of
their organizations, with bank holding companies increasingly
adding a vast array of nonbank subsidiaries (Avraham, Selvaggi,
and Vickery 2012; Copeland 2012).7
Intermediation services are no longer necessarily housed
in a single, one-stop-shop, general-purpose entity. Instead,
highly specialized entities work in parallel and in sequence to
fulfill the functions of the traditional intermediary. For example, asset managers provide liquidity services and products
that are close substitutes for demandable deposits; specialty
lenders originate loans independent of deposit liabilities;
issuers and underwriters guarantee packaging into securities
and market placement; and brokers and dealers manage the
funding and collateral pledging that are at the center of the
securities markets (Kirk et al. 2014).
While this model of intermediation is usually said to allow
for a more efficient allocation of risk and for a solution to
some of the associated agency frictions (such as the asymmetric information between borrowers and lenders or banks
and depositors), it also creates new frictions across the newly
emerging specialized intermediaries (extensively documented,
for instance, in Ashcraft and Schuermann [2008]). Hence,
we argue that while the model allowed for the emergence
of specialized intermediaries, their organization as separate
subsidiaries within a common hierarchy internalized some
of these frictions by sharing sources of intermediation information, coordinating deal flow, benefiting from cross-guarantees
within different parts of the organization, and centralizing the
credit standing of the organization in its entirety.8 Adapting to
this new industrial environment, the complex holding company structures offered key advantages by collecting specialists
together under one corporate organization. Our hypothesis is
6
The volume The Evolution of Banks and Financial Intermediation is available
at http://www.newyorkfed.org/research/epr/2012/EPRvol18n2.pdf.
that those economic advantages drove the emergence of complex bank holding companies. This conglomeration underpins
the value-creation part of complexity.
3. Acquisitions in the Financial
Sector
How does the structure of the intermediation industry evolve
over time? Which entities (whether banks or nonbanks)
undertake significant organizational transformation? How
diffuse is this process in the cross-section? When does it take
place? We address these questions using the SNL Financial
Mergers and Acquisitions (SNL M&A) database.
SNL captures the universe of U.S. financial acquisition
deals starting in 1983 and continuing to the present using
many sources, including press releases, public filings, participant surveys, adviser surveys, and news searches. SNL’s
coverage tracks new financial players involved in M&A
activity, allowing us to track sector-wide growth in size and
complexity.
We start by compiling a panel data set of acquisitions. For
each deal, SNL provides information on the buyer name,9 the
target name, the buyer industry, the target industry, the value,
and the completion date. Because the database lacks a unique
entity identifier, we work with entity names.10 We use SNL’s
general industry-type variable to bin entities by industry. SNL
classifies entities by the Standard Industrial Classification code
sourced from the Securities and Exchange Commission or the
Federal Deposit Insurance Corporation. When such information is missing or ambiguous, SNL internally assigns an
industry code based on major sources of revenue or underwriting operations. It reports the nominal value of the deal,
defined as the total consideration paid to the seller, when that
information is available.
The SNL M&A raw database has over 37,000 deals. We
restrict our analysis to whole-entity acquisitions completed
before 2013. We drop a few observations that we found to
have uninformative participant names, such as “private
investor,” “management group,” or “mortgage banking.” We
also filter out acquisitions in which a participant is not in the
financial sector. Ten industry types remain: bank, asset manager, broker-dealer, financial technology, insurance broker,
insurance underwriter, investment company, real estate,
7
In 2011, for instance, bank holding companies controlled about 38 percent
of the assets of the largest (top twenty) insurance companies, roughly 41
percent of total money market mutual fund assets, and approximately 93
percent of the assets of the largest (top thirty) brokers and dealers (Cetorelli 2012).
8
This argument follows directly from Stein (2002) and Rajan and Zingales (2000).
88
Evolution in Bank Complexity
9
SNL lists the ultimate parent of the actual acquirer as the buyer.
10
To make sure that names are unique within an entity and to reduce
potential coding errors, we clean all names by removing all special characters
and capitalizing all letters.
savings bank/thrift/mutual,11 and specialty lender. Finally,
taking advantage of the fact that some entities appear multiple
times, we fill in the missing fields of an entity if those fields are
unique and available elsewhere in the data set.
In total, 19,532 deals meet these criteria. The data span
23,451 unique U.S. entities (7,893 unique banks), with a total
of 6,507 unique buyers, 18,402 unique targets, and 19,486
unique buyer-target pairs.
Deal value is available when disclosed, as happens with all
public acquisitions. These make up 58 percent of the acquisitions in our data set. For calculation purposes, we set the value
to zero if it is missing. We rely on SNL to convert all non-dollar-denominated values to U.S. dollars using exchange rates
at the completion date, although this conversion is infrequent
because of the U.S.-only nature of the SNL M&A database.
We also normalize all deal values to 2012 dollars using the
consumer price index (CPI) for all urban consumers, all items,
not seasonally adjusted. Since the CPI is available only monthly
but our acquisition data are daily, we linearly interpolate to get
an estimate of the CPI at the deal completion date.
To measure the total acquisition activity of entities, we
construct two aggregates across all acquisitions in which the
entity acts as the buyer. The first consists of the raw number of
deals, while the second consists of the total sum of deal values.
4. Data Construction
Up until now, we have focused on acquisitions. However, this
limits our ability to answer questions on the cumulative effects
of acquisition activity. We therefore extend our analysis to
studying entire organizations, or families, themselves. We consider a family to be the complete picture of a self-owned entity
and all of its subsidiaries.
The term family lends itself to a host of other relevant
terms for the structure of organizations. The exhibit on this
page illustrates an example of a “family tree.” An entity within
a family may have an “immediate parent,” the direct owner,
and an “ultimate parent,” the highest owner in the family tree.
For example, in Tree 1 at Time 0, A is the immediate parent of
B and the ultimate parent of both B and C.
We use our information on acquisitions to assemble a family-level
panel data set. In our earlier data set, an observation is
an acquisition, such as “A buys D.” Our family-level data set
looks at an entire tree as an observation, such as “Tree 1
at Time 0.”
Tree 1
Time 0
Time 1
A
A
B
B
C
C
D
We start with market data using the Center for Research
in Security Prices (CRSP) U.S. Stock Database, provided by
Wharton Research Data Services. A key variable from this
data set is the PERMCO, a unique entity identifier that is
consistent through time. To bring our earlier discussion to the
data, we define a family as any group of entities that share a
PERMCO, thus restricting our sample to public families. We
add in regulatory accounting data from the Board of Governors of the Federal Reserve System, Consolidated Financial
Statements for Bank Holding Companies (FR Y-9C), a quarterly
regulatory report filed by BHCs. To match to these databases,
we add in four more linking identifiers available from the
SNL M&A data set: the ticker symbol of the entity’s primary
exchange stock, the Committee on Uniform Security Identification Procedures code (CUSIP) of the entity’s primary
exchange security, the Federal Reserve Research, Statistics, Supervision and Regulation, and Discount and Credit Database
identifying number (RSSD ID) of the entity,12 and the RSSD
ID of any BHC parent.
A fundamental insight that informed our data construction
is that a family tree requires knowledge only of the immediate
parent of each entity in the family. For instance, in the above
picture, we need only “A owns B” and “B owns C” to identify
“Tree 1 at Time 0.” To construct our panel data set, we exploit
this principle by creating a separate “dictionary” data set that
lists the universe of unique entities in the cleaned SNL M&A
data set. We then create two new variables that track each
entity’s ownership—one for the immediate parent and one for
the ultimate parent. This new data set allows us to “look up”
entities at different points in time, using the immediate and
ultimate parent variables to build a snapshot of the family tree.
12
11
Note the separation of banks and thrifts.
A unique identifier assigned by the Federal Reserve System to all financial
institutions, main offices, and branches. RSSD IDs are the primary identifier
for the FR Y-9C.
FRBNY Economic Policy Review / December 2014
89
Because we lack information on family structure before
SNL’s acquisition coverage, we set the baseline owner of each
entity to itself at the beginning of our data process. Further,
in defining our family structures, we include only entities that
are involved in an acquisition at some point in our sample
period. In other words, our data limitations anchor our results
to changes in complexity relative to our baseline and through
the acquisition channel exclusively; we capture neither the
structure before the start of the SNL M&A data set nor changes through de novo entity creation.
Our primary algorithm updates the dictionary data set
by sequentially reading from the acquisition-level data set
described in section 3. As acquisitions occur, we replace the
target’s parent variables in the dictionary data set. We first
replace the immediate parent with the name of the buyer,
reflecting the change in ownership.13 We assume that whole
acquisitions carry all previously acquired entities, and thus we
replace the immediate parent of all subsidiaries of the target.14
Finally, we update the ultimate parent variable by tracing the
path of immediate parents.
To illustrate our approach, consider again the family
tree above. In the dictionary data set at Time 0, “A owns
B,” “B owns C,” and some entity (perhaps itself) owns D.
When we read the deal “A buys D,” we change D’s immediate parent to A. At Time 1, we have “A owns B,” “B owns C,”
and “A owns D.” To identify the ultimate parent, we simply
trace all entities back to A.
At each quarter-end, we sum the dictionary data set
from entity level to ultimate parent level, constructing a
profile of variables that count the number of subsidiaries
in each industry for each ultimate parent. We append all
quarter-specific cross-sections to form the basis of our
panel data set.
Since we capture changes in organizational structure only
through the acquisition channel, we may be concerned with
important missing links across ultimate parents that do not
appear in our data. To resolve this potential issue, we match
all owners to their CRSP PERMCO and FR Y-9C RSSD ID
at each quarter. This match restricts our sample to public
FR Y-9C filers but ensures a time-consistent and regulatory-based definition of a banking family. As noted above,
our data from SNL include neither PERMCO identifiers
nor RSSD ID identifiers of the top regulatory filer. However, the SNL and CRSP data sets share ticker and CUSIP
variables, allowing a direct match to the PERMCO. Similarly, we use the other SNL-provided RSSD ID variables to
match to the top regulatory filer of the FR Y-9C. As a last
13
Note that in replacing the previous immediate parent, we also capture sales.
14
In all subsidiaries, we include subsidiaries of subsidiaries, subsidiaries of
subsidiaries of subsidiaries, and so on.
90
Evolution in Bank Complexity
layer of robustness, we rely on the PERMCO-RSSD ID link
data set provided by the Federal Reserve Bank of New York
to ensure proper identification of families.15 We then sum
any families with the same PERMCO as before, creating
our final panel data set.
To make sure our algorithm works as intended and
correctly captures important acquisitions, we do a variety of
hand inspections using the raw SNL M&A database and the
National Information Center (NIC) website.16 For instance,
because of its size and acquisition history, Bank of America
offers a rich case study. We look at its history in detail, from
NationsBank’s buy of C&S/Sovran, Fleet’s buy of Shawmut,
BankAmerica’s buy of Security Pacific, and NationsBank and
BankAmerica’s consolidation to the name we know today. Our
database accurately covers all of these important acquisitions.
Among other firms checked are Allco, BNY Mellon, Countrywide, Key, Regions, and Washington Mutual.
Although NIC is the natural choice as the information
center of BHCs, two problems prevent NIC data from helping
our understanding of this evolution when we compare SNL
with NIC, particularly with respect to de novo entity creation.
First, NIC focuses on the regulated banking industry, covering
nonbank financial firms only insofar as they link to regulated
entities. Therefore, unlike SNL, NIC lacks information on
deal-level analysis at the broadest levels of the financial sector.
We cannot see changes in the structure of nonbank financial
firms unless they are already underneath the umbrella of a
BHC. Further, we cannot find out how nonbank financial
firms come under the control of a BHC, such as M&A as opposed to de novo creation. Second, NIC is extremely different
from SNL in its scope of coverage; it is very detailed within
the banking dimension but classifies many other financial
subsidiaries as “domestic entity other,” a catch-all type that
includes some things we care about (asset management
subsidiaries) and some things we do not (collateralized debt
obligations, special-purpose vehicles, and the like). This group
is extremely difficult to disentangle. Conversely, SNL focuses
on specific entity types that are relevant to the asset securitization chain and is thus more useful for our purposes.
Note that the mapping from SNL’s bank-type industry
variable to FR Y-9C filers is not one to one. Of the 1,028
unique RSSD IDs in our family-level data set, about 85 percent are banks and 15 percent are thrifts. Wells Fargo achieves
the highest bank consolidation in fourth-quarter 2008, totaling
15
If any of the identifier matches disagree, we use the link that appears most
often. We have confirmed by hand that this reduces error more than throwing
away data when links are ambiguous.
16
For example, to check for possible conceptual errors in our primary
algorithm, we go through a similar exercise as in our family-tree illustration
with ABN AMRO.
361 banks. By the end of the sample, Regions Financial
Corporation maintains the highest measure at 193.
Our final data set consists of 1,013 families spanning
first-quarter 1988 to fourth-quarter 2012. This sample captures 22 percent of all FR Y-9C filers and 79 percent of all
entities with a PERMCO-RSSD ID link. To give a picture of
size, in fourth-quarter 2010, our sample totals 71 percent of
the book value of equity from the FR Y-9C.
5. Analysis
As premised above, we operationalize bank complexity by
measuring the extent to which a BHC expands its “horizontal”
structure, acquiring entities operating in different industries
of the financial sector. We must stress that our approach allows us to capture only incremental levels of complexity from
acquisition dynamics. We cannot capture organic growth in
complexity (de novo entity creations), nor entities acquired
before the start of our sample period, nor the purpose of the
acquisitions. That said, the quality checks on our constructed
family-level data show that we capture a significant extent
of the overall evolution in organizational structure of the
largest BHCs.
5.1 Sector-Wide Dynamics
We begin by illustrating some of the characteristics of the
original SNL Financial M&A database. As mentioned above,
we partition the data into ten industry types within the financial sector.
Table 1 presents basic information about the acquisitions that take place over the sample period. The far-left
column lists each of the ten industries within our data
set. The “total unique” column presents the total number
of unique entities across buyers and targets. The “unique
buyers” (“unique targets”) column presents the total number of unique buyers (targets).
The database allows us to identify 23,451 unique entities
that appear at least once in acquisitions as buyers or targets
over our sample period. Among industries, commercial banks
account for about 34 percent of the unique entities, followed
by insurance firms, thrifts, and specialty lenders. Of all these
entity types, banks are by far the most involved in buying: 45
percent of unique buyers are banks, and 37 percent of banks
act as buyers at least once in our sample. They are also the
largest industry represented as unique targets, although to a
smaller extent. Table 1 gives a flavor of the overall scope of the
Table 1
Unique Entities in Acquisitions Data Set
Total
Unique
Unique
Buyers
Unique
Targets
Bank
7,893
2,904
5,843
Asset manager
1,648
374
1,306
Broker-dealer
1,387
361
1,070
Financial technology
1,989
426
1,621
Insurance broker
3,682
504
3,237
Insurance underwriter
2,193
793
1,514
64
40
27
Industry
Investment company
Real estate
Savings bank/thrift/mutual
Specialty lender
Total
229
87
150
2,352
676
1,927
2,014
342
1,707
23,451
6,507
18,402
Source: Authors’ calculations, based on information in the SNL Financial
Mergers and Acquisitions database.
database and the related dynamics in acquisitions. However,
it cannot offer direct insights into the process of horizontal
organizational expansion; in referring to buyers and targets,
the database does not indicate whether the underlying participants were from the same or from different industries.
Table 2 takes a different look at the same acquisition activity.
It illustrates the extent to which each industry consolidates
(same-type entity deals) or expands (different-type entity
deals). Panel A displays the total number of acquisitions; panel
B displays the total real value of acquisitions. We organize
each panel as a two-way matrix. The rows show the industry
of the buyer, while the columns show the industry of the
target. Hence, the on-diagonal numbers represent same-industry
consolidation, while the off-diagonal numbers represent
cross-industry expansion.
We capture 19,532 acquisition events in our data set. As
indicated by the total number of on-diagonal events (13,070),
the financial sector overall experiences a substantial amount
of same-industry consolidation. Banks account for almost
half of these transactions. Likewise, banks also capture the
lion’s share of off-diagonal acquisition activity; their 3,742
acquisitions constitute about 60 percent of the 6,462 total
off-diagonal acquisitions. For some industries, banks
outperform same-industry entities in number of acquisitions.
For example, banks acquire 519 asset managers, while asset-manager
entities acquire only 459 other asset managers. Regardless of
the target industry, the proportion of acquisitions by banks is
high. For instance, banks are buyers in about 40 percent of all
asset-manager acquisitions, 26 percent of all broker-dealer acquisitions, and 37 percent of all specialty-lender acquisitions.
This summary table suggests the significance of how much
bank organizational structure has transformed over time. It also
FRBNY Economic Policy Review / December 2014
91
Table 2
Entity Industries in Consolidation and Expansion
Panel A: Types in Acquisitions, by Number
Target Industry
Buyer Industry
Bank
Savings
Bank/Thrift/ Asset
BrokerMutual
Manager Dealer
Bank
Financial Insurance Insurance Investment
Technology
Broker Underwriter Company
Real
Estate
Specialty
Lender
Total
6,076
1,305
519
292
164
759
38
3
1
653
9,810
Savings bank/
thrift/mutual
359
705
45
28
8
115
21
-
2
138
1,421
Asset manager
2
1
459
38
110
27
24
6
17
51
735
Broker-dealer
6
6
127
613
78
59
9
4
9
42
953
Financial technology
2
-
13
23
1,123
60
8
-
-
13
1,242
Insurance broker
4
1
31
12
35
1,762
18
-
-
6
1,869
14
18
138
55
126
533
1,451
-
4
54
2,393
2
1
19
4
4
4
2
11
4
42
93
130
Insurance underwriter
Investment company
Real estate
Specialty lender
Total
1
1
3
3
-
-
1
-
111
10
19
21
10
26
20
11
5
3
2
769
886
6,485
2,059
1,364
1,094
1,668
3,330
1,577
27
150
1,778
19,532
Real
Estate
Specialty
Lender
Total
Panel B: Types in Acquisitions, by Value (Millions of U.S. Dollars)
Target Industry
Buyer Industry
Bank
Savings bank/
thrift/mutual
Asset manager
Savings
Bank/Thrift/ Asset
BrokerMutual
Manager Dealer
Bank
Financial Insurance Insurance Investment
Technology
Broker Underwriter Company
1,405,983
203,243
43,512
173,952
18,083
3,297
16,783
1,127
333
276,048
2,142,361
18,982
54,333
3,359
119
74
165
3,409
-
86
15,165
95,691
0
17
68,463
7,812
46,776
2,575
1,692
416
70,405
29,347
227,504
6,099
2,665
19,461
106,443
4,302
1,467
970
1,921
15,183
9,463
167,975
Financial technology
25
-
3,813
1,784
91,225
437
1,284
-
-
733
99,301
Insurance broker
10
11
41
41
5,346
21,359
244
-
-
1
27,054
-
2,284
22,354
740,825
2,657
4,669
4,120
12,276
136,014
93
136,921
Broker-dealer
Insurance underwriter
124,460
785
28,783
15,605
10,929
8,032
527,592
Investment company
0
19
654
18
6
129
5
Real estate
0
78
599
3
--
--
133
110
848
1,904
2,006
1,884
62
1,824
393
416
73,561
83,008
1,555,669
261,999
170,590
307,784
178,625
37,524
553,935
6,514
229,390
430,885
3,732,916
Specialty lender
Total
Source: Authors’ calculations, based on data from SNL Financial.
92
Evolution in Bank Complexity
-
hints at how the structure has changed with respect to entities in
separate but related industries. Our conclusions are even more
striking if we restrict our attention to the dollar value of these
transactions (Table 2, panel B). Indeed, off-diagonal acquisitions
performed by banks are more than 80 percent of the total value
of all off-diagonal acquisitions.
Who are the top buyers over the period? How much are they
buying? Tables 3 and 4 show the top fifty buyers by number and
value of acquisitions, r espectively. The top entities by number
of acquisitions are three of the now largest insurance brokers:
Arthur J. Gallagher, Brown & Brown, and Hub International.
As Table 3 shows, they acquired hundreds of entities, although
almost exclusively consolidating within their own industry.
Banks follow in the ranking, also displaying very large numbers
of acquisitions but with a more balanced distribution between
bank and nonbank targets. Many of the banks at the lower end
of the list fell in the mass of acquisition activity after geographic
deregulation. This consolidation may have set the stage for future
expansion, as banks developed the scale and size necessary for
later expansions in complexity.
Interestingly, banks dominate the ranking by value. Table 4
captures the most active firms over time, irrespective of when the
activity took place and whether the entities are still in operation.
This time-independence is the reason NationsBank is second
on the list, despite its current incarnation as Bank of America.
The artifacts of bank acquisition activity show a compounding
and progressive industry buildup. For instance, although Bank
of America is highly diverse today, it inherited the results of the
earlier evolution of NationsBank and Merrill Lynch. Likewise,
Citigroup inherited part of its diversity from the previous activity
of Travelers Group. The same holds for Wells Fargo from Wachovia (originally First Union) and Norwest, and JPMorgan Chase
from Bank One, Chase Manhattan, and Washington Mutual.
It is important to note that the phenomenon of horizontal
expansion is not confined to a small handful of entities. As the
tables show, below the top-ranked acquirers, we see a significant
number of cross-industry acquisitions.
Next, we offer documentation on the dynamics of acquisitions. Chart 2 shows the composition of industries in four-year
periods within our sample. Although the database shows mainly
banks (and thrifts) as buyers in the late 1980s, variation in buyer
type steadily increases over time. By the second half of the 1990s,
all industry types perform acquisitions. Likewise, the variety in
target types increases gradually over time, with nonbank targets
already representing the large majority in the second half
of the 1990s.
Chart 3 illustrates that the share of the dollar value of acquisitions reflects the gradual process of expansion in industry types, although the relative prevalence of each industry
by value differs somewhat from prevalence by number. For
instance, there is a relatively large number of insurance broker
entities that are either buyers or targets of acquisitions, but
they account for a much smaller share of the overall value.
Conversely, there are relatively fewer insurance underwriters
involved in acquisitions, but they account for a larger share.
Charts 4 and 5 combine the number of acquisitions within
and across industries. While the process of same-industry
consolidation is important in itself, for our purposes, we want
to keep our focus on organizations expanding into other
industries within the financial sector. To this end, it is useful
to report the breakdown of acquisition activity (for buyers
and targets), separating same-industry and cross-industry
deals. Chart 4 shows that same-industry consolidation is
quite diffusive across the various industries. Although banks
dominated the activity during the geographic deregulation of
the mid-1990s, there is sizable consolidation across the other
industries as well, continuing into the present.
Chart 5 confirms and reinforces the message of the previous ones, which is that during our sample period the entire
financial sector was reorganizing. Banks were buying nonbanks, but not to the exclusion of substantial cross-industry
acquisitions of other entity types. Moreover, targets were not
concentrated in any particular industry, suggesting that no
particular industry-specific factors drove the development.
Rather, it indicates a diffused transformation of the intermediation industry, with a progressive expansion of the organizational boundaries of intermediation firms.
5.2 Bank-Specific Dynamics
We shift our focus to banks themselves and follow their evolution. We start with a specific examination using the same deal
data as above. Later in the paper, we present details of bank
evolution at the family (or BHC) level.
Chart 6 goes into the specifics of the cross-industry evolution in bank organizational structure. Besides the extensive
acquisition of thrifts in the early part of the period, the data
denote how banks gradually expanded their footprint. Banks
proceeded first by acquiring entities that were arguably closer
to their traditional mode of operations—specialty lenders and
asset managers, both specialized intermediaries that increased
their roles once securitization-based intermediation became
more prevalent. The expansion progressed naturally, with
banks incorporating brokers and dealers later in the sample
period. These entities rose in importance with the trading of a
progressively increasing stockpile of securities created through
asset securitization (Cetorelli and Peristiani 2012). Moreover,
the process continued with the incorporation of insurance and
financial technology firms, which offer payment-related services.
FRBNY Economic Policy Review / December 2014
93
Table 3
Top Fifty Buyers, by Number
Value (Millions of U.S. Dollars)
Rank
Name
Industry
All
Consolidation
Count
Expansion
All
Consolidation
Expansion
1
Arthur J. Gallagher & Co.
Insurance broker
3,314
3,249
65
249
245
4
2
Brown & Brown
Insurance broker
2,029
2,011
18
236
234
2
3
Hub International
Insurance broker
834
832
2
159
156
3
4
BB&T
Bank
19,989
15,291
4,697
142
23
119
5
Wells Fargo
Bank
50,566
48,577
1,989
138
34
104
6
Norwest
Bank
64,191
55,112
9,079
123
86
37
7
National Financial Partners
Corporation
Insurance broker
739
731
8
95
62
33
8
Bank of New York
Bank
29,062
22,661
6,401
76
4
72
9
Regions Financial Corporation
Bank
27,951
26,154
1,797
74
50
24
10
Union Planters
Bank
9,564
7,672
1,893
69
53
16
11
First American Corporation
Insurance underwriter
5,738
171
5,566
66
4
62
12
U.S. Bancorp
Bank
12,146
5,151
6,995
64
17
47
13
First Union
Bank
72,837
61,532
11,305
64
29
35
14
Stewart Information Services
Insurance underwriter
40
40
0
63
4
59
15
Goldman Sachs
Broker-dealer
13,725
10,020
3,705
60
10
50
16
SouthTrust
Bank
2,450
1,539
910
60
46
14
17
Marsh & McLennan Companies
Insurance broker
6,757
6,635
122
58
49
9
18
Compass Bancshares
Bank
2,524
2,375
149
55
41
14
19
Bank One Corporation
Bank
70,781
56,069
14,712
55
36
19
20
Citigroup
Bank
100,742
2,530
98,212
54
2
52
21
Community First Bankshares
Bank
1,004
983
21
53
26
27
22
Hibernia Corporation
Bank
2,006
1,678
327
51
40
11
23
First American Corporation
Insurance underwriter
178
175
3
50
3
47
24
PNC Financial Services
Bank
34,106
28,577
5,529
47
17
30
25
KeyBank
Bank
12,518
9,648
2,870
46
20
26
94
Evolution in Bank Complexity
Table 3 (continued)
Top Fifty Buyers, by Number
Value (Millions of U.S. Dollars)
Rank
Name
Industry
All
Consolidation
Count
Expansion
All
Consolidation
Expansion
26
USI Holdings Corporation
Insurance broker
546
527
19
45
43
2
27
Wachovia
Bank
67,562
23,837
43,726
45
11
34
28
Zions Bancorporation
Bank
5,591
5,463
129
45
35
10
29
First Banks
Bank
1,141
801
340
43
31
12
30
American International Group
Insurance underwriter
59,147
58,330
817
42
22
20
31
Colonial Bancgroup
Bank
2,970
2,348
622
42
31
11
32
SunGard
Financial technology
1,942
1,795
148
42
38
4
33
Fifth Third Bank
Bank
18,416
14,189
4,227
41
18
23
34
Synovus
Bank
2,503
1,994
509
41
29
12
35
Old National Bank
Bank
1,641
1,319
322
39
24
15
36
Aon plc
Insurance broker
8,359
3,297
5,063
39
31
8
37
JPMorgan Chase
Bank
85,253
75,001
10,251
38
2
36
38
Marshall & Ilsley
Bank
8,380
4,661
3,720
38
17
21
39
HCC Insurance Holdings
Insurance underwriter
1,339
811
528
37
10
27
40
Comerica
Bank
6,033
5,947
87
36
27
9
41
Fidelity National Financial
Insurance underwriter
6,857
2,145
4,712
36
8
28
42
FNB Corporation
Bank
2,135
1,883
252
36
17
19
43
Fiserv
Financial technology
6,533
5,992
541
35
28
7
44
Mercantile Bancorporation
Bank
7,078
4,910
2,169
35
23
12
45
National City Corporation
Bank
26,288
20,778
5,509
34
11
23
46
Hilb, Rogal & Hobbs Company
Insurance broker
380
380
0
34
33
1
47
LandAmerica Financial Group
Insurance underwriter
1,172
971
201
33
2
31
48
Commerce Bancshares
Bank
49
Willis Group
Insurance broker
50
Royal Bank of Canada
Bank
990
924
67
33
30
3
1,920
1,888
32
33
32
1
12,409
5,530
6,879
33
4
29
Source: Authors’ calculations, based on data from SNL Financial.
Notes: Consolidation captures acquisitions in which the buyer and target have the same type. Expansion captures acquisitions in which the buyer and target
have different types.
FRBNY Economic Policy Review / December 2014
95
Table 4
Top Fifty Buyers, by Value
Value (Millions of U.S. Dollars)
Rank
Name
Industry
Count
All
Consolidation
Expansion
All
Consolidation
Expansion
100,364
16
3
13
11
1
Bank of America
Bank
187,572
87,208
2
NationsBank
Bank
138,702
135,166
3,535
23
12
3
Travelers Group
Insurance underwriter
137,466
5,892
131,573
8
1
7
4
Citigroup
Bank
100,742
2,530
98,212
54
2
52
5
JPMorgan Chase
Bank
85,253
75,001
10,251
38
2
36
6
First Union
Bank
72,837
61,532
11,305
64
29
35
7
Bank One Corporation
Bank
70,781
56,069
14,712
55
36
19
8
Wachovia
Bank
67,562
23,837
43,726
45
11
34
9
Capital One
Bank
66,804
22,434
44,370
12
2
10
10
Norwest
Bank
64,191
55,112
9,079
123
86
37
11
Blackstone Group
Asset manager
61,048
1,271
59,776
19
4
15
12
American International Group
Insurance underwriter
59,147
58,330
817
42
22
20
13
Chase Manhattan
Bank
58,120
45,275
12,845
26
4
22
14
Wells Fargo
Bank
50,566
48,577
1,989
138
34
104
15
Washington Mutual
Bank
50,347
320
50,027
27
4
23
16
Firstar Corporation
Bank
44,430
43,827
602
21
15
6
17
Fleet Financial Group
Bank
43,867
37,165
6,702
26
15
11
18
Berkshire Hathaway
Insurance underwriter
35,792
35,029
763
24
19
5
19
PNC Financial Services
Bank
34,106
28,577
5,529
47
17
30
20
HSBC
Bank
32,703
11,053
21,650
10
2
8
21
MetLife
Insurance underwriter
32,523
31,912
612
17
8
9
22
Toronto-Dominion Bank
Bank
29,866
14,567
15,299
21
5
16
23
Bank of New York
Bank
29,062
22,661
6,401
76
4
72
24
Kohlberg Kravis Roberts
Asset manager
29,002
0
29,002
6
0
6
25
Regions Financial Corporation
Bank
27,951
26,154
1,797
74
50
24
96
Evolution in Bank Complexity
Table 4 (continued)
Top Fifty Buyers, by Value
Value (Millions of U.S. Dollars)
Rank
Name
Industry
All
Consolidation
Count
Expansion
All
Consolidation
Expansion
26
BlackRock
Asset manager
26,847
26,847
0
9
7
27
Anthem Incorporated
Insurance underwriter
26,360
26,360
0
2
2
2
0
28
National City Corporation
Bank
26,288
20,778
5,509
34
11
23
29
St. Paul Companies
Insurance underwriter
25,074
24,063
1,012
12
7
5
30
SunTrust Banks
Bank
24,070
23,019
1,051
32
13
19
31
Chemical Bank
Bank
23,610
23,610
0
13
11
2
32
ING Group
Insurance underwriter
23,270
16,628
6,642
20
4
16
33
UBS
Bank
22,775
0
22,775
17
0
17
34
Morgan Stanley
Broker-dealer
21,216
0
21,216
21
1
20
35
Credit Suisse
Bank
20,110
0
20,110
13
0
13
36
BB&T
Bank
19,989
15,291
4,697
142
23
119
37
UnitedHealth Group
Insurance underwriter
18,476
17,897
579
23
16
7
38
Fifth Third Bank
Bank
18,416
14,189
4,227
41
18
23
39
Deutsche Bank
Bank
18,398
13,055
5,342
13
1
12
40
Aegon
Insurance underwriter
18,274
17,923
352
10
7
3
41
First Bank System
Bank
17,646
16,123
1,523
22
14
8
42
Swiss Re
Insurance underwriter
17,108
16,967
140
16
14
2
43
Merrill Lynch
Broker-dealer
16,182
4,761
11,422
25
17
8
44
Conseco
Insurance underwriter
15,583
4,253
11,331
16
7
9
45
Banco Bilbao Vizcaya Argentaria
Bank
15,499
15,499
0
9
7
2
46
Dean Witter Discover
Broker-dealer
15,390
15,390
0
2
2
0
47
Household International
Specialty lender
14,610
14,421
189
13
6
7
48
Monte dei Paschi di Siena
Bank
13,898
13,898
0
1
1
0
49
Equity Office
Real estate
13,813
13,813
0
3
3
0
50
Goldman Sachs
Broker-dealer
13,725
10,020
3,705
60
10
50
Source: Authors' calculations, based on data from SNL Financial.
Notes: Consolidation captures acquisitions in which the buyer and target have the same type. Expansion captures acquisitions in which the buyer and target
have different types.
FRBNY Economic Policy Review / December 2014
97
Chart 2
Types in All Acquisitions, by Number
Buyers
Targets
1989-92
1993-96
1997-2000
1989-92
1993-96
1997-2000
2001-04
2005-08
2009-12
2001-04
2005-08
2009-12
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
Source: Authors’ calculations, based on data from SNL Financial.
Chart 3
Types in All Acquisitions, by Value
Buyers
Targets
1989-92
1993-96
1997-2000
1989-92
1993-96
1997-2000
2001-04
2005-08
2009-12
2001-04
2005-08
2009-12
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
Source: Authors’ calculations, based on data from SNL Financial.
98
Evolution in Bank Complexity
Chart 4
Types in Same-Industry Acquisitions, by Number
1989-92
1993-96
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
1997-2000
2001-04
2005-08
2009-12
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
Source: Authors’ calculations, based on data from SNL Financial.
Note: Same-industry acquisitions represent deals in which the buyer and target have the same type.
Chart 5
Types in Cross-Industry Acquisitions, by Number
Buyers
Targets
1989-92
1993-96
1997-2000
1989-92
1993-96
1997-2000
2001-04
2005-08
2009-12
2001-04
2005-08
2009-12
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
Source: Authors’ calculations, based on data from SNL Financial.
Note: Cross-industry acquisitions represent deals in which the buyer and target have different types.
FRBNY Economic Policy Review / December 2014
99
Chart 7
Chart 6
Nonbank Targets of Bank Buyers (Number)
Nonbank Targets of Bank Buyers (Share)
100
Percent
250
80
200
60
150
40
100
Number
50
20
0
0
1990
1995
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
2000
2005
2010
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
1990
1995
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
2000
2005
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
Source: Authors’ calculations, based on data from SNL Financial.
Source: Authors’ calculations, based on data from SNL Financial.
Note: Vertical cross-sections illustrate the average share of targets
by type in a given quarter.
Note: Vertical cross-sections illustrate the average number of targets
by type in a given quarter.
Chart 7 instead displays the number, not the share, of
acquisition types through time. It shows that the process of
expansion remained active throughout the period, perhaps
slowing down only in the post-crisis years.
5.3 Evolution in Bank Families, or
Organizational Changes in BHCs
The entity-level analysis in the previous subsection already
hints at the evolution in complexity of U.S. banking firms.
However, maintaining the focus on individual entities actually
understates the extent to which bank organizational boundaries really expanded. Entity-level analysis misses the process
of merging, changes in names, and branching into multiple
levels of affiliation. As a result, entity, rather than family,
analysis leaves us blind to the actual size and composition
of entity families. For example, in Table 4, Bank of America
and NationsBank are the first- and second-highest ranked
entities by acquisition value. However, these entities are truly
the same; most of NationsBank’s history folded into Bank of
America upon creation. Within this new entity are many enti-
100
Evolution in Bank Complexity
2010
ties acquired along the way, perhaps representing a diversified
portfolio or a focused industry giant. To track complexity
accurately through time, we need a picture of the same entity’s
organization before and after the deal.
As explained in section 4, our methodology allows us to
combine and track overall complexity, as captured by the
amount and type of performed acquisitions (and sales). This
buildup takes place within the walls of a banking family,
defined by aggregating the information of individual entities
under a common highest-holder identifier.
What does the typical BHC family look like? How does its
structure evolve over time? Chart 8 addresses these questions
by depicting the evolution of organizational profiles in our
sample. The typical BHC changed appreciably over time. A
BHC family was identified by having mostly commercial
bank and thrift subsidiaries in the early 1990s. However, the
organizational boundaries expanded significantly starting in
the mid-1990s, as BHCs began adding an increasing number
of nonbank subsidiaries.
The process that we are able to pick up through the data
on acquisition matches well the data on total assets of BHCs,
depicted earlier in Chart 1, which shows the increasing
Chart 9
Chart 8
Nonbanks and Securitization
Organizational Evolution
100
Percentage share
60
80
50
60
40
Percent
Billions of U.S. dollars
12,000
10,000
Nonbanks in family
organization profiles
8,000
6,000
40
30
20
Asset securitization
outstanding
20
0
1990:Q1
1995:Q1
Bank
Asset manager
Financial technology
Insurance underwriter
Real estate
2000:Q1
2005:Q1
2010:Q1
Savings bank/thrift/mutual
Broker-dealer
Insurance broker
Investment company
Specialty lender
10
1990:Q1
4,000
2,000
1995:Q1
2000:Q1
2005:Q1
0
2010:Q1
Sources: Authors’ calculations, based on data from SNL Financial;
Securities Industry and Financial Markets Association.
Notes: The black line illustrates the share (by count) of nonbanks in family
organizational profiles. The green line illustrates asset securitization
outstanding in billions of U.S. dollars.
Source: Authors’ calculations, based on data from SNL Financial.
Note: Vertical cross-sections illustrate the average share of types
within a bank family in a given quarter.
contribution of nonbank subsidiaries to the total assets of
their organizations. This evolution in BHCs’ organizational
footprint also coincides closely with the concurrent evolution
in asset-securitization activity. Chart 9 shows the time series
of the ratio of nonbank subsidiaries to total subsidiaries of all
the BHCs in our sample, together with the time series of total
asset securitization outstanding. As the chart suggests, the
organizational expansion of BHCs tracks quite closely the rise
in securitization activity observed from the mid-1990s up to
the financial crisis.
Table 5 shows snapshots of family complexity taken in a
given year, capturing the number of both bank and nonbank
entities amassed through the acquisition channel by the top
fifty BHC families (ranked by total assets) up to that year.
BHCs in the early 1990s were relatively simple in organizational structure. Among the top ten in 1990, only BankAmerica
Corporation, back then a holding company headquartered
in San Francisco, California, had performed ten nonbank
acquisitions, and Security Pacific Corporation had performed
seven. Among the remaining top fifty, Bank One and Barnett had
performed five nonbank acquisitions each. Five years later, the
picture was already quite different. The number of acquisitions
was much higher, both within and across industries. Some
families from 1990 had disappeared from the subsequent list
as surviving ones absorbed them (BankAmerica, for instance,
acquired Security Pacific).
The BHC organizational profiles only increase in c omplexity
as time goes by, with very large numbers of entities wrapped
under common ownership and control. Moreover, the lists
show that the process takes place across institutions, and it is
not a phenomenon confined to just the largest entities.
Another way to capture the sector-wide transformation is
to look at time-series metrics of BHC structures. Chart 10, for
instance, displays the average number of commercial banks
acquired and kept within a family in a given year. This n
umber,
not surprisingly, steadily increases, again reflecting the process
of geographic deregulation and consequent consolidation.
The number of nonbank acquisitions in Chart 9
could still fail to show true expansion across industries. For
instance, BHCs could have performed many acquisitions
concentrated in just one nonbank industry. In order to capture
the extent of broad horizontal expansion, we calculate a
Herfindahl-Hirschman Index (HHI) of industrial concentration. This index is 1 if the BHC has only commercial banks
and smaller than 1 if the BHC acquires nonbank subsidiaries.
Furthermore, it progressively decreases as the acquisition
profile among the ten industries becomes more “diverse.” In
the same chart, we report the average HHI of BHC families
over time. The steady downward trend shows a push toward
broad expansion in organizational boundaries.
FRBNY Economic Policy Review / December 2014
101
Table 5
Top Fifty Families by Size and Time, 1990-2000
1990
Rank
Name
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
Citi
BankAmerica
Chase Manhattan
J. P. Morgan
Security Pacific Corporation
Chemical Banking
NCNB
Bankers Trust New York
Manufacturers Hanover
C&S/Sovran
First Interstate Bancorp
First Chicago
PNC Financial
Bank of New York Company
Banc One
First Union
SunTrust Banks
Bank of Boston
Fleet/Norstar Financial
Barnett Banks
Norwest
First Fidelity Bancorp
Mellon Bancorp
Continental Bank
NBD Bancorp
Society
National City
Shawmut National
CoreStates Financial
Midlantic
Bank of New England
Key
First Bank System
Boatmen’s Bancshares
First of America Bank
Comerica
UJB Financial
Manufacturers National
Meridian Bancorp
Crestar Financial Corporation
Huntington Bancshares
42
43
44
45
46
47
48
49
50
Northern Trust
State Street Boston
Signet Banking
Ameritrust
Michigan National
Bancorp Hawaii
Valley National
Dominion Bancshares
BayBanks
102
1995
Banks Nonbanks
6
3
1
1
10
18
5
1
1
1
7
3
4
3
16
15
1
2
4
4
10
2
2
3
1
2
2
3
2
4
1
8
8
2
5
13
2
5
2
2
4
1
10
0
0
7
2
0
0
0
0
0
0
0
0
5
1
0
1
1
5
2
1
0
0
0
0
1
3
0
0
0
1
3
1
3
4
0
1
1
2
1
1
1
1
1
4
3
1
1
2
1
0
0
1
1
1
0
0
0
Name
2000
Banks Nonbanks
Name
Citi
BankAmerica
NationsBank
J. P. Morgan
Chemical Banking
First Chicago NBD
Bankers Trust New York
First Union
Banc One
Fleet Financial Group
PNC Bancorp
Norwest
Key
First Interstate Bancorp
Bank of New York Company
National City
Bank of Boston
SunTrust Banks
Barnett Banks
Mellon Bancorp
Comerica
First Bank System
Boatmen's Bancshares
CoreStates Financial
State Street Boston
First of America Bank
SouthTrust
Southern National
Huntington Bancshares
Northern Trust
Firstar
Crestar Financial Corporation
AmSouth Bancorp
Fifth Third Bancorp
Mercantile Banc
UJB Financial
BanPonce
Meridian Bancorp
GreenPoint Financial
Integra Financial
Regions Financial
5
16
17
1
18
1
1
22
60
25
14
65
26
23
5
12
11
6
7
5
24
27
29
6
1
7
28
6
20
4
1
4
10
9
16
3
4
9
0
2
11
2
28
3
1
7
0
0
25
13
21
10
18
12
4
1
6
9
2
9
8
10
10
7
5
2
11
6
31
8
3
1
16
8
7
9
4
2
4
3
3
8
Citigroup
JPMorgan Chase
Bank of America
Wells Fargo
Bank One
First Union
FleetBoston Financial
SunTrust Banks
U. S. Bancorp
Key
Firstar
Bank of New York Company
PNC Financial Services Group
State Street
BB&T
Mellon Financial
Fifth Third Bancorp
SouthTrust
Regions Financial Corporation
Comerica
Summit Bancorp
AmSouth Bancorp
MBNA
Charles Schwab
Northern Trust
Union Planters Corporation
Charter One Financial
M&T Bank
Huntington Bancshares
Popular
Old Kent Financial
Zions Bancorp
Compass Bancshares
First Tennessee National
Banknorth Group
Hibernia
National Commerce
GreenPoint Financial
Provident Financial
North Fork Bancorp
Pacific Century Financial
MBNA
Bancorp Hawaii
First Security
First Tennessee National
BayBanks
Old Kent Financial
First Empire State
Union Planters Corporation
Signet Banking
0
3
14
12
5
4
3
33
3
1
2
4
10
2
3
4
7
2
Associated Banc-Corp
Colonial BancGroup
People’s Mutual Holdings
Centura Banks
TCF Financial Corporation
Commerce Bancshares
First Citizens Bancshares
FirstMerit
BOK Financial Corporation
Evolution in Bank Complexity
Banks Nonbanks
1
17
104
194
74
73
45
11
77
26
0
5
14
1
55
20
27
47
83
25
6
33
0
3
6
78
4
14
33
14
14
35
45
10
25
44
18
0
4
5
8
37
25
77
80
20
77
47
23
53
20
1
32
16
8
89
19
39
12
28
9
7
17
3
14
6
33
15
18
11
5
12
9
4
15
16
12
25
6
12
15
2
14
26
2
18
2
28
6
3
13
1
11
5
17
8
5
15
9
2
Table 5 (continued)
Top Fifty Families by Size and Time, 2005-10
2005
Bank
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
Name
Citigroup
Bank of America
JPMorgan Chase
Wachovia
Wells Fargo
MetLife
U.S. Bancorp
SunTrust Banks
Countrywide Financial Corporation
National City
BB&T
Fifth Third Bancorp
Bank of New York Company
State Street
Key
PNC Financial Services Group
Capital One Financial Corporation
Regions Financial Corporation
MBNA
North Fork Bancorp
Comerica
Northern Trust
AmSouth Bancorp
Popular
Charles Schwab
Zions Bancorp
Mellon Financial
Commerce Bancorp
First Horizon National
Huntington Bancshares
Compass Bancshares
Synovus Financial Corporation
New York Community
Associated Banc-Corp
Colonial BancGroup
First Bancorp
Webster Financial
Doral Financial
Mercantile Bancshares
BOK Financial Corporation
W Holding Company
Sky Financial Group
First Citizens
South Financial Group
Commerce Bancshares
TCF Financial Corporation
Valley NBC
Fulton Financial
Investors Financial
Cullen/Frost Bankers
2010
Banks
6
114
75
138
211
1
116
12
1
31
105
47
5
1
28
21
45
158
0
8
23
6
30
15
3
50
23
4
9
34
45
27
2
24
31
7
24
1
16
18
2
12
9
25
28
2
9
22
1
15
Nonbanks
59
113
65
117
119
9
83
34
4
54
161
53
59
15
27
26
22
70
6
17
10
9
75
9
19
17
36
14
28
13
14
14
7
7
13
6
30
1
10
2
1
17
16
15
6
10
12
11
3
8
Name
Bank of America
JPMorgan Chase
Citigroup
Wells Fargo
Goldman Sachs
Morgan Stanley
MetLife
U. S. Bancorp
PNC Financial Services Group
Bank of New York Mellon
Capital One Financial Corporation
SunTrust Banks
State Street
BB&T
American Express Company
Regions Financial Corporation
Fifth Third Bancorp
Key
Northern Trust
M&T Bank
Discover Financial
Comerica
Huntington Bancshares
CIT Group
Zions Bancorp
Marshall & Ilsley
New York Community
Popular
Synovus Financial Corporation
First Horizon National
BOK Financial Corporation
Associated Banc-Corp
First Niagara Financial
First Citizens Bancshares
East West Bancorp
TCF Financial Corporation
Webster Financial
Cullen/Frost Bankers
SVB Financial Group
Fulton Financial
First Bancorp
Valley National Bancorp
FirstMerit
Wintrust Financial Corporation
Susquehanna Bancshares
BankSouth
Bank of Hawaii
PrivateBancorp
UMB Financial Corporation
Franklin Resources
Banks
Nonbanks
117
81
5
305
0
0
1
126
69
6
54
25
2
112
0
191
55
30
6
27
0
23
51
0
55
32
5
18
29
9
20
25
8
14
10
2
21
19
2
26
9
16
4
10
16
29
8
5
19
0
166
97
108
244
89
25
22
96
117
98
41
41
26
190
12
163
69
31
11
34
3
10
34
21
17
34
12
10
15
29
2
7
34
16
3
11
32
12
4
13
8
14
11
10
27
18
2
2
14
13
Source: Authors’ calculations, based on data from SNL Financial; Federal Reserve System, Form FR Y-9C, Schedule HC.
FRBNY Economic Policy Review / December 2014
103
Chart 10
Concentration and Diversification
Number
HHI
1.0
12
10
0.9
8
0.8
6
4
0.4
2
0
1990:Q1
1995:Q1
2000:Q1
0.6
2010:Q1
2005:Q1
Source: Authors’ calculations, based on data from SNL Financial.
Notes: The black line illustrates the average number of banks acquired
and kept within a family in a given year. The green line represents the
average Herfindahl-Hirschman Index (HHI) calculated across the ten
10
ni 2
types of bank families. For each family, HHI is defined as
N
(i=1)
where ni represents the count of subsidiaries of type i and N represents
the total count of subsidiaries.
Σ
( )
6. Conclusion
Three key observations can summarize the evolution in the
structure of financial firms. First, bank holding companies
have become less bank-centric by expanding the types of their
subsidiaries. Second, this phenomenon was very widespread,
as financial firms other than bank holding companies also
expanded their scope. Finally, bank holding companies
expanded by adding more banks to their firms in the earlyand mid-1990s. As we noted earlier, there are several hypotheses
that might be consistent with those observations. First, it
seems that the geographic deregulation of banking in the
United States led to significant changes in the structure of
banking markets (while not covered in our paper, this
phenomenon has been studied extensively) and bank holding
companies. This expansion and consolidation positioned bank
holding companies to take advantage of later regulatory
changes to increase their complexity. Second, and along these
lines, GLB may have also allowed bank holding companies to
expand into activities from which they were previously
excluded, such as brokering and dealing.
While deregulation or firms’ attempts to evade existing
regulation may have allowed firms to evolve in the ways we
describe, these rationales unlikely explain fully the evolution.
The acquisitions we see in the data are among firms still in the
104
Evolution in Bank Complexity
regulated sector, and many of these firms organize themselves
as bank holding companies, which the Federal Reserve
supervises at the consolidated level.
Instead, some other changes in financial intermediation
seem to be required to explain such widespread and profound
shifts in the industry. Here again, there are several possible
candidates. For instance, it may be that the more geographically expansive nature of business enterprises gave rise to an
increased demand for cross-border banking, both within the
United States and overseas. That could have provided an
impetus for the early wave of bank acquisition we see in our
sample. An alternative hypothesis is that specialized firms,
whose contributions to finance are to add value along a chain
of financial engineering that operates externally to any
particular firm, are now more efficient than generalist firms,
which build an integrated value chain internally.
This hypothesis could be supplemented to account for the
acquisitions of specialist firms by increasingly large BHC
conglomerates. For example, information and credit frictions
may be more difficult to overcome for isolated specialist firms
but more manageable with help from internal capital markets
in larger firms. Our results are consistent with this move
toward a model of finance more oriented toward securitization. The hypothesis itself may be dependent on the long-term
and ongoing revolutions in information technology and
communications that have allowed more quantification of
financial information and have improved the ability to
communicate and manage that information. In that sense,
for banking firms to stay viable in a changing industry,
complexity is a necessary adaptation.
The changes documented in this paper refine our understanding of bank complexity across a number of dimensions.
First, they highlight the expanded scope and complexity of
individual firms. Second, they suggest that the industrial
organization of finance is changing profoundly: Market
interactions among more numerous and more specialized
firms have displaced the earlier organization of generalized
firms, which engaged in most stages of finance by using
internal resources. Third, bank holding companies have
become increasingly less bank-centric, increasing the importance of consolidated supervision by the cooperative effort of
a larger set of functional regulators. Given these findings,
design of informed regulation of complex banking organizations presents a key challenge going forward.
The financial crisis of 2007-09 raises concerns about the
very existence of supersized institutions. Why does society
need incredibly large and complex banking institutions when
they are a potential cause of systemic disruption? Possible
“subsidies” from explicit or perceived government guarantees
may distort incentives in failure resolution. Size and complexity
may also lead to complicated and ineffective monitoring, such
as duplication of rules or regulation that is too strict (or
too weak).
Our documentation of the evolving structure of banks
offers potential insights for the evaluation of policy solutions
to these bank complexity problems. For instance, blunt fixes
such as reinstating GLB might artificially impose breakups,
fragmenting the intermediation industry and trading large
and complex holding companies for shadow entities outside
the scope of oversight. If complex conglomerate structures
are the result of an adaptation to technological and financial
advances, then tractable policies such as enhanced capital
requirements, effective resolution plans, and stress tests may
reduce systemic risk while retaining intermediation synergies,
such as reducing informational frictions across links in the
intermediation chain.17
17
For a discussion of this policy trade-off, see Federal Reserve Bank of New
York President Bill Dudley’s speech: “Global Financial Stability—The Road
Ahead,” February 26, 2014. Available at http://www.bis.org/review/r140226b.htm.
FRBNY Economic Policy Review / December 2014
105
References
Adrian, T., A. B. Ashcraft, and N. Cetorelli. 2013. “Shadow Bank
Monitoring.” Federal Reserve Bank of New York Staff Reports,
no. 638, September.
Ashcraft, A. B., and T. Schuermann. 2008. “Understanding the
Securitization of Subprime Mortgage Credit.” Foundations and
Trends in Finance 2, no. 3 (July): 191-309.
Avraham, D., P. Selvaggi, and J. Vickery. 2012. “A Structural View of
U.S. Bank Holding Companies.” Federal Reserve Bank of New
York Economic Policy Review 18, no. 2 (July): 65-82.
Carpenter, D., and M. Murphy. 2010. “Permissible Securities
Activities of Commercial Banks under the Glass-Steagall Act
(GSA) and the Gramm-Leach-Bliley Act (GLBA).” CRS Reports
R41181. Washington, D.C.: Library of Congress, Congressional
Research Service, April 12.
Cetorelli, N. 2012. “A Principle for Forward-Looking Monitoring
of Financial Intermediation: Follow the Banks!” Federal
Reserve Bank of New York Liberty Street Economics blog,
July 23. Available at: http://libertystreeteconomics.newyorkfed.
org/2012/07/a-principle-for-forward-looking-monitoring-offinancial-intermediation-follow-the-banks.html.
Cetorelli, N., B. H. Mandel, and L. Mollineaux. 2012. “The Evolution
of Banks and Financial Intermediation: Framing the Analysis.”
Federal Reserve Bank of New York Economic Policy Review
18, no. 2 (July): 1-12.
Cetorelli, N., and S. Peristiani. 2012. “The Role of Banks in Asset
Securitization.” Federal Reserve Bank of New York Economic
Policy Review 18, no. 2 (July): 47-63.
Federal Reserve Bank of New York. 1986. “Recent Trends in
Commercial Bank Profitability: A Staff Study.”
Fein, M. 2004. Securities Activities of Banks. New York: Aspen
Publishers.
Isaac, W. M., and M. L. Fein. 1988. “Facing the Future—Life without
Glass-Steagall.” Catholic University Law Review 37, no. 281
(Winter).
Kirk, A., J. McAndrews, P. Sastry, and P. Weed. 2014. “Matching
Collateral Supply and Financing Demands in Dealer Banks.”
Federal Reserve Bank of New York Economic Policy Review
20, no. 2 (December): 127-51.
Omarova, S., and M. Tahyar. 2011. “That Which We Call a Bank:
Revisiting the History of Bank Holding Company Regulation in
the United States.” Review of Banking and Financial Law 31:
113-203.
Pavel, C. 1986. “Securitization.” Federal Reserve Bank of Chicago
Economic Perspectives 10, no. 4 (July): 16-31.
Pozsar, Z., T. Adrian, A. Ashcraft, and H. Boesky. 2010. “Shadow
Banking.” Federal Reserve Bank of New York Staff Reports,
no. 458, last revised February 2012.
Rajan, R. G., and L. Zingales. 2001. “The Influence of the Financial
Revolution on the Nature of Firms.” American Economic
Review 91, no. 2 (May): 206-11.
Stein, J. C. 2002. “Information Production and Capital Allocation:
Decentralized versus Hierarchical Firms.” The Journal of
Finance 57, no. 5 (October): 1891-1922.
Coase, R. H. 1937. “The Nature of the Firm.” Economica 4, no. 4
(November): 386-405.
Copeland, A. 2012. “Evolution and Heterogeneity among Larger
Bank Holding Companies: 1994 to 2010.” Federal Reserve Bank
of New York Economic Policy Review 18, no. 2 (July): 83-93.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
106
Evolution in Bank Complexity
Nicola Cetorelli and Linda S. Goldberg
Measures of Global Bank
Complexity
l
Although the complexity of global banking
institutions is generally thought to contribute
to the risk of systemic disruptions, no single
accepted metric for complexity exists.
l
To address this gap, this study introduces two
broad measures: Organizational complexity
captures the number and geographic spread
of an institution's affiliates, as well as the
levels of ownership linking affiliates; business
complexity captures the range of activities
conducted within an institution's walls.
l
Using these measures, the authors assess the
complexity of a sample of 170 global banking
organizations. They find that complexity
cannot be equated with institution size;
although affiliate counts are correlated with
size, no close relationship exists with other
complexity measures.
l
In addition, the authors conclude that the
institutions differ greatly in the number of their
affiliates, the complexity of their ownership
trees, and the degree of diversification in their
business activities.
Nicola Cetorelli is an assistant vice president and Linda S. Goldberg
a vice president at the Federal Reserve Bank of New York.
nicola.cetorelli @ny.frb.org; linda.goldberg@ny.frb.org
1. Introduction
The increasing size and complexity of financial institutions has
received renewed attention in recent years—prompted in part
by the debate over the issue of too-big-to-fail entities. How the
size of failing institutions might contribute to systemic disruption is well understood. Complexity, however, is a thornier, less
easily defined concept, although it is a natural subject of policy
concern given the systemic implications of resolving failing
institutions. Resolvability requires successfully executing an
orderly liquidation in the event of an organization’s distress and
default; in the case of complex institutions—many with global
reach—such liquidations may be more difficult because a large
number of legal entities or legal systems are involved.
Concerns over the potential systemic repercussions of disruptions to complex organizations have inspired a number of
ideas for preemptive “fixes,” including capping of size, breakup
and separation of the institution along business lines, organizational restructuring to limit the cross-border dimension of
complexity (this last remedy captured in a proposed Federal
Reserve rule to strengthen the oversight of U.S. operations of
foreign banks),1 and efforts to make organizations more robust, including the already-implemented enhanced capital and
liquidity requirements for systemically important financial
1
For details, see http://www.federalreserve.gov/newsevents/press/
bcreg/20121214a.htm.
The authors gratefully acknowledge the excellent data work of Arun Gupta,
Meru Bhanot, Samuel Stern, and Rose Wang, as well as input from Philip
Strahan and from colleagues at the Federal Reserve Bank of New York who
participated in a broader initiative on understanding size and complexity
in financial institutions. The views expressed in this article are those of the
authors and do not necessarily reflect the position of the Federal Reserve
Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
107
institutions. Other approaches to resolution include the
FDIC’s Title II Orderly Liquidation Authority approach under
the Dodd-Frank Act, whereby financial organizations operating in the United States would do so with a “single entry”
strategy intended to reduce system spillovers from resolution
as well as the fiscal consequences of such events.2
In the context of these initiatives, we note that there is no
single accepted metric for complexity and that analysis of this
issue across broad groups of financial firms is relatively scarce.
It is well known that banks have developed broader networks
of affiliated banking and nonbanking entities at home and
abroad. Herring and Santomero (1990) were among the first
to predict such an expansion of financial conglomerates, arguing that it would arise from synergies in the production of financial services and in the consumption of financial services.3
Twenty years later, Herring and Carmassi (2010) documented
how far this trend toward consolidation and conglomeration
in financial services had progressed, observing that, by the
middle of this century’s first decade, large complex financial
institutions had hundreds or thousands of subsidiaries.4 At
least half a dozen top U.S. bank holding companies (BHCs)
had more than a thousand subsidiaries in 2012, in contrast to
a single firm with such numbers in 1990, as shown in Chart 1
(Avraham, Selvaggi, and Vickery 2012). The organizational
evolution of U.S. BHCs followed an intense process of industry consolidation and substantial acquisitions of nonbank
subsidiaries (Cetorelli, McAndrews, and Traina 2014). On
the international side, the extent of banking’s globalization
through the establishment of affiliates in other parts of the
world has been documented in numerous studies, including
a recent broad overview by Claessens and van Horen (2013).
These studies have been revealing, but the complexity of these
organizations has not been documented comprehensively.
Despite the centrality of the bank complexity issue, no
shared consensus has emerged just yet on what complexity
might mean in the context of banking, or at least what might
be the agreed-upon dimensions of our analysis of complexity.
Concentrating on global banks adds many layers to considerations of complexity, so a focus on global banking organizations is bound to yield a more exhaustive take on the issue
than an examination of purely domestic banking entities.
2
See http://www.fdic.gov/about/srac/2012/2012-12-10_title-ii_orderly
-liquidation-authority.pdf.
3
Herring and Santomero (1990) were also prescient in anticipating some of
the policy concerns that would arise from the growth of institutional size and
complexity.
4
Herring and Carmassi (2010) discuss some potential consequences, but
primarily argue that complexity increases systemic risk, worsens information
and incentive problems within organizations, and impedes timely regulatory
intervention and disposition of financial firms.
108
Measures of Global Bank Complexity
Chart 1
Number of Subsidiaries in U.S. Top Fifty
Bank Holding Companies
3,500
2012
1990
2012
1990
3,000
2,500
2,000
1,500
1,000
500
0
1
2
3
4
5
6
7 10 20 30 40 50
Rank of U.S. bank holding company by total assets
Source: Avraham, Selvaggi, and Vickery (2012).
Accordingly, we turn our attention to financial institutions
from around the world that have operations within the United
States and financial institutions from the United States that
have branches or subsidiaries abroad.
We adopt two broad measurement concepts. We introduce
“organizational” complexity metrics to indicate the degree
to which the organization is structured through separate
affiliated entities. Organizational complexity also encompasses a related dimension specific to global entities—namely,
geographic complexity, as captured by the span of the organization’s affiliates across different regions or countries. In addition, we introduce “business” complexity, a concept referring
to the type and variety of activities that may be conducted
within the walls of a given institution. Organizational measures have a more direct fit with the main concerns typically
associated with complexity, such as resolution, fragmentation, cross-border systemic risk, internal liquidity dynamics,
managerial agency frictions, and “too big to fail.” Business
complexity concepts may speak more to the diversification
and fragmentation of the type of production undertaken by
organizations. Neither metric adequately captures the systemic nature of the distress resulting from potential failures;
for this, the metric would need to incorporate insights on the
criticality of the functions performed in the organization.
Since our focus is on global banking organizations, we
pay careful attention to the fact that these are structured to
encompass affiliates worldwide. The number of affiliates can be
relatively few or in the thousands. This pattern of complexity
reflects the broader growth in global banking over recent
decades, as international financial markets in general have
grown more interconnected. Foreign banks now represent
over a third of the banks in most countries, often accounting
for more than half of banking assets (Claessens and van Horen
2013). In the case of the United States, these shares are slightly
smaller but still quite significant. For instance, foreign banks
account for about 25 percent of total banking assets, and five
of the ten largest broker-dealers are foreign owned.
We selected our sample of global banking organizations by
considering the universe of financial institutions with operations
in the United States.5 For non-U.S. entities, our sample includes
small financial organizations and most of the financial organizations designated as G-SIFIs (global systemically important
financial institutions).6 These institutions support a broad range
of real activities in the United States and around the world,
including traditional lending, securities underwriting, loan
syndicate participation, and funds collection for local or parent
operations. We provide comparative analysis by also considering
U.S. institutions with a global footprint. We measure complexity
for each financial institution (U.S. or non-U.S.) by using detailed
data on the counts of affiliates organized under common ownership and control, and we use this information to document a
substantial heterogeneity across global institutions along all of
the alternative dimensions of complexity. Finally, we show the
relationship between different measures of complexity and the
size of banking organizations.
The analysis yields a number of interesting observations.
First, global banking organizations are highly diverse in terms
of size and the correlated metric of absolute counts of affiliates
around the world. These affiliates span multiple levels of ownership through an organizational tree. Second, within these
organizations, the counts of nonfinancial affiliated entities
are generally many times the counts of affiliated banks. Third,
business-type complexity within these organizations—measured with Herfindahl index constructs—shows different tendencies according to the economic geography of the financial
institutions’ parent organizations, with large compositional
distinctions across firms by parent nationality.
Details on the location of affiliates of each parent organization add another important dimension of complexity. We
observe very large differences in the patterns of geographic
complexity among institutions across countries and regions
and even within country of origin. For example, global
5
In particular, we consider which foreign banking organizations operate
branches in the United States.
6
The Financial Stability Board’s November 2012 update of G-SIFIs is discussed
at http://www.financialstabilityboard.org/publications/r_121031ac.pdf.
banking organizations with Japanese parentage are the least
geographically diverse in terms of affiliate locations (that is,
they are more likely to be located within Japan), while these
same organizations tend to have lower overall numbers of
affiliated entities. By contrast, financial organizations with
parents in the euro area tend to be larger in number, have
more affiliates on average, and are more differentiated in terms
of the geographic diversity of affiliate locations. The U.K.
financial organizations are fewer in number, but have large
numbers of affiliates and high geographic diversity.
Finally, we consider whether organizations’ complexity and
size are comparable concepts that can be used interchangeably in discussions of size premia and too-big-to-fail debates.
We find a strong correlation between the complexity of large
financial organizations—as measured by affiliate counts—and
the organizations’ size . However, this tight link disappears
with the other measures of complexity we have described.
2. The Sample of Global Banks
and Available Data for
Measuring Complexity
Perspectives on the complexity of an organization start with
access to detailed data describing that organization’s structure.
All U.S. banks, as well as all branches and subsidiaries of foreign
banks within the United States, file regulatory reports in the
United States. These reports provide information on the structure
of the organization that the reporting entities belong to, but primarily report data on the components within the United States.
For a more complete picture of the entire parent or bank holding
company, we supplement the information from regulatory
reports with metrics of foreign bank organizational structure and
size that are drawn from reporting available through the Bureau
van Dijk’s Bankscope database. We focus our attention on the
subset of foreign-owned global institutions that are the ultimate
parents of the U.S. branches of the foreign organizations.7
Since our focus is on global banks, we also look at those
banks of U.S. parentage that have affiliates outside of the
United States. This information on U.S. global banks is drawn
7
Foreign banking organizations are present in the United States also through
ownership of U.S.-chartered bank subsidiaries. We could include these
entities in our analysis of global complexity. However, branches are a direct
emanation of a foreign-located parent, while subsidiaries (and, if existing,
their U.S. holding company parents) are locally capitalized and under direct
control of the U.S. regulator. In that sense, the implications associated with
complexity of the parent organizations are quite distinct. For our purposes, we
choose to focus our attention on the organizations that operate in the United
States through bank branches, recognizing that some of these organizations
may also have other U.S. subsidiaries, which can be banks and/or nonbanks.
FRBNY Economic Policy Review / December 2014
109
Table 1
Foreign Banking Organizations with U.S. Branches, by Highholder Region
As of Fourth-Quarter 2012
Highholder Data
U.S. Branch Data
Number
of Highholders
Highholder
Total Assets
(Billions of Dollars)
Number of G-SIFIs
G-SIFI Asset Share
(Percent)
Number
of U.S. Branches
Branch Total Assets
(Billions of Dollars)
29
21,379
8
64
46
596
United Kingdom
4
6,855
3
78
11
143
Japan
8
6,163
3
78
18
440
China
6
9,312
1
20
11
53
Switzerland
2
2,621
2
100
8
134
Highholder Region
Euro area
Canada
7
3,375
0
0
20
396
Other Americas
19
1,477
0
0
22
47
Other Asia
37
4,114
0
0
59
61
Other
23
5,644
1
16
27
217
135
60,940
18
48
222
2,089
35
12,568
8
81
—
—
All foreign
United States
Sources: Federal Financial Institutions Examination Council (FFIEC), Report of Assets and Liabilities of U.S. Branches and Agencies of Foreign Banks, 002
regulatory filing; Bureau van Dijk, Bankscope database.
Notes: Highholder region information for the U.S. branches of foreign banking organizations filing with the FFIEC was matched from Bankscope's
Ownership Module. We initially matched 140 highholders—that is, ultimate owners—in Bankscope. Of the 140, 3 were dropped because we could not find
an ownership tree; 2 were dropped because they did not meet our criteria for complexity (that is, they did not have an ownership share exceeding 50 percent
in their affiliates). “Other Asia” comprises Hong Kong, India, Indonesia, Malaysia, Pakistan, the Philippines, Singapore, South Korea, Taiwan, and Thailand.
“Other Americas” comprises Argentina, Bermuda, Brazil, Chile, Colombia, Costa Rica, Ecuador, Guam, Panama, Puerto Rico, Uruguay, and Venezuela.
“Other” comprises Australia, Bahrain, Egypt, Israel, Jordan, Kuwait, Nigeria, Norway, Saudi Arabia, Sweden, Turkey, and the United Arab Emirates. G-SIFI
asset share is defined as the percentage of the region's total assets that are associated with a global systemically important financial institution.
from regulatory reporting in the United States and serves
as a reference point for comparisons with the complexity of
foreign financial organizations operating in the United States.
For all global banks, we provide metrics of organizational
structure as well as various descriptive statistics obtained
using these metrics. Our analysis primarily examines data on
organizational structure in place at the end of 2012.
2.1 Foreign Organizations with U.S. Branches
As part of our criteria for defining a sample of global banks,
we begin with information pertaining to the 222 branches of
foreign banking organizations that filed regulatory reports
in the United States at the end of 2012.8 As shown in Table 1,
overall these branches belong to a total of 135 foreign banking organizations (FBOs). Asia as a whole (Japan, China,
and “other Asia”) accounts for the largest number of parent
organizations from a single region, but euro-area organizations dominate from the perspective of total assets. The total
worldwide assets of these euro-area FBOs exceed $21 trillion.
A number of the foreign banking organizations in the
United States have G-SIFI status—a sign of their significant
global footprint. In terms of geographical distribution, most
G-SIFIs are originally from Europe. While European FBOs
are the largest worldwide, their U.S.-specific presence, measured by the asset size of their bank branches, is not dissimilar to that of FBOs originating in other regions. Branches
8
In the fourth quarter of 2012, 230 U.S. branches of foreign banks filed
regulatory reports. Of these, we were able to match only 222 to complete
highholder data from Bankscope.
110
Measures of Global Bank Complexity
themselves follow heterogeneous business models (this information is not reported in the table). For example, many smaller
branches often lend to nonresident borrowers and support
trade finance. Most of the larger branches instead conduct
trade finance and also provide short- and long-term lending
to support customers from their home country as well as U.S.
business clients. Many of the foreign organizations use their
branches to help manage the liquidity of the larger entity.
Finally, the largest FBOs have many activities that extend beyond lending, including sales and trading, corporate finance,
and asset management. Some of these activities are conducted
outside the branches and through affiliated U.S. subsidiaries.
The final row of Table 1 provides some comparable statistics on U.S. global financial institutions that engage in banking
activity. A total of 35 U.S. financial institutions have branches
or subsidiaries outside of the United States and are considered
global banks by these criteria. Eight of these institutions are
classified as G-SIFIs, representing 81 percent of the $12.6 trillion in total assets across all U.S. global banking organizations.
2.2 Parents and Their Affiliates
Measurement of the complexity of global banking organizations
requires multiple steps. Typically, the immediate owner of the
U.S. branch is a commercial bank, but that entity can have a
different ultimate owner. Indeed, there can be many intermediate ownership links, with ownership shares that vary all along
the levels of ownership in an organizational tree. Determining
the ultimate owner, or “highholder,” of an organization requires
climbing up the ladder of an organization’s ownership.
A number of issues concerning ownership of the organization must be resolved before we can generate useful metrics of
complexity. First, within financial firms, legal and regulatory
distinctions are made between related institutions, those with
majority ownership, and those that are controlled. For our purposes, we seek to capture a level of ownership that is sufficient
to constitute affiliation from an economic perspective—that is,
where control can be presumed. Second, we confront the question of how to deal with multiple levels of ownership trees under
an ultimate parent, since most parents own entities that have
stakes in other entities. Third, we recognize the difficulty in constructing metrics that aggregate over affiliates of different sizes
and types. While some methods of aggregation best demonstrate
the dimensionality of the organization, and perhaps are most
useful for indicating potential frictions in a firm-resolution
scenario, other methods might be more useful for systemic risk
discussions. The latter point raises the issue of whether ideal
complexity constructs would show which entities serve some
“critical function” from the vantage point of the organization’s
production function, in the sense of having the potential to
significantly disrupt some part of the organization’s business in
the event of their absence.9 Moreover, while recognizing these
important conceptual issues, we confront the practical issue
of whether all this relevant information is available. Below we
outline the approach followed based on these considerations and
data availability, addressing only some of these issues.
Our parent concept is the ultimate parent organization that
presides over the U.S. branch, its commercial bank owner, and
the structures above these entities. The full vertical ownership
and vertical affiliate structure are available in regulatory reports filed in the United States for the banks and bank holding
companies with a U.S. parent. We use these data to measure
the complexity of U.S. organizations, as also examined in
Avraham, Selvaggi, and Vickery (2012). However, the ownership structure reported above the particular banking entity in
the United States generally does not capture the full structure
for the whole foreign parent organization, particularly for
larger and more complex organizations.
For foreign parents, we follow Herring and Carmassi
(2010) and use Bankscope’s Ownership Module to extract
relevant organizational structure.10 For each organization, the
data sources contain information on affiliate names, percentage of ownership by the immediate parent or a related control
categorization, geographic location, and type. Information on
the size or balance sheet data of affiliates is less consistently
available. The data are available in levels of direct ownership
from the parent—meaning, for example, that a level 1 affiliate
is directly owned by the ultimate parent entity. Level 2 entities
are owned by level 1 entities, and so on down through level 10
of an ownership tree. Each affiliated entity is tied to its direct
parent with information provided on the quantitative level or
a percentage grouping of ownership, as well as with information on the entity type, industry, and size.11 The structural
9
For a discussion of critical functions, see Annex 3 of the Financial Stability
Board’s work on recovery and resolution, available at https://www
.financialstabilityboard.org/publications/r_121102.pdf. In practice, such
determinations are made at the level of specific products and services.
10
In terms of procedure, we begin with the regulatory reports filed in the
United States. These provide information on “entity” names and identification
codes that are then hand-matched with names of organizations reported
in Bankscope. We then cull information on the organizational structure of
the foreign parent. We were able to match approximately 97 percent of all
reporting U.S. branches of foreign banking organizations to a foreign parent,
which represented 98 percent of all FBO branch assets in the United States in
the fourth quarter of 2012. The missing entities are typically smaller branches
that have been in the overall sample for shorter periods of time; they are less
likely to be in organizations with multiple branches in the United States.
11
Not all fields of data are equally well populated. We include the foreign par
ent itself as an affiliate in the organizational structure and assign it to level 0.
Suppose a bank headquartered in Germany had one affiliate in France. This
organization is intuitively more complex than a bank headquartered in France
FRBNY Economic Policy Review / December 2014
111
Exhibit 1
Sample Organizational Structure of a Simple Foreign Banking Organization
Banks
Insurance
Mutual fund
Other financial
Nonfinancial
United Bank
for Africa Plc
18 other banks
and 5 other affiliates
at level 1
UBA Ltd
Kampala, UG
UBA Asset Mgmt Ltd
Lagos, NG
Level 0
UBA Metropolitan Life Insurance Ltd
Lagos, NG
Level 1
Level 2
Level 3
Source: Bureau van Dijk, Bankscope database.
Notes: Highholder structure information is drawn from Bankscope’s Ownership Module from 2012:Q4. Highholder and affiliates shown are selected to
illustrate a foreign banking organization with subsidiaries only at the first level. “Other financial” includes the following Bankscope entity types: financial
company, private equity firm, venture capital firm, and hedge fund. “Nonfinancial” includes the following Bankscope entity types: industrial company,
foundation/research institute, and self-owned firm.
information available from Bankscope is typically the most
recently reported. For the details reported below, we use
information contained in Bankscope as of the end of 2012. We
follow Herring and Carmassi (2010) and sort these affiliates
into broad buckets: banks, insurance companies, mutual and
pension funds, other financial subsidiaries, and nonfinancial
subsidiaries. Bankscope defines “other financial” as consisting
of four Bankscope categories: “financial companies,” “private
equity” firms, “venture capital” firms, and “hedge funds,” with
“financial company” not separately defined. We restrict our
analysis to include only those entities in which a parent has
50 percent or more ownership. Thus, to be included in our
affiliate counts, an ultimate parent organization has an affiliate
below it (at level 1) if the ownership threshold is at least
50 percent, and if the level 1 organization has an ownership
stake of at least 50 percent in the level 2 organization, and so
on all the way to level 10, which is the furthest distance from
the ultimate parent that we found recorded within Bankscope.
Given these conditions, all statistics provided present a
conservative view of the ownership and complexity of the organizations. We have performed the analysis using ownership
shares of both 25 percent and 50 percent and have generated
quite similar results for both cutoff levels.
Footnote 11 (continued)
with one affiliate in France. Adding the foreign parent as an affiliate noticeably
alters the complexity measures only in cases where the parent has few affiliates.
112
Measures of Global Bank Complexity
To understand these structures, consider Exhibits 1 and 2,
which show the types of organizational trees that emerge
from the data. The entity depicted in Exhibit 1 has a relatively
simple organizational structure. In this case, United Bank for
Africa Plc is a parent organization with only level 1 affiliates
in the hierarchy, and most of the affiliates are classified as
commercial banks. This structure contrasts sharply with that
provided in Exhibit 2, which shows a small part of the organization under parent Deutsche Bank AG. This organization
is highly complex, encompassing a broad range of affiliates of
different types cascading down the various levels of the tree.
For example, the highholder has numerous direct ownership
positions shown in level 1, spread across types of entities as
the color coding indicates. These level 1 affiliates have their
own ownership positions in entities captured as level 2 affiliates, also across a range of bank and nonbank types.
Some caveats apply to the results. All affiliate counts should
be considered illustrative as opposed to definitive, because
our approach has potential shortcomings. First, we match a
U.S. branch to its ultimate highholder and then match that
highholder to a Bankscope entity, thus introducing a risk of
mismatch. Second, we examine the most recent organizational tree under a highholder as reported in the Bankscope
Ownership Module, but we do not view the longer history of
organizational trees. While we expect considerable inertia in
the organizational structure and counts, structures potentially
Exhibit 2
Sample Organizational Structure of a Complex Foreign Banking Organization
Banks
Insurance
Mutual fund
Other financial
Nonfinancial
Deutsche
Bank AG
51 banks and
932 other affiliates
at level 1
1,436 affiliates
at levels 2
through 10
Primelux Insurance
Luxembourg, LU
76 affiliates
at levels 2
through 10
Level 0
DB Capital Markets
(Deutschland) GmbH
Frankfurt, DE
Norisbank GmbH
Berlin, DE
IZI Düsseldorf InformationsZentrum Immobilien
GmbH & Co. KG
Düsseldorf, DE
DB Delaware Holdings UK Ltd
London, GB
Level 1
7 affiliates
at levels 2
through 10
Level 2
DB Capital Markets
Asset Mgmt Holding GmbH
Frankfurt, DE
Deutsche GrundbesitzAnlagegesellschaft
mbH & Co
Frankfurt, DE
DB Capital &
Asset Mgmt
KAG mbH
Köln, DE
Level 3
Source: Bureau van Dijk, Bankscope database.
Notes: Highholder structure information is drawn from Bankscope’s Ownership Module from 2012:Q4. Highholder and affiliates shown are selected to
illustrate a multilevel foreign banking organization. “Other financial” comprises the following Bankscope entity types: financial company, private equity firm,
venture capital firm, and hedge fund. “Nonfinancial” comprises the following Bankscope entity types: industrial company, foundation/research institute, and
self-owned firm.
could change dramatically over time. Third, we make specific
assumptions about the ownership share that warrants inclusion in our counts. Since lower ownership shares could also be
associated with valid affiliates, our counts likely understate the
total number of affiliates under control of an ultimate parent.
3. Evidence on Complexity
3.1 Measures of Complexity
We construct a number of complexity metrics, each with a
different value depending on the economic issues to be
addressed, including activities during the life of the organization or during periods of extreme stress and resolution. While
finer measures could potentially be constructed using more
detailed supervisory or regulatory data, the measures we
present have the advantage of being available for a wide cross
section of entities and therefore are useful for cross-country and
broad conceptual discussions. For example, we can consider the
complexity of a firm’s organizational structure, which maps into
the issues normally raised when the terminology of complexity
is used in policy circles. For instance, a firm organized with
multiple separate legal entities is likely to pose greater challenges for those executing an orderly liquidation, thus potentially increasing the risk of systemic repercussions. Likewise, we
can consider the fragmentation of business activities across
different entity types, which is relevant for policy in that it may
increase the challenges in conducting effective monitoring and
regulation if, for instance, the separate subsidiaries are under
the oversight of separate regulatory agencies.12 For global firms,
12
U.S. bank holding companies are a good example of this. These
organizations as a whole are subject to the supervision of the Federal Reserve,
but the activities of certain subsidiaries are under the direct regulation of
other agencies (for example, the SEC for broker-dealers and funds, and
state and federal insurance bodies for insurance subsidiaries). This issue
is amplified for global organizations with subsidiaries located in foreign
countries that are subject to local regulatory jurisdictions.
FRBNY Economic Policy Review / December 2014
113
Table 2
Complexity Metrics
Type
Name
Construction
Comments
Organizational
Count
Number of 50+% owned affiliates
under a parent organization
Organizational
CountNBtoB
Number of 50+% owned nonbank
affiliates/number of 50+% owned
bank affiliates
Business
Business complexity
(
))
(
The normalized Herfindahl index is based on affiliate types given in
Bankscope, grouped into 1) banks, 2) insurance companies, 3) mutual
and pension funds, 4) other financial subsidiaries, and 5) nonfinancial
subsidiaries. Output values range from 0 to 1, where 0 is lowest
complexity and 1 is highest complexity.
i
2
T
___
T-1
1-∑Ti=1 ________
count i
totalcount
where T is the number of types
Geographic
Geographic complexity
(
(
))
countr 2
R
___
R-1
1 - ∑rR=1 ________
totalcountr
where R is the number of regions
an organizational footprint that spans multiple countries also
adds to the challenges of oversight and resolution.
Table 2 provides the set of measures—organizational,
business, and geographic—that we construct for each of
the global financial firms. The standard measure of organizational complexity, count, is the total number of affiliates—including the ultimate parent—that satisfy the percent
ownership criteria we apply in constructing the metric. This
measure is especially relevant for thinking about organizational fragmentation and resolution planning. A second
organizational measure, countNBtoB, is computed as the
ratio of counts of nonbank affiliates to bank affiliates. This
indicator is more relevant for potential discussions about
the relationship between bank and nonbank affiliates and for
discussions about the pattern of liquidity flows between the
commercial banks and the rest of the organizational structure.13 The two other metrics, introduced to capture business
and geographic complexity, are constructed as Herfindahl
concentration indexes. The business complexity measure
gauges the diversity of the affiliates in terms of the types of
business they conduct, with types divided into five buckets;
the geographic complexity measure assesses the diversity of
the affiliates in terms of geographic location, with locations
divided into thirteen regions.
13
See Cetorelli and Goldberg (2013).
114
Measures of Global Bank Complexity
The affiliate count includes the parent itself as an affiliate.
The normalized Herfindahl index is based on affiliate regions given in
Bankscope, grouped into 1) euro area, 2) United Kingdom, 3) Japan,
4) South Korea, 5) China, 6) Canada, 7) United States, 8) Taiwan,
9) Middle East, 10) other Americas, 11) other Europe, 12) other Asia,
13) other. Output values range from 0 to 1, where 0 is lowest complexity
and 1 is highest complexity.
3.2 Organizational Complexity
of Non-U.S. Global Banks
We begin by describing the findings for those organizations
owned by parents outside the United States. The statistics for
these institutions are constructed using the Bankscope database, as noted earlier. We later turn to the statistics that are
computed for U.S. financial institutions and that are based on
the U.S. regulatory reporting by those entities.
Consider first the patterns in our broadest metric of organizational complexity, which is the total count of affiliates under
a highholder with U.S. branches and where at least 50 percent
ownership of an affiliate is required at each level of the
organization. Chart 2 provides total counts for highholders.
Those organizations with more than 100 affiliates are shown
in the top panel, and those organizations with fewer than 100
affiliates are presented in the bottom panel. Each vertical bar
represents a separate highholder.14 Among these highholders,
twenty-four have more than 250 affiliates and fifteen have
more than 500 affiliates; the highholder with the highest count
has 2,729 affiliates (top panel). Most of the foreign organizations have fewer than 100 affiliates (bottom panel).
14
We do not focus on the specific factors driving the establishment of a
given legal entity. In some cases, tax or regulatory arbitrage may be factors
explaining the existence of a subsidiary, more so than actual business
activities. However, such entities still contribute to more difficult monitoring
and regulation, more complex resolution, and perhaps a denser network of
interconnections within the organization.
Chart 2
Foreign Banking Organizations: Number of Affiliates, by Level
3000
Level 1
Total number of affiliates
Level 2
Level 3
Level 4 to 10
Foreign Banking Organizations with More Than 100 Affiliates
2500
2000
1500
1000
500
0
100
Foreign Banking Organizations with Fewer Than 100 Affiliates
80
60
40
20
0
Source: Bureau van Dijk, Bankscope database.
Notes: Each bar in the chart panels represents a separate foreign banking organization. Highholder structure information is drawn from Bankscope’s
Ownership Module. Level 1 is the first level down from the highholder. An affiliate is considered to be owned by the highholder if a series of 50-plus
percent ownership links exists between it and the highholder. Highholders in the figures are sorted by number of affiliates. Data shown do not include
level 0, which contains one affiliate that represents the highholder itself. Data capture organizational structures as of end-2012.
The color segments within each vertical bar of Chart 2 show
how many affiliates are captured at each level of the organizational tree, from level 1 through level 10. We provide buckets
of levels to keep this information visually accessible, showing
counts for affiliates at level 1, level 2, level 3, and levels 4 and beyond. It is noteworthy that Herring and Carmassi (2010) use the
pattern and counts of only level 1 affiliates to capture complexity.
Our decomposition shows that studies limiting the analysis
to level 1 affiliates, while informative, will not present the full
richness and diversity of affiliate structures. Level 1 affiliates
dominate the structures for entities with fewer than 100 affili
ates, but even these lower-complexity organizations appear
quite different when levels 2 through 4 are added to the metrics
of organizational structure. The role of the multiple levels of
FRBNY Economic Policy Review / December 2014
115
Chart 3
Foreign Banking Organizations: Breakdown of Affiliates by Type
Total number of affiliates
3,000
Bank
Insurance
Mutual fund
Other financial
Nonfinancial
Foreign Banking Organizations with More Than 100 Affiliates
2,500
2,000
1,500
1,000
500
0
100
Foreign Banking Organizations with Fewer Than 100 Affiliates
80
60
40
20
0
Source: Bureau van Dijk, Bankscope database.
Notes: Each bar in the chart panels represents a separate foreign banking organization. “Other financial” includes the following Bankscope entity types:
financial company, private equity firm, venture capital firm, and hedge fund. “Nonfinancial” includes the following Bankscope entity types: industrial
company, foundation/research institute, and self-owned firm. Data capture organizational structures as of end-2012.
ownership is especially important in the organizations depicted
in the top panel. The level 1 affiliates would capture only a
small fraction of affiliates for many of these large players. While
most of the counts of affiliate ownership are within three levels
from the top of the organization, a sizable share of affiliates are
further from the ultimate parent, at levels 4 through 10. Level 1
affiliates are the largest group of affiliates across these global
banking organizations. There are more than 7,000 level 1 affiliates, 9,000 level 2 affiliates, and more than 6,000 level 3 affiliates,
so the total number of affiliates down to and including level 10
is well in excess of 29,000 for the 100-plus foreign parents.
116
Measures of Global Bank Complexity
These non-U.S. global bank affiliates can also be sorted by
types of activities. As previously noted, affiliates owned are classified as belonging to one of five types of primary activity: bank, insurance, mutual fund, other financial, and nonfinancial. Chart 3
recasts the organizations shown in Chart 2 using delineation
by types of activity rather than level in the reporting structure.
The counts of nonfinancial affiliates are generally many times
the counts of banks. Insurance companies are least pervasive at
each level, followed by banks and then mutual funds.
The second organizational complexity metric captures
the extent to which the structure of the organization goes
15
We use the International Monetary Fund’s 2012 definitions to define the
euro area and the Middle East. We then categorize the remaining countries
using the geoscheme created by the United Nations Statistics Division, with
African and Oceanian countries making up the “other” countries category.
16
The list of countries in each region is reported in the footnote of Chart 4.
Chart 4
Foreign Banking Organizations:
Number of Highholders and Affiliates, by Region
30
Total Number of Highholders by Highholder Region
25
20
15
10
5
0
1,000
Average Number of Affiliates
by Highholder Region
800
600
400
200
re
a
Ch
in
Ca a
na
da
T
M aiw
i
O ddl an
th
e
er
E
Am ast
O
er
th
i
er cas
Eu
ro
O
th pe
er
As
ia
O
th
er
n
pa
ut
h
Ko
m
do
Ja
So
ite
d
Ki
Eu
ro
ng
ar
ea
0
Un
beyond banks. The median ratio of nonbank affiliate counts
to bank affiliates across the smaller (fewer than 100 affiliates)
organizations is 3.5, while the median ratio across the more
complex (more than 100 affiliates) organizations is 19. If these
ratios are taken as a metric of activity levels (as opposed to
just fragmentation for other reasons), we would conclude that
nonbank activity rises as organizations become more complex.
Business and geographic complexity metrics for the foreign
organizations also provide interesting insights. To make this
comparison most informative, we break down the parentage of
the foreign organizations by country or region.15 As reported
in Table 1, Asia as a whole accounts for the largest number of
foreign banking organizations with U.S. branches. The euro area
ranks second in terms of counts.16 However, euro area banks
are significantly larger in terms of overall asset size. The average
number of affiliates per parent also differs substantially across
regions (Chart 4, bottom panel). Highholders in the United
Kingdom have the largest number of affiliates by far, with euro
area highholders coming in second. Next, we supplement this
information with descriptive statistics on the business complexity and geographic complexity of the organizations by parentage
(that is, by the country or region of the ultimate owner).
The measure of business complexity is constructed as a
Herfindahl-type index. The index is 0 for organizations with low
complexity—which in practice means that the organization is
exclusively composed of commercial banks—and 1 for organizations with the highest business complexity. In the latter case, the
affiliate counts would be equal across the five categories of types:
banks, insurance companies, mutual and pension funds, other
financial subsidiaries, and nonfinancial subsidiaries. Chart 5
presents the business complexity measure in two ways: by composition into types (bottom panel) and by Herfindahl readings
(top panel), shown as box-and-whiskers plots. The whiskers
show the full range of Herfindahl readings constructed across
the organizations from each country or region. The box shows
the median degree of diversity and the lower and upper quartiles
of diversity across all institutions from that country or region.
The box portions in the top panel differ in length, indicating that the scope of differences from the mean by parent
geography is limited for the U.K., South Korean, and Canadian
parents, but broader for parents from Taiwan, the Middle East,
other Asia, and the euro area. The range of differences is par
ticularly high for parents from other Asia. The type breakdowns
in the lower panel show that South Korean organizations have
Source: Bureau van Dijk, Bankscope database.
Notes: “Middle East” comprises Bahrain, Egypt, Jordan, Kuwait,
Saudi Arabia, Turkey, and the United Arab Emirates. “Other
Americas” comprises Argentina, Bermuda, Brazil, Chile, Colombia,
Costa Rica, Ecuador, Guam, Panama, Puerto Rico, Uruguay, and
Venezuela. “Other Europe” comprises Norway, Sweden, and
Switzerland. “Other Asia” comprises Hong Kong, India, Indonesia,
Malaysia, Pakistan, the Philippines, Singapore, and Thailand. Data
capture organizational structures as of end-2012.
the heaviest relative concentration of banks, followed by Chinese organizations. South Korean organizations also have the
heaviest concentration of mutual fund affiliates. European organizations, whether from the euro area, the United Kingdom, or
the rest of Europe, have the heaviest concentration of affiliates
categorized as “other financial firms.” The affiliates of Taiwanese
parents are the most evenly distributed across types.
FRBNY Economic Policy Review / December 2014
117
Chart 5
Foreign Banking Organizations: Business Complexity of Affiliates
Business Complexity by Highholder Region
1.0
0.8
0.6
0.4
0.2
th
er
th
er
A
O
a
si
e
ro
p
Eu
er
O
er
th
th
id
M
So
U
O
O
Am
e
er
ic
as
Ea
st
an
dl
C
Ta
iw
an
ad
a
na
hi
C
re
a
ut
h
Ko
Ja
pa
n
ni
te
d
Ki
Eu
ro
ng
do
ar
ea
m
0
Bank
Insurance
Mutual fund
Other financial
Nonfinancial
Business Type Breakdown of Affiliates by Highholder Region
Euro area
United Kingdom
Japan
South Korea
China
Canada
Taiwan
Middle East
Other Americas
Other Europe
Other Asia
Other
0
20
40
Percent
60
80
100
Source: Bureau van Dijk, Bankscope database.
Notes: The top panel summarizes business complexity, which is described in Table 2. The top whisker identifies a region’s maximum
business complexity, the top line of the box is the 75th percentile, the line inside the box is median complexity, the bottom line of the box is
the 25th percentile, and the bottom whisker identifies the minimum business complexity (excluding outliers). Outliers are identified by
points using the conventional formula 1.5 * interquartile range. For both panels, “Middle East” comprises Bahrain, Egypt, Jordan, Kuwait,
Saudi Arabia, Turkey, and the United Arab Emirates. “Other Americas” comprises Argentina, Bermuda, Brazil, Chile, Colombia, Costa
Rica, Ecuador, Guam, Panama, Puerto Rico, Uruguay, and Venezuela. “Other Europe” comprises Norway, Sweden, and Switzerland. “Other
Asia” comprises Hong Kong, India, Indonesia, Malaysia, Pakistan, the Philippines, Singapore, and Thailand. Business type breakdown is
consistent with reporting conventions in Bankscope’s ownership module. Data capture organizational structures as of end-2012.
The geographic complexity measure incorporates information on the geographic location of each parent organization’s
affiliates. For this construction, affiliate locations are broken
down into thirteen groups: euro area, United Kingdom, Japan,
South Korea, China, Canada, United States, Taiwan, Middle
118
Measures of Global Bank Complexity
East, other Americas, other Europe, other Asia, and “other”
(Chart 6). The panels of the chart are constructed similarly to
those already discussed for the business complexity measures.
Very large differences exist across banks by country or region,
and within country of origin, in the patterns of geographic
Chart 6
Foreign Banking Organizations: Geographic Complexity of Affiliates
Geographic Complexity by Highholder Region
1.0
0.8
0.6
0.4
0.2
U
th
er
si
th
er
A
O
a
e
O
Eu
er
th
er
th
O
O
Am
e
dl
id
M
So
ro
p
er
ic
as
Ea
st
an
C
Ta
iw
an
ad
a
na
hi
C
ut
h
Ko
re
a
Ja
pa
n
ni
te
d
Eu
Ki
ro
ng
do
ar
ea
m
0
Home
United States
Rest of world
Geographic Breakdown of Affiliates by Highholder Region
Euro area
United Kingdom
Japan
South Korea
China
Canada
Taiwan
Middle East
Other Americas
Other Europe
Other Asia
Other
0
20
40
Percent
60
80
100
Source: Bureau van Dijk, Bankscope database.
Notes: The top panel summarizes geographic complexity, which is described in Table 2. The top whisker identifies a region’s maximum
geographic complexity, the top line of the box is the 75th percentile, the line inside the box is median complexity, the bottom line of the
box is the 25th percentile, and the bottom whisker identifies the minimum geographic complexity (excluding outliers). Outliers are
identified by points using the conventional formula 1.5 * interquartile range. For both panels, “Middle East” comprises Bahrain, Egypt,
Jordan, Kuwait, Saudi Arabia, Turkey, and the United Arab Emirates. “Other Americas” comprises Argentina, Bermuda, Brazil, Chile,
Colombia, Costa Rica, Ecuador, Guam, Panama, Puerto Rico, Uruguay, and Venezuela. “Other Europe” comprises Norway, Sweden, and
Switzerland. “Other Asia” comprises Hong Kong, India, Indonesia, Malaysia, Pakistan, the Philippines, Singapore, and Thailand. Data
capture organizational structures as of end-2012.
diversity of their affiliates. The banks with Japanese parentage
are in organizations that are among the least geographically
diverse in terms of the average affiliate structure and that also
have lower overall numbers of affiliates. The euro area organizations are large in number and large in their average number
of affiliates. The U.K. organizations are fewer in number, but
they also have large numbers of affiliates.
The lower panel of Chart 6 provides an additional perspective on geographic diversity by distinguishing affiliates
that are located in the home country/region from those in
FRBNY Economic Policy Review / December 2014
119
3.3 Organizational Complexity
of U.S. Global Banks
Table 3
Pearson and Spearman Correlations
of Complexity Measures
Pearson Correlations
Ln
Count
CountNBtoB
Business
Complexity
Ln count
1
CountNBtoB
0.67*
Business complexity
0.03
-0.33*
1
Geographic complexity
0.31*
-0.14
0.29*
Geographic
Complexity
1
1
Spearman Rank Correlations
Ln
Count
Ln count
CountNBtoB
Business complexity
Geographic complexity
CountNBtoB
Business
Complexity
Geographic
Complexity
1
0.68*
-0.02
0.32*
1
-0.24*
1
-0.07
0.28*
1
Source: Bureau van Dijk, Bankscope database.
Note: Complexity measures are constructed using end-2012 data from
Bankscope’s Ownership Module.
*Significant at the 5 percent level.
the United States and from those in the rest of the world. It is
interesting that most countries/regions have more than half of
their affiliates in their home market. Having a U.S. presence
in total affiliates is strongest for organizations from Canada,
Japan, and other Americas (which includes Mexico). Organizations from other countries might have branches and a small
number of affiliates in the United States, but about 95 percent
of their legal entities are typically located elsewhere.
Overall, these metrics of complexity address different
dimensions of the business make-up and geographical reach
of global organizations with branches in the United States.
Note that the metrics are not always significantly or positively
correlated with each other. As reported in Table 3, counts
are positively correlated with the ratios of nonbank to bank
affiliates. The correlation between affiliate counts and the
measures of geographic complexity is statistically significant.
Business complexity and geographic complexity are positively
correlated, but both are negatively correlated with the nonbank-to-bank-count ratios.
U.S. banks and their organizations can also be highly complex,
as evidenced by U.S. legislative actions addressing recovery
and resolution planning in the aftermath of the Great Recession. To illustrate this complexity and provide an appropriate
comparison with foreign organizations in the United States,
we start with the top-fifty U.S. bank holding companies in
2013—similar in size to the larger FBOs—and limit our discussion to U.S.-owned organizations with global banking activities. To meet the global banking criterion, an organization
must have some branch or subsidiary outside of the United
States and must file a report indicating exposure to foreign
countries.17 In this way, we can compare U.S. organizations
that have global banks with foreign organizations that have
global banks.18 As reported in Table 1, these criteria generate a
sample of thirty-five organizations with U.S. owners.
For information related to organizational complexity, we
start with a database that collects FR Y-10 reports, the “Report
of Changes in Organizational Structure” filed by each institution.19 The “structure data” use Regulation Y definitions of
control and include affiliates that are controlled and regulated
by the bank holding company. The database contains information on the geography of each affiliate, as well as information
on the type of affiliate as captured by the U.S. NAICS (North
American Industry Classification System) codes. We can
clearly differentiate between banks (NAICS 5221), insurance
companies, nonfinancial firms, and other financial firms. We
do not have a readily available mapping that cleanly separates
the mutual funds from other financial firms, a division that
would allow for a direct correspondence with the categories
drawn from the Bankscope data for foreign organizations. We
use the most current structure as of the fourth quarter of 2012.
The counts of subsidiaries under the parent organization
exceed 3,000 for three of the organizations, total more than
1,000 for another three, and are below 100 for many of the
other U.S. banking organizations (Chart 7, top panel). The
U.S. organizations are similar to their foreign global counterparts in that banking entities represent only a small share
17
Instructions for the preparation of the FFIEC 009 Country Exposure Report are
provided at http://www.ffiec.gov/PDF/FFIEC_forms/FFIEC009_201103_i.pdf.
18
Because our analysis is ultimately motivated by the potential implications
for the United States of the existence of complex global banking organizations,
it makes sense to identify U.S. global organizations by looking at entities that
have either branches or subsidiaries abroad. This is not inconsistent with our
approach to analyzing foreign global families, identified as those having only
branch operations in the United States (see footnote 7).
19
120
Measures of Global Bank Complexity
See http://www.federalreserve.gov/reportforms/forms/FR_Y-1020121201_i.pdf.
Chart 7
U.S. Global Banks: Breakdown of Affiliates by Type and Regional Composition
3,500
Number of affiliates
Affiliates by Type for U.S. Highholders
Bank
Insurance
Other financial
Nonfinancial
3,000
2,500
2,000
1,500
1,000
500
0
3,500
Affiliates by Region for U.S. Highholders
United States
Euro area
United Kingdom
Japan
Canada
Other Americas
Other
3,000
2,500
2,000
1,500
1,000
500
0
Source: Board of Governors of the Federal Reserve System, FR Y-10 and FR Y-6 reporting forms.
Notes: Each bar in the chart panels represents a separate U.S. global bank. Highholder structure information is provided by the Federal Reserve Bank of
New York’s Statistics Function, sourced from the Federal Reserve Board’s reporting forms. Data capture organizational structures as of end-2012. We first
define the euro area and Middle East using the IMF’s 2012 definitions. We then categorize the remaining countries using the U.N. Statistics Geoscheme.
“Other Americas” comprises the following countries: Argentina, Aruba, the Bahamas, Barbados, Bermuda, Brazil, the British Virgin Islands, the Cayman
Islands, Chile, Colombia, Costa Rica, the Dominican Republic, Ecuador, El Salvador, Guam, Guatemala, Haiti, Honduras, Jamaica, Mexico, the Netherlands Antilles, Nicaragua, Panama, Paraguay, Peru, Puerto Rico, Trinidad and Tobago, the Turks and Caicos Islands, Uruguay, and Venezuela. “Other”
includes the following regions: the Middle East (Bahrain, Egypt, Lebanon, Oman, Qatar, Saudi Arabia, Turkey, and the United Arab Emirates), other
Europe (Channel Islands, Croatia, Czech Republic, Denmark, Gibraltar, Hungary, Iceland, Liechtenstein, Monaco, Norway, Poland, Russia, Sweden,
Switzerland, Ukraine, and Yugoslavia), and other Asia (Bangladesh, Brunei, Hong Kong, India, Indonesia, Kazakhstan, Malaysia, Pakistan, the Philippines,
Singapore, Thailand, and Vietnam). “Other” also includes the following countries: Australia, Botswana, Cameroon, Côte d’Ivoire, the Democratic
Republic of the Congo, Gabon, Israel, Kenya, Mauritania, Mauritius, Morocco, Namibia, New Zealand, Nigeria, Senegal, Somalia, South Africa, Swaziland,
Tanzania, Tunisia, Uganda, Zambia, Zimbabwe, and the Virgin Islands.
of the subsidiaries. Other financial entities and nonfinancial
entities account for the vast majority of affiliates. As for the
geographic location of the affiliates (Chart 7, bottom panel),
U.S. global organizations exhibit considerable variation in
the extent of home bias in their affiliates’ locations. The mean
share of affiliates within the United States is 83.2 percent,
while the non-U.S. affiliates are concentrated in the euro area,
the United Kingdom, and other Americas.
4. Is Organizational Size Analogous
to Complexity?
Discussions of complexity often treat fragmentation of
the organization—and the number of affiliates—as a concept
analogous to the size of the organization. In this section, we
consider the relationship between our alternative complexity metrics and the size of the highholder organization as
FRBNY Economic Policy Review / December 2014
121
Chart 8
Foreign Banking Organizations: Relationship between Size and Complexity
8
Subsidiary count (Ln count)
80
Panel A
7
Panel B
70
Slope of fit line = 0.723***
R 2 = 0.480
6
50
4
40
3
30
2
20
1
10
0
20
22
24
26
28
Ln 2012 total assets of highholder
0
20
30
Business complexity
1.0
Panel C
0.8
0.8
0.6
0.6
0.4
0.4
Slope of fit line = 0.013*
R 2 = 0.025
0.2
0
20
22
24
26
28
Ln 2012 total assets of highholder
Slope of fit line = 2.256***
R 2 = 0.100
60
5
1.0
Nonbank-to-bank ratio (CountNBtoB)
22
24
26
28
Ln 2012 total assets of highholder
30
Geographic complexity
Panel D
Slope of fit line
= 0.064***
R 2 = 0.148
0.2
0
20
30
22
24
26
28
Ln 2012 total assets of highholder
30
Sources: Bureau van Dijk, Bankscope database; Board of Governors of the Federal Reserve System, FR Y-7Q reporting form.
Notes: Complexity measures are constructed using end- 2012 data from Bankscope’s Ownership Module. “Total assets” data are drawn from Bankscope
and the Federal Reserve Board’s FR Y-7Q reporting form.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
reflected in asset valuation. Overall, we find that the straight
measures of affiliate counts are positively correlated with
size of the highholder organization, such that the larger
organizations have more affiliates. However, other measures
of complexity that use information on type, organizational
structure, and regional placement of affiliates are not as tightly
correlated with the size of the overall institutions.
4.1 Complexity and Size for Foreign Global
Organizations
Chart 8 provides plots and regression fits between measures of
complexity and size. Panel A shows the relationship between
the (logarithm of) counts of affiliates and the (logarithm of)
asset size of the foreign global organizations.20 The slope of the
20
The size of the parent organization (in terms of assets) is, however, strongly
correlated with the size of its branches within the United States.
122
Measures of Global Bank Complexity
Chart 9
U.S. Global Banks: Relationship between Size and Complexity
10
Subsidiary count (ln)
800
Panel A
Nonbank-to-bank ratio
Panel B
700
Slope of fit line = 1.024***
R 2 = 0.786
8
Slope of fit line = 50.787**
R 2 = 0.172
600
500
6
400
300
4
200
100
2
23
1.0
24
25
26
27
Ln 2012 total assets of highholder
28
29
0
23
Business complexity
0.8
0.8
0.6
0.6
0.4
0.4
Slope of fit line = -0.013**
R 2 = 0.016
0
23
24
25
26
27
Ln 2012 total assets of highholder
25
26
27
Ln 2012 total assets of highholder
28
29
28
29
Geographic complexity
1.0
Panel C
0.2
24
Panel D
Slope of fit line = 0.093***
R 2 = 0.226
0.2
28
29
0
23
24
25
26
27
Ln 2012 total assets of highholder
Source: Board of Governors of the Federal Reserve System, FR Y-10 and FR Y-9C reporting forms.
Notes: Complexity measures are constructed using end-2012 data from the FR Y-10 reporting form. “Total assets” data are drawn from the FR Y-9C
reporting form.
* Significant at the 10 percent level.
** Significant at the 5 percent level.
*** Significant at the 1 percent level.
regression line is significant, and about half the cross-sectional
variation in counts is explained by size. An organization that
is twice as large as another is likely to have 70 percent more
affiliates. If resolution of failing institutions is a concern, this
relationship shows that the larger—and often more systemi
cally important—institutions may have more complex and
numerous affiliate structures, suggesting that resolution costs
increase with size.
Consider next the concepts that might be relevant for
understanding the business models of the global banking
organizations. The ratio of nonbank affiliate counts to bank
affiliate counts is positively correlated with size (Panel B),
but size explains less than 10 percent of cross-sectional
variation. Additionally, the relationship between size and the
diversity of affiliate types is close to zero as organizational
size increases (Panel C), making size a poor predictor of
affiliate-type diversity.
Similar observations pertain to the metrics of affiliates’
geographic complexity (Panel D). Recall that we presented
evidence of significant home bias in the affiliate locations
FRBNY Economic Policy Review / December 2014
123
for these organizations. Some organizations, regardless of
size, have all of their legal entities in their home markets.
Other organizations, regardless of size, are broadly diversified
geographically. Overall, the relationship between size and
diversity by region is highly diffuse, even if positively sloped.
4.2 Complexity and Size of U.S. Global
Financial Institutions
For U.S. global financial institutions, the tight relationship
between size and complexity is a feature only of the count
metric, which is the number of affiliated entities under the
parent organization. As shown in Panel A of Chart 9, the (log)
count of affiliates rises one-for-one with the (log) size of the
overall organization, a tighter and more linear fit than that
observed for organizations with foreign parents.21
For all other measures, the correlations with size—even
when statistically significant—are decidedly weaker. The ratio
of nonbank to bank counts, shown in Panel B, shows a weak
relationship to organizational size in U.S. global organizations,
as it did for the foreign organizations, with a regression fit of
only 17 percent. There is little relationship between size and
the diversity of affiliate types (here consisting of four types,
instead of the five types identified for the measures relating to
the non-U.S. entities), which have a slope of essentially zero
and explain only 2 percent of the cross-sectional variation in
these values (Panel C). The relationship between geographic
diversity and size is positive but also weak (Panel D). Smaller
U.S. entities in our sample are more likely to have affiliates
located exclusively in the United States. Otherwise, geographic
dispersion is not related to the size of the organization.
5. Conclusion
Our examination of the complexity of global banking organizations—both foreign institutions that have operations in
the United States and U.S. institutions that have branches or
subsidiaries abroad—has produced a number of significant
findings. Above all, we have documented that there is more
to complexity than just organizational size. Global entities
can differ tremendously in their organizational complexity,
business complexity, and global footprint.
It is not clear what might be driving the buildup in
bank complexity. Complexity may result in part from firms
growing larger as they attempt to achieve economies of
scale and scope. Managerial motives (empire building,
entrenchment) or rent seeking (monopoly power, acquisition of too-big-to-fail status) may also be contributing
factors. Geographic diversification and the development of
complex affiliate structures might reflect taxation regimes
and efforts to avoid business transparency and achieve less
restrictive regulation across markets (Baxter and Sommer
2005).22 Moreover, some of the growth in complexity may
be an endogenous response to an evolving intermediation
technology that favors the growth of organizations incorporating, under common ownership and control, the many
financial entities (specialty lenders, asset managers, finance
companies, brokers and dealers, and others) that have
increasingly become essential to the financial intermediation process (see, for example, Poszar, Adrian, Ashcraft, and
Boesky [2010], Cetorelli, Mandel, and Mollineaux [2012],
and Cetorelli and Peristiani [2012]).
Whatever the main causes of complexity may be, our analysis of global banking organizations—which are arguably the
most complex among banking institutions in general—reveals
a substantial degree of diversity in the forms that complexity takes. Banking organizations may display relatively few
entities that are in their immediate control but, under that
first layer of organizational complexity, many more affiliates
may be connected indirectly to the same common highholder
through multiple rounds of ownership. Alternatively, banking
organizations may display a relatively narrow business scope,
but still operate through a large number of entities broadly
located across the globe. Or it could be that the organizations
display a relatively narrow geographic focus but engage in a
wide variety of business activities.
There is substantial room for further research to clarify
the positive and negative consequences of business, organizational, and geographic complexity for individual financial
organizations and the financial systems they inhabit. For
instance, a bank that is part of a complex organization, spanning multiple sectors and countries, may benefit from larger
and more diversified internal capital markets. Likewise, it may
22
21
This finding is consistent with the evidence in Avraham, Selvaggi, and
Vickery (2012), which showed that organizational size was the only significant
determinant of this count measure of complexity, and that no role was played
by an industry concentration index, geographical concentration indexes, or
shares of domestic commercial bank assets.
124
Measures of Global Bank Complexity
Desai, Foley, and Hines (2006) examine U.S. multinational firms and
show that they establish operations in tax haven countries as part of their
international tax-avoidance strategies. Rose and Spiegel (2007) argue that,
while activities in offshore financial centers are likely to encourage bad
behavior in some countries, they may also have positive effects, such as
providing competition for the domestic banking sector.
gain access to external markets and benefit from the credit
standing of the broader organization. In addition, there may
be benefits from business synergies such as product complementarities, information flows, and cost savings on common
resources. If these working hypotheses are correct, the mode
of operation of a bank may differ in accordance with the complexity of its family.
Complexity may alter balance sheet management strategies,
affecting decisions about funding models, liquidity policies,
and investment and lending strategies. Hence, organizational
complexity may have broad economic implications not just
during episodes of financial distress but also in normal times.
These observations suggest the importance of achieving a
fuller understanding of the drivers and forms of complexity—
and of using this knowledge to assess the positive and negative
externalities that complexity generates.
FRBNY Economic Policy Review / December 2014
125
References
Avraham, D., P. Selvaggi, and J. Vickery. 2012. “A Structural View
of U.S. Bank Holding Companies.” Federal Reserve Bank of
New York Economic Policy Review 18, no. 2 (July): 65-81.
Claessens, S., and N. van Horen. 2014. “Foreign Banks: Trends and
Impact.” Journal of Money, Credit, and Banking 46, no. s1
(February): 295-326.
Baxter, T., and J. Sommer. 2005. “Breaking Up Is Hard to Do:
An Essay on Cross-Border Challenges in Resolving Financial
Groups.” In Systemic Financial Crises: Resolving Large
Bank Insolvencies, 175-91. Singapore: World Scientific.
Desai, M., C. F. Foley, and J. R. Hines, Jr. 2006. “The Demand for Tax
Haven Operations.” Journal of Public Economics 90, no. 3
(February): 513-31.
Cetorelli, N., and L. Goldberg. 2013. “Size, Complexity, and Liquidity
Management: Evidence from Foreign Banks in the United States.”
Unpublished manuscript, Federal Reserve Bank of New York.
Herring, R., and J. Carmassi. 2010. “The Corporate Structure of
International Financial Conglomerates: Complexity and Its
Implications for Safety and Soundness.” In A. Berger,
P. Molyneux, and J. Wilson, eds., The Oxford Handbook
of Banking. Oxford: Oxford University Press.
Cetorelli, N., B. Mandel, and L. Mollineaux. 2012. “The Evolution
of Banks and Financial Intermediation: Framing the Analysis.”
Federal Reserve Bank of New York Economic Policy Review
18, no. 2 (July): 1-12.
Herring, R., and A. Santomero. 1990. “The Corporate Structure of
Financial Conglomerates.” Journal of Financial Services
Research 4, no. 4 (December): 471-97.
Cetorelli, N., J. McAndrews, and J. Traina. 2014. “Evolution in Bank
Complexity.” Federal Reserve Bank of New York Economic
Policy Review 20, no. 2 (December): 85-106.
Pozsar, Z., T. Adrian, A. Ashcraft, and H. Boesky. 2010. “Shadow
Banking.” Federal Reserve Bank of New York Staff Reports,
no. 458.
Cetorelli, N., and S. Peristiani. 2012. “The Role of Banks in Asset
Securitization.” Federal Reserve Bank of New York Economic
Policy Review 18, no. 2 (July): 47-63.
Rose, A., and M. Spiegel. 2007. “Offshore Financial Centres: Parasites
or Symbionts?” Economic Journal 117, no. 523 (October):
1310-35.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
126
Measures of Global Bank Complexity
Adam Kirk, James McAndrews, Parinitha Sastry, and Phillip Weed
Matching Collateral
Supply and Financing
Demands in Dealer Banks
• The 2008 failure and near-collapse of some
of the largest dealer banks underscored the
complexity and vulnerability of the industry.
• A study of dealer banks finds that their unique
sources of financing are highly efficient in
normal times, but may be subject to marked
and abrupt reductions in stressful times.
• Dealer banks’ sources of financing include
matched-book repos, internalization, and
collateral received in connection with overthe-counter derivatives trading.
• Under some conditions, U.S. accounting rules
allow dealer banks to provide financing for
more positions than are reflected on their balance sheets. Rules that permit netting of certain
collateralized transactions may not yield a true
economic netting of dealer banks' exposures.
• A prudent risk management framework
should acknowledge the risks that inhere in
collateralized finance.
Adam Kirk is a risk analytics associate, James McAndrews an executive
vice president and the director of research, Parinitha Sastry a former senior
research associate, and Phillip Weed a risk analytics associate at the Federal
Reserve Bank of New York.
1. Introduction
B
anks are usually described as financial institutions that
accept deposits of dispersed savers and use the deposited
funds to make loans to businesses and households. This
description is accurate but incomplete, as banks also engage in
other types of intermediation that finance economic activity.
Some banks act as dealers in markets, providing liquidity
and supporting price discovery by buying and selling financial instruments, helping to facilitate trade in markets. Banks
also perform prime brokerage services—a role that involves
providing financing to investors along with many ancillary
services, such as collateral management, accounting, and
analytical services. The banks that engage in these activities, which we call dealer banks, facilitate the functioning of
financial markets.
To conduct their business, dealer banks rely on varied
and, in some cases, unique sources of funding. In most cases,
dealer banks’ lending is collateralized by securities or cash. As
in a standard bank, funding for a loan made by the bank may
come from the bank’s own equity or from external sources,
that is, from parties that are not borrowers from the bank.
Unlike a standard bank, however, dealer banks can employ
internal sources to fund a customer loan, either by taking a
trading position that offsets that of the customer receiving the
The authors thank Tobias Adrian, Darrell Duffie, Steven Spurry, and James
Vickery for helpful comments. The views expressed in this article are those of
the authors and do not necessarily reflect the position of the Federal Reserve
Bank of New York or the Federal Reserve System.
Correspondence: jamie.mcandrews@ny.frb.org
FRBNY Economic Policy Review / December 2014
127
loan or by utilizing an offsetting position taken by another
customer. For example, the bank may make a “margin loan” to
one customer, lending cash to finance the customer’s security
purchase, with the customer offering the purchased security as
collateral for the bank loan. Another customer may request to
borrow the same security to establish a short position, offering
cash to the bank as collateral for the loan. The two customers’
pledges of collateral provide the bank with the resources to
fulfill both customers’ demands for borrowing. That dealer
banks can in some cases use the collateral pledged by one customer to lend to another, or to fund a trade made by the bank,
confers a cost advantage since internal sources of funding are
generally less expensive than external market sources. Dealer
banks also maintain specialization in collateral valuation and
management, which reinforces the aforementioned financing
cost advantages. Consequently, such collateralized lending to
investors is concentrated in dealer banks.
The interdependence of the financing for the borrowing of
one customer and the collateral posted by another customer
makes the sources of funding for dealer banks vulnerable in
ways that are different from those of standard banks. Consider
that in a standard bank, when a borrower repays a loan, the
bank can often redeploy the repaid funds as a loan to another
borrower or as payment to a deposit holder. In contrast, when
a borrower repays the dealer bank, the borrower also reclaims
the collateral it posted to the bank. If the dealer has repledged
this collateral to finance another customer’s position, it must
find a substitute for the reclaimed collateral returned to the
borrower. In other words, the dealer must scramble to find an
alternative source of the collateral in order to meet its obligations. In times of financial market stress, external parties may
be reluctant to lend to the dealer bank, even against collateral,
so it can be costly and difficult for the bank to seek funding
externally. This vulnerability of dealer banks, though similar
to that faced by standard banks when depositors withdraw,
differs in that it occurs instead when borrowers repay their
loans, reflecting the profound interdependence between the
bank’s customers, their borrowing, and their pledges of collateral. Of course, not all of the dealer bank’s funding is internally
generated and so, like standard banks, dealer banks engage in
maturity transformation and thus are also susceptible to rapid
withdrawals of external sources of funding.
This article aims to provide a descriptive and analytical
perspective on dealer banks and their sources of financing. In
reviewing the methods by which dealer banks reuse collateral,
we consider various concepts related to collateralized finance,
many of which have been discussed in Duffie (2010, 2011),
Stigum and Crescenzi (2007), and Committee on the Global
Financial System (2013). We conclude that this type of financing yields high levels of efficiency in normal times, but may be
128
Matching Collateral Supply and Financing Demands
subject to significant and abrupt reductions in stressful times,
relative to the external financing sources upon which other
banks rely. That conclusion raises many issues about how
policy should address this type of financial sector vulnerability, which we briefly discuss. In addition, the limitations of
existing sources of data on the extent of the use of collateral by
dealer banks leads us to recommend more extensive reporting
of dealer banking financing arrangements.
First, we create an analytical and stylized framework of
dealer banks to outline their major collateralized finance activities. Under certain circumstances, U.S. accounting rules allow
the dealer to provide financing for more positions than are reflected on its balance sheet. Dealer banks can take advantage of
netting rules when calculating the size of their balance sheets.
For example, under both U.S. and international accounting
standards, the exposure of a dealer bank to a customer that has
offsetting collateralized positions with the dealer bank can be
reported as the net economic claim on the dealer bank by the
customer. Consider the following (extreme) example. Suppose,
as outlined above, Customer A borrows cash and provides
a security as collateral to the dealer bank; suppose, furthermore, that Customer B borrows the security and provides cash
collateral to the dealer bank. The dealer bank uses the collateral
provided by one customer to satisfy the borrowing demands of
the other. Now suppose that, later, Customer B borrows cash
and proffers a different security to the dealer bank as collateral,
and Customer A borrows that security and supplies cash to the
dealer bank as collateral. Then because each customer’s exposure may be eligible to be net on the balance sheet of the dealer,
the dealer may be able to report assets and liabilities equal
to $0, even though it had provided financing in substantial
amounts to the two customers. Consequently, a dealer bank’s
balance sheet captures only a portion of its gross provision of
financing to customers.
As a result of this fact, we present both a stylized balance
sheet and a stylized collateral record that together allow for
a better representation of how dealers provide collateralized
financing. We then apply this stylized framework to explain
how dealer banks perform key intermediation functions and
discuss the various methods by which dealer banks can reuse
collateral provided by customers. We review three of them in
detail: matched-book financing, internalization of collateral
financing, and pledging of collateral received in over-thecounter (OTC) derivatives trading. The nature of these activities allows dealer banks to derive efficiencies in their use of
collateral and assist in the performance of financial markets.
We also use and apply data in firms’ public disclosures
to our stylized framework, to the extent that dealer banks'
activities are reflected in such disclosures, and attempt to
measure the degree to which firms economize and optimize on
their collateral resources. To determine how much financing
a dealer bank provides to customers, one must examine the
“collateral record” of a dealer bank, which can be found in its
10-K and 10-Q public disclosures. However, the nature of the
reporting is not standardized across dealer banks; as a result,
we are forced to restrict ourselves to a small number of banks.
We choose to focus on Bank of America, Citigroup, Goldman
Sachs, JP Morgan Chase, and Morgan Stanley—the bank holding companies with the largest broker-dealer subsidiaries.1 As
the largest dealer banks, their data capture the majority of such
activity. We also include Lehman Brothers for its historical relevance to the crisis. These data allow us to provide a consistent
aggregate view of the amount of collateral received, collateral
pledged, and the size of the dealer banks' collateralized liabilities, for the very largest dealer banks. These data portray how
these aggregate amounts have changed across time, especially
during the period of the financial crisis and its aftermath.
In our review, we rely on two notions of efficiency employed by dealer banks. First, we define “collateral efficiency”
as the percentage of a dealer bank’s collateral received that is
rehypothecated. This concept is one indicator that focuses on
how extensively the dealer bank uses its customer-provided
collateral resources. It is likely, and in our sample we verify, that
this measure is increasing with the size of the dealer’s collateral
pool, as a larger portfolio of collateral will contain securities
that match more customer demands than would a smaller
portfolio. Other factors that we conjecture would increase
collateral efficiency include the number and mix of customers,
the operational capacity of the dealer, and other economic
features of the dealer firm, such as its creditworthiness, that
make it a good counterparty.
The second concept of efficiency captured by dealer banks,
“collateralized financing efficiency,” is a broad economy.
Dealer banks seek to optimize their use of collateral to reduce
their costs of serving customers’ demand for borrowing. This
concept differs from the previous one in that collateralized
financing efficiency refers to all the economic benefits reaped
by dealer banks in their allocation of firm and customer
collateral. By rehypothecating the collateral that secures
dealer banks’ loans to customers, the dealer bank can provide to customers lower-cost financing, or increase its own
profit margins. This lower cost is a reflection of two potential
1
Broker-dealers are firms that participate in markets by buying and selling
securities on behalf of themselves and their clients. They must register with
the Securities and Exchange Commission (SEC), and are often a subsidiary of
a larger bank holding company. Any securities purchased by the firm for its
account can be sold to clients or other firms, or can become part of the firm’s
own holdings. Our definition of dealer banks includes activities performed
by broker-dealers, but also includes OTC derivative dealing activities, which
are often conducted in the affiliated depository institution subsidiary of the
parent holding company (rather than the broker-dealer subsidiary).
benefits captured in the collateralized financing arrangements
in which dealer banks specialize. First, in a violation of the
Miller-Modigliani theorem and framework, dealer banks can
attract funding more cheaply by pledging collateral, rather
than borrowing on an uncollateralized basis; a fortiori, the
dealer bank can obtain funds for an even lower cost if those
funds themselves are provided as collateral when a customer
borrows a security held by the dealer bank.2 Second, by using
collateral of one customer to satisfy the borrowing demand of
another customer, the dealer can in certain instances minimize the amount of economic and regulatory capital and
liquidity needed to support its financing activities. In our review, we provide a measure of gross collateral received relative
to assets recorded on the balance sheet, which can provide a
gauge of the efficiency of collateralized finance provided by
dealer banks. Those economies, which we will discuss in more
detail below, also lead to a lower cost of provision of financing
services by the dealer bank.
Additionally, like banks of all types, dealer banks engage
in maturity and credit transformation; however, dealer banks
also engage in the transformation of customer collateral. For
example, a dealer bank can lend to a customer for a specific
maturity, and then obtain funds by pledging the collateral provided by the customer but at a shorter maturity; that sort of
maturity transformation is just one way by which dealer banks
provide additional value to customers. Various types of credit
transformations are also made by dealer banks as they seek to
satisfy the demands of different customers. This includes collateral substitution, in which the dealer bank effectively lends
one type of security while the customer provides the dealer
bank collateral of a different type.
To the extent that dealer banks capture efficiencies from
collateralized finance, we would expect that they would dominate this form of finance as they could provide these services
at lower cost than alternative approaches. It is important to
keep in mind that notwithstanding the presence of collateralized financing efficiencies, the dealer bank is subject to
significant risks that may offset the lower costs provided by
this form of finance in normal times, in a full consideration of
social costs and benefits.
In particular, the dependency of the funding available to
dealer banks sourced from collateral provided by customers
was clearly evident in the financial crisis of 2007-09. As we will
see, the amount of funding available to dealer banks shrank
2
In the Miller-Modigliani framework, firms and households are risk-neutral
and markets are complete, so borrowing on a collateralized or uncollateralized basis is essentially equivalent, and would yield the same interest rate.
However, in a framework in which information about the extent of borrowing
by the firm is not known by the lender, lenders are risk-averse and markets
are incomplete; collateralized borrowing rates may be below uncollateralized
borrowing rates.
FRBNY Economic Policy Review / December 2014
129
precipitously in the wake of the bankruptcy of Lehman Brothers
Holdings International. Further, the gross amount of collateral received by the other dealer banks in our sample, and the
amount that these dealer banks in turn pledged as collateral, fell
even more precipitously, indicating that the collateral provided
by customers, when used as a secondary source of funding by
the bank itself, is subject to greater withdrawal than the net
claims or obligations as reported on-balance-sheet.
A limitation of our analysis lies in the way that dealer banks
report their activities in providing collateralized finance. Because of the aforementioned interdependencies, dealer banks
report their holdings and uses of collateral in ways that are
open to alternative interpretations. As a result, it is not always
clear how best to describe their balance sheet in a way that is
consistent across firms. The reporting is heterogeneous and,
consequently, not fully comparable across firms. This places
severe limitations on the number of firms whose financing
arrangements we review in this article.
Our study is organized as follows. Section 2 begins by defining the businesses of dealer banks, and follows by constructing
some stylized balance sheets that clearly depict the sources
and uses of funding for the major dealer banks. In section 3,
we describe the main types of dealer financing arrangements,
including those that allow the banks to utilize internal sources
of funding for their lending, using our stylized frameworks so
that comparisons can be made across institutions. In section 4,
we use the public disclosures to provide measures of the stylized
balance sheets and collateral record we introduce in section 2
for the firms, measuring the relative importance and evolution
of the sources of financing over time. Section 5 concludes.
2. An Overview of Dealer Banks
Dealer banks are active in the intermediation of many markets,
either in their role as dealers or in their role as prime brokers
where they provide financing to investors. Dealer banks are
financial intermediaries that make markets for many securities and derivatives by matching buyers and sellers, holding
inventories, and buying and selling for their own account when
buyers and sellers approach the dealer at different times, for
different quantities, or are clustered on one side of the market.
Many banks with securities dealer businesses also act in the
primary market for securities as investment banks, underwriting issues to sell later to investors. Services typically provided
by dealers include buying and selling the same security simultaneously, extending credit and lending securities in connection with transactions in securities, and offering account
services associated with both cash and securities.
130
Matching Collateral Supply and Financing Demands
Many dealers carry out their activities in a broker-dealer
subsidiary of a bank holding company. For most derivatives
trades, dealers are one of the two counterparties, with many
dealers recording their derivative exposures at their affiliated
bank, the depository institution subsidiary of the parent company. Prime brokers are the financing arm of the broker-dealer,
offering advisory, clearing, custody, and secured financing
services to their clients, which are often large active investors,
especially hedge funds. Prime brokers can conduct a variety of
transactions for their customers, including derivatives trading,
cash management, margin lending, and other types of financing transactions.
Dealer banks, like other for-profit businesses, strive to
minimize the cost of providing financing to customers, which
often need cash or particular securities. They can do this in
part through a strategy of meeting their clients’ needs without relying wholly on costlier sources of external funding.
Sometimes this is accomplished if the dealer bank itself has
an offsetting position, or at other times another customer’s
position. By fulfilling the collateral needs of one party (either
in the form of cash or securities) with an already existing
source of that collateral, the dealer bank can avoid additional
financing transactions. This maximizes its income directly by
eliminating a borrowing cost, as well as indirectly by minimizing costs associated with larger balance-sheet sizes.
2.1 Stylized Framework for Dealer Banks
Our stylized framework consists of two components: a balance sheet and a collateral record.3 While a complete representation of a dealer bank’s financial reporting is out of the scope
of this article, we describe conceptually how certain financing
activities appear on the balance sheet and the collateral record.
By examining both the balance sheet and the collateral record,
we can, to some extent, trace how much the firm is relying on
internal sources of collateralized financing, that is, financing
provided either by the dealer’s own trading activity or by other
customers’ activities, and how much is sourced externally.
In Table 1, we present a simplified (and reduced) version
of the official balance sheets reported by our sample of dealer
banks, focusing on the parts most oriented toward their dealer
banking business. We intend to use this simplification of the
3
The collateral record can be thought of as analogous to a balance sheet,
in that it records all sources and uses of collateral by the dealer bank. Like
the balance sheet, it is an accounting concept, but it reflects underlying
commitments made by the dealer bank. As such, it can also be thought of as a
commitment schedule of the firm to receive/deliver collateral or cash from/to
customers under specific conditions.
Table 1
Stylized Balance Sheet
Assets
Liabilities and Equity
Cash
Equity
Instruments owned
Instruments sold but not yet owned
Reverse repo/securities borrowing
Repo/securities lending
Brokerage receivables
Brokerage payables
• Brokerage receivables are economically similar to reverse
repos/securities borrowing, but are generally related to
other forms of collateralized lending, such as brokerage
customer margin loans and collateral posted in connection
with derivatives.
Liabilities and equity are grouped into the categories outlined above and typically reflect a “source of ” or “obligation to
return” cash.
• Equity reflects all balance-sheet equity accounts, such as
earnings and stock issuance.
balance sheet to illuminate those dealer-bank-specific and
unique financing activities. Some categories are excluded
because they are less relevant to the collateralized finance
business unique to dealer banking, while others are grouped
together because they are economically similar. This allows us
to apply a single framework consistently across firms whose
reporting disclosures are not always homogenous.
Assets are grouped into the categories outlined above and
typically reflect a “use of ” or “claim to” cash.
• Cash will generally include the dealer’s own funds that are
held in an account with a bank, such as a deposit with a
bank within the same bank holding company, a Federal
Reserve Bank, or a third-party bank. Cash will also include
funds deposited with a bank that are fully segregated on
behalf of a customer of the dealer.
• Financial instruments owned will reflect the fair value of
risky positions owned by the bank, such as securities, physical commodities, principal investments, and derivative
contracts. In concept, the fair value reflects the cash that
could be obtained upon sale of the instrument.
• Reverse repurchase agreements (reverse repo)/securities borrowing generally reflects a cash outlay and a receipt of a financial
instrument as collateral, such as a security.4 The reverse repo is
recorded on the balance sheet as the value of the cash outlay,
not the collateral. These collateralized transactions are governed by specific SIFMA5 forms. (For a more detailed discussion of these transactions, see Adrian et al. [2011].)
• Instruments sold but not yet owned reflect the dealer’s own
short positions in a financial instrument, such as a security,
physical commodity, or derivative contract.
• Repurchase agreements (repos)/securities lending generally
reflects a cash receipt and a pledge of a financial instrument,
such as a security. These are similar to the reverse repo/securities borrowing transactions described above, but in these
the dealer bank takes the opposing side of the trade.
• Brokerage payables are economically similar to repos/securities lending, but are generally related to other collateralized borrowings, such as brokerage customer credit balances and collateral received in connection with derivative
transactions.
While the balance sheet represents an accurate snapshot
of the net economic claims on and obligations of the dealer
relative to those counterparties from an idealized simultaneous settlement of all claims in default, it does not necessarily
reveal an accurate view of the dealer bank’s actual collateral
sources and uses in real time, nor of the total amount of
financing that the dealer bank is providing to customers. In
this way, the balance sheet and the collateral record offer alternative insights into the financing and funding conditions of
the firms. Combining the information from the balance sheet
and the collateral record allows us to glimpse some of the
collateral efficiencies and “collateralized financing” efficiencies
experienced by the dealer bank.
The collateral record is divided into two categories, total
collateral received that can be repledged, and total collateral
pledged (Table 2). The collateral record reflects sources and
4
A repurchase agreement, or repo, is an agreement to sell a security with
a commitment to repurchase it at a specified date in the future, usually the
next day, for a stated price. The economic function of these agreements is
essentially equivalent to a short-term secured loan, and usually the value of
the securities purchased is greater than the cash outlay, with the difference
referred to as a haircut. For more details, see Copeland, Martin, and Walker
(2010). For the party on the opposite side of the transaction, the agreement is
called a reverse repo.
5
Repurchase and reverse repurchase agreements are typically governed by a
master repurchase agreement (MRA) or global master repurchase agreement
(GMRA). Securities borrowing and securities lending are typically governed
by a master securities lending agreement (MSLA).
Table 2
Stylized Collateral Record
Collateral Received
Collateral Repledged
-
-
-
-
FRBNY Economic Policy Review / December 2014
131
uses of collateral broadly, including on a gross outstanding
basis, and does not conform to specific guidance under U.S.
generally accepted accounting principles (GAAP).
Dealer banks receive cash and securities as collateral in
connection with reverse repos, securities borrowing, and
brokerage receivables.
While these transactions may also be reflected on the
stylized balance sheet, the reported numbers will differ from
the collateral record for several reasons. First, the balance
sheet does not fully reflect the use of collateral in the transaction. For example, a dealer may extend a $100 margin loan
to a brokerage customer to purchase a security, which will
be recorded as a $100 brokerage receivable on our stylized
balance sheet. In this case, the dealer may have received (and
was permitted to repledge) $140 of the brokerage customer’s
security. The collateral received can be delivered or repledged
in connection with repos, securities lending, and brokerage
payables. In this example, the dealer could repledge the $140
of the client’s securities in a repurchase agreement; the movement of the client’s securities would show up in the dealer’s
collateral record, but the stylized balance sheet would only
reflect the margin loan and repurchase agreement.
Crucially, U.S. GAAP allows for the netting of receivables
(for example, reverse repo, securities borrowing, and brokerage receivables) and payables (for example, repo, securities
lending, and brokerage payables) when:
of the business day. It must be probable that the associated
banking arrangements will provide sufficient daylight overdraft or other intraday credit at the settlement date for each
of the parties.
f. The enterprise intends to use the same account at the
clearing bank or other financial institution at the settlement date in transacting both 1) the cash inflows resulting
from the settlement of the reverse repurchase agreement
and 2) the cash outflows in the settlement of the offsetting
repurchase agreement.7
As a result, U.S. GAAP netting has the effect of reducing
the size of the balance sheet relative to the collateral record.
3. Review of Select Activities
at Dealer Banks
The following sections outline specific activities or transactions
that dealer banks conduct in carrying out financial intermediation, focusing on three in particular: matched-book dealing,
internalization, and derivatives collateral. While not exhaustive,
these activities are representative of the activities inherent in the
dealer’s business model, which are accompanied by a unique set
of risks that are not faced by standard banks.
a. The repurchase and reverse repurchase agreements are
executed with the same counterparty.
b. The repurchase and reverse repurchase agreements have
the same explicit settlement date specified at the inception
of the agreement.
c. The repurchase and reverse repurchase agreements are executed in accordance with a master netting agreement (MNA).6
d. The securities underlying the repurchase and reverse repurchase agreements exist in book-entry form and can be
transferred only by means of entries in the records of the
transfer system operator or securities custodian.
e. The repurchase and reverse repurchase agreements will be
settled on a securities transfer system (for which specific
operational conditions are described) and the enterprise
must have associated banking arrangements in place (also
described in detail). Cash settlements for securities transferred are made under established banking arrangements
that provide that the enterprise will need available cash on
deposit only for any net amounts that are due at the end
3.1 Matched-Book Dealing
Dealer banks often refer to a balance sheet where repurchase
agreements finance offsetting reverse repurchase agreements
as a “matched book.” The dealer bank’s business model relies
on optimizing its uses and sources of collateral. In essence,
this means some clients demand cash and possess securities,
while others demand securities and possess cash. In a typical
matched-book transaction, a client provides a security as
collateral in exchange for cash and grants the dealer the right
to repledge this collateral. The dealer repledges this security
to another client to source the cash. As a result, the dealer’s
balance sheet does not reflect any security owned. This can
be an efficient method to finance securities for customers if
the dealer has better access to repo markets generally, and the
dealer can earn a slight interest rate spread in the difference in
the interest paid to lenders and the rate it charges its borrowers. This incremental spread is one form of the “collateralized
financing” efficiency exploited by dealers.
6
A master netting agreement in effect allows all transactions covered by the
MNA between the two parties to offset each other, aggregating all trades on
both sides and then replacing them with a single net amount (International
Swaps and Derivatives Association 2012).
132
Matching Collateral Supply and Financing Demands
7
Financial Accounting Standards Board Interpretation no. 41, “Offsetting of
Amounts Related to Certain Repurchase and Reverse Repurchase Agreements” (FIN 41).
Exhibit 1
Matched-Book Dealing
Transaction 1: Customer A lends Dealer $1,020 in Security Q and receives $1,000 in cash.
Transaction 2: Dealer lends Customer B $1,020 in Security Q and receives $1,000 in cash.
Customer A
(Broker-Dealer)
$1,020 Security Q
Dealer
$1,000 Cash
$1,020 Security Q
$1,000 Cash
Stylized Collateral Record
Stylized Dealer Balance Sheet
Category
Customer B
(Mutual Fund)
Beg. Balance
Transaction 1 Transaction 2 End. Balance
Cash
—
(1,000)
1,000
—
Instruments owned
—
—
—
—
Reverse repo/securities borrowed
—
1,000
—
1,000
Brokerage receivables
—
—
—
—
Total assets
—
Repo/securities loaned
Instruments sold,
but not yet owned
Brokerage payables
Total liabilities
—
—
1,000
1,000
—
—
—
—
—
—
—
—
—
1,000
Total equity
—
Transaction
Transaction 1
Transaction 2
Collateral
Received
Collateral
Repledged
1,020
1,020
1,000
Dealers can run a matched book using various types of
transactions. For illustrative purposes, we focus on the simplest example, described above, of offsetting repos and reverse
repos. Exhibit 1 presents a dealer that starts with no balance
sheet, but is then approached by another broker-dealer, Customer A, which is looking for a $1,000 overnight cash loan
and offers a $1,020 security as collateral. The dealer enters
into a matched-book trade by simultaneously executing an
overnight reverse repo with Customer A (Transaction 1) and
an overnight repo with Customer B (Transaction 2), a mutual
fund willing to invest its excess cash overnight.
The dealer’s balance sheet reflects a symmetrical increase in
both a claim to $1,000 cash and an obligation to return $1,000
cash. Although the dealer acted as principal, the balance sheet
reflects no position in Security Q. However, the collateral
record shows that the dealer received and acquired the right
to repledge or sell $1,020 of Security Q, of which it actually
repledged $1,020.
If the dealer had been unable to use Customer A’s collateral
to secure a loan from Customer B, it might have had to borrow on an unsecured basis to source the cash or, alternatively,
encumber some of the bank’s own collateral. As a result, the
—
transaction might have become uneconomical from the dealer’s perspective. In this example, the dealer passed the haircut
required by Customer B (approximately 2 percent) entirely
on to Customer A. As a result, in the example the dealer reaps
efficiencies to the extent that it can borrow from Customer B
at a lower cost than it can lend to Customer A.
Furthermore, there are cases where the dealer bank executes matched-book transactions in a way that can provide it
a net funding source. Consider a modification to our example, in which the dealer is able to demand a higher degree of
overcollateralization on the reverse repo. Suppose the dealer
required Customer A to deliver $1,060 worth of securities as
collateral for the cash borrowed, and Customer B still required
only $1,020 of the securities from the dealer in exchange for
its cash. Here, the dealer retains an additional $40 of securities
that it could potentially pledge to additional financing transactions. The dealer, in charging a higher haircut than the one
it pays, generates an additional financial capacity as a result
of its intermediation activities. In turn, these extra efficiencies—we might call them a “collateral haircut margin”—allow
the dealer to provide prime brokerage and lending services at
lower costs. Whether the haircut margin reflects a transfer to
FRBNY Economic Policy Review / December 2014
133
dealer banks, or whether competition among dealer banks for
the profits provided by this haircut margin results in lower financing costs for customers—and therefore provides a benefit
to society—depends on the level and nature of the competition between dealer banks.
Maturity, Credit, and Collateral Transformation
In the original example, the final maturity of both transactions
was the following day. However, a matched book does not
always involve executing offsetting repurchase and reverse
repurchase agreements that are “perfectly matched” in terms
of the final maturity date or the credit quality of the involved
counterparties. That is, dealer banks engage in maturity and
credit transformation.
First, dealers can borrow cash through repo at shorter
maturities than those at which they lend through reverse repo.
Maturity mismatches expose the dealer to some interest rate
risk, should short-term borrowing rates spike before maturity.
In an extreme event, the dealer is exposed to “rollover risk,”
in which it could prove difficult for the dealer to roll over its
borrowings, while still being required to fund the lending on
longer-term reverse repos.
Second, dealers can borrow from more creditworthy investors and lend to less creditworthy borrowers, which introduces
an element of credit risk, although this risk is mitigated by
requiring collateral and charging haircuts accordingly. Generally, these risks are common to most financial intermediaries,
including traditional banks.
U.S. GAAP Netting and Collateral Transformation
The matched-book examples thus far have been presented
as two transactions from the dealer’s perspective, each with
a different counterparty. In practice, dealers will often have
multiple transactions executed with a single counterparty. Under U.S. GAAP, repos and reverse repos can be reported on a
net basis with a single counterparty if executed in accordance
with a master netting arrangement and if the agreements have
the same explicit settlement date, as well as some additional
operational requirements.8
Importantly, offsetting repurchase agreements are not required to be collateralized by the same securities to be eligible
for U.S. GAAP netting. In essence, this means a dealer can
deliver $100 cash in exchange for a U.S. Treasury security
and, separately, borrow $100 cash and pledge a corporate
8
Refer to International Swaps and Derivatives Association (2012).
134
Matching Collateral Supply and Financing Demands
bond, and offset these two transactions on its balance sheet
as long as the other required conditions are met. This form of
collateral transformation presents the dealer with more opportunities to optimize its sources and uses of collateral with
clients without enlarging, or “grossing up,” its balance sheet.
However, this also introduces an additional layer of complexity in analyzing the dealer’s collateral position, particularly in
periods when market clearing conditions for different types of
securities diverge.
3.2 Internalization of Trading Activities
Dealers achieve yet another source of collateralized financing
efficiency by “internalizing” their trading activities, that is,
by using offsetting trading positions between two clients or
between clients and the dealer bank to “finance” each other.
Similar to the concept of matched book, opportunities to
“internalize” can arise via the provision of funds by the dealer
bank collateralized by client securities. Those securities are
then reused and delivered into another transaction as a means
of financing the client position. Its name refers to the concept that the bank, in some cases, can source financing for
a customer internally, without the need to attract additional
funding from the external marketplace for funds.
Though internalization exhibits certain similarities with
matched book as a financing mechanism, it differs in the degree
of cost advantage, in its ability to minimize the size of the balance sheet, and in its flexibility to generate financing for dealer
bank trading positions. While these differences generally suggest
that internalization is a low-cost and flexible form of financing
for dealer banks, internalization is vulnerable to a unique set
of risks, as it relies on the market positioning of customers.
As conditions in markets change, owing to a significant price
move, for example, either one side or the other might rapidly
exit its financing position from the dealer, forcing the dealer to
quickly replace securities or cash from external markets.
Exhibit 2 depicts one example of internalization, with the
prime brokerage business of a dealer bank facilitating opposing transactions for two separate hedge fund clients. In this
example, the dealer bank lends to a hedge fund client on margin and uses a portion of the securities purchased to fund the
original margin loan (Transaction 1b). Internalization occurs
when a separate client has sold short the same security, and
therefore the collateral backing the margin loan is rehypothecated and delivered into the short position (Transaction 2b).
In this example, the dealer bank starts with a balance sheet
of zero. Customer A deposits $500 of cash into its brokerage account (Transaction 1a) and then borrows $500 from the dealer
Exhibit 2
Customer-to-Customer Internalization
Transaction 1a: Customer A deposits $500 in Cash into its brokerage account.
Transaction 1b: Dealer lends Customer A $500 in Cash to purchase $1,000 of security Q, receiving $700 of rehypothecatable collateral.
Transaction 2a: Customer B deposits $350 in Cash into its brokerage account.
Transaction 2b: Customer B sells short $700 of security Q, posting the cash proceeds to the Dealer as collateral.
End. Balances: Dealer holds the residual $550 of cash in a segregated lock-up account.
Customer A
(Hedge Fund)
$700 Security Q
$500 Cash
Dealer
(Prime Broker)
$700 Security Q
$700 Cash
Stylized Dealer Balance Sheet
Stylized Collateral Record
Beg.
Balance
Transaction
1a
Transaction
1b
Cash (including
segregated
lock-up)
Instruments
owned
Reverse repo/
securities
borrowed
Brokerage
receivables
—
500
(1,000)
350
700
550
—
—
—
—
—
—
—
—
—
—
—
—
—
—
500
—
—
500
Total assets
—
500
(500)
350
700
1,050
Repo/securities
loaned
Instruments
sold, but not
yet owned
Brokerage
payables
—
—
—
—
—
—
—
—
—
—
—
—
—
500
(500)
350
700
1,050
Total liabilities
—
500
(500)
350
700
1,050
Total equity
—
Category
Customer B
(Hedge Fund)
Transaction Transaction
2a
2b
bank to acquire a $1,000 long position in Security Q
(Transaction 1b), using $500 of the funds deposited in
Transaction 1a to make the purchase. Customer A pledges the
acquired securities as collateral for the loan. As Customer A
purchases the securities on margin, the dealer gains rehypothecation rights over the collateral posted in the amount of
140 percent of the margin loan, which is $700 of Security Q in
this example. The remaining $300 of Security Q is segregated
and placed off the dealer’s books.
Separately, hedge fund Customer B, intending to open a
short position in the same security, first deposits $350 of cash
End.
Balance
Transaction
Transaction 1b
Transaction 2b
Collateral
Received
Collateral
Repledged
700
—
—
700
—
into its brokerage account (Transaction 2a) and then borrows
$700 of Security Q from the dealer bank (Transaction 2b),
pledging and depositing a total of $1,050 with the dealer bank
($700 in cash collateral and the $350 in its brokerage account).
Here we assume both clients hold margin accounts governed
by Regulation T,9 which generally allows a client to borrow up
to 50 percent of the value of a security pledged as collateral (in
this case, $500 for Customer A) and requires clients to maintain
9
The Federal Reserve Board’s Regulation T relates to cash accounts held by
customers and limits the amount of credit that dealers may extend to customers for the purchase of securities.
FRBNY Economic Policy Review / December 2014
135
margin in the amount of 150 percent of the market value of
open short positions (in this case, $1,050 for Customer B).
The dealer settles Customer B’s short sale by using the
securities pledged by Customer A for its margin loan, effectively internalizing the two positions. The dealer’s ending
balance sheet will reflect a segregated cash balance of $550, a
brokerage receivable in the amount of the $500 margin loan to
Customer A, and a brokerage payable to Customer B equivalent to $1,050.10
Differences between Internalization
and Matched Book
This example highlights a key difference with matched-book
financing—as the name implies, internalization eliminates the
need for external sources of financing, and represents a form
of both “collateral” and “collateralized financing” efficiency.
Absent the ability to internalize these positions, the dealer
would need to engage in two additional external transactions
to satisfy both clients’ positions. First, the margin loan would
require financing, which the dealer bank would most likely
obtain from the repo market. Second, the dealer bank would
have to source the security to satisfy the client’s short position, likely through a securities borrowing transaction. Both
of these external transactions would resemble our example of
matched book, in that the dealer bank would seek to earn a
small spread based on its superior access to repo and securities borrowing markets. Instead, the dealer bank furnished
its clients with a total of $1,200 in credit (the $500 margin
loan and $700 short position), earning interest and fees on
that level of credit, but has a balance sheet of only $1,050.
Internalization allows the dealer to generate potential income
from finding and matching, among its own customers, natural
buyers and sellers of the same security. Importantly, internalization also presents regulatory advantages from a capital
and leverage perspective; eliminating the need to engage in
external repo and securities borrowing transactions minimizes the size of the balance sheet and enables the dealer bank
to increase other client activity.
A second substantive difference from matched book lies
in the dealer bank’s ability to finance its own positions with
10
This amount is a function of both clients’ “net equity,” as per SEC rule 15c3-3,
and is not accessible to the dealer as a source of funding for other activities.
The “locked-up” amount reflects the difference between the value of collateral
rehypothecated from Customer A’s margin account and the receivable from
the margin loan ($700 - $500 = $200), plus the difference between Customer
B’s credit balance (that is, the original cash deposit plus the proceeds from the
short position) and the market value of the short position ($1,050 - $700 =
$350). Therefore, the total locked-up cash balance is $200 + $350 = $550.
136
Matching Collateral Supply and Financing Demands
client activity. A dealer bank may be naturally long a security
as a part of its market-making inventory, as a hedge, or as an
investment. Under circumstances where a client sells short
that same security, the dealer bank can deliver its own inventory into the short sale, or in other words internalize the two
positions. Again, the dealer benefits significantly from this
form of internalization as it earns a fee on the client’s short,
and saves on the financing cost of its own inventory, although
it does not achieve the same degree of balance-sheet reduction
observed in the case of internalization between two clients.
Risks Associated with Internalization
The internalization of client and firm trading activities affords
the dealer bank distinctive cost and income advantages; however, it engenders a unique set of risks.
Unlike the traditional banking model, a dealer bank’s client
assets and liabilities tend to have an undefined set of maturities. The maturity of offsetting client positions is therefore
difficult to predict precisely. Short-term imbalances in the duration of client or dealer positions that have been internalized
against each other pose significant risks to the dealer. During
a period of market or firm-specific stress in particular, a dealer
may need to replace one side of an internalized transaction.
For example, a client may liquidate its account by repaying its
margin loan, resulting in a cash inflow to the dealer; however,
the dealer may have already rehypothecated the underlying
collateral for the margin loan to deliver into another client’s
short sale. In this event, the dealer bank may need to source
a hard-to-borrow security in an illiquid market in order to
settle the sale of the margined long position. Similarly, a client
may “buy back” a short position that was previously financing another client’s long position, which may force the dealer
to resort to the external market to seek additional funds in
a potentially illiquid repo market. While these imbalances
between long and short positions might resolve themselves
over a period of time, they can be temporarily destabilizing,
requiring the dealer bank to increase its balance sheet to
finance positions externally or, if that were to prove difficult,
to sell assets or close client positions quickly.
In a similar vein, dealers can look to any unused capacity to
internalize trading positions when wholesale funding markets
experience temporary dislocations. This residual capacity in
certain cases could function as a buffer, allowing dealer banks
to shift from external sources of financing to internal ones
during a short-lived period of market stress. Importantly,
however, the ability to internalize is likely correlated with
the relative liquidity of a given position. In other words, the
least liquid positions—those with the greatest probability of
becoming unfundable during a period of stress—would have
the fewest opportunities for internalization. Alternatively,
more common securities, such as exchange-listed asset classes,
would likely present more opportunities for internalization
as they would be present in greater abundance, offering more
opportunity for matching with other client positions. This inventory of client positions, then, allows the dealer bank to use
internalization, where possible, as a potential cushion against
the cost of finding more expensive funding or tapping into
liquidity reserves to replace existing wholesale sources.
Internalization and Financial Reporting
Internalization is an important source of financing for dealer
banks. However, under current standards for financial reporting, the degree to which dealer banks internalize trading
activities or maintain available but untapped capacity to
internalize positions is, at best, unclear. Since internalization
results from the optimization of trading activities visible only
through a dealer bank’s collateral record, it is neither directly
nor quickly observable given current standards of public
financial disclosure. The “leveraging effect” of client-to-client
internalization largely occurs off-balance-sheet, with only
an imperfect record appearing in the footnotes to the firms’
reported financial statements, where repledged collateral
received from margin lending is aggregated with repledged
collateral received through other secured transactions.
Moreover, U.S. GAAP accounting allows dealers to net long
and short exposures within individual client margin accounts,
which further augments the balance-sheet efficiency of internalized transactions, but by extension increases the disparity
between the gross positions financed and the net exposures
reported on-balance-sheet.
3.3 Derivatives Collateral Received
The final category of dealer bank financing examined in this
article is collateral received or posted in relation to secured
derivatives transactions. These transactions generate or
use cash through receiving or posting initial margin (IM)
and variation margin (VM), which serve to offset the risks
associated with current and potential future exposure, respectively.11 In principle, the collateral and collateralized financing
11
See Basel Committee on Banking Supervision and International Organization of Securities Commissions (2013, p.10). Here, exposure refers generally
to the replacement cost should the derivative counterparty default. Current
exposure (CE) is a function of the current mark-to-market value of the
efficiencies gained through derivatives transactions are similar
to those arising from matched-book transactions or internalization. That is, a dealer bank that has sold a derivative to
a client can purchase an equal and opposite exposure from
another dealer bank, using the collateral received from one
transaction to satisfy the collateral requirement on the second,
while capturing a small income spread.
Unlike other secured transactions addressed in this article,
however, the derivatives transactions as defined here do not
entail the exchange of cash for securities, but rather the posting or receipt of collateral to secure an economic claim. Derivatives are collateralized according to contractual terms stipulated in the Credit Support Annex (CSA) of an International
Swaps and Derivatives Association (ISDA) master agreement,
which establishes the types of acceptable collateral, among
other rules. Cash tends to be favored in this context because it
is operationally easier to exchange and attains a greater degree
of balance-sheet efficiency through the cash collateral netting
provisions granted under U.S. GAAP and IFRS.
Firms can offset their derivative assets against derivative
liabilities when:
a. Each of two parties owes the other determinable amounts.
b. The reporting party has the right to set off the amount
owed with the amount owed by the other party.
c. The reporting party intends to set off.
d. The right of setoff is enforceable at law.
Additionally, cash collateral received or paid in connection
with a derivatives contract can be net against the fair value of
the contract if executed under a master netting arrangement.12
Net Financing and Efficiencies
Asymmetries in contractual terms covering the extent of
collateralization may give rise to situations in which dealer
banks receive more collateral than they post, generating
net financing possibilities to the extent that this excess can
be repledged.
transaction, whereas potential future exposure (PFE) reflects certain aspects
of the contract itself (for example, revaluation/margining period) and the
prospective volatility of the underlying instrument.
12
“Without regard to the condition in paragraph 5(c), a reporting entity may
offset fair value amounts recognized for derivative instruments and fair value
amounts recognized for the right to reclaim cash collateral (a receivable) or
the obligation to return cash collateral (a payable) arising from derivative
instrument(s) recognized at fair value executed with the same counterparty
under a master netting arrangement.” Financial Accounting Standards Board
Interpretation no. 39, “Offsetting of Amounts Related to Certain Contracts”
(FIN 39).
FRBNY Economic Policy Review / December 2014
137
Exhibit 3
Asymmetric Collateral Terms on Matched Derivatives
Transaction 1: Customer A purchases a Total Return Swap long position, for which Customer A pays $1,000
in initial margin (IM).
Transaction 2: Dealer sells the same exposure to Customer B, another dealer, but is required to post only $500 in IM.
Transaction 3: Customer A’s contract value appreciates $100, requiring the Dealer to post $100 in collateral to Customer A.
Transaction 4: Dealer’s contract with Customer B depreciates, requiring Customer B to post $100 in collateral to the Dealer.
Transaction 1
$1,000 Cash
Customer A
(Pension Fund)
Transaction 2
$500 Cash
Dealer
Transaction 3
$100 Cash
Transaction 4
$100 Cash
Stylized Dealer Balance Sheet
Stylized Collateral Record
Transaction 1
Transaction 2
Transaction 3
Transaction 4
End.
Balance
1,000
(500)
(100)
100
500
—
—
—
—
—
—
—
—
—
—
—
500
—
—
500
1,000
—
(100)
100
1,000
Repo/securities
loaned
Instruments sold,
but not yet owned
—
—
—
—
—
—
—
—
—
—
Brokerage payables
1,000
—
—
—
1,000
Total liabilities
1,000
—
—
—
1,000
—
—
(100)
100
—
Category
Cash
Instruments owned
Reverse repo/
securities
borrowed
Brokerage
receivables
Total assets
Total equity
The size of this potential net financing pool is linked to a
variety of factors specific to the dealer bank and the nature of
the derivatives transactions. Much like other forms of secured
financing, the dealer’s relative credit quality and market access
will influence its ability to negotiate preferential margining
terms. In general, the tendency to margin on a portfolio basis
suggests that large active dealers would benefit from economies of scale, minimizing their requirements to post collateral
on interdealer transactions, while reinforcing their ability to
command greater amounts from smaller or nondealer counterparts. Forthcoming rules governing the margining of OTC
derivatives may limit this benefit by establishing minimum
levels for the calculation of IM and VM; however, it is unlikely
that the benefits of scale would be eliminated entirely.
138
Customer B
(Dealer)
Matching Collateral Supply and Financing Demands
Transaction
Transaction 1
Transaction 2
Transaction 3
Transaction 4
Collateral
Received
Collateral
Repledged
1,000
—
—
100
—
500
100
—
Exhibit 3 uses our stylized framework to illustrate how
matched collateralized derivatives transactions can both generate net financing for a dealer and minimize leverage through
balance-sheet netting provisions. In this example, the dealer
engages in matched derivatives transactions, remaining market-risk-neutral, but establishing preferential terms for IM. At
inception, the offsetting transactions are reflected in the dealer
bank’s cash position, a brokerage receivable representing the
IM paid on the hedging transaction, and a brokerage payable
associated with the dealer’s obligation to return IM received
from Customer A. Notably, we assume that the fair values of
each transaction will be fully collateralized by cash VM such
that they qualify for netting treatment, and therefore the contract exposures will not be reported on-balance-sheet.
Irrespective of market movements in the underlying position, the dealer will retain the net funding gained through
the receipt of IM. Furthermore, margin deposits tend to
earn a short-duration money market yield, rendering this an
inexpensive form of financing for dealer banks. Thus, because
of the cash/collateral netting and portfolio margining imposed
by the dealer bank, the dealer reaps collateralized financing
efficiencies. The netting here is not bilateral customer-to-dealer
netting, but netting by the dealer bank itself. “Rehypothecating” cash is effectively netting by the dealer of collateral
received and collateral posted.
Potential Risks
Balance-sheet and cost advantages aside, the stock of net collateral received by a dealer bank is exposed to certain vulnerabilities that call into question its overall durability as a means
of financing, even under circumstances where the offsetting
transactions are matched in terms of market risk and level of
collateralization.
First, in a traditional sense, these transactions are subject to
the same rollover risk considerations as other dealer financing
arrangements. At the maturity of a swap transaction, unless
the position is rolled over, the collateral received would need
to be returned to the original client. If a dealer offsets a position with one of shorter duration, or if a dealer obtains some
amount of net collateral received on transactions of matched
duration, at maturity it faces a financing gap in the amount of
the margin posted to the offsetting transaction.
Second, from the contractual perspective, transactions are
often embedded with certain credit rating downgrade triggers
requiring the posting of additional collateral or imposing
more constraining restrictions on rights of rehypothecation.
Other contractual risks exist as well, such as the potential for
a client to replace existing collateral posted with a currency or
security that cannot readily be reposted to a matched derivative position, however, this risk would only be present to the
extent that the dealer bank takes a sort of contractual basis
risk by accepting divergent collateral types on matched trades.
Finally, dealer banks may be beholden to reputational
considerations in periods of stress. While they may have
contractual rights over the use of client collateral, they may
nevertheless honor client requests to segregate collateral or
close out trades preemptively in the spirit of preserving their
franchise. It is this element of uncertainty and contingent risk
that undermines the durability of net collateral received in
relation to derivatives as a source of dealer financing.
Table 3
Net Current Credit Exposure of OTC Derivatives
March 31, 2013
Rank
Holding Company
Total OTC
Derivatives
(Billions of Dollars)
1
Goldman Sachs Group, Inc., The
152,679
2
JPMorgan Chase & Co.
144,490
3
Bank of America Corporation
110,506
4
Morgan Stanley
103,813
5
Citigroup Inc.
93,816
6
Wells Fargo & Company
15,015
7
HSBC North America Holdings Inc.
12,238
8
Bank of New York Mellon Corporation, The
12,021
9
State Street Corporation
6,802
10
PNC Financial Services Group, Inc., The
3,547
11
Suntrust Banks, Inc.
2,521
12
Fifth Third Bancorp
1,663
13
Capital One Financial Corporation
1,417
14
TD Bank US Holding Company
1,385
15
Northern Trust Corporation
1,154
16
KeyCorp
1,067
17
Unionbancal Corporation
1,036
18
RBS Citizens Financial Group, Inc.
951
19
Deutsche Bank Trust Corporation
834
20
Regions Financial Corporation
732
21
Ally Financial Inc.
661
22
BB&T Corporation
579
23
BancWest Corporation
456
24
M&T Bank Corporation
450
25
BBVA USA Bancshares, Inc.
426
Total for industry
673,018
Sources: OCC; FR Y-9C, Schedule HC-L.
Note: Total OTC Derivatives is the sum of all net current credit exposures
(Line 15(a)).
4. Data
Five bank holding companies—Bank of America, Citigroup,
Goldman Sachs, J.P. Morgan Chase, and Morgan Stanley—
represent more than 95 percent of the domestic banking
industry’s net current credit exposure for over-the-counter
derivatives, which totaled $673 billion in 2013:Q1 (Table 3).
These five banks are the major derivatives dealers, so we focus
on these companies. We also include Lehman Brothers for
its relevance to the crisis. By including these firms, we can
FRBNY Economic Policy Review / December 2014
139
Chart 1
Select Funding Sources of Major Derivatives Dealers
December 31, 2012
700
Brokerage payables
Oblig. to return secs. rec’d. as collateral
Instruments sold but not yet owned (shorts)
Collateralized funding
Billions of U.S. dollars
600
500
400
(26)
(62)
(62)
(79)
(23)
300
200
100
0
Citi
MS
GS
ML
JPM
quently, Citigroup and J.P. Morgan Chase have a much larger
proportion of deposits as a share of their liabilities, and so the
select liabilities we display in Chart 1 reflect a lower percentage
of their total liabilities than for Goldman Sachs, Merrill Lynch,
and Morgan Stanley. In many of our reported figures below, we
concentrate our analysis on Goldman Sachs, Merrill Lynch, and
Morgan Stanley, for two reasons. First, disclosures from these
firms are primarily oriented toward dealer banking—more
so than for the universal banks of Citigroup and J.P. Morgan
Chase, which have large deposit franchises and corporate
and household lending businesses, in addition to their dealer
banking activities. Because our stylized balance sheet excludes
the deposit-taking part of standard banking, we more closely
approximate our stylized balance sheet by focusing on the three
former investment banks. Second, the reporting of the collateral
record is least consistent, among the five BHCs reported above,
for Citigroup and J.P. Morgan Chase, with some elements only
available at the annual frequency or not reported in an equivalent manner, as the other banks.
Source: Company 10Q/10K filings.
Note: The figures in parentheses represent the sum of the select
liabilities as a percentage of total liabilities.
examine the importance of the unique forms of financing we
outline as well as how variable they were through the crisis.
To begin to gauge the size and importance of different
funding sources for dealer banks, we show in Chart 1 the select liabilities of our candidate firms as of December 31, 2012,
excluding unsecured borrowings and deposits in accordance
with our stylized balance sheet.13 Each of the five firms whose
liabilities we display is a bank holding company (BHC) that
performs the more standard banking activities of deposit taking and lending to households and commercial firms, as well
as the activities we group and display under dealer banks.
With the exception of dealing in OTC derivatives, most of
the dealer bank activities are concentrated in the broker-dealer
subsidiaries of the BHCs. Goldman Sachs and Morgan Stanley
were “stand-alone” investment banks in 2008 prior to their conversion to BHCs in September 2008, so their businesses remain
more concentrated in dealer banking and prime brokerage than
those of Bank of America, J.P. Morgan Chase, and Citigroup,
reflected by the high portion of their total liabilities represented by select dealer banking funding sources. Merrill Lynch,
a subsidiary of Bank of America, filed its 10-Q and 10-K reports
separately from Bank of America up until 2013:Q1. Conse13
Recall that Lehman Brothers Holdings International filed for bankruptcy on
September 15, 2008, so that firm is not shown in the chart.
140
Matching Collateral Supply and Financing Demands
4.1 Data Sources
In the following sections, we use data from the firms’ 10-Q/
10-K filings to analyze their balance sheet and collateral
records using our stylized framework. The components of our
stylized balance sheet are calculated directly from the firms’
consolidated balance sheets. We can estimate the firms’ collateral record by exploiting self-reported data that appear either
in parentheses on the balance sheet or in textual footnotes. We
focus on the 10-Q/10-K data in this article because they offer
the most consistent measures of the balance sheet and collateral record for dealer banks (see the data appendix).
Firms report collateral received from counterparties in
connection with certain brokerage activities, such as reverse
repurchase agreements, securities borrowing, and derivatives,
as well as the amount of the collateral received that was subsequently repledged by the firm. The firms in our sample separately report the portion of their financial instruments owned
that they have pledged as collateral that can be repledged, as
well as financial instruments that have been pledged that cannot be repledged; taking the sum of these two numbers gives
us the amount of financial instruments owned by the firm that
it has pledged.
Firms also specifically state the amount of cash collateral
posted and received in connection with derivatives activity
that qualifies for U.S. GAAP netting.
4.2 Quantifying the Collateral Record
Reconstructing the collateral record as described above can
shed light on the efficiencies captured by dealer banks through
their secured activities. Although we are limited in our ability
to fully quantify the sources of financing examined in our stylized framework—in particular, the internalization of trading
activities—we assess two aspects of the collateral record that
are indicative of the benefits dealer banks realize through the
intermediation of secured transactions.
First, the level of collateral received that has been rehypothecated indicates firms’ reliance on “customer collateral”
generated through secured lending activities and derivatives
to raise financing, from both internal and external sources;
these data allow us to directly measure the banks' “collateral
efficiency,” as we have defined it. Second, the total stock of
collateral held and the total stock of collateral pledged relative
to the balance sheet can be used to indicate the degree of “collateralized financing efficiency” achieved by the dealer banks.
In both cases, we examine with particular attention the
financial crisis period of 2008-09 characterized by significant
balance-sheet deleveraging. With respect to levels of rehypothecation, we draw upon the example of Lehman Brothers
to illustrate the magnitude of contraction in a case that ultimately ended in bankruptcy and liquidation.
Finally, we attempt to decompose the level of collateral
efficiencies achieved into its transactional sources, that is,
for the three types of activities described earlier—matchedbook, internalization, and derivatives. Although this falls
short of fully quantifying the amount of financing generated
by the methods examined through our stylized framework,
it provides some insight into the relative materiality of each
source. Moreover, it allows us to observe a rough trend during
the period of the crisis, raising important questions about the
systemic risk effects of each activity.
Collateral Efficiency
We first display a measure of “collateral efficiency,” which we
earlier defined as the percentage of a dealer bank’s collateral
received that is rehypothecated. Recall that it is likely that
collateral efficiency is increasing with the size of the dealer’s
collateral pool, as a larger portfolio of collateral will contain
securities that match more customer demands than would a
smaller portfolio. Indeed, this correlation was positive and
significant at the aggregate level for the sample of banks we
examine in this article (Chart 2). Further, a simple regression
using the panel data with entity-fixed effects confirms this
positive and significant correlation between collateral efficiency
Chart 2
Collateral Efficiency and the Dealer’s Collateral Pool
90
Collateral efficiency (percent)
85
80
75
70
1,000
1,500
2,000
2,500
3,000
Collateral received that can be repledged (collateral pool)
(billions of U.S. dollars)
Source: Company 10Q/10K filings; includes GS, MS, and ML.
and the collateral pool for these three firms. Other factors that
we conjecture would increase collateral efficiency include the
number and mix of customers, the operational capacity of the
dealer, and other economic features of the dealer firm, such as
its creditworthiness, that make it a good counterparty.
The collateral efficiency achieved by dealer banks underlies
the more expansive collateralized financing efficiencies that
pervade dealer banking. We examine this efficiency by comparing the size of the collateral pledged by dealer banks and
the size of their on-balance-sheet assets and liabilities.
Collateralized Financing Efficiency:
A Liability Perspective
Recall that we defined collateralized financing efficiency as all
the economic benefits reaped by dealer banks in their allocation
of firm and customer collateral. To provide indicators of this
efficiency, we first display the total stock of collateral pledged by
the dealer banks relative to on-balance-sheet transactions that
consume collateral, namely, their total secured liabilities. This is
consistent with our stylized balance sheet for the dealer banks,
where we focus on their secured financing activities.
The difference between the amount of collateral pledged by
the dealer bank and its level of secured on-balance-sheet liabilities
highlights the netting and other balance sheet economies that
enable dealer banks to gain collateralized financing efficiency.
This provides a measure of collateral financing efficiency.
FRBNY Economic Policy Review / December 2014
141
Chart 4
Chart 3
Secured Liabilities Relative to Collateral Pledged
3,000
Billions of U.S. dollars
Collateral Efficiency and Total Primary Dealer Repo
90
88
Collateral pledged
2,500
Percent
Billions of U.S. dollars
86
2,000
1,500
Scale
82
Payables to clients
1,000
4,000
Primary dealer
repo outstanding
84
Rehypothecated
collateral
Firm shorts
0
2007Q4
2008
2009
2010
Secured funding
2011
2012 2013Q1
3,500
3,000
80
2,500
78
2,000
76
500
5,000
4,500
1,500
Collateral efficiency
74
1,000
Scale
72
70
500
0
2005
06
07
08
09
10
11
12 13
Source: Company 10Q/10K filings; includes GS, MS, and ML.
Source: Company 10K/10Q filings, Federal Reserve Bank of
New York (FR 2004).
Notes: Collateral efficiency includes GS, MS, and ML; shaded
area highlights rapidly declining collateral efficiency.
In Chart 3, we also measure the amount of rehypothecated
collateral, recalling that this is the numerator in our measure for collateral efficiency. The chart illustrates the strong
dependency on reuse of collateral received to finance dealer
bank intermediation of cash and securities. Conceptually, in the
event that all secured borrowers of the bank were to demand
segregation of their collateral or fully restrict rehypothecation
rights, the amount of rehypothecated collateral represents the
total amount of financing that a dealer bank would need to raise
from its own collateral or from the unsecured debt market to
maintain its existing secured lending activities. The chart shows,
therefore, just how important the amount of rehypothecation is
to the dealer bank in achieving its efficiencies.
Chart 3 illustrates the trend between 2007:Q4 and 2013:Q1
of secured funding, including repo and securities lending
transactions (blue area), firm shorts14 (dark blue area), and
payables to clients (light blue area), again restricting our
purview to Goldman Sachs, Merrill Lynch, and Morgan
Stanley for comparability. The thick blue line indicates the
total amount of collateral pledged, with the thin black line
representing the portion of collateral pledged that was sourced
from other secured transactions and has been rehypothecated.
The difference between the two lines represents the amount of
the firms' own collateral pledged to secured transactions.
14
Includes securities sold and not yet purchased. Excludes on-balance-sheet
derivatives transactions, as the fair value of derivative liabilities reported
on-balance-sheet generally refers to unsecured derivatives. This introduces a
certain amount of error into our discussion of liabilities requiring collateral to
be posted, as certain derivatives are collateralized by cash or securities that do
not qualify for netting.
142
Matching Collateral Supply and Financing Demands
The importance of rehypothecation and the matching of
sources and uses of collateral are emphasized by the level of
rehypothecation relative to secured liabilities and total collateral pledged, and in normal times represents how efficient the
dealer banks are in economizing on collateral. Total secured
liabilities peaked in 2008:Q1 at just under $2 trillion, or
68 percent of the balance sheet, evidence of their materiality
as a source of dealer funding. At that time, the level of collateral that had been rehypothecated and repledged exceeded
the total secured liabilities reported on-balance-sheet by
$156 billion, which indicates a very high level of collateralized
financing efficiency.
The subsequent crisis-era period between 2007:Q4 and
2008:Q4 depicts a decline in total collateral pledged of nearly
$1.5 trillion, or a 55 percent decrease. At the same time, onbalance-sheet secured liabilities declined by a much lower
amount—$897 billion, or a 47 percent drop. In addition,
the level of collateral rehypothecation fell by $1.2 trillion, or
57 percent, over the same period. The accelerated decline of
the collateral stock pledged and the level of rehypothecation
suggest a sort of collateral scarcity that particularly affects
dealer banks. As the dealer banks’ collateral efficiency plummeted, as shown directly in Chart 4, they had to supply more
of their own collateral to secure funding as well as rely on
uncollateralized funding or increases in equity.
Why did such a precipitous drop in collateralized financing
occur during the crisis? Duffie (2013) provides a case study of
some of the factors leading to the decline by examining features of Morgan Stanley’s experience during the days following
the bankruptcy of Lehman Brothers International Holding Co.
In that case, Duffie reports that “[t]he dominant source of loss
in liquidity was through an effective run by Morgan Stanley’s
prime-brokerage clients.” As prime-brokerage clients exited
their positions with Morgan Stanley, the firm lost access to
securities those clients had posted as collateral which, because
of overcollateralization of client positions, in turn reduced the
amount of financing Morgan Stanley could raise using those
securities as collateral for its own borrowing. At the same time,
however, many clients continued to have high and immediate
demand for funding, which Morgan Stanley worked to fulfill;
denying client requests would send a very negative signal about
the firm’s ability to meet its other obligations, potentially crippling the firm. Consequently, Morgan Stanley had to rely on
other and in general more costly external sources of financing
to fulfill the demands it faced, which tended to expand its balance sheet, all else equal. These strains led to a significant drop
in the levels of collateral received and pledged by the dealer
banks and a decline in the efficiency of the activity, to which
we turn next. It is likely that the drop reflected a combination
of decreases in both demand and supply of this type of financing, as hedge funds and other clients reduced their risk profile
and cut back on risky positions that required financing, and
as dealer banks faced much higher costs of financing as they
relied on more costly external sources of financing, including
relying on higher levels of equity.
Collateralized Financing Efficiency: An Asset
Perspective
We turn next to the asset side of our stylized balance sheet and
the total stock of collateral managed by a dealer bank, including
the stock that has been pledged or encumbered as well as what
remains unencumbered and available. Assets generally reflect
a firm’s earning potential; however, a simple balance sheet represents net economic claims and, as a result, can understate the
earning potential discussed in our stylized framework, stemming from the reuse of collateral that appears in the collateral
record. Collateralized financing efficiency can be measured as
the difference between the total collateral stock as viewed from
the collateral record relative to the reported balance sheet. This
in turn signals how much gross financing the dealer bank has
extended in its activities, relative to the amount of lending it
reports on-balance-sheet. However, as discussed earlier, we
cannot adequately capture the full extent of the dealer’s collateralized financing efficiency using the data that we have because
Chart 5
Collateral Stock Relative to Total Assets
4,500
Billions of U.S. dollars
4,000
Collateral stock range
3,500
3,000
2,500
2,000
1,500
Total assets
1,000
500
0
2005
06
07
08
09
10
11
12 13
Source: Company 10K/10Q filings; includes GS, MS, and ML.
Note: Bottom line of range excludes cash (lower bound); top line
includes cash balances (upper bound).
we do not know the “opportunity cost” the dealer avoided by
not seeking funding from external markets. Thus, this estimate
of collateralized financing efficiency will inherently be an underestimate, and will not reflect the full economy involved that
would otherwise include the lower costs of internal sources of
funds as well as the economization of capital and liquidity.
To arrive at this figure, we approximate the total stock of
collateral held as the sum of the collateral received in relation to
secured transactions reported in the footnotes of firms’ SEC filings and the cash and financial instruments owned that appear
on-balance-sheet. Notably, we do not include intangible assets,
traditional loans, and certain investments (such as investments
in subsidiaries) in the total stock of collateral, as these are not
typically pledged as collateral for secured transactions.
Additionally, the linkage between the collateral record and
the balance sheet is ambiguous, in particular, as it relates to
cash. We expect that cash received in relation to derivatives or
other secured transactions would appear both on the collateral
record and the balance sheet; however, current disclosures do
not provide enough detail to distinguish the overlap. Therefore,
we approximate the potential range of total collateral received,
using a lower bound that excludes on-balance-sheet cash (unsegregated and segregated) and an upper bound that includes
all on-balance-sheet cash. Separately, and later in this article, we
examine the net receipt of cash collateral in relation to OTC derivatives. Again, we restrict our view to Goldman Sachs, Merrill
Lynch, and Morgan Stanley—the most comparable banks.
FRBNY Economic Policy Review / December 2014
143
Chart 5 illustrates the trend of total assets (blue area) and
our estimated range of the total collateral stock (blue band).
Empirically, we can observe that dealer banks tend to generate stocks of collateral in excess of their total balance sheet
during periods of stable market conditions, although this
spread contracts acutely in response to market disruption.
The stock of collateral for our sample firms peaked in 2008:Q1
at approximately $3.9 trillion–$4.2 trillion in notional terms
(116–125 percent when measured relative to total on-balancesheet assets). That is, the firms extended financing to customers in excess of their on-balance-sheet reported lending by
approximately 16–25 percent.
In 2009:Q1, the stock of total collateral fell to $2.0 trillion–$2.2 trillion, or 92–103 percent of total on-balance-sheet
assets. The drop in the total stock of collateral outpaced
balance sheet deleveraging in both notional and percentage
terms, falling $1.9 trillion–$2.0 trillion, or 47–50 percent,
versus a $1.2 trillion, or 38 percent, decline in total onbalance-sheet assets—a fact that illustrates the outsized effect
of deleveraging on the collateral record.
Balance-sheet declines understate the amount of contraction
in secured financial activity for dealer banks, and the collateralized financing efficiencies exploited by dealer banks disappear
rapidly during periods of stress. This disparity between the net
economic claims or obligations on-balance-sheet and the gross
collateral flows is an important concept, particularly when collateral sources and uses are allocated across different customers
or a customer and the dealer bank. Dealer banks, in an effort to
preserve their franchise, do not necessarily unwind positions on
a net basis. Recall that deleveraging can occur asymmetrically,
resulting in large funding gaps in the interim.
The average collateral stock as a percentage of total assets
was between 118 and 125 percent between 2005 and 2010;
this number fell in the 2010-12 period to 107–116 percent.
Two reasons for this decrease in firms’ collateral efficiency and
collateralized financing efficiency likely relate to increased
regulatory restrictions and increased risk aversion by dealer
counterparties.
First, there have been several proposed regulatory changes
that indirectly limit the levels of collateral rehypothecation.
Recent liquidity regulations, such as the proposed Liquidity
Coverage Ratio (LCR), could reduce these levels by requiring
dealer banks to hold a buffer of unencumbered high-quality
liquid assets as a reserve against short-term market and idiosyncratic liquidity risk in the future. Additionally, recent rules
from the Commodity Futures Trading Commission would
require that the execution and clearing of standardized swap
contracts be shifted to central counterparties. These changes
might decrease the stock of rehypothecatable collateral held by
dealer banks for two reasons: 1) it may disintermediate dealer
144
Matching Collateral Supply and Financing Demands
banks as market participants interact directly through the
central counterparty or exchange and 2) it will likely concentrate a pool of collateral at the central counterparty, reducing
the overall supply of collateral held by dealer banks. Proposals
for an international supplemental leverage ratio would also
indirectly limit the extent of rehypothecation, in particular, in
terms of matched-book activity. Industry commentary suggests that the proposed rules discourage the use of matchedbook repo, in particular reducing incentives for the use of
repo backed by highly liquid securities. Additionally, the
SEC has proposed placing new requirements on the segregation of collateral received in relation to centrally cleared
“security-based” swaps, which could, on net, reduce levels of
dealer rehypothecation. The Financial Stability Board and the
European Commission have also countenanced the idea of
outright constraints on rehypothecation, although these proposals are much further from tangible implementation.
Second, risk-averse dealer counterparties may be placing
increased contractual limits on collateral rehypothecation to
mitigate counterparty risk exposure.
4.3 Composition Analysis
We now attempt to decompose the stock of collateral received
in relation to secured activities into its component pieces,
linking them where possible to our stylized framework.
This view cannot fully explain the “collateralized financing
efficiency” or the extent of matched financing that takes place
between the sources and uses of collateral; however, it reflects
a perspective of all sources of collateral received, including
their relative materiality and durability. Chart 6 shows this
decomposition of the stock of collateral received, which
excludes financial instruments owned, into collateral received
from reverse repo and securities borrowing (dark blue area),
derivatives dealing (light blue area), and brokerage activities
(blue area).
Estimates of Collateral Received by Activity
For this chart, we attempt to estimate the amount of collateral received in connection with reverse repo and securities
borrowing transactions using the total amount reported on
the firms’ balance sheets. However, because of the collateralized financing efficiencies discussed earlier, the balance-sheet
number will severely underestimate the actual amount of
collateral received in such transactions, as this proxy does not
account for counterparty netting or haircuts, when lenders ask
Table 4
Chart 6
2013:Q1 Disclosures from Morgan Stanley,
Goldman Sachs, and Merrill Lynch
Collateral Received from Specific Activities
and Total Assets
3,500
Billions of U.S. dollars
Amount
Billions of
Dollars
Percent
1,702
100
Reverse repo/securities borrowing
reported on balance sheet
816
48
Reverse repo/securities borrowing
counterparty netting
under U.S. GAAP
268
16
Derivatives cash collateral netting
under U.S. GAAP
3,000
Total collateral
received
Total assets
2,500
2,000
From:
1,500
Brokerage activity
1,000
Derivatives dealing
500
Matched-book dealing
0
2008
09
10
11
12
13
184
11
Derivatives other collateral
27
2
Remaining residual amount
407
24
Source: Company 10K/10Q filings; includes GS, MS, and ML.
borrowers to pledge collateral in excess of the value of the secured transaction. This omission is sizable; as of August 2013,
median haircuts in the U.S. tri-party repo market ranged from
2 percent for U.S. government and agency securities to 8 percent for some noninvestment-grade securities.
In addition, as described earlier, U.S. GAAP allows dealer
banks to net down all secured funding transactions with a single
counterparty. Recent disclosures by dealer banks provide us with
insight into the magnitude of this divergence between the firms’
collateral record and the size of their balance sheets because
firms have begun reporting netting amounts. As of 2013:Q1,
counterparty netting reduced the amount of reverse repo and
securities borrowing reported on-balance-sheet by $268 billion,
which is approximately 16 percent of the total stock of collateral
received (Table 4). In sum, this proxy estimates all collateral
received from reverse repo and securities borrowing transactions,
a portion of which will be delivered into repo and securities lending transactions to form a “matched book.”
Additionally, the amount of collateral received in connection with derivatives transactions is reported directly by our
sample of dealer banks, but this number only includes cash
collateral received that qualifies for netting treatment under
U.S. GAAP. This proxy will also underestimate the actual
collateral received from derivatives, since dealers receive cash
collateral that does not qualify for netting as well as noncash
collateral. Again, more recent disclosures from our sample
of banks show that about $27 billion of other collateral is
received in connection with derivatives, or about 2 percent of
the total stock of collateral received (see Table 4). A section
below furthers expands upon cash collateral received from
derivatives transactions.
The amount of collateral received from brokerage activity
is estimated as the total amount of collateral received less the
amounts attributed to matched-book dealing and derivatives.
This will include collateral received from margin lending, of
which a portion was likely delivered into customer or firm
short positions, or “internalized.” However, this residual
balance will also include the amount of residual error resulting from the proxies used above, or specifically the amount
of counterparty netting for repo/reverse repo transactions,
aggregate repo haircuts, and the amount of cash and noncash
collateral received from derivatives that does not qualify for
netting under U.S. GAAP. This is reported in Table 4 to amount
to $407 billion, or 24 percent of the total collateral received by
the dealer banks, indicating that margin and securities lending
and other brokerage activities are a powerful source of collateralized financing efficiency for the dealer banks.
Prior to the crisis, brokerage activities contributed the
majority of collateral to the total stock received, with reverse
repo and securities borrowing contributing a close second (see
Chart 6). At the height of the crisis and around the point of
Lehman’s failure, our sample of dealer banks were all deleveraging at a breakneck pace. As observed in Charts 3 and 5, this
entailed major reductions in assets and wholesale secured
liabilities. We highlight above how this had an outsized effect
on the stock of collateral as reported on the collateral record.
Chart 6 attempts to attribute that contraction to the various
FRBNY Economic Policy Review / December 2014
145
Chart 7
Net Derivatives Cash Collateral Position
140
Billions of U.S. dollars
120
100
GS
80
60
MS
JPM
40
20
0
-20
-40
2007Q4
C
ML
08
09
10
11
12
13
Source: Company 10K/10Q filings.
activities of the dealer banks. Gorton and Metrick (2012) and
Duffie (2011) discuss the runs occurring in dealer banks, with
Gorton and Metrick focusing specifically on repo runs. Our
analysis suggests that there may also have been runs affecting
margin and securities lending, and brokerage services, in
addition to those on repos.
Based on our approximations, the collateral received in
relation to brokerage activities collapsed dramatically in a
remarkably short time frame, accounting for most of the
aggregate decline. This may suggest that prime brokerage
clients were selling assets or closing out short positions in
an attempt to withdraw funds and limit credit exposure to
the dealer banks. At its most severe, this likely would result
in a significant loss of internalization, as discussed through
our stylized framework. At a minimum, it implies that the
collateral financing efficiencies obtained by dealer banks are
fleeting, and that they can disappear during a period of severe
market disruption.
Cash Collateral in Relation to OTC Derivatives
Cash collateral received in relation to derivatives transactions exceeds that posted for most of the dealer banks in our
sample. For derivatives collateral, we can now expand our
sample of firms to include Citigroup and J.P. Morgan Chase, as
the reporting conventions are more homogenous among the
five banks. In Chart 7, we see that the dealer banks, especially
Goldman Sachs, typically receive in cash much more from
derivatives counterparties than the amount they post to their
146
Matching Collateral Supply and Financing Demands
counterparties. This excess cash usually earns a short-term
money market rate for the counterparty that posted it, and so
to the extent that it can be rehypothecated, represents a very
low-cost source of funding for the dealer banks.15 Chart 7
shows that derivatives cash collateral is a significant source of
low-cost funding for the largest OTC derivatives dealer banks.
The time series of net derivatives cash collateral position
shows a trend that is significantly at odds with the reductions
in collateral received with the other dealer bank activities
during the financial crisis. In Chart 7, we see that during the
crisis period, the levels of net funding generated from derivatives activities actually increased. A closer look at individual
firms reveals that Goldman Sachs was by far the primary
beneficiary, but that Morgan Stanley and J.P. Morgan also
benefited. While this is at first blush surprising, it reflects a
broad shift toward increased collateralization in the wake of
the failure of Lehman Brothers, and the near-failure of AIG
may underlie these figures. In other words, the prior deficiencies in collateralization of derivatives positions may have been
corrected as the crisis developed.
Performance under Severe Duress: A Lehman
Brothers Case Study
The asset and liability perspectives both highlight how the
financing and collateral efficiency and the gross stock of rehypothecatable collateral evaporate during a period of market
stress; however, in this case Lehman serves as a more targeted
example. Although data limitations prevent a similar asset and
liability analysis, we can observe changes in Lehman’s levels
of collateral pledged, collateral received, and the percentage
of collateral received that had been rehypothecated. Depicted
in Chart 8, all levels fell dramatically between 2008:Q1 and
2008:Q2, presumably surrounding the market turmoil related
to J.P. Morgan’s acquisition of Bear Stearns. Of note, these
levels declined well in advance of Lehman’s actual failure, with
the last disclosure as of May 31, 2008. Additionally, Lehman’s
levels of rehypothecation were elevated relative to the rest
of our sample at 92 percent in 2008:Q1 versus an average of
83 percent, which may have contributed to its downward
spiral of deleveraging and asset fire sales.
15
As discussed above, it is not fully clear in the financial reporting of the
dealer banks whether cash is included or excluded from the collateral record.
As a result, in the collateral record it is not clear if cash collateral is always
included. That is why we provide the range of possible collateral in Chart 3,
in one case excluding from the collateral record the derivatives cash collateral
figures reported below. Consequently, one should not necessarily conclude
that the net derivatives cash position reported in Chart 7 is an additional
collateralized financing efficiency, over and above that reported in Chart 3.
Chart 8
Lehman and the Collateral Record in the Run-Up
to Collapse
1,200
Billions of U.S. dollars
Percent
Rehypothecated collateral
95
Scale
1,000
90
800
85
Collateral pledged
600
80
Collateral received
400
2005
06
07
08
75
Source: Company 10K/10Q filings.
5. Conclusion
The economies of the activities undertaken by dealer banks
relate intrinsically to the way these banks source and use
collateral. In this article, we describe three types of activities—matched-book financing, internalization, and derivatives collateral received in excess of posted—that allow dealer
banks to reap efficiencies by reusing collateral provided by
customers. Additionally, we discuss how netting accounting
rules, excess collateralization, cheaper internal sources of cash
and securities, and other collateral efficiencies allow them to
finance customer demands in excess of their own liabilities.
We attempt to measure these sources using publicly disclosed
data in 10-Q and 10-K filings, illustrating how these sources
of financing have evolved over time, including during the
financial crisis of 2007-09. The data reveal that, while efficient
in normal times, such financing drastically and abruptly dries
up during times of financial stress.
In particular, we describe two types of efficiencies gained by
dealer banks: collateral efficiency and collateralized financing
efficiency. First, dealer banks realize “collateral efficiencies”
by rehypothecating collateral they have received from their
customers. This ability to rehypothecate collateral allows them
to “internalize” their sources of collateral and cash, finding uses
for them among their other customers, or for their own trading.
Collateral efficiency is likely related to the scale of the dealer
bank’s activity and the distribution of securities pledged as
collateral by its customers. Second, dealer banks reap “collater-
alized financing efficiencies,” which allow them to engage in a
larger amount of collateralized lending than is reported on their
balance sheets. A dealer bank’s collateralized financing efficiency is related to the amount of netting allowed by U.S. and
international accounting standards; the accounting treatment of
brokerage activities, such as shorts; the differential between the
cost of internal sources of funding and external ones; and the
fees/income earned on lending activities. To determine the level
of a firm’s collateralized financing efficiency, an analyst must
consult the collateral record of the banks, which is embedded
within the text of firms’ 10-Q and 10-K reports.
Unsurprisingly, we find that the experience of the financial
crisis was especially troubling for dealer banks. The collateral
they had received from customers disappeared when customers
exited positions that the dealer bank had financed. Because
dealer banks had heavily utilized the customer-provided collateral, they were forced to source collateral and cash externally to
manage and meet their obligations at the same time that markets were most disturbed. Notably, the dealer banks’ brokerage
receivables were most affected by the crisis, plummeting significantly more than the firms’ other sources of collateral and much
more than the balance sheet assets of the firms. This likely
is the result of the significant moves in markets, including the
equity markets, which at the height of the crisis led customers
to exit leveraged bets (such as margin loans) on those markets
as quickly as possible. The dealer banks were heavily exposed
to this source of risk in their financing profile. In contrast, the
dealer banks received more collateral in connection with derivatives during and after the financial crisis. This likely reflects a
widespread undercollateralization of derivative positions prior
to the crisis, as well as a renewed focus on counterparty credit
risk during the crisis as many dealer bank counterparties experienced credit rating downgrades.
Our observations raise the question of whether the risk of
dealer financing, which is more comprehensively, although
still imperfectly, reflected in a bank's collateral record than in
its balance sheet, is managed appropriately. That the amount
of financing extended by dealer banks, as measured by the collateral record, fell further and more swiftly than the amount
measured by the banks’ balance sheets suggests that a prudent
risk management framework would acknowledge the risks
inherent in collateralized finance, and allocate both capital
and liquidity to be available to address any shortfalls that
would arise in a risk event. Our observations reflect the fact
that reputational and other economic considerations provide
incentives to dealer banks to roll over one side of a customer’s
trade, while the other side is extinguished, which brings the
exposure on-balance-sheet. Accounting netting in this case
does not reflect true economic netting of risk exposures. This
line of reasoning leads us to suggest that the recent consulta-
FRBNY Economic Policy Review / December 2014
147
tive document of the Basel Committee on Bank Supervision
(2013) that outlined a revision to the Basel III leverage ratio
framework which, when measuring securities financing transactions, excludes any recognition of accounting netting, may
be warranted as a measurement approach.16 We do not advocate that a binding leverage requirement for capital should
be applied, as this would essentially equalize the risk weights
for different types of risk exposures, opening the window for
dealer banks to increase the risk of their positions while not
increasing their required regulatory capital.17 However, some
capital and liquidity charge (as, for example, is the case with
the Liquidity Coverage Ratio) for financing transactions that
are currently subject to accounting netting treatment, and are
therefore off-balance-sheet, does seem warranted.
Our measures of collateral and collateralized financing efficiencies have declined in the aftermath of the financial crisis.
16
Note that this is also the approach taken by the Liquidity Coverage Ratio,
which requires the reporting of gross contractual obligations for secured
transactions that mature within the thirty-day period. This is impactful for
the collateral swap/optimization trades we discuss in the context of matched
book, which allow the dealer to transform its collateral profile without
expanding its balance sheet. Under circumstances where dealers exchange
less liquid collateral for highly liquid collateral, they must hold liquidity in an
amount equivalent to the difference in stressed run-off rates applied to each
class of collateral. Furthermore, in case of cash brokerage internalization, the
Liquidity Coverage Ratio requires liquidity to be held for instances where
offsetting customer-to-customer or customer-to-firm exposures are used to
finance one another.
17
Darrell Duffie makes this point most clearly in his October 13, 2013, Brookings
Institution presentation, “Capital Requirements with Robust Risk Weights.”
148
Matching Collateral Supply and Financing Demands
That trend likely reflects some greater regulatory limitations
on collateral rehypothecation, and some greater restrictions
put in place by customers on the reuse of their collateral.
Nonetheless, the size and importance of the financing and
collateral efficiencies we describe in this study remain large for
the dealer banks.
It is important to acknowledge the limitations of the data
used in this study, which reflect the inadequate reporting
requirements for collateral used by dealer banks. More regular, frequent, and standardized public disclosures on asset
encumbrance—including the level of unencumbered assets
relative to unsecured liabilities, overcollateralization levels,
and received collateral that can be rehypothecated—would
allow for more reliable measurements of these activities. Such
data could provide a fuller picture of the financial condition
and vulnerabilities of dealer banks.
Data Appendix
In our study, we choose to focus on data from firms’
10-Q/10-K filings to examine dealer banks’ financing transactions. The 10-Q balance sheet data are firm specific and
consolidated at the bank holding company level, which
include U.S. and U.K. broker-dealer subsidiaries. However, as
we discuss in the study, using 10-Q/10-K data restricts which
firms we can analyze; of the major OTC derivatives dealers,
only three institutions report their balance sheet and collateral
record in a consistent way.
There are other data sources that potentially offer similar
insights, namely, the Federal Reserve Board’s Flow of Funds
data and the Federal Reserve Bank of New York’s data on
primary dealers. In this appendix, we describe the advantages
and disadvantages of these alternative data sources and offer a
robustness check for some of our main findings.
The Flow of Funds data aggregate all broker-dealer quarterly
balance sheets from their FOCUS regulatory submissions to the
SEC and provide industry-wide data on those activities of dealer
banks. However, the Flow of Funds data are inadequate for our
analysis in four respects. First, U.K. broker-dealer subsidiaries
of U.S. bank holding companies are not included in the Flow
of Funds data. This omission could have a sizable effect on the
balance-sheet data since the U.K. dealer banks are particularly
large prime brokers and are not bound by SEC rules 15c3-3, allowing them to rehypothecate securities to a larger extent than
their U.S. counterparts. Second, the Flow of Funds only reports
a combined number for repurchase agreements and federal
funds, and this number represents a “net” amount, that is, total
fed funds and repo borrowing less fed funds and repo lending.
Third, the Flow of Funds does not offer data on firms’ collateral
sources and uses, such as collateral received or pledged by the
dealer. Fourth, the Flow of Funds does not report firm shorts,
that is, securities sold but not yet purchased. As a result, we
can only proxy one component of our stylized framework—the
stylized balance sheet—using the Flow of Funds, and we cannot
isolate all the components of dealer banks’ secured funding or
proxy the sources of dealer banks’ collateral received.
The Federal Reserve Bank of New York provides at a
weekly frequency aggregate data on primary dealers’ positions, financing, and settlement activities across asset classes,
collected from their FR 2004 regulatory filings. Financing data
are split into two categories that together represent a view of
the collateral record: securities received as collateral by the
dealer from its counterparties (“securities in”) and securities
pledged by the dealer as collateral (“securities out”). This
includes collateral received and pledged in connection with
securities lending, repurchase agreements, and margin loans,
Chart a1
Collateral Received Compared with
Total Financial Assets
6,000
Billions of U.S. dollars
5,000
4,000
Collateral stock
3,000
2,000
Collateral received
by primary dealers
1,000
0
Total financial assets
2004
05
06
07
08
09
10
11
12 13
Sources: FR 2004 Financing Data for Primary Dealers; Flow of
Funds Data for Securities Brokers and Dealers.
and is reported on a gross basis. Additionally, the FR 2004
data report the portion of securities pledged and received in
connection with repos and reverse repos, which is a useful
proxy for dealers’ matched-book transactions. However, these
data exclude collateral received and pledged in connection
with derivatives activity and do not distinguish between
firms’ own collateral that was pledged and collateral received
that the dealer rehypothecated. As a result, the FR 2004 data
underestimate collateral pledged/received and preclude any
measure of dealers’ “collateral efficiency” as we have defined
it. Additionally, like the Flow of Funds data, the FR 2004 data
exclude financing activities of U.K. broker-dealer subsidiaries
of U.S. bank holding companies.
Though these data are limited for the reasons described
above, they do allow for a rough check for some of our main
conclusions. We combine Flow of Funds balance sheet data
for U.S. broker-dealers with FR 2004 data describing the
primary dealers’ quarter-end collateral record to obtain a
time-consistent series.18
Chart A1 confirms that dealers generate stocks of collateral
in excess of their balance sheet (see the similarity to Chart 5
18
We recognize that these data sources cover different populations and,
unlike the sources we use in the study, do not cover collateral associated with
derivatives.
FRBNY Economic Policy Review / December 2014
149
Data Appendix (continued)
in our study). We proxy the collateral stock by taking the sum
of the primary dealers’ “Securities In” (from the FR 2004)
and dealer banks’ credit market instruments owned (from
the Flow of Funds); here, we measure the balance sheet using
broker-dealers’ total financial assets (from the Flow of Funds).
The drop in the collateral stock greatly exceeds that of the
balance sheet during the crisis, suggesting that balance-sheet
declines do not fully reflect the reduced provisioning of secured financing by the dealers. The collateral stock peaked at
$5.1 trillion in 2008:Q1 and subsequently fell 46 percent (or
$2.3 trillion) to a trough of $2.7 trillion in 2009:Q3. In contrast, the balance sheet dropped from $3.2 trillion in 2008:Q1
to $2 trillion in 2009:Q3, representing a fall of 36 percent (or
$1.2 trillion). In sum, these data indicate that dealers’ collateralized financing efficiencies can vanish precipitously during
periods of market disruption.
It is important to note that these two data sources represent different samples, with the Flow of Funds representing
all U.S. broker-dealers and the FR 2004 data representing just
primary dealers, a subset of the total industry. This means our
measure of the dealers’ balance sheet encompasses the entire
(domestic) broker-dealer industry, while our measure of the
collateral stock only includes the primary dealers. It is likely
that the collateralized financing activities are concentrated at
the largest broker-dealers, and so the primary dealers’ collateral record could be expected to represent the vast majority
of all dealer banks’ collateral received. That said, our measure
of the collateral stock would tend to underestimate the total
for all broker-dealers, meaning that the level of the collateral
stock for all broker-dealers would be even higher and thus
more in excess of the broker-dealers’ aggregate balance sheet.
This result, then, would likely be even stronger with aggregate
collateral record data from the full broker-dealer industry.
Chart A2 plots the composition of the collateral received
by primary dealers as is reported in the FR 2004. These data
confirm the significant contraction in matched-book and other
150
Matching Collateral Supply and Financing Demands
Chart A2
Collateral Received by Primary Dealers
(“Securities In”)
Billions of U.S. dollars
5,000
4,000
Collateral sourced from
other financing activities
3,000
2,000
Collateral sourced
from reverse repos
1,000
0
2008
09
10
11
12
13
Source: FR 2004 Financing Data for Primary Dealers.
sources. Additionally, and perhaps more importantly, “other
sources” declined as a percentage of total collateral received.
These observations are consistent with the firm-specific data
from the SEC disclosures we present in Chart 6 in the study,
and indicate that both brokerage activities and matched-book
dealing were significant sources of dealer collateral precrisis,
and that both of these activities plunged dramatically during
the crisis, along with total collateral received. In contrast to
Chart 6, Chart A2 attributes a larger portion of total collateral
received to matched book. This could be a result of accounting
idiosyncrasies of the FR 2004; for example, primary dealers
may have included collateral received in connection with
other secured transactions (such as margin loans) with that
sourced from reverse repos in their FR 2004 reports.
References
Adrian T., B. Begalle, A. Copeland, and A. Martin. 2011. “Repo and
Securities Lending.” Federal Reserve Bank of New York Staff
Reports, no. 529, December.
———. 2013. “Replumbing Our Financial System—Uneven Progress.”
International Journal of Central Banking 9, supplement 1:
251-80. Available at http://www.ijcb.org/journal/ijcb13q0a12.pdf.
Basel Committee on Banking Supervision. 2013. “Revised Basel
III Leverage Ratio Framework and Disclosure Requirements.”
Consultative document.
Gorton, G., and A. Metrick. 2012. “Securitized Banking and the Run
on Repo.” Journal of Financial Economics 104, no. 3: 425-51.
Basel Committee on Banking Supervision and International
Organization of Securities Commissions. 2013. “Margin
Requirements for Non-Centrally Cleared Derivatives.”
Consultative document.
Committee on the Global Financial System. 2013. “Asset Encumbrance,
Financial Reform, and the Demand for Collateral Assets.”
Copeland, A., A. Martin, and M. Walker. 2010. “The Tri-Party Repo
Market before the 2010 Reforms.” Federal Reserve Bank of New
York Staff Reports, no. 477, November.
Duffie, D. 2010. “The Failure Mechanics of Dealer Banks.” Journal
of Economic Perspectives 24: 51-72.
International Swaps and Derivatives Association. 2012. “Netting and
Offsetting: Reporting Derivatives under U.S. GAAP and under IFRS.”
King, M., 2008. Are the Brokers Broken? New York: Citibank
Global Markets Group.
Singh, M., and J. Aitken. 2009. “Deleveraging after Lehman—
Evidence from Reduced Rehypothecation.” IMF Working Paper
no. 09/42, March.
Squam Lake Working Group on Financial Regulation. 2010. “Prime
Brokers and Derivatives Dealers.” Council on Foreign Relations
working paper.
Stigum, M., and A. Crescenzi. 2007. The Money Market. 4th ed.
New York: McGraw Hill.
———. 2011. How Big Banks Fail: And What To Do About It.
Princeton, N.J.: Princeton University Press.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2014
151
Phoebe White and Tanju Yorulmazer
Bank Resolution Concepts,
Trade-offs, and Changes
in Practices
• As the 2007-08 financial crisis demonstrated,
the failure or near-failure of banks entails
heavy costs for customers, the financial
sector, and the overall economy.
• Methods used to resolve failing banks range
from private-sector solutions such as mergers
and acquisitions to recapitalization through
the use of public funds.
• The feasibility and cost of these methods
will depend on whether the bank failure is
idiosyncratic or part of a systemic crisis, and
on factors such as the size, complexity, and
interconnectedness of the institution in distress.
• This study proposes a simple analytical
framework—useful to firms and regulators
alike—for assessing these issues and
determining the optimal resolution policy
in the case of particular bank failures.
Phoebe White is a former research associate and Tanju Yorulmazer a former
research officer at the Federal Reserve Bank of New York.
Correspondence: research.publications@ny.frb.org
1. Introduction
During the recent crisis, some of the world’s largest and
most prominent financial institutions failed or nearly failed,
requiring intervention and assistance from regulators. Measures
included extended access to lender-of-last-resort facilities, debt
guarantees, and injection of capital to mitigate the distress.1
Chart 1 shows some of the largest financial institutions
that failed and/or received government support during
the recent crisis. As we can see, these institutions were
large and systemically important. For example, for a brief
period in 2009, Royal Bank of Scotland (RBS) was the
largest company by both assets and liabilities in the world.
Table 1 summarizes the interventions and resolutions of
major financial institutions that experienced difficulties
during the recent crisis. The chart and the table indicate
the extraordinary levels of distress throughout the system
and the unprecedented range of actions taken by resolution
1
For a discussion of the disruptions and the policy responses during the
recent crisis, see Yorulmazer (2014).
The authors thank Hamid Mehran and seminar participants at the Federal
Reserve Bank of New York and the Dutch National Bank for very helpful
comments. The views expressed in this article are those of the authors and
do not necessarily represent the position of the Federal Reserve Bank of
New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
153
Chart 1
Some of the Largest Institutions that Failed and/or Received Government Intervention
during the Recent Crisis
4,000
Assets, in billions of dollars
3,744.5
3,500
3,000
2,500
2,000
1,705.3
896.3
ng
le
y
rti
s
Bi
&
or
d
ad
f
Br
nc
e
&
Ro
104.2
130.7
Fo
153.7
Le
ic
es
te
r
ck
227.4
lia
ut
ua
l
te
ar
n
W
as
h
Be
ar
S
an
Le
hm
s
Br
ot
he
rs
De
xi
a
G
AB
N
AI
Am
ro
S
BO
H
Ll
oy
ds
S
RB
309.7
Al
399.0
or
th
er
n
639.4
500
0
M
1,049.9
N
1,000
gt
on
1,302.2
in
1,359.0
1,500
Source: Public filings as of period before resolution.
authorities, since many countries lacked an efficient
framework for resolving large and systemically important
financial institutions (SIFIs).
In the United States, prior to the passage of the
Dodd‑Frank Wall Street Reform and Consumer Protection
Act, insolvent nondeposit-taking institutions were dealt
with under the Bankruptcy Code, as opposed to the special
resolution regime administered by the Federal Deposit
Insurance Corporation (FDIC). Chart 2 shows the largest
corporate bankruptcies in U.S. history; Lehman Brothers was
by far the greatest. In the absence of an orderly resolution
regime, the failure of Lehman led to unprecedented
disruptions in financial systems globally. While many
counterparties to Lehman suffered direct losses, others
experienced distress owing to information contagion and
fire‑sale externalities from a sell-off in assets.
One of the most significant effects was on the money
market mutual fund industry, where the Reserve Primary
Fund, the oldest money market fund, “broke the buck”
because of its exposure to Lehman Brothers debt securities
and had to be liquidated, marking only the second such
episode in history. This event led to a run on the money
154 Bank Resolution Concepts, Trade-offs, and Changes in Practices
market mutual fund industry, a development that adversely
affected the shadow banking industry.2 Regulators attempted
to contain the disruptions in financial markets with
extraordinary interventions including capital injections,
debt guarantee programs, and many lending facilities.
Financial intermediaries and banks perform important
roles for the efficient functioning of the economy, such as
channeling funds from savers to investors and providing
payment services, and their liquid liabilities can act as money.
As a result, failure of these institutions can pose significant
disruptions, and corporate bankruptcy may not be the
2
On September 19, 2008, the Federal Reserve announced the institution of
the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity
Facility (AMLF). The AMLF provided nonrecourse loans to commercial
banks to purchase eligible asset-backed commercial paper from money
market mutual funds (MMFs). The U.S. Treasury also provided a temporary
guarantee on the share price of MMFs through the Temporary Guarantee
Program for Money Market Funds and the Federal Reserve announced
another lending program, the Money Market Investor Funding Facility
(MMIFF), as a complement to the AMLF intended to provide nonrecourse
loans to money market funds. However, no loans were made under the
MMIFF. The facility was closed on October 30, 2009.
Table 1
Major Interventions and Resolutions during the Recent Financial Crisis
Institution
Date
Resolution Method/Support
ABN Amro
October 2007
The private acquisition by a consortium consisting of the Royal Bank of Scotland (RBS), Banco Santander,
and Fortis marked the largest worldwide acquisition of a bank and the second largest European
cross-border transaction. When Fortis and RBS ran into trouble, their holdings of ABN Amro’s assets
were nationalized by the Dutch and U.K. governments, respectively.
ING Group
October 2008
Received a €10 billion capital injection from the Dutch government in exchange for securities and veto
rights on major operational changes and investments. The injection was also conditional on ING
divesting certain operations.
Fortis
September 2008
to May 2009
The Netherlands, Belgium, and Luxembourg provided a capital injection of €11.2 billion on
September 28, 2008, each taking a 49 percent stake in Fortis’s banking operations in their respective
countries. Fortis was sold in parts, with a majority stake sold to BNP Paribas on May 13, 2009.
Dexia
September 2008
Dexia was recapitalized by the French and Belgian governments through a capital injection of €3 billion,
and it received a state guarantee in order to regain access to wholesale funding markets.
Northern Rock
September 2007
to February 2008
In September 2007, the Bank of England provided a liquidity support facility and government guarantee
of certain liabilities. In February 2008, the bank was nationalized by the British government.
Alliance & Leicester
July 2008
Private acquisition by Banco Santander for £1.26 billion
Bradford & Bingley
September 2008
The U.K. government nationalized the institution on September 29, 2009, selling the savings unit and
branches to Banco Santander.
HBOS
September 2008
to January 2009
The terms of a takeover by Lloyds TSB were agreed to in September 2008. In October 2008, the
U.K. Treasury injected new capital amounting to £17 billion, or a 43 percent equity stake in the
combined Lloyds TSB and HBOS. In January 2009, HBOS was acquired by Lloyds TSB.
UBS
December 2007
to October 2008
In December 2007, the bank received a capital injection from the Government of Singapore Investment
Corporation. In October 2008, UBS sold CHF 60 billion of its troubled assets to a special purpose vehicle
acting as a "bad bank" entity, a transaction that was funded by a CHF 6 billion capital injection from the
Swiss government and a CHF 54 billion loan from the Swiss National Bank.
Anglo Irish Bank
January 2009
Nationalized when the Irish government determined that recapitalization would not be enough to save the bank.
Allied Irish Bank
February 2009
Received capital injection of €3.5 billion
Bank of Ireland
February 2009
Received capital injection of €3.5 billion
Bankia SA
May 2012
Bank was partly nationalized through a €19 billion recapitalization by Spain.
Bear Stearns
March 2008
The bank was sold to JPMorgan Chase with assistance from the Federal Reserve in the form
of a nonrecourse loan of $29 billion.
Lehman Brothers
September 2008
Lehman filed for chapter 11 bankruptcy. It was the largest bankruptcy filing in U.S. history.
AIG
September to
November 2008
On September 16, 2008, the Federal Reserve extended a credit facility of $85 billion, secured by stock in the
form of warrants for a 79.9 percent equity stake. The loan was restructured in November in coordination
with the U.S. Treasury, which extended the facility and lowered its rate. AIG also received $40 billion in a
capital injection under the Troubled Asset Relief Program (TARP).
Washington Mutual
September 2008
On September 25, 2008, Washington Mutual was seized by the Office of Thrift Supervision and placed in
receivership with the Federal Deposit Insurance Corporation. The banking subsidiaries were sold through
purchase and assumption to JPMorgan Chase, while the holding company filed for chapter 11 bankruptcy.
Citigroup Incorporated
October 2008 to
January 2009
Received two capital injections through TARP: $25 billion in October 2008 and an additional $20 billion in
January 2009. Also in January 2009, Citigroup separated its core and noncore assets in a good bank–bad bank
split (Citicorp and Citi Holdings).
Wells Fargo & Company
October 2008
Received $25 billion capital injection under TARP
State Street Corporation
October 2008
Received $2 billion capital injection under TARP
Bank of America
Corporation
October 2008 to
January 2009
Received two capital injections through TARP: $25 billion in October 2008 and an additional
$20 billion in January 2009
JPMorgan Chase & Company
October 2008
Received a $25 billion capital injection under TARP
Morgan Stanley
October 2008
Received $10 billion capital injection under TARP
Goldman Sachs Group
October 2008
Received $10 billion capital injection under TARP
Bank of New York Mellon
October 2008
Received $3 billion capital injection under TARP
Wachovia
September 2008
The Federal Reserve provided Citigroup with liquidity to aid in purchase of Wachovia.
Ultimately, the bank was acquired by Wells Fargo.
FRBNY Economic Policy Review / December 2014
155
Chart 2
Largest Public Company Bankruptcy Filings
1980–Present
Assets, in billions of dollars
800
691.1
700
600
500
400
300
200
103.9 91.1 80.5 65.5 61.4 40.5 39.3 36.5 36.2
100
Enro
n (20
01)
Cons
eco (
2002
MF G
)
lobal
Hold
ings
(2011
)
Chry
sler (
2009
Thor
)
nbur
g Mo
rtgag
e (20
09)
Pacifi
c Ga
s&E
lectr
ic (20
01)
(2009
)
CIT G
roup
otors
(2009
)
2002
)
ral M
om (
World
C
Gene
Lehm
an B
rothe
rs
(2008
)
0
Source: BankruptcyData.com.
appropriate resolution regime.3 Hence, authorities use various
methods to resolve failed banks, ranging from full or partial
private-sector resolution methods, such as the sale of a bank to
a healthy bank via merger and acquisition (M&A), the transfer
or sale of all or parts of the assets and liabilities to another
bank via purchase and assumption (P&A), or government
intervention using public funds to recapitalize banks.
This paper provides a discussion of the costs associated
with different resolution methods. Furthermore, we provide
a simple framework to analyze the optimality of resolution
methods. We show that private resolution methods, such as
M&A and P&A, are preferred options since they minimize the
costs associated with bank failures and their resolution.
The availability of resolution options depends on the
characteristics of the failed bank. For example, when the
losses in the failed bank are large, there may not be a ready
buyer for the bank without assistance. Furthermore, if the
failed bank is large and complex or if failure occurs during
3
Section 3 provides a discussion of the resolution methods used by
authorities. DeYoung, Kowalik, and Reidhill (2013) highlight the importance
of resolution technologies showing that the limited set of failed bank
resolution technologies can leave regulators with little choice but to bail
out systemically important banks.
156 Bank Resolution Concepts, Trade-offs, and Changes in Practices
a systemic crisis that causes many banks to experience
distress, it may not be feasible to find a healthy bank to
acquire the failed bank, and the regulators may need to
employ alternative resolution methods such as liquidation or
recapitalization. In this case, resolution is more challenging
since it entails trade-offs between disruptions arising from a
disorderly liquidation and the fiscal costs and moral hazard
resulting from using public funds for recapitalization. Hence,
regulators need to employ a “state-contingent” resolution
policy that depends on whether failure occurs in an
idiosyncratic failure state or in a systemic-crisis state.
Empirical evidence on the timing of bank failures suggests
that failures are not uniformly distributed over time; instead,
they are clustered. So when banks fail, they tend to fail
together around the same time. Charts 3 and 4 show the
number of failed banks in the United States and the size of
their assets and deposits, respectively.
The pattern of bank-failure clustering in systemic crises
makes the resolution of failed banks more challenging for
authorities, since in such states of the world, the availability of
preferred resolution options is limited, which is the primary
theme of the article.
The article is organized as follows: Section 2 discusses
the corporate bankruptcy regimes in the United States and
the costs associated with bank failures and their resolution.
Section 3 examines the resolution methods used by
authorities. Section 4 discusses the trade-offs associated
with resolution of failed banks and provides an analytical
framework to develop an optimal resolution regime, which
would depend not only on the failed institution itself, but
also on its macro environment. Section 5 reviews recent
steps taken by authorities to improve resolution regimes, and
section 6 presents concluding remarks.
2. Bankruptcy Regimes and
Costs of Bank Failures
In this section, we provide a brief summary of the corporate
bankruptcy regime in the United States. Many aspects of
corporate insolvency proceedings have proved problematic in
the case of a bank failure, which we address in the subsequent
discussion of costs.
Bankruptcy can be initiated voluntarily by the debtor
or involuntarily by the petitions of creditors whose claims
are in default. The initiation of the process automatically
prevents (or “stays”) creditors from collecting on their
claims, therefore providing the bankruptcy court with time
for review. Importantly, all creditors have “standing” to be
Chart 3
Number of Bank Failures in the United States
Number of failures
600
500
400
300
200
100
0
1980
82
84
86
88
90
92
94
96
98
2000
02
04
06
08
10
12
94
96
98
2000
02
04
06
08
10
12
Source: Federal Deposit Insurance Corporation.
Chart 4
Assets of Failed Banks in the United States
Billions of dollars
2,500
Total deposits
Total assets
2,000
1,500
1,000
500
0
1980
82
84
86
88
90
92
Source: Federal Deposit Insurance Corporation.
represented in the proceedings, and often their consent is
required in a number of areas.
In the United States, two common forms of bankruptcy
are Chapter 7 liquidation and Chapter 11 reorganization. In
Chapter 7 liquidation, the firm is taken over by a receiver
who liquidates the assets and distributes the proceeds to the
creditors. Alternatively, in Chapter 11 reorganization, the
firm’s management typically acts as trustee and leads the
creation of the reorganization plan, which must ultimately
be approved by the creditors; otherwise, the parties can seek
an alternative plan under a newly appointed trustee. The
creditors are typically paid in securities of the reorganized
firm. Furthermore, during the reorganization proceedings,
the firm can arrange for debtor-in-possession (DIP)
financing to continue operations.
In Chapter 7 liquidation, bankruptcy courts usually
adhere to the priority schedule of claims, with secured
creditors experiencing higher recovery rates on their
claims than unsecured creditors. The priority of claims
is more likely to be renegotiated, however, in the case of
Chapter 11 reorganization.
Resolving a failed bank through general insolvency
proceedings is difficult for a number of reasons. First, banks
are characterized by significant financial fragility owing
to their unique structure. Their liabilities are primarily
composed of liquid deposits, redeemable at par, whereas
their assets are usually long-term loans which are often
illiquid. Bank assets are also typically less transparent,
which would make DIP financing expensive or unattainable.
Furthermore, as banks perform essential roles in the
FRBNY Economic Policy Review / December 2014
157
functioning of financial markets and the economy, their
failures can have considerable costs and externalities. Thus,
the primary objective of a resolution regime should be to
minimize these costs.4 Prompt action, as opposed to the
delayed and lengthy administrative bankruptcy process, is
important for resolving these institutions effectively while
maintaining public confidence.
Next, we explore in detail the costs associated with bank
failures and their resolution. We put these costs into four
broad categories: disruptions to the customers of the bank,
disruptions to other financial institutions through contagion,
fiscal costs associated with the resolution of failed banks, and
distorted incentives and moral hazard.
2.1 Disruptions to the Failed
Bank’s Customers
On the asset side, banks have loans through which they
channel funds from savers to the firms that invest in
profitable projects. Firms that use bank financing and have
an established relationship with their bank may find it
difficult and costly to find other sources of financing when
their bank fails.5 On the liability side, banks have liquid
liabilities that act as money. Therefore, a bank’s failure can
disrupt payment services for the depositors and creditors,
resulting in significant welfare losses (Kahn and Santos 2005;
Gorton and Huang 2004, 2006).
(Allen and Gale 2000).6 Furthermore, these losses can create
distress for the affected institutions and may lead to their
failure, resulting in knock-on effects and further rounds of
failures and potential system-wide distress.
Another important channel through which a financial
institution’s difficulties can affect other institutions is created
by information contagion, which occurs when creditors of
other banks perceive the institution’s difficulties as a negative
signal about the health of their own bank (Chen 1999;
Acharya and Yorulmazer 2008). While such actions can be a
rational response of creditors, they can lead to “wrong runs”
where even healthy institutions can experience a creditor
run.7 Such runs are more likely when financial institutions are
opaque and when creditors do not have detailed information
about the health of their financial institution.
As more prominently observed during the recent crisis,
contagion can also arise through fire-sale externalities, where
the sales of assets of the institution in distress can depress asset
prices (Shleifer and Vishny 1992; Allen and Gale 1994, 1998)
and the value of the assets of other institutions, thereby possibly
triggering additional asset sales leading to a fire-sale spiral.8
2.3 Fiscal Costs
Resolution of failed banks is usually associated with fiscal
costs that can arise from payments through a deposit
6
2.2 Contagion
The failure of a bank can have adverse effects on other
banks and financial institutions. This contagion can arise
through various channels such as direct exposures through
interlinkages, information contagion, and fire-sale
externalities, to list a few.
Banks and financial institutions in general have direct
exposure to each other through borrowing and lending. When
a bank fails, other institutions can experience direct losses
4
For more discussion on costs associated with bank failures, see Bliss and
Kaufman (2006) and Hüpkes (2004). On the resolution of failed banks, see
Santomero and Hoffman (1998), Basel Committee on Banking Supervision
(2002), Hoggarth, Reidhill, and Sinclair (2004), and Beck (2011), to cite a few.
5
For a discussion of relationship banking, see Boot (2000) and the
references therein.
158 Bank Resolution Concepts, Trade-offs, and Changes in Practices
See also Leitner (2005). Rochet and Tirole (1996) provide a model where
banks monitor each other (peer monitoring) through cross-holdings. A series
of papers, Sheldon and Maurer (1998) for Switzerland, Furfine (1999) for the
United States, Upper and Worms (2002) for Germany, Wells (2002) for the
United Kingdom, and Elsinger, Lehar, and Summer (2006) for Austria, to cite
only a few, provide empirical analyses of contagion through interlinkages.
Nier et al. (2007) provide a theoretical model and simulation results to
analyze contagion through interlinkages.
7
Saunders and Wilson (1996) examine deposit flows in 163 failed and
229 surviving banks over the Depression era of 1929-33 in the United States.
For the years 1929 and 1933, they find evidence of “flight to quality” where
withdrawals from failed banks were associated with deposit increases in
surviving banks. However, they observe a decrease in deposits in both failed
and surviving banks for the period 1930-32. One possible explanation for these
events is that the depositors may not have had accurate information about each
bank and may have based their decisions on publicly available information,
such as the overall state of the economy or even the number of recent bank
failures. Therefore, imperfect information can lead to runs on healthy banks.
8
Cifuentes, Ferrucci, and Shin (2005) simulate a model where banks are
interconnected through cross-holdings and sales by distressed institutions
depress the market price of assets. An initial shock may force some banks
to liquidate some of their illiquid assets to satisfy the regulatory solvency
constraints. Marking to market of the asset book can induce more asset sales,
depressing prices further and inducing even more sales. Therefore, contagious
failures can result from small shocks through asset prices.
insurance fund when available cash in the fund has been
exhausted, from recapitalization of distressed banks, and
from administrative costs associated with restructuring or
liquidating the failed bank. These costs are exacerbated when
governments need to intervene and come up with funds
quickly; that is, immediacy can entail further costs.
The fiscal costs of providing funds with immediacy
can be linked to a variety of sources, most notably: 1) the
distortionary effects of tax increases and 2) the likely effect
of government deficits on the country’s exchange rate,
manifested in the fact that banking crises and currency crises
have often occurred in tandem in many countries (especially
in emerging market countries). Ultimately, immediacy can
result in further fiscal costs: Government expenditures and
inflows during the regular course of events are smooth,
relative to the potentially rapid growth of off-balance-sheet
contingent liabilities, such as deposit insurance funds and the
costs of bank bailouts.9
2.4 Incentives
During times of systemic crises regulators may feel compelled
to provide assistance to banks that experience difficulties. This
assistance may be in the form of access to lender-of-last-resort
facilities, guarantees for the bank’s debt, and capital injections.
This safety net provided by regulators may create incentives for
banks to take excessive risk, leading to moral hazard. Hence,
during any regulatory intervention, the potential costs of moral
hazard should be taken into account.
An important issue is that regulatory actions may entail
time inconsistency, where ex ante regulators would like to be
tough to prevent incentives for excessive risk-taking. However,
during a systemic crisis, the costs associated with not assisting
(such as the costs of liquidation) can be so high that regulators
may feel compelled to provide help (Mailath and Mester 1994;
Acharya and Yorulmazer 2007, 2008).
9
See, for example, the discussion on fiscal costs associated with banking
collapses and bailouts in Calomiris (1998). Hoggarth, Reidhill, and Sinclair
(2004) find that the cumulative output losses have amounted to an astounding
15 to 20 percent of annual GDP in the banking crises of the past twenty-five
years. Caprio and Klingebiel (1996) argue that the bailout of the thrift industry
cost $180 billion (3.2 percent of GDP) in the United States in the late 1980s.
They also document that the estimated cost of bailouts, as a share of GDP,
were 16.8 percent for Spain, 6.4 percent for Sweden, and 8 percent for Finland.
Honohan and Klingebiel (2000) find that countries spent 12.8 percent of their
GDP to clean up their banking systems whereas Claessens, Djankov, and
Klingebiel (1999) set the cost at 15 to 50 percent of GDP.
3. Resolution Methods
When a bank experiences difficulties or eventually
fails, regulators use various resolution methods. A brief
description of the widely used methods follows, with Table 1
providing examples of the various resolution methods used
in the most recent crisis.
• Mergers and acquisitions: A bank that experiences
difficulties can be acquired by a healthy bank. Even though
the distressed bank may be approaching insolvency, it may
still be an attractive target for other banks due to its franchise
value, which derives from its customer base and established
relationships. This private-sector resolution technique does
not require any public-sector intervention or administration.
• Purchase and assumption: The failing institution enters
receivership and its charter is terminated. In a P&A
transaction, all or part of the bank’s assets and liabilities are
transferred to another institution. In the United States, the
FDIC pays to the successor the gap in value between assets
and liabilities transferred, and the receivership liquidates any
assets not transferred. For example, Washington Mutual, after
being placed in FDIC receivership, was sold through P&A to
JPMorgan Chase in 2008 without government assistance.10
While P&A is still a private-sector resolution, it may require
the use of some public funds as we explain below.
• P&A with assistance: In an assisted P&A transaction,
authorities provide guarantees, including loss-sharing
agreements or put options to sell the assets back to the
authority. An early and large transaction of this type in the
United States took place in 1991, when the FDIC’s resolution
of Southeast Banking Corporation included a provision
to reimburse acquirers for 85 percent of net losses on the
acquired assets. More recently, the acquisition of Bear Stearns
by JPMorgan Chase was facilitated by assistance from the
Federal Reserve.
• Bridge bank: A new bank, called the bridge bank, is set up
in order to maintain banking operations until a permanent
solution can be implemented. Typically, only a portion of
the assets would be transferred to the bridge bank, while
the remaining assets would be passed to the receiver for
liquidation. The ultimate aim is to sell the bridge bank
through a P&A transaction. An example of this method
was seen in the resolution of Bank of New England in 1991,
when the FDIC created a bridge bank for each of Bank of
New England's three subsidiary banks, all of which were
ultimately sold to Fleet/Norstar Financial Group.
10
While the Washington Mutual transaction was regarded as a private
resolution, it has been argued that it would not have been successful
without the receivership powers of the FDIC.
FRBNY Economic Policy Review / December 2014
159
• Good bank–bad bank separation: The bank in distress is split
in two: a “good bank” that retains the performing assets, and
a “bad bank” that receives the remaining assets that would be
restructured or liquidated. Often a trust or asset management
company structure is used. This is a more general method that
could also be used in conjunction with a restructuring and
recapitalization. A good example is the resolution of banks
during the Swedish Financial Crisis, which is discussed as a
case study (see box).
• Liquidation and deposit payoff: In liquidation, the institution
is closed and the assets are placed in a liquidating receivership.
The liquidation value of the assets is used to repay creditors.
In the United States, the FDIC pays insured depositors
either directly or through an acquiring institution serving
as a paying agent. An insured deposit payoff was used in the
failure of Penn Square Bank, N.A., in 1982.11 More than half
of the bank’s deposits were uninsured, including significant
funds of other banks, which led to serious adverse effects on
the banking industry.
• Recapitalization: The institution is kept open through public
assistance. This can be done in a number of ways, including
a restructuring, a “bail-in” that forces creditors to write off
some of their claims, an outright nationalization in which
shareholders are wiped out and management is replaced,
or a capital injection in which shareholders are diluted but
remain and management does not change.12 Table 1 lists many
examples of recapitalizations and capital injections from the
recent crisis.
Each of the resolution options discussed comes with
certain trade-offs and imposes, to varying degrees, some or all
of the costs outlined previously. Furthermore, the availability
and the relative costs of the resolution methods depend on the
state of the world we are in (whether facing an idiosyncratic
bank failure or a systemic crisis), and on factors such as the
size, complexity, and interconnectedness of the institution
in distress. In the next section, we provide a framework
to analyze the feasibility and optimality of the resolution
methods and the trade-offs that may arise.
4. Feasibility and Trade-Offs
So far, we have discussed the costs associated with the failure
and resolution of banks and the methods authorities use
to resolve failed banks. In this section, we analyze the costs
11
Managing the Crisis: The FDIC and RTC Experience, Part II, Chapter 2,
in FDIC (1998).
12
See Philippon and Schnabl (2013) for an analysis of efficient
recapitalization of banks.
160 Bank Resolution Concepts, Trade-offs, and Changes in Practices
associated with different resolution methods and try to
formalize an optimal resolution policy.
A private-sector resolution, through which the failed bank
is acquired by a healthy bank, imposes the least cost, since
the franchise value is preserved, there is no disruption to the
bank’s customers or the payment system itself, and there are
no fiscal costs.13 However, the feasibility of such an option
depends on the size and complexity of the failed bank, as well
as the state of the world. When a private-sector resolution is
not feasible, the authorities resort to methods such as assisted
sales, liquidation, and recapitalization, each of which entails
certain trade-offs and higher costs. Next, we provide a simple
analytical framework to analyze these issues formally.
4.1 An Analytical Framework
Suppose we have the following framework involving
two banks that are identical to start. The banks have the
following balance sheet:
Assets
Liabilities
Risky assets (a)
Insured deposits (id )
Uninsured debt (d )
Equity (e)
The bank finances itself with insured deposits (insurance
is provided by the FDIC), uninsured debt, and equity capital,
where id + d + e = 1. The bank has one unit of the risky
investment (a = 1), which has a random return with the
high return R > 1 and the low return r < id. So, when the
return is high, the bank is solvent and does not require any
intervention. However, when the return is low, the bank’s
capital is wiped out, so the bank becomes insolvent and needs
to be resolved.
To keep the framework simple, we first focus on the
following resolution methods: 1) whole-bank purchase and
assumption, 2) liquidation, and 3) recapitalization. Next, we
analyze the costs associated with different resolution methods
and the optimal choice in different states of the world.
Along the lines of our earlier discussion, we assume
that the bank’s assets are specific so that sale of the assets
to another bank (via P&A) and liquidation can result in
13
In evaluating the costs of resolution methods, we should take into
account the potential effects on size and complexity of the institutions
resulting from a private transaction. For example, these institutions may
become larger and more complex and therefore more difficult to resolve
in the case of future distress.
A Good Example: Lessons from the Resolution of the Swedish Financial Crisis
Sweden experienced a twin crisis in the early 1990s, which marked
the first systemic crisis in industrialized countries since the 1930s.
It is usually argued that this episode can be regarded as a good
example of a swift, effective, and low-cost resolution of banking
crisis. However, the Swedish experience has some unique features
that may be difficult to replicate in all crises.a
Crisis and intervention: After deregulation of the credit
markets in 1985, low interest rates, lax supervision, and the
credit expansion contributed to an overheating property
market.b Finance companies were less regulated compared to
banks and were financed by a new type of commercial paper
called “marknadsbevis” guaranteed by banks. When one of
these companies folded in September 1990, the market for these
securities dried up and banks had to keep funding the companies
since they were closely linked.
In the early stages, no comprehensive framework existed and
the government tackled problems case by case. By the fall of 1991,
two of the six largest financial institutions, Forsta Sparbanken
and Nordbanken, had inadequate capital. The state guaranteed
a loan for Forsta and took over Nordbanken injecting capital to
own 77 percent of its shares and split Nordbanken by transferring
nonperforming loans to an asset management company (AMC)
called Securum. Within a year, Gota Bank experienced difficulties
and was also taken over by the government and split into a good
bank and an AMC, called Retrieva.c
While there were no significant banks runs, the banks’
foreign creditors started to cut their credit lines, and the Swedish
authorities needed to restore confidence. In December 1992,
Sweden guaranteed all bank deposits and creditors of the nation’s
114 banks, but not the shareholders. The parliament passed
the Bank Support Act authorizing the government to provide
support in the form of loan guarantees, capital contributions,
and other appropriate measures.d Overall, to resolve the crisis,
Swedish authorities forced banks to write down their losses, used
methods such as capital injections (both private and public), and
separated troubled institutions into “good banks” and “bad banks,”
employing AMCs to restructure and divest the assets of the bad
banks. Banks were told to write down their losses promptly. Bank
owners were invited to inject capital, or let the Swedish authorities
intervene, which implied wiping out shareholders.
Exit: Exit from the guarantees and the divesting of assets was
smooth with low cost. In 1996, Sweden rescinded the guarantees,
replacing them with a bank-financed depositor-protection scheme.
Securum sold its real estate assets in 1995 and 1996, when the
market had started to recover, and was dissolved at the end of 1997
much faster than originally envisaged.e
Sweden shelled out 4 percent of its GDP to rescue its financial
system. After the recovery from asset sales, the cost ended up
being less than 2 percent. It is argued that factors such as political
consensus, decisiveness, and transparency surrounding the
management of the crisis contributed to restoring confidence
and to the eventual success of the resolution. As well as the right
policies, various other factors that may not be present in all crises
have an influence on this favorable outcome.
Complexity of financial instruments: The assets that were
resolved mostly involved those related to real estate and were
not very complex, factors that made the resolution easier and
less costly. However, over time, the financial industry and
financial contracts became much more complex. An important
feature of the recent crisis was the difficulty of assessing complex
financial instruments and structures, as well as off-balance sheet
commitments and bank-related vehicles such as structured
investment vehicles and conduits. These complex instruments,
valuation issues, and institutional arrangements make it more
difficult for analysts and counterparties to understand a bank’s
financial position, adding to the difficulties of the resolution.
Macroeconomic factors helped recovery in Sweden: Sweden
had a fixed exchange rate before the crisis. Once the krona peg
had been abandoned and the currency depreciated, Swedish
goods regained competitiveness in export markets. Furthermore,
a quick rebound in the Swedish economy stemmed from an
increase in economic growth in Europe. The strong international
recovery helped push up real estate values in Sweden and
improved the balance sheet of banks, which played an important
role in the recovery process. While Sweden is a small economy
compared to the rest of the world, slowdowns in big industrial
countries such as the United States and those in Europe can
themselves drag the global economy down and such an export-led
recovery may not be feasible, especially when countries are in a
currency union, such as in Europe.
a
This discussion of Sweden’s experience builds on Yorulmazer (2009).
b
From 1987 to 1990, credit rose from 90 to 140 percent of GDP and prices
of commercial real estate doubled.
c
During 1993, Nordbanken and Gota bank were merged, retaining the
name Nordbanken, and becoming Sweden’s fourth largest bank. The
bank was operationally restructured and partially sold to the private
sector. Their respective AMCs—Securum and Retrieva—were merged in
December 1995.
d
The parliament gave the Bank Supervisory Authority the power to decide
and manage support operations.
e
Several factors contributed to the AMCs’ success. AMCs could rely on an
efficient judicial system, which allowed them to force most of their debtors
into bankruptcy when their operations did not prove economically viable.
The restructuring of the assets was also facilitated by the fact that most
of the assets transferred were related to real estate and were not like the
complex assets seen in the most recent crisis.
FRBNY Economic Policy Review / December 2014
161
Table 2
Costs Associated with Different Resolution Methods
Purchase and assumption (P&A)
P&A plus liquidation
Assisted P&A
Liquidation
Recapitalization
a
Cost to FDICa
Fiscal Cost
Moral Hazard
id - (r - ΔPA)
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
f (d )
m
id - (r - (αΔPA + (1 - α)ΔL))
id - (r - Δ'PA ) + β
id - (r - ΔL)
id - r
The cost to the Federal Deposit Insurance Corporation (FDIC) incorporates customer and market disruptions.
misallocation costs. However, we assume that this cost is
lower under P&A compared with liquidation since the assets
stay with the banking system, which helps preserve their
value. In particular, we assume that when the assets are sold to
another bank, they generate a value of r - ΔPA, whereas when
the assets are liquidated they generate a value of r - ΔL with
0 ≤ ΔPA < ΔL.
Let p be the price at which the assets are sold by the FDIC.
Suppose that the assets can be sold at their fair value so that
p = r - ΔPA under P&A and p = r - ΔL under liquidation.
Note that the difference between the value of insured deposits
and the value of the asset recovery needs to be covered by
the FDIC. Hence, the cost to the FDIC is c = id - p, with the
cost to the FDIC under liquidation being higher than the cost
under P&A. Therefore, the FDIC prefers P&A to liquidation.
Note that in both P&A and liquidation, shareholders are
wiped out so moral hazard is not a concern.
The other alternative is to recapitalize the failed bank. While
there can be many variations of a recapitalization in terms
of which stakeholders receive how much (discussed below),
here we focus on the case where insured depositors and debt
holders are paid in full, but the shareholders are wiped out. The
recapitalization will result in fiscal costs but help keep the bank
open and preserve its going-concern value so that the assets
generate a return of r. In this case, in addition to the shortfall
(id - r) that will come from the FDIC, the government needs
to come up with d to pay debt holders. This would result in
a cost of f (d ). Hence, the additional costs beyond the loss of
the FDIC in this case would be f (d ) + m, where m represents
the costs associated with adverse incentives arising from
recapitalization. (In this case, the adverse incentives refer to
those of debt holders since shareholders are wiped out.) We
assume that ΔPA < f (d ) + m so that the aggregate resolution
cost under P&A is lower than the cost of recapitalization.
162 Bank Resolution Concepts, Trade-offs, and Changes in Practices
Within this framework, P&A results in the lowest resolution
cost and is the preferred option, where the comparison between
liquidation and recapitalization depends on the relative costs
of ΔL and f (d) + m, respectively. Table 2 summarizes the costs
associated with different resolution methods.
Next, we focus on different states of the world and the
feasibility of each option. In an “idiosyncratic” failure
state, only one bank fails, while the other stays healthy.
In an “aggregate” failure state, both banks fail, resulting
in a systemic crisis. P&A would be available only in an
idiosyncratic failure state, where there are available buyers.
Hence, in an aggregate failure state, the regulators face the
trade-off between a disorderly liquidation with the cost of ΔL
and recapitalization with the cost of f (d ) + m.
The framework is kept simple on purpose to illustrate
the primary trade-offs regulators face, particularly during
systemic crises. However, it can easily be extended to analyze
a wider range of resolution options discussed earlier. For
example, when we analyzed P&A above, we assumed that all
the assets were being sold to the healthy bank. However, in
practice, only a fraction of the assets can be transferred while
the rest is liquidated. Let α be the fraction of assets sold under
P&A and (1 - α) be the remaining fraction that is liquidated.
In that case, the cost would be αΔPA + (1 - α)ΔL. Note that
the cost is decreasing in the fraction of assets that have been
sold through P&A.
While passing a greater amount of assets in P&A typically
lowers the cost to the FDIC, large and complex assets held
by the failed institution may lead to lower bids by potential
successors, who incorporate large discounts to compensate
for the uncertain asset value. This, in turn, increases the loss
in value by ΔPA. In this case, rather than accepting a high cost
to the FDIC associated with the low bids, or the alternative
option of passing only the most transparent assets and
liquidating the rest, the resolution authority may face a lower
cost by assisting the P&A through a loss-sharing agreement.
Suppose that with this type of assistance, an acquirer will
purchase all assets instead at a cost of Δ'PA < ΔPA, since the
loss-sharing agreement provides insurance for the acquirer.
However, assistance can increase the cost to the FDIC since
the FDIC may have to absorb a portion of the acquirer’s
losses.14 Let β be the expected cost of the assistance. While the
assistance (such as in the form of guarantees) can weaken the
incentives of the acquirer to exert effort to generate the full
return from the acquired assets—in turn, increasing β—an
assisted P&A can still be a better option than liquidation if the
cost of a disorderly liquidation is significant (high ΔL ) and/or
the expected cost of the assistance is not very high.
Another important issue is that during a recapitalization,
different stakeholders can suffer varying levels of costs. In
the benchmark case above, we assumed that uninsured debt
holders are paid in full. However, uninsured debt holders
can suffer some losses as well, resulting in a bail-in of the
bank (discussed later in detail). In general, the uninsured
debt holders can be paid an amount x ϵ [0, d]. In that case,
the fiscal cost of the recapitalization would be f (x). Since
debt holders suffer some losses, they would have incentives
to monitor the banks properly so that the cost of moral
hazard m would decrease to m' < m. In other versions of
recapitalization, it is also possible that the shareholders are not
wiped out completely. In this case, the fiscal cost as well as the
cost of moral hazard would increase.
Various other factors such as size and complexity affect the
cost of resolution and the feasibility of resolution options. One
would expect that, as the assets get more complex, they would
be harder for the acquirers to value and even manage, regardless
of whether it is a P&A agreement or liquidation. Hence, as
assets become more complex, ΔPA and ΔL would increase.
The size of the failed institution would also have an
important effect on the resolution. In our simple framework,
suppose that one bank is large, whereas the other is relatively
small. If the small bank fails, the large bank, if healthy, can
acquire the small bank. However, if the large bank fails, the
small bank may not have the means to acquire the large
bank and may not have the expertise to run the assets of the
large bank efficiently, especially since, in most cases, size and
complexity go hand in hand. Hence, when a large bank fails,
the result would be a systemic crisis even though the small
bank is healthy, and the private resolution options such as
P&A may not be available. Hence, bank size can lead to a
systemic crisis on its own.
Our simple framework can easily be extended to model
a wide range of resolution options, such as the use of a
bridge bank or an asset management company (AMC). In
certain cases, when immediate P&A would be too disorderly
and entail high costs, regulators may resort to methods
that would allow them to restructure the failed institution
and increase the feasibility of a P&A agreement in the
future—for example, the creation of a bridge bank. While
the bridge bank can create administrative costs, setting
one up can provide other institutions with time to conduct
due diligence and evaluate asset values without inhibiting
operations or disrupting payment systems and loan creation.
The authorities should compare the premium over market
value that could be expected from the eventual sale with the
additional administrative costs arising from the bridge bank.
Hence, a bridge bank is a preferable option if it leads to a
profitable P&A down the road net of any administrative costs.
Furthermore, the bridge bank can facilitate the resolution of
multiple failures at once, where the failed banks merge into
the bridge bank.
Regulators also use other methods such as a
good bank–bad bank separation followed by the setting-up
of an AMC. First, the bad assets of the bank are separated from
the good assets so that confidence can be restored in the good
and it can continue operation. Then, the AMC can focus on
restructuring or liquidating the bad assets. This method can
have various advantages over market-based solutions such as
liquidations, including 1) economies of scale in administering
workouts and in forming and selling portfolios of assets,
2) benefits from special powers to expedite loan resolution,
3) allowing the good bank to focus on normal banking
business such as issuing loans, and 4) enabling the AMCs,
which have longer horizons, to recover more compared with
an immediate liquidation of assets. Table 3 summarizes the
options for resolution and their relative costs, and Chart 5
illustrates the decision process taken by resolution authorities
along the lines of our analytical framework.
4.2 Evidence from the FDIC
We have pointed out the many costs associated with certain
resolution methods, although quantifying and comparing
the magnitude of each component empirically across varying
time horizons and failure periods is challenging. However,
data provided by the FDIC’s Historical Statistics on Banking
(HSOB) allow us to compare various resolution methods
14
In the United States, loss sharing typically provides for the FDIC to cover
up to 80 percent of losses on specific assets, while offering even greater loss
protection “in the event of financial catastrophe.”
FRBNY Economic Policy Review / December 2014
163
Table 3
A Summary of Options for Failure Resolution and Relative Costs
Costs
Option
Mergers and
acqusitions
Feasibility
Disruptions
to Customers
Disruptions
to System
Fiscal
Moral Hazard
Not feasible when
there are no willing,
healthy buyers
None
None
None
None
Not feasible when
there are no willing,
healthy buyers
The smaller the amount
of assets and liabilities
transferred to the
acquirer, the greater
the disruptions
The smaller the amount
of assets transferred
to the acquirer, the
more assets need to be
liquidated, leading to
fire-sale externalities
When recovery
from the transfer
or sale of assets is
lower compared with
transferred liabilities,
the greater are the
fiscal costs
Moral hazard
introduced if uninsured
deposits and any
additional debt claims
are transferred,
requiring payment
from public sources
that is not recovered
If losses are not
shared appropriately
between acquirer
and the authorities,
guarantees can distort
acquirer's incentives
to maximize the value
from the assets
Purchase and
assumption (P&A)
Without assistance
There may be willing
buyers with assistance
(next option)
With assistance
The smaller the amount
of liabilities transferred
to the acquirer, the
greater the direct losses
to the creditors
Not feasible when
there are no willing,
healthy buyers
Assistance may
facilitate the transfer
of a greater portion
of assets and liabilities,
A bridge bank may help reducing disruptions
facilitate transaction
(next option)
Assistance may
facilitate the transfer
of a greater portion of
assets and liabilities,
reducing disruptions
Higher potential costs
due to guarantees
Bridge bank
A bridge bank may
The smaller the amount
facilitate a restructuring of assets and liabilities
and P&A in the future
transferred to the
bridge bank, the greater
Not a preferred option
the disruptions
if the bridge bank will
not increase asset value
A bridge bank may
prevent the disorderly
liquidation of assets
and provide time for an
orderly restructuring
Setting up a bridge
bank can increase
administrative costs
Moral hazard
introduced if creditor
losses are covered
using public funds
Liquidation
Not a preferred option
if disruptions arising
from liquidation are
too great
Disorderly liquidation
is likely to lead to
fire-sale externalities,
greater direct losses to
the creditors, and loss
of confidence
Fiscal costs may be
high if low recovery
from disorderly
liquidation does not
cover payout of insured
deposit claims
Moral hazard
is very low, as
liquidation promotes
market discipline
Going-concern value
and customer/bank
relationships are
destroyed
Potential disruptions
to payment services
But assistance may
facilitate transfer of
greater assets and
liabilities reducing
fiscal costs
Recapitalization
through private
bail-in
(shareholders
wiped out)
Not a feasible option if
creditors do not agree
Creditors suffer
some losses but
going concern and
customer/bank
relationships
are preserved
This option prevents
disorderly liquidation,
although there are
some direct losses
to the creditors
Bail-in helps lower
fiscal costs
Mitigates moral hazard
since recapitalization is
done through private
rather than public funds
Recapitalization
using public funds
(shareholders
wiped out)
Not a feasible
(or preferred)
option if government
does not have funds
to recapitalize
Mitigates disruptions
as going-concern value
and customer/bank
relationships
are preserved
Mitigates disruptions
as direct losses are
limited and fire-sale
externalities are avoided
High fiscal costs
Moral hazard is created
since creditors do not
suffer losses
Recapitalization
using public funds
(shareholders diluted
but retain some
stake in firm)
Not a feasible
(or preferred)
option if government
does not have funds
to recapitalize or
moral hazard would
be too great
Mitigates disruptions
as going-concern value
and customer/bank
relationships
are preserved
Mitigates disruptions
as direct losses are
limited and fire-sale
externalities are avoided
High fiscal costs
Moral hazard is highest
since even shareholders'
losses are limited
164 Bank Resolution Concepts, Trade-offs, and Changes in Practices
empirically in terms of the cost to the FDIC.15 The estimated
losses to the fund are available for most bank failures since
1986, although it is important to note that the processes used
by the FDIC have evolved over time.16 Generally, when a
failing institution is taken into receivership, the FDIC solicits
bids from acquirers to purchase all or part of the assets and
assume all or part of the liabilities (P&A). However, prior to
the passage of the Federal Deposit Insurance Corporation
Improvement Act (FDICIA) in 1991, bids were accepted from
potential acquirers for the assumption of all deposits only.
The passage of the FDICIA imposed a number of
provisions, including requirements for prompt corrective
action (PCA) and least-costly resolution methods. Under PCA,
a conservator or receiver must be appointed within ninety days
of an institution becoming critically undercapitalized; that is,
its tangible equity falling to (or below) 2 percent of total assets.
Further, while it has access to a number of resolution tools, the
FDIC is required to perform a least-cost test when deciding
how to resolve the institution. However, the “systemic risk
exception” allows the FDIC to bypass the least-cost method if it
would have serious adverse effects on financial stability.
It wasn’t until after the FDICIA that bids were also accepted
for insured deposits only. Table 4 shows that, on average, P&A
transactions in which only insured deposits are transferred are
less costly to the FDIC. If a bid is for all deposits, the premium
offered by the acquirer—reflecting the value of relationships—
has to be at least as much as the amount of uninsured deposits
in order for the transaction to be less costly than an (insured )
deposit payoff by the FDIC.
The authority for the FDIC to establish a bridge bank,
chartered by the Office of the Comptroller of the Currency,
was provided by the Competitive Equality Banking Act
(CEBA) of 1987. Before a failed bank enters a bridge, the
FDIC must apply the least-cost test, considering the premium
over market value that could be expected from the eventual
sale compared with an immediate liquidation of assets. The
least-cost test is applied again at the final sale resolution of the
bridge bank before a sale can be made.
As shown in Table 4, P&A transactions implemented
after setting up a temporary bridge bank, have, on average,
led to lower costs to the FDIC; over the period from 1987 to
2012, losses to the FDIC in an insured-deposits-only P&A
transaction represented 14.8 percent of bank assets when
a bridge bank was established, compared to 19.9 percent
of assets without the use of a bridge bank. Note that losses
were considerably higher if a bridge bank was set up and no
effective P&A transaction was available.
15
The data are available at http://www2.fdic.gov/hsob/.
16
FDIC (1998) provides a history of bank failure resolutions from 1980-94.
Last, the data show that, when liquidation was used by
the FDIC, it was very costly; however, liquidation was used
when P&A was not feasible (or more costly) and the failure
did not trigger the systemic risk exception to use open bank
assistance. The costs associated with assisted transactions
are slightly more difficult to evaluate, although on average,
the FDIC recovered most of the funds, resulting in losses of
only 8 percent of bank assets. The 115 assisted transactions
included in the table all occurred prior to 1993, when an
amendment to the Federal Deposit Insurance Act of 1950
prohibited “the use of insurance fund monies in any manner
that benefits any shareholder of an institution that had
failed or was in danger of failing.” (Eighty of the 115 assisted
transactions occurred in 1988.)
In interpreting these results, we find our analytical
framework very helpful. One of the interesting empirical
results from the FDIC data is the striking difference
between the cost associated with liquidation and that
of other resolution methods. As our framework shows,
everything equal, liquidation is more costly than P&A,
and would therefore only be used when options such as
P&A are not available. To start with, the banks that were
liquidated may have been in worse shape or may have
failed in a systemic crisis if a ready buyer was not available.
These two factors together help explain the high costs of
liquidation shown in the data.
5. Recent Developments
During the recent crisis, we witnessed the failure or near
failure of some of the most prominent financial institutions
around the globe. Recent experience highlighted some of
the shortcomings of the regulatory framework to resolve
financial institutions and the need for a special resolution
regime for systemically important institutions in cases
where bankruptcy is not an effective option. The crisis led
to a revision of the current regulatory framework to deal
with distressed institutions. In this section, we review recent
developments in the United States, the United Kingdom, and
the European Union.
5.1 United States
In the United States, the FDIC possesses expansive powers
to resolve failed federally insured depository institutions
under the statutory objective to maximize the institution’s
FRBNY Economic Policy Review / December 2014
165
Chart 5
Resolution Decision Tree
Is M&A available?
Private resolution
No
Is P&A without
assistance available?
Costs increasing
Yes
M&A
No
Yes
Would P&A
with assistance
perform better?
No
P&A without
assistance
Is P&A with
assistance available?
Yes
No
P&A with
assistance
Would P&A be available
after bridge bank?
Yes
No
Bridge bank
Is recapitalization
less costly
than liquidation?
P&A
Yes
No
Is it feasible to
replace management,
wipe out shareholders,
and bail-in creditors?
Liquidation
Yes
No
Recapitalization,
stakeholders
are punished
Recapitalization,
stakeholders are
not fully punished
return on assets and minimize costs to the insurance fund.
In contrast with corporate bankruptcy proceedings, the
FDIC, acting as receiver of a failed institution, is not subject
to court supervision, and assumes the rights and powers
of the institution’s stockholders, directors, and parties with
contractual rights. This authority includes the power to merge
the institution with another insured depository institution
without the need for consent.
The failure of a number of firms such as Lehman Brothers
during the recent crisis proved that U.S. regulatory agencies
did not have adequate tools for resolving systemically
important nonbank institutions. Below we discuss two
recent developments that resulted from Dodd-Frank: 1) the
resolution and recovery plans of the act's Title I, and 2) the
Orderly Liquidation Authority (OLA) of its Title II.
166 Bank Resolution Concepts, Trade-offs, and Changes in Practices
Living Wills
Title I of Dodd-Frank requires all bank holding companies
with total consolidated assets greater than $50 billion and
all nonbank financial companies designated as systemically
important by the Financial Stability Oversight Council to
submit resolution plans, or “living wills,” to the Federal Reserve
and the FDIC.17 Each plan must provide a strategic analysis
of the institution’s rapid and orderly resolution in the event of
material financial distress or failure, through a reorganization
or liquidation under the Bankruptcy Code.
17
The final rule was effective November 30, 2011. See “Resolution Plans Required,”
76 Federal Register (November 1, 2011). The final rule also applies to a foreign bank
or company treated as a bank holding company under the International Banking
Act of 1978 that has total consolidated assets greater than $50 billion.
Table 4
Summary of Costs to the FDIC under Various Resolution Methods, 1986-2012
Resolution Method
Number of
Institutions
Average Assets
(Millions of U.S. dollars)
Average Cost-to-Assets Ratio
(Percent)
Purchase and assumption (P&A)
Insured deposits only
112
293.31
19.9
1,263
587.00
23.7
P&A-insured only
26
3,324.11
14.8
P&A-all-deposits
499
667.78
19.2
Liquidation
256
229.15
51.7
115
165.52
8.4
Insured deposit transfer
106
157.60
30.1
Deposit payoff (direct)
160
66.53
27.8
All-deposits transfer
Bridge bank
a
b
Assisted transactions
Liquidation
Source: Federal Deposit Insurance Corporation, Historical Statistics on Banking.
Notes: The table only includes resolutions for which estimated costs were available and excludes transactions where it was not determined if all deposits
or insured deposits only were transferred in P&A. Additionally, the table excludes thirty-seven transactions where the Federal Savings and Loan Insurance
Corporation took over management and generally provided assistance and one reprivatization transaction.
a
Bridge banks also include thrift conservatorships.
Assisted transactions include open bank assistance transactions and assisted whole-bank P&A transactions.
b
As firms conduct their strategic analyses of orderly
resolution, the assumptions made concerning economic
conditions at the time of failure are critical for determining
the availability of tools and techniques, as we set forth in
our framework. For their initial resolution plans, filers were
provided with a set of baseline economic conditions to use
in their analysis, although subsequent submissions will need
to create a plan for resolution under “adverse” and “severely
adverse” economic conditions.18 Our framework shows that the
availability of options for resolution depends not only on the
institution in distress but also the health of other institutions.
Hence, any resolution and recovery plan should have a
macroprudential view and should not treat the institution
in distress in isolation. At least the “adverse” and “severely
adverse” scenarios should take into account the possibility of a
systemic crisis in cases where many banks experience distress
at the same time, huge fire-sale discounts are commonplace,
and certain resolution options are not available.
Orderly Liquidation Authority
The OLA, established in 2010 under Title II of Dodd‑Frank,
expands the FDIC’s authority to resolve failing banks
by including systemically important nonbank financial
institutions (SIFIs), which previously would have been
resolved through corporate bankruptcy.19 Further, for banks
that are consolidated under a bank holding company, Title II
acts under a “single point of entry” framework to facilitate
continuity of critical services and reduce costs.
In resolving a failed institution, the FDIC would assign losses
to shareholders and unsecured creditors of the holding company
and transfer sound subsidiaries to a new solvent entity. As
receiver, the FDIC can raise funds (up to a limit) through a line
of credit from the U.S. Treasury, but Title II includes a provision
that prohibits the use of taxpayer funds to cover the cost of
resolution; therefore, all funds must be recovered.
19
18
Conditions developed pursuant to Section 165(i)(1) of the Dodd-Frank Act
may be referenced.
See “Certain Orderly Liquidation Authority Provisions under Title II of the
Dodd-Frank Wall Street Reform and Consumer Protection Act, Final Rule,”
76 Federal Register (July 15, 2011). Additionally, in a speech to the U.S. House
of Representatives’ Committee on Financial Services, Osterman and Wigand
(2013) explore the application of OLA in resolutions.
FRBNY Economic Policy Review / December 2014
167
Before a firm can enter orderly liquidation proceedings, the
Treasury secretary must receive a written recommendation
based on a two-thirds vote from the Board of Governors
of the Federal Reserve System and another regulator, and,
in consultation with the U.S. president, determine that the
financial institution is in danger of default and that failure
would have “serious adverse effects on the financial stability of
the United States.” It must also be determined that there is no
viable private sector alternative available.
While Title II takes steps towards outlining viable
alternatives to the bailout of a private institution, it has been
argued that the legislation can be further improved. Plosser
(2013) contends that it affords significant discretion to
regulators, and that the complicated procedure to invoke the
OLA may take time, increasing costs and limiting options. Still,
the expanded powers of the FDIC to take into receivership
those SIFIs that otherwise would have relied on the bankruptcy
process for resolution should significantly reduce the costs
associated with failure that we have outlined in our framework.
5.2 United Kingdom
The failure of Northern Rock in 2007 was a wake-up call for
regulators and since then there have been wide reforms of
financial regulation in the United Kingdom. Prior to 2008,
the British legal system did not distinguish between banks
and other failing companies, and therefore authorities did
not have the ability to take Northern Rock into receivership.20
The Banking (Special Provisions) Act was passed in 2008
as a temporary measure, giving the U.K. Treasury powers
to facilitate orderly resolution through directed transfers of
property, rights, and claims of a failed depository institution.
The Banking Act of 2009 replaced the temporary regime
and created a Special Resolution Regime (SRR) for failing
banks, influenced by the U.S. approach. The Financial Services
Authority (FSA), the regulator of financial firms at the time,
was given the right to trigger the SRR. Under the SRR, the
U.K. authorities have powers similar to the FDIC in resolving
a failed institution, and the choice of method would also
involve a cost test.21
However, the regime set up under the Banking Act of 2009
did not cover nondeposit-taking financial firms. To address
this flaw and improve financial supervision generally, further
reforms were implemented in April 2013. Under the new
regulatory regime, the FSA ceased to exist, and the Prudential
Regulation Authority (PRA) was formed as part of the Bank
of England to regulate deposit-takers, insurers, and major
investment firms. Firms will assist the PRA and the SRR in
assessing resolvability and drawing up recovery and resolution
plans. The PRA, in consultation with the Bank of England
and the Treasury, makes the decision to initiate the SRR for a
failing institution.
In addition, the publication of the Report of the Independent
Commission on Banking led by John Vickers (known as the
“Vickers Report”) made formal recommendations for further
reform in 2011.22 The focus of the Vickers Report is the
notion that banks should “ring-fence” retail and commercial
banking operations by establishing a separate legal entity
to carry out these activities. The purpose is to protect these
operations from the riskier wholesale and investment banking
services. The Vickers Report also recommends that large
U.K. ring-fenced retail banks hold a greater amount of capital
than what is proposed under Basel III in order to improve their
“loss absorbency.” Many of the recommendations outlined in
the Vickers Report have been incorporated in the Banking
Reform Act of 2013, which is being implemented in 2014.
This legislation gives the new PRA power to enforce the full
separation of banking activities.
5.3 European Union
More recently, in response to the financial crisis,
European Union (EU) authorities have worked to improve
the framework of banking regulation within the European
Economic and Monetary Union. Prior to the crisis, many EU
countries relied on insolvency (bankruptcy) proceedings to
deal with bank failures, which is suboptimal for a number of
reasons we have already outlined. The European Commission
has taken steps under the Bank Recovery and Resolution
Directive to establish a common set of rules for national
authorities to follow when winding down failed banks.
In 2012, the European Central Bank (ECB) proposed
the creation of a European Banking Union, which would
involve the establishment of the Single Supervisory
Mechanism (SSM), the Single Resolution Mechanism, and
a common system of deposit protection. Under the SSM
proposal, the ECB supervises banks in the euro area and
other member states, and, when a bank is in severe stress, it
informs the Single Resolution Board, which would oversee
20
For a discussion of the Northern Rock episode, see Shin (2009) and
Goldsmith-Pinkham and Yorulmazer (2010).
21
See Brierley (2009).
168 Bank Resolution Concepts, Trade-offs, and Changes in Practices
22
The report is available at http://bankingcommission.independent.gov.uk/.
the resolution.23 The Single Resolution Authority (SRA)
will have access to a privately funded European Resolution
Fund, generated by levies on the private sector, replacing the
national resolution funds of the euro area states. The fund will
need to cover 0.8 percent of the total insured deposits in any
given country. The SRA will be expected to choose the leastcost resolution method, as practiced by the FDIC, but it will
require access to the European Stability Mechanism as a fiscal
backstop in case a systemic crisis develops.
5.4 Bail-In Debt
The resolution directive proposed by the EU is focused
on the idea that the shareholders and creditors must
face losses before a failing bank can receive any taxpayer
bailouts. It proposes that shareholders, unsecured creditors,
and uninsured depositors (with deposits greater than
100,000 euros), in that order, would be forced to cover at
least 8 percent of the institution’s total liabilities before the
resolution fund provides any support. Power to carry out
bail-in within resolution is listed as one of the “key attributes”
of effective resolution regimes for financial institutions by
the Financial Stability Board (FSB 2011), which the Federal
Reserve and the FDIC helped to develop and which G-20
leaders endorsed in 2011. In general, this method could
include writing down and/or converting to equity any or all
unsecured and uninsured creditor claims in a manner that
respects the hierarchy of the claims. Importantly, it would
provide a capital buffer for distressed firms that would
otherwise have difficulty raising new equity.
In the United States and elsewhere, requirements for
contingent convertible bonds (CoCos) and bail-in debt have
been proposed.24 CoCos are loss-absorbing instruments which
are converted to equity if a predetermined trigger, based
on regulatory capital levels, is hit. The United Kingdom is
working to include bail-in measures in its resolution regime.25
Meanwhile, Swiss authorities support bail-ins of a range of
creditors, including shareholders, holders of CoCos, and
other bondholders, especially for the country’s largest banks,
23
UBS and Credit Suisse.26 In general, while a number of issues
will need to be addressed, a bail-in resolution method may
come with significant advantages relative to the costs we
have outlined; it can provide capital during times of distress
and reduce moral hazard and disruptions to customers and
markets in the case of a systemic failure.
5.5 Cross-Border Issues in Resolution
Another important issue emerging from the recent crisis was
the lack of a framework for resolving banks with cross-border
operations. For example, the failure of Lehman Brothers had
widespread repercussions given its operations across fifty
countries. Indeed, the FSB’s key attributes state that institutionspecific cooperation agreements should be in place between
the home and host authorities for all global SIFIs (G-SIFIs).
The United States has been one of the first countries to
incorporate cross-border planning into its statutory regime
as it is home country to eight of the twenty-eight global
systemically important banks identified by the FSB.27 OLA
requires the FDIC to coordinate with the foreign regulatory
authorities in resolving G-SIFIs. In addition to resolution
planning, the United States has taken steps to improve the
supervision of U.S. operations of foreign banks, and last
year the Federal Reserve sought comment on its proposal to
require large foreign banking organizations to organize their
U.S. subsidiaries under an intermediate holding company,
subject to the requirements of U.S. bank holding companies.
Owing to the connections between financial institutions
in the United States and the United Kingdom, the bilateral
relationship is perhaps the most significant with regard
to the resolution of G-SIFIs, especially given the need to
prevent disruptive forms of ring-fencing of the host country’s
operations of a failed firm. Working relationships will also be
established with the European Union, Switzerland, and Japan,
which also host a number of G-SIFIs. As resolution regimes
are developed internationally to address cross-border issues
explicitly, the feasibility of an orderly and timely resolution that
minimizes disruptions and panic should improve, although
there is still considerable work to be done in most jurisdictions.
See European Commission (2013).
24
For analysis of contingent capital, see Sundaresan and Wang (forthcoming),
Bank of Canada (2010), Calomiris and Herring (2011), Flannery (2002, 2009),
Glasserman and Nouri (2012) and Pennacchi (2010), to cite a few.
26
25
27
Lloyds Banking Group was the first to issue CoCo bonds in 2009, which
included the terms that the security would be converted to ordinary shares if
the Tier I capital ratio fell below 5 percent.
A recent CoCo deal issued by Credit Suisse included terms that holders
of the security stood to lose the whole investment if the bank breached its
5 percent Tier I capital ratio.
In a speech given in 2013, the Federal Reserve’s Michael Gibson reviews
the steps taken by the United States to formalize cross-border resolution
planning. See Gibson (2013).
FRBNY Economic Policy Review / December 2014
169
6. Conclusion
Bank failures entail costs for bank customers, for the
financial sector, and the overall economy. Hence, efficient
resolution of financial institutions in distress is an extremely
important issue.
This article provides a discussion of the costs associated
with bank failures and the methods authorities use to resolve
banks. While regulators can employ various methods ranging
from private-sector resolution in the form of M&A and
P&A to government intervention and recapitalization of
banks using public funds, we have shown that some of these
methods may not be feasible in certain states of the world.
In particular, although private-sector resolution is a
preferred option in terms of minimizing costs associated with
170 Bank Resolution Concepts, Trade-offs, and Changes in Practices
bank failures, it may not be a feasible one when the failing
institution is large and complex or when its failure occurs
during a systemic crisis. When many banks experience
distress simultaneously, there may not be a ready buyer
for the failed bank. Hence, when the preferred option is
not available, the authorities face certain trade-offs, as
they choose from second-best options such as disorderly
liquidation and the use of public funds to resolve banks.
Thus, systemic crises always entail higher aggregate
resolution costs and trade-offs.
The optimal design of regulation and a resolution regime
needs to take into account the fact that certain preferred
options may not be available during systemic crises. Further, it
should aim to minimize the probability of systemic crises and
the costs associated with resolving failures in those scenarios.
References
Acharya, V., and T. Yorulmazer. 2007. “Too Many to Fail: An Analysis
of Time Inconsistency in Bank Closure Policies.” Journal of
Financial Intermediation 16, no. 1 (January): 1-31.
Calomiris, C., and R. Herring. 2011. “Why and How to Design a
Contingent Convertible Debt Requirement.” Unpublished paper,
University of Pennsylvania.
———. 2008. “Cash-in-the-Market Pricing and Optimal Resolution
of Bank Failures.” Review of Financial Studies 21, no. 6
(November): 2705-42.
Caprio, G., and D. Klingebiel. 1996. “Bank Insolvencies: Cross
Country Experience.” World Bank Policy Research Working
Papers, no. 1620, July.
———. 2008. “Information Contagion and Bank Herding.” Journal
of Money, Credit, and Banking 40, no. 1 (February): 215-31.
Chen, Y. 1999. “Banking Panics: The Role of the First-Come,
First-Served Rule and Information Externalities.” Journal of
Political Economy 107, no. 5 (October): 946-68.
Allen, F., and D. Gale. 1994. “Liquidity Preference, Market
Participation and Volatility of Asset Prices.” American
Economic Review 84, no. 4 (September): 933-55.
———. 1998. “Optimal Financial Crises.” Journal of Finance 53,
no. 4 (August): 1245-84.
———. 2000. “Financial Contagion.” Journal of Political
Economy 108, no. 1 (February): 1-33.
Cifuentes, R., G. Ferrucci, and H. S. Shin. 2005. “Liquidity Risk
and Contagion.” Journal of the European Economic
Association 3, no. 2-3 (April-May): 556-66.
Claessens, S., S. Djankov, and D. Klingebiel. 1999. “Financial
Restructuring in East Asia: Halfway There?” World Bank
Financial Sector Discussion Papers, no. 3, September.
Bank of Canada. 2010. “Contingent Capital and Bail-In Debt: Tools
for Bank Resolution.” Financial System Review, December.
DeYoung, R., M. Kowalik, and J. Reidhill. 2013. “A Theory of Failed
Bank Resolution: Technological Change and Political Economics.”
Journal of Financial Stability 9, no. 4 (December): 612-27.
Basel Committee on Banking Supervision. 2002. “Supervisory
Guidance on Dealing with Weak Banks.” Report of the Task Force
on Dealing with Weak Banks, March.
Elsinger, H., A. Lehar, and M. Summer. 2006. “Risk Assessment
for Banking Systems.” Management Science 52, no. 9
(September): 1301-14.
Beck, T. 2011. “Bank Resolution: A Conceptual Framework.” In
P. Delimatsis and N. Herger, eds., Financial Regulation
at the Crossroads: Implications for Supervision,
Institutional Design, and Trade. Alphen aan de Rijn:
Kluwer Law International.
European Commission. 2013. “An Important Step Towards a
Real Banking Union in Europe.” Statement by Commissioner
Michel Barnier following the Trilogue Agreement on the Creation
of the Single Supervisory Mechanism for the Eurozone, Brussels,
Belgium, March 19.
Bliss, R. R., and G. G. Kaufman. 2006. “U.S. Corporate and Bank
Insolvency Regimes: A Comparison and Evaluation.” Bepress
Legal Series, no. 1390.
Federal Deposit Insurance Corporation. 1998. Managing the Crisis:
The FDIC and RTC Experience 1980-1994. Washington, D.C.
Boot, A. 2000. “Relationship Banking: What Do We Know?” Journal
of Financial Intermediation 9, no. 1 (January): 7-25.
Brierley, P. 2009. “The U.K. Special Resolution Regime for Failing
Banks in an International Context.” Bank of England Financial
Stability Paper no. 5, July.
Financial Stability Board. 2011. “Key Attributes of Effective
Resolution Regimes for Financial Institutions.” October.
Available at http://www.financialstabilityboard.org/publications/
r_111104cc.pdf.
Flannery, M. 2002. “No Pain, No Gain? Effecting Market Discipline
via Reverse Convertible Debentures.” Unpublished paper,
University of Florida.
Calomiris, C. 1998. “Blueprints for a Global Financial Architecture,
Reform Proposals.” International Monetary Fund.
FRBNY Economic Policy Review / December 2014
171
References (Continued)
———. 2009. “Stabilizing Large Financial Institutions with
Contingent Capital Certificates.” Unpublished paper,
University of Florida.
Nier, E., J. Yang, T. Yorulmazer, and A. Alentorn. 2007. “Network
Models and Financial Stability.” Journal of Economic
Dynamics and Control 31, no. 6 (June): 2033-60.
Furfine, C. H. 1999. “Interbank Exposures: Quantifying the Risk of
Contagion.” BIS Working Papers, no. 70, June.
Osterman Jr., R. J., and J. Wigand. 2013. “Who Is Too Big to Fail?
Examining the Application of Title I of the Dodd-Frank Act
before the Subcommittee on Oversight and Investigations.”
Remarks delivered to the Committee on Financial Services,
U.S. House of Representatives, Washington, D.C., April 16.
Gibson, M. S. 2013. “Cross-Border Resolution.” Remarks delivered
to the Subcommittee on National Security and International
Trade and Finance, Committee on Banking, Housing, and Urban
Affairs, U.S. Senate, Washington, D.C., May 15.
Pennacchi, G. 2010. “A Structural Model of Contingent Bank Capital.”
Unpublished paper, University of Illinois College of Business.
Glasserman, P., and B. Nouri. 2012. “Contingent Capital with
a Capital-Ratio Trigger.” Management Science 58, no. 10
(October): 816-33.
Philippon, T., and P. Schnabl. 2013. “Efficient Recapitalization.”
Journal of Finance 68, no. 1 (February): 1-42.
Goldsmith-Pinkham, P., and T. Yorulmazer. 2010. “Liquidity, Bank
Runs and Bailouts: Spillover Effects during the Northern Rock
Episode.” Journal of Financial Services Research 37, no. 2-3
(June): 83-98.
Plosser, C. 2013. “Can We End Too Big to Fail?” Remarks delivered
at the Fourth Annual Simon New York City Conference "Reform
at a Crossroads: Economic Transformation in the Year Ahead,"
New York City, May 9.
Gorton, G., and L. Huang. 2004. “Liquidity, Efficiency and Bank
Bailouts.” American Economic Review 94, no. 3 (June): 455-83.
Rochet, J., and J. Tirole. 1996. “Interbank Lending and Systemic
Risk.” Journal of Money, Credit, and Banking 28, no. 4
(November): 733-62.
———. 2006. “Bank Panics and the Endogeneity of Central Banking.”
Journal of Monetary Economics 53, no. 7 (October): 1613-29.
Hoggarth, G., J. Reidhill, and P. Sinclair. 2004. “On the Resolution of
Banking Crises: Theory and Evidence.” Bank of England Working
Paper no. 229, September.
Honohan, P., and D. Klingebiel. 2000. “Controlling Fiscal Costs of
Banking Crises.” World Bank Working Paper no. 2441, September.
Santomero, A. M., and P. Hoffman. 1998. “Problem Bank Resolution:
Evaluating the Options.” Wharton Financial Institutions Center
Working Paper Series, no. 98-05, October.
Saunders, A., and B. Wilson. 1996. “Contagious Bank Runs:
Evidence from the 1929-1933 Period.” Journal of Financial
Intermediation 5, no. 4 (October): 409-23.
Hüpkes, E. 2004. “Insolvency: Why a Special Regime for Banks?”
Mimeo, Swiss National Bank.
Sheldon, G., and M. Maurer. 1998. “Interbank Lending and Systemic
Risk: An Empirical Analysis of Switzerland.” Swiss Journal of
Economics And Statistics 134, no. 4: 685-704.
Kahn, C., and J. Santos. 2005. “Endogenous Financial Fragility
and Bank Regulation.” Unpublished paper, Federal Reserve
Bank of New York.
Shin, H. S. 2009. “Reflections on Northern Rock: The Bank Run That
Heralded the Global Financial Crisis.” Journal of Economic
Perspectives 23, no. 1 (Winter): 101-19.
Leitner, Y. 2005. “Financial Networks: Contagion, Commitment,
and Private Sector Bailouts.” Journal of Finance 60, no. 6
(December): 2925-53.
Shleifer, A., and R. W. Vishny. 1992. “Liquidation Values and Debt
Capacity: A Market Equilibrium Approach.” Journal of
Finance 47, no. 4 (September): 1343-66.
Mailath, G. and L. Mester. 1994. “A Positive Analysis of Bank Closure.”
Journal of Financial Intermediation 3, no. 3 (June): 272-99.
Sundaresan, S., and Z. Wang. Forthcoming. “On the Design of
Contingent Capital with Market Trigger.” Journal of Finance.
172 Bank Resolution Concepts, Trade-offs, and Changes in Practices
References (Continued)
Upper, C., and A. Worms. 2002. “Estimating Bilateral Exposures in
the German Interbank Market: Is There a Danger of Contagion?”
Deutsche Bundesbank Discussion Papers, no. 09/02.
Wells, S. 2002. “U.K. Interbank Exposures: Systemic Risk
Implications.” Bank of England Financial Stability
Review, December.
Yorulmazer, T. 2009. “Lessons from the Resolution of the Swedish
Financial Crisis.” Unpublished paper, Federal Reserve Bank of
New York. Available at http://www.newyorkfed.org/research/
economists/yorulmazer/Lessons-Swedish-Crisis1.pdf.
———. 2014. “Case Studies on Disruptions during the Crisis.” Federal
Reserve Bank of New York Economic Policy Review 20, no. 1
(February) 17-28.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / December 2014
173
Michael J. Fleming and Asani Sarkar
The Failure Resolution
of Lehman Brothers
The experience of resolving Lehman in the
bankruptcy courts has led to an active debate
about the effectiveness of U.S. Chapter 11
proceedings for complex financial institutions.
Lehman’s poor pre-bankruptcy planning
may have substantially reduced the value of
Lehman’s estate and contributed to many
ensuing disputes with creditors.
For over-the-counter (OTC) derivatives
transactions, where much of the complexity
of Lehman’s bankruptcy resolution was
rooted, creditors’ recovery rate was below
historical averages for failed firms comparable
to Lehman.
The settlement of OTC derivatives was a long
and complex process, occurring on different
tracks for different groups of derivatives
creditors.
Some of the losses borne by Lehman
investors stemmed from the manner in which
Lehman failed and could have been avoided
in a more orderly process.
Michael J. Fleming is a vice president and Asani Sarkar an assistant vice
president at the Federal Reserve Bank of New York.
michael.fleming@ny.frb.org; asani.sarkar@ny.frb.org
1. Introduction
L
ehman Brothers Holdings Inc. (LBHI) filed for Chapter 11
bankruptcy on September 15, 2008, while its subsidiaries
did so over the subsequent months (see Exhibit 1 for Lehman’s
organizational structure).1 With 209 registered subsidiaries in
twenty-one countries, Lehman’s Chapter 11 filing was one of
the largest and most complex in history. Creditors filed about
$1.2 trillion of claims against the Lehman estate (LBHI,
“The State of the Estate,” September 22, 2010), which was
party to more than 900,000 derivatives contracts at the time
of bankruptcy.
Several bodies of law applied to Lehman’s various corporate
entities (Exhibit 2):
• The U.S. Bankruptcy Code applied to LBHI and its
subsidiaries.
• The Securities Investor Protection Act (SIPA) regime
applied to the insolvent broker-dealer, Lehman
Brothers Inc. (LBI).
• More than eighty jurisdictions’ insolvency laws applied to
the non-U.S. Lehman Brothers entities, such as Lehman’s
U.K.-based broker-dealer Lehman Brothers International
(Europe) (LBIE).
1
When referring to LBHI and all its subsidiaries as an ensemble, we use
“Lehman.” Otherwise, when referring to the holding company (subsidiary),
we use “LBHI” (the subsidiary name). Appendix A lists the acronyms and
initialisms used in the article.
The authors thank Tobias Adrian, Wilson Ervin, Sahil Godiwala, Anna Kovner,
Lisa Kraidin, Antoine Martin, James McAndrews, H
amid Mehran, João
Santos, Joseph Sommer, and Emily Warren for helpful discussions and/
or comments on earlier drafts as well as Samuel Antill, Weiling Liu, and
Parinitha Sastry for excellent research assistance. The views expressed in this
article are those of the authors and do not necessarily reflect the position of
the Federal Reserve Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
175
Exhibit 1
Organization Chart for Lehman’s U.S. and European Subsidiaries
Lehman Brothers
Holdings Inc.
(LBHI)
Lehman Brothers
Bancorp Inc.
Lehman Brothers
Commercial Corp.
(LBCC)
Lehman Brothers
Commercial Bank
Lehman Brothers
Bank, FSB
Lehman Brothers
Inc. (LBI)
Neuberger
Berman
Lehman Brothers
Holdings Plc, U.K.
Lehman Brothers
International Europe Ltd.
(LBIE)
Lehman Brothers
OTC Derivatives Inc.
(LOTC)
Other Lehman
European
subsidiaries
Aurora Loan
Services, LLC
Lehman Brothers
Derivatives Products Inc.
(LBDP)
Lehman Commercial
Lehman Brothers
Paper Inc.
Financial Products Inc.
(LCPI)
(LBFP)
Lehman Brothers
Special Financing Inc.
(LBSF)
Lehman Brothers
Commodity Services Inc.
(LBCS)
Sources: Derived from Valukas (2010).
Notes: The exhibit shows the organizational structure for Lehman Brothers’ U.S. and major European subsidiaries.
• The Federal Deposit Insurance Act applied to its state-
chartered bank and federally chartered thrift.
• U.S. state insurance laws applied to its insurance subsidiaries.
The failure of Lehman Brothers was associated with
substantial losses for its equity holders and creditors. The
experience of resolving Lehman in the bankruptcy courts
has since led to an active debate regarding the effectiveness
of U.S. Chapter 11 proceedings for complex financial
institutions. Some economists have suggested a modification
of C
hapter 11, called Chapter 14, to apply to all financial
companies exceeding $100 billion in consolidated assets
(Jackson 2012). In contrast, Title II of the Dodd-Frank Wall
Street Reform and Consumer Protection Act, passed in 2010,
creates an alternative resolution mechanism, the Orderly
Liquidation Authority, that expands the reach of the Federal
176
The Failure Resolution of Lehman Brothers
Deposit Insurance Corporation (FDIC) to resolve large nonbank financial institutions such as Lehman.
In this article, we examine the resolution of Lehman in
the U.S. Bankruptcy Court proceedings2 with a view toward
understanding the sources of complexity in its resolution to
thereby inform the debate on appropriate resolution mechanisms for complex financial institutions. Below are the main
steps involved in Lehman’s bankruptcy process (Exhibit 2):3
2
While this article focuses on the application of the U.S. Bankruptcy Code
to Lehman, we include two appendixes on the settlement of centrally cleared
derivatives (Appendix B) and the resolution of LBI under the SIPA regime
(Appendix C). Moreover, in a companion article, we discuss the value
destruction resulting from the Lehman bankruptcy (Fleming and Sarkar 2014).
3
At various points during the bankruptcy proceedings, the Lehman estate
also brought a number of motions and adversary proceedings to facilitate the
case, to determine liabilities, and to recover or sell assets, as shown in Exhibit 2.
Exhibit 2
Chapter 11 Bankruptcy Process for Lehman Brothers
Exhibit 2
Exhibit 2
Chapter
11 Bankruptcy Process for Lehman Brothers
Chapter
Bankruptcy
Process for Lehman Brothers
Holding11
company
LBHI
Holding company LBHI
Holding company LBHI
Unregulated legal entities
LBSF
LBDP
LBDP
Unregulated legal entities
Unregulated legal entities
LCPI
LBFP
LBSF
LBSF
LBDP
LBDP
LBDP
LBDP
LCPI
LBFP
Pre-bankruptcy
LBFP planning LCPI
Adversary
proceedings
(Lehman
Adversary
debtors,
proceedings
Adversary
creditors,
(Lehman
proceedings
Barclays,
debtors,
(Lehman
clearing
creditors,
debtors,
agents,
etc.)
Barclays,
creditors,
clearing
Barclays,
agents,
etc.)
clearing
agents, etc.)
Chapter 15
proceeding
Chapter 15
proceeding
Chapter
15
proceeding
LOTC
LOTC
LOTC
Pre-bankruptcy planning
Pre-bankruptcy
planning
Debtors file
for Chapter 11
bankruptcy
Debtors file
for
Chapter
Debtors
file11
bankruptcy
for
Chapter 11
bankruptcy
Automatic
First
stay for all
day
assets except
Automatic
motions
QFCs
First
stay
for all
Automatic
day
First
assets
except
stay for
all
motions
day
QFCs
assets
except
motions
Plan of
QFCs
reorganization
Plan of
reorganization
Plan of
reorganization
Confirmation
of plan
Confirmation
of plan
Confirmation
of plan
U.S.
brokerdealer LBI
U.S.
brokerU.S.
brokerdealer
SIPALBI
dealer LBI
proceeding
or special
SIPA
provision
proceeding
SIPA 7
of
Chapter
or special
proceeding
provision
or
special
ofprovision
Chapter 7
of Chapter 7
Commodity
broker LBI
Closing
and netting
of QFCs
Closing
and
netting
Closing
of QFCs
and
netting
of QFCs
Ladder of unsecured
creditor priorities:
• DIP financers
Ladder of unsecured
• Taxing authorities
creditor
Ladder
of priorities:
unsecured
• Unsecured
creditors
•creditor
DIP financers
priorities:
• Shareholders
• Taxing
authorities
DIP financers
• Unsecured
creditors
Taxing authorities
• Shareholders
Unsecured creditors
• Shareholders
LBI sold in 363 sale
to Barclays
Assets
liquidated
LBI sold in 363 sale
to Barclays
LBI sold
in 363 sale
to Barclays
Assets
liquidated
Assets
liquidated
Commodity
broker LBI
Commodity
broker LBI
Special
provision of
Chapter 7
Special
provision
Specialof
Chapter of
7
provision
Chapter 7
Foreign
brokerdealer
Foreign
LBIE
brokerForeign
dealer
brokerLBIE
dealer
LBIE
Insolvency
or regulatory
process in
Insolvency
United
orInsolvency
regulatory
Kingdom
in
orprocess
regulatory
United in
process
Kingdom
United
Kingdom
Insured
depository
institution
Insured
depository
Insured
institution
depository
institution
FDIC process
FDIC process
FDIC process
Insurance
company
Insurance
company
Insurance
company
State
resolution
process
State
resolution
State
process
resolution
process
Source: Derived from U.S. Government Accountability Office (2011).
Notes: The exhibit shows the bankruptcy process for Lehman Brothers and its affiliates. LBHI is Lehman Brothers Holdings Inc.; LBSF is Lehman Brothers
Special Financing;
LBDP
Lehman Brothers
Derivatives
Products;
LBFP is Lehman Brothers Financial Products; LCPI is Lehman Commercial Paper Inc.;
Source:
Derived from
U.S.isGovernment
Accountability
Office
(2011).
LOTC
isDerived
Lehman
Brothers
OTC
Derivatives;
LBI for
is Lehman
is Lehman
Brothers
International
(Europe);Inc.;
SIPALBSF
is Securities
Source:The
from
U.S.the
Government
Accountability
OfficeBrothers
(2011). Inc.;
Notes:
exhibit
shows
bankruptcy
process
Lehman
Brothers
andLBIE
its affiliates.
LBHI
is Lehman
Brothers Holdings
is Lehman Brothers
Investor
Protection
Act;
FDIC
is
Federal
Deposit
Insurance
Corporation;
QFCs
are
qualified
financial
contracts;
DIP
is
debtor
in
possession.
Chapter 15
SpecialThe
Financing;
LBDP the
is Lehman
Brothers
Derivatives
Products;
LBFP
Lehman
Brothers
LCPI is Lehman
Commercial
Inc.;
Notes:
exhibit shows
bankruptcy
process
for Lehman
Brothers
andisits
affiliates.
LBHI isFinancial
LehmanProducts;
Brothers Holdings
Inc.; LBSF
is LehmanPaper
Brothers
of
the Bankruptcy
Code governs
judicial cross-border
coordination.
Sale
the company,
inBrothers
whole orInternational
in part, is commonly
called
Section
363 sale because
LOTC
isFinancing;
Lehman Brothers
Derivatives;
is Lehman
Brothers
Inc.;ofisLBIE
is Lehman
(Europe);
SIPA aisCommercial
Securities
Special
LBDP is OTC
Lehman
Brothers LBI
Derivatives
Products;
LBFP
Lehman
Brothers
Financial Products; LCPI
is Lehman
Paper Inc.;
that
is
the
section
of
the
Bankruptcy
Code
that
applies
to
sales
that
are
free
and
clear
of
creditor
claims.
Investoris Protection
Act; FDIC
is Derivatives;
Federal Deposit
qualifiedBrothers
financialInternational
contracts; DIP
is debtor
in possession.
LOTC
Lehman Brothers
OTC
LBI Insurance
is LehmanCorporation;
Brothers Inc.;QFCs
LBIE are
is Lehman
(Europe);
SIPA
is SecuritiesChapter 15
of
the Bankruptcy
governs
judicial Deposit
cross-border
coordination.
Sale of
the company,
in whole
or in
part, is commonly
called
a Section 363
sale because
Investor
ProtectionCode
Act; FDIC
is Federal
Insurance
Corporation;
QFCs
are qualified
financial
contracts;
DIP is debtor
in possession.
Chapter
15
that
the section of
the Bankruptcy
Codecross-border
that applies coordination.
to sales that are
free
clear of creditor
claims.
of theis Bankruptcy
Code
governs judicial
Sale
of and
the company,
in whole
or in part, is commonly called a Section 363 sale because
that is the section of the Bankruptcy Code that applies to sales that are free and clear of creditor claims.
• Pre-bankruptcy planning, including searching for potential
buyers and preparing for filing of a bankruptcy petition;
• Closing and netting out qualified financial contracts
(QFCs);
• First-day-of-bankruptcy motions to obtain funding in order
to operate businesses during bankruptcy and permission
to use cash collateral on which secured creditors had claims;
• Section 363 asset sales;4
4
Sale of the company, in whole or in part, is commonly called a Section 363 sale
because this section of the Bankruptcy Code applies to sales that are free and
clear of creditor claims. Asset sales also occur as part of the confirmation plan.
FRBNY Economic Policy Review / December 2014
177
• Establishing the total amount owed to creditors through
the claims process, by providing reports on the debtor’s
financial condition and reviewing (and objecting to, if
necessary) creditor claims;
• Filing a plan of reorganization5 after negotiations with
significant creditors, along with a disclosure statement to
inform creditors about the plan;
• Confirming the plan to settle creditor claims through
voting by creditors and a confirmation hearing; 6 and
• Making payments to creditors under the plan.
We discuss Lehman’s pre-bankruptcy planning, its
funding sources during bankruptcy, the settlement of
QFCs, the claims process, and the amounts recovered by
different creditor groups. The bulk of our study is devoted
to the settlement of Lehman’s creditor and c ounterparty
claims, especially those relating to over-the-counter
(OTC) derivatives. We focus on derivatives because we
find that much of the complexity of Lehman’s b
ankruptcy
was rooted in the settlement procedures for its OTC
derivatives positions. Moreover, derivatives receive special
treatment under the U.S. Bankruptcy Code through exemptions or “safe harbor” from several provisions of the
code (for example, exemption from the automatic stay; see
Appendix D for a more complete discussion of safe harbor
provisions). However, questions have been raised regarding the desirability of providing these exceptions. For example, Andrew Gracie, the executive director of the Bank
of E
ngland’s special resolution unit argues that the onset
of a bank resolution should not, by itself, be considered
an event of default that allows counterparties to quickly
terminate derivative contracts, as happened with Lehman.7
By providing a detailed description of the use of safe harbor provisions and other derivatives settlement procedures
in the Lehman bankruptcy, our study may help inform the
discussion on the role of derivatives in bankruptcy.
5
In Lehman’s case, the reorganization plan resulted in liquidation of the
company. There are advantages to using Chapter 11, rather than Chapter 7,
for liquidation (for example, the debtor, rather than a trustee, has control over
the sale process). However, failed Chapter 11 cases are often converted to
Chapter 7 cases.
6
Lehman was also involved in Chapter 15 cases, which were ancillary to
the U.S. bankruptcy case and involved cross-border insolvency. Such cases
allowed Lehman’s foreign creditors (who had claims against a Lehman
foreign subsidiary in a foreign judicial or administrative proceeding) to be
recognized by the U.S. Bankruptcy Court and to participate in Lehman’s
U.S. bankruptcy case. See http://www.uscourts.gov/FederalCourts/
Bankruptcy/BankruptcyBasics/Chapter15.aspx. In this article, we do not
cover cross-border issues, although to the extent that the resolution of
Lehman’s U.K. broker-dealer affected the SIPA proceedings, these are discussed
in Appendix C.
7
See http://www.bloomberg.com/news/2014-03-04/boe-seeks-derivativespact-to-prevent-a-repeat-of-lehman-cascade.html.
178
The Failure Resolution of Lehman Brothers
The payout ratio to Lehman’s creditors was initially
estimated to be about 21 percent on estimated allowable
claims of $362 billion, implying a loss to creditors and
counter- parties of roughly $286 billion. Actual distributions to date appear to have exceeded initial estimates,
although some of the amount distributed has gone to
other Lehman creditors rather than third-party creditors.
Comparison with historical experience indicates that
the recovery rate for LBHI’s senior unsecured creditors
has been below average so far, even after accounting for
possible mitigating factors (for example, the state of the
economy and the credit cycle). However, recovery rates
varied across creditor groups. Creditors of three Lehman
derivatives entities received full recovery on their claims,
and counterparties of centrally cleared securities were
mostly made whole. In contrast, many of Lehman’s OTC
derivatives’ counterparties suffered substantial losses.
Some of the losses borne by Lehman investors e manated
from the manner in which Lehman failed and could have
been avoided in a more orderly liquidation process. The
bankruptcy was poorly planned, for example, which may
have s ubstantially reduced the value of Lehman’s estate
(Valukas 2010, p. 725) and contributed to ensuing litigation
with creditors.
Creditor losses would have been more substantial without
the ability of LBI, the U.S. brokerage subsidiary of LBHI and
subsequently of Barclays Plc, to finance positions through
the Federal Reserve’s (Fed) liquidity facilities. Such financing
was critical to the relatively smooth transfer of LBI customer
accounts to Barclays and the preservation of firm value.
Since then, the Dodd-Frank Act has circumscribed the ability
of the Fed to act as lender of last resort to the same extent
that it did during the financial crisis.
We assess the effectiveness of the settlement procedures
with respect to their speed, predictability, and transparency.
We find that the speed of resolution varied across claimant
groups. Retail OTC derivatives counterparties of Lehman
terminated their contracts within weeks of LBHI’s bankruptcy filing under the safe harbor provisions, but final
settlement of their claims remains incomplete.8 In contrast,
derivatives c ontracts of large, institutional counterparties
(which constituted a small share of Lehman’s derivative
contracts by number, but a significant share by value) took
several years to terminate, let alone finally settle.
Regarding the predictability of the settlement process, while existing case law provided a useful starting
point for the Lehman resolution, the court provided new
8
As explained in Appendix D, while termination is the first step in settling an
OTC derivatives position, final settlement of terminated derivatives contracts
requires further steps, such as valuing transactions.
i nterpretations of provisions in the Bankruptcy Code
(regarding, for example, some aspects of the safe harbor
provisions for derivatives). In part, this reflected the importance of complex financial s ecurities to which Lehman
was a party. The bankruptcy court had to analyze these
securities for the first time and s ometimes came out with
controversial judgments that s urprised many observers.
Finally, regarding transparency, we find that while the
L ehman estate provided substantial ongoing i nformation
on the progress of resolution, the information was
s ometimes either incomplete or reported in a piecemeal
manner that made it difficult to obtain an integrated view
of bankruptcy outcomes.
In the remainder of the article, we discuss the effectiveness of Lehman’s pre-bankruptcy planning (Section 2),
funding during the first week of bankruptcy (Section 3),
the settlement of financial contracts with an emphasis on
QFCs (Section 4), and creditors’ recovery rates under
Chapter 11 (Section 5). Section 6 summarizes our findings.
2. Pre-Bankruptcy Planning
Companies facing potential bankruptcy find it
advantageous to consult a Chapter 11 attorney early so
that there is more time to put together a plan and
assemble a team of professionals (such as counsel and
financial advisors) to work with the company. An important goal of pre-petition planning is to maintain the
operations of the business during the bankruptcy process
(for example, by arranging for funding and preparing an
operating budget to conserve cash).
The Lehman bankruptcy was considered disorderly, in
part because the institution did not plan sufficiently for
the p
ossibility of bankruptcy. Indeed, Lehman’s actions
were not those of a company husbanding resources in
anticipation of bankruptcy. For example, Lehman continued to repurchase shares at the beginning of 2008 and
decided against h
iring bankruptcy counsel in August 2008
(Valukas 2010, p. 718). Management did not seriously consider bankruptcy until a few days before filing, and Lehman did not try to sell its subsidiaries until the week before
its collapse (U.S. Government Accountability Office 2011).9
Lehman consciously avoided bankruptcy planning owing
to continuing interest from strategic partners and its belief
that such planning would be a self-fulfilling prophecy
9
Lehman had discussions with Bank of America (for a proposed merger
between the two companies) in July 2008 and again in September 2008,
when U.S. Treasury Secretary Henry Paulson urged Bank of America to buy
Lehman (Valukas 2010, p. 697).
(Valukas 2010, p. 718).
The three or four days prior to LBHI’s bankruptcy filing
were filled with confusion and indecision. Lehman engaged
bankruptcy counsel on September 10, 2008, and preparation
for filing of the bankruptcy petition began the following day
(Valukas 2010, p. 719). At the same time, however, Lehman
continued to believe that it would be rescued. Indeed, as late
as September 14, 2008, Lehman contemplated a six-month
period to unwind its positions, during which it would
employ many people (Valukas 2010, p. 371).
A key step in planning for a Chapter 11 bankruptcy
filing is to have certain “first day” motions and orders
ready so that the judge can consider them at the beginning of the case. These orders facilitate the operational
aspects of the b ankruptcy filing and contribute toward
a prompter and more orderly resolution (Wasserman
2006). LBHI and its b ankruptcy counsel initially filed few
of the typical first-day motions that seek the bankruptcy
court’s authorization to carry on the many facets of “business
as usual” that otherwise would be prohibited by various
Bankruptcy Code provisions (for example, maintain accounts
and current cash management systems, affirm clearinghouse
contracts, and so on; see Azarchs and Sprinzen [2008]).
Similarly, LBHI’s affidavit accompanying its bankruptcy
petition was unusually brief. Typically, these affidavits set out
in some detail the debtor’s business rationale for its first-day
motions and provide the outlines of its Chapter 11 strategy. In
Lehman’s case, other than “preserve its assets and maximize
value for the benefit of all stakeholders,” little was set out
(Azarchs and Sprinzen 2008). The lack of first-day motions
and the sparseness of the debtor’s affidavit suggest a lack of
preparedness for bankruptcy.
The abruptness of LBHI’s filing is reported to have
reduced the value of Lehman’s estate by as much as $75 billion
(Valukas 2010, p. 725). For example, 70 percent of derivatives
receivables worth $48 billion were lost that could otherwise
have been unwound.10 The lack of planning also contributed
to many ensuing disputes with creditors.
10
An alternative view is that the Lehman estate did not suffer any substantial
loss on its derivatives position since LBHI’s counterparties initially overstated
some of their claims, which were subsequently overturned by the bankruptcy
court (U.S. Government Accountability Office 2013).
FRBNY Economic Policy Review / December 2014
179
3. Funding in the First Week
of Bankruptcy
Unlike LBHI, LBI did not file for bankruptcy on September 15,
2008, because it expected to conduct an orderly liquidation by
unwinding its repos and matched books while attempting to
find a buyer (Valukas 2010, p. 2117). Ownership of LBI’s assets
was transferred to Barclays on September 22. However, in
order to remain a going concern, LBI needed liquidity between
September 15 and 22. Absent such liquidity, the sale would
have failed, further impairing the value of Lehman’s e state.
At, and just after, the time of LBHI’s bankruptcy filing,
LBI’s cash position was precarious (“Trustee’s Preliminary
Investigation Report and Recommendations,” August 25,
2010). More than 90 percent of LBI’s assets had been composed
of reverse repos, stock borrowing agreements, and financial
instruments owned. Reverse repos and securities loans had
declined since May 2008 (Panel A of Table 1). Tri-party repo
funding in particular had dropped from $80 billion on May 31,
2008, to $650 million on September 19, 2008. Failed transactions and the failure of counterparties to return margin posted
by LBI harmed its cash position. Finally, customer and prime
broker accounts moved to other broker-dealers, while clearing
firms required additional collateral, deposits, and margins.11
In order to operate until its sale was completed, LBI had
to rely on other funding sources, including the Fed’s liquidity
facilities and advances by Barclays and LBI’s clearing agents.
3.1 Post-Petition Financing of LBI by the Fed
In connection with LBHI’s preparations for bankruptcy
petition, the Fed, acting in its capacity as lender of last resort,
advised Lehman that it would provide up to two weeks of
overnight secured financing through the Primary Dealer
Credit Facility (PDCF) to facilitate an orderly unwind of LBI
(Valukas 2010, p. 2118). Without Fed funding, LBI’s c ustomers
would have faced long delays in accessing their accounts
while their claims were resolved in the SIPA proceedings (as
discussed further in Appendix C).
11
An additional factor, noted by Duffie, Li, and Lubke (2010), is the use of
novations by LBHI’s counterparties (whereby they would exit their positions
by assigning them to other dealers) in the days before bankruptcy. These
novations depleted LBHI’s cash reserves and, effectively, those of LBI (since
LBHI was the main source of LBI’s funding).This occurred because when
Lehman’s original dealer counterparty, through novation, transferred its
position to another dealer, Lehman lost the associated “independent amount”
of collateral (which functions similar to an initial margin). The collateral was
not replaced because initial margins are not posted in dealer-to-dealer trades.
180
The Failure Resolution of Lehman Brothers
On September 14, 2008, the Fed expanded the set of
collateral acceptable at the PDCF to include all tri-party-
eligible collateral.12 Under the PDCF, the Fed extended
between $20 billion and $28 billion per day to LBI from
September 15 to September 17, 2008 (Panel B of Table 1).
However, the Fed limited the collateral LBI could pledge to
what it had in its clearance box at JPMorgan Chase (JPMC)
on September 12 and also imposed higher haircuts on LBI
than on other dealers (Valukas 2010, p. 2119).13 Nevertheless,
LBI borrowed against a wide variety of collateral, such as
asset-backed securities and equity (Panel B of Table 1).
In addition to the PDCF, the Fed had introduced the
Term Securities Lending Facility (TSLF) and single-tranche
term repurchase agreements in March 2008 to address the
liquidity pressures in secured funding markets.14 While LBI
had outstanding borrowing of $18.5 billion from the TSLF at
the time of bankruptcy, it did not undertake new borrowing
from the TSLF after bankruptcy. Similarly, LBI had singletranche term repos outstanding of $2 billion at the time of
bankruptcy, but did not undertake new borrowing through
the program after bankruptcy.
3.2 Post-Petition Financing of Lehman
by Barclays
On September 17, 2008, the Fed and Barclays formally agreed
that Barclays would replace the Fed as a source of secured
funding for LBI (Valukas 2010, p. 2162). On September 18,
in exchange for $46.2 billion in cash, the Fed delivered LBI
collateral to Barclays and advised it of the option to finance
the collateral at the PDCF (Valukas 2010, p. 2165). Between
September 18 and September 22, 2008, Barclays borrowed up
to $48 billion from the PDCF and $8 billion from the TSLF
(Panel C of Table 1).
12
Eligible collateral originally comprised Fed-eligible collateral plus
investment-grade corporate securities, municipal securities, mortgage-backed
securities, and asset-backed securities. See http://www.federalreserve.gov/
newsevents/press/monetary/20080914a.htm.
13
Clearance box assets are securities that were held in LBI’s “clearing box
accounts” at JPMC. These assets facilitated securities trading by providing
collateral against which open trading positions could be secured.
14
For the Fed’s announcement of the TSLF program, see http://federalreserve
.gov/newsevents/press/monetary/20080311a.htm. Under single-tranche
repurchase agreements, the Fed’s Open Market Trading Desk lent money in
the form of term twenty-eight-day repurchase agreements against Treasury,
agency debt, or agency mortgage-backed securities. Dealers could borrow
against all three types of collateral, which constituted a single tranche, as
opposed to the Desk’s conventional repurchase arrangements whereby each
type of collateral constitutes a separate tranche. See http://www.newyorkfed
.org/markets/operating_policy_030708.html for further details.
Table 1
Funding for Lehman around the First Week of Its Chapter 11 Filing
Panel A: Short-term Assets of LBI, May 31–September 19, 2008
Billions of dollars
May 31, 2008
August 31, 2008
September 19, 2008
141.2
143.5
11.1
87.7
68.2
41.8
228.8
211.6
121.9
Reverse repos
Securities loans
Repos plus securities loans
Panel B: Borrowing by LBI, September 15-17, 2008
Share of Collateral Pledged (Percent)
Source of
Funding
Type of
Funding
Amount
(Billions of
Dollars)
UST/
Agency
Securities
Agency
MBS
09/15/2008b
Fed
PDCF
28.0
13.1
7.0
5.0
41.8
10.0
09/16/2008c
Fed
PDCF
19.7
6.5
0.0
9.6
53.9
1.7
09/17/2008d
Fed
PDCF
20.4
16.4
13.3
2.1
31.5
0.4
09/15/2008
Barclays
Tri-party
repo
15.8
Not known
09/16/2008
Barclays
Tri-party
repo
15.8
Not known
09/17/2008
Barclays
Tri-party
repo
15.8
Not known
Loan Date
Private-Label Corporate Municipal
MBS
Bonds
Bonds
ABS
Equity
Othera
12.2
7.6
3.3
14.9
9.0
4.4
16.6
18.6
1.1
Equity
Othera
Panel C: Borrowing by Barclays, September 18-22, 2008
Share of Collateral Pledged (Percent)
Source of
Funding
Type of
Funding
09/18/2008
09/18/2008
Fed
Fed
09/19/2008e
09/19/2008e
Fed
Fed
09/22/2008f
Fed
PDCF
TSLF
Schedule 2
PDCF
TSLF
Schedule 1
PDCF
Loan Date
Amount
(Billions
of Dollars)
UST/
Agency
Securities
Agency
MBS
47.9
5.0
14.7
0.0
52.8
45.4
5.9
20.1
7.4
0.0
0.5
0.0
3.1
34.4
15.5
NE
0.1
NE
16.0
2.7
0.8
0.0
10.3
2.0
11.9
35.6
20.3
27.4
2.7
0.0
3.4
34.4
50.4
NE
0.3
NE
16.0
0.4
11.0
10.4
20.4
2.7
3.7
51.1
0.3
Private-Label Corporate Municipal
MBS
Bonds
Bonds
ABS
Sources: Trustee’s Preliminary Investigation Report and Recommendations (2010), Valukas (2010), and http://federalreserve.gov/newsevents/
reform_transaction.htm.
Notes: LBI is Lehman Brothers Inc.; UST is U.S. Treasury; MBS is mortgage-backed securities; ABS is asset-backed securities; Fed is Federal Reserve;
PDCF is Primary Dealer Credit Facility; TSLF is Term Securities Lending Facility; NE is not eligible.
For PDCF, the “other” category includes international securities (securities issued by non-U.S. entities, government, and private sources, including
supranational agencies) and other eligible collateral.
b
Lehman and Barclays begin to negotiate sale of LBI’s business and assets to Barclays.
a
c
Lehman and Barclays execute Asset Purchase Agreement, providing for sale to Barclays of selected Lehman assets.
d
Lehman asks Bankruptcy Court to schedule sale hearing and establish sale procedures.
Bankruptcy Court holds sale hearing to consider proposed sale of LBI to Barclays.
f
Barclays buys LBI, and sale transaction is closed. Almost all of LBI assets and employees are transferred to Barclays.
e
FRBNY Economic Policy Review / December 2014
181
Barclays also provided overnight funding to LBI of
$15.8 billion through tri-party repo transactions between
September 15 and September 17 (Panel B of Table 1).
And on September 17, Barclays provided $450 million in
debtor-in-possession financing to LBHI secured by LBHI’s
assets in Neuberger Berman (Azarchs and Sprinzen 2008).
Funds under the facility helped sustain LBHI’s businesses
pending the completion of LBI’s sale.
3.3 Post-Petition Financing by LBI’s
Clearing Agents
JPMorgan Chase and Citibank advanced credit to LBI
after the bankruptcy of LBHI, allowing LBI to clear trades
and obtain funding. For example, at the urging of the Fed
and LBHI, JPMC made clearing advances to unwind LBI’s
outstanding tri-party repos worth $87 billion on September 15
and substantial additional amounts on the following day to
“avoid financial market disruption” (LBHI, “Debtors versus
JPMorgan Chase Bank, N.A.,” April 19, 2012). LBI was a party
to tri-party term repos that continued to perform, and it
obtained overnight funding through general collateral finance
(GCF) repos (Valukas 2010, p. 2124).
JPMC and Citibank were faced with requests for advances
after the bankruptcy filing of LBHI. Although they may
have had pre-petition secured claims against LBHI under
its g uarantees, these guarantees were cut off by the filing
and would not cover later events. The court confirmed that
their new, post-petition advances would continue to benefit
from the pre-petition guarantees under securities contracts
and thereby allowed LBI to continue clearing and settling
securities trades until its sale.15
3.4 Sale of LBI to Barclays
The Section 363 sale of LBI to Barclays (Exhibit 2) illustrates
the complexities of an expedited sale of a large financial
institution during bankruptcy under the Bankruptcy Code.
For example, the Fed had to finance LBI temporarily and then
arrange for Barclays to replace it, as discussed previously.
Later, Barclays argued that it had not agreed to purchase some
of the collateral that it was being asked to finance, leading
to disputes with its clearing agent JPMC and also with LBI
that persisted and threatened to derail the transaction during
the weekend following September 19, 2008 (when the sale of
LBI to Barclays closed). Eventually, a resolution was reached
with the help of the Fed and with the Depository Trust and
Clearing Corporation (DTCC) agreeing to clear LBI trades
for less than the required collateral (Valukas 2010, p. 2197).16
Even after the sale closed, unsecured creditors tried to get the
sale order overturned.
4. Settlement of Lehman’s OTC
Derivatives Positions
Lehman traded in equities, fixed-income securities, and
derivatives in U.S. and international markets. In the
United States, many of these securities (such as equity, listed
corporate and municipal bonds, U.S. government debt, and
certain derivatives contracts) are centrally cleared, and their
settlement occurred outside of the Chapter 11 bankruptcy
process. Where Lehman acted as a broker on behalf of
retail or wholesale clients and the securities were centrally
cleared, the central clearinghouse was the client’s counterparty. Accordingly, the central counterparties (CCPs) acted
on behalf of the clients to either close out or transfer their
accounts to third-party brokers. Where Lehman acted for
its own account, the CCPs were Lehman’s counterparty,
and they generally closed out Lehman’s house (proprietary)
positions. Since our focus is on Lehman’s resolution under
the U.S. Chapter 11 Code, we relegate discussion of Lehman’s
centrally cleared positions to Appendix B.
The remainder of this section describes the settlement of
Lehman’s OTC derivatives contracts (for example, interest
rate swaps) that were bilaterally cleared (Exhibit 3).17 Prior
to bankruptcy, Lehman’s global derivatives position was
estimated at $35 trillion in notional value, accounting
for about 5 percent of derivatives transactions globally
16
Specifically, DTCC agreed to clear LBI trades even though the available
collateral was $6 billion less than what it had previously required.
17
15
The court also denied the rights of other parties, such as Bank of America
and Swedbank AB, a Swedish bank and creditor to LBHI, to set off Lehman’s
pre-petition obligations against its cash deposit accounts, thus allowing
Lehman to preserve cash. Swedbank sought to offset Lehman’s payment
obligations under pre-petition swaps with deposits Lehman had made at
Swedbank post-petition. Bank of America seized Lehman’s account funds,
which were unrelated to safe harbor transactions.
182
The Failure Resolution of Lehman Brothers
A derivatives contract is an International Swaps and Derivatives Association
(ISDA) Master Agreement, supplemented with a schedule. The Master
Agreement and schedule collectively set forth the fundamental contractual
terms of all derivatives transactions that are executed between the parties.
Each individual transaction is documented with a confirmation. There may be
several confirmations (corresponding to individual derivatives transactions)
under a single Master Agreement and schedule (Durham 2010). Hence, there
will typically be multiple trades associated with each derivatives contract.
Exhibit 3
Lehman’s Derivatives Settlement Procedures
All derivatives positions
OTC positions that are not
centrally cleared
Centrally cleared exchangetraded and OTC positions
Contract terminates
automatically or CP chooses
to terminate early
CP defers
termination
CCPs suspend or limit
market access to
Lehman
Lehman reconciles
claims and values
each transaction
Court approves Alternative
Dispute Resolution
mechanism for settling claims
CCPs net and
liquidate
positions
CP makes
or receives
payment
Court approves
expedited settlement
procedures
Lehman and big bank CPs
negotiate derivatives
claims framework
Lehman assigns claims to
third parties or arranges for
mutual termination
Net amount due to or
from CP determined
Net amount due to or
from CP determined
Net amount due to or
from CP determined
CP makes or
receives payment
CP makes or
receives payment
CP makes or
receives payment
Source: Authors’ compilation.
Notes: The exhibit shows the detailed settlement procedure for derivatives contracts of Lehman Brothers. OTC is over-the-counter; CP is counterparty;
CCP is central counterparty.
(Summe 2012).18 Its OTC derivatives positions represented
96 percent of the net worth of its derivatives-related entities
(Panel A of Table 2). The settlement of these contracts under
18
Outside of the United States, derivatives transactions were executed
through LBIE.
the Chapter 11 provisions proved challenging, partly owing
to the inherent complexity of these procedures and to the
presence of large and global derivatives counterparties, as
discussed below.
Concern over the size of Lehman’s OTC derivatives positions led to a special trading session on September 14, 2008,
FRBNY Economic Policy Review / December 2014
183
Table 2
Settlement of Lehman’s Over-the-Counter Derivatives Contracts
Panel A: Lehman Derivative Positions, at Time of Bankruptcy
Net Wortha (Billions of Dollars)
Share of Net Worth (Percent)
All positions
21.0
100.0
OTC positions
20.3
96.2
0.8
3.8
Exchange-traded positions
Panel B: Termination of Derivative Claims
Contract
Number
Initial position
Not Terminated (Percent)
> 6,000
b
Transactions
Number
100.0
Terminated as of Nov. 13, 2008
Not Terminated (Percent)
> 900,000
100.0
—
733,000
23.8
Not terminated as of Jan. 2, 2009
2,667
43.6b
18,000
2.0b
Not terminated as of June 17, 2009
1,068
16.9
5,858
0.5b
b
Panel C: Timeline of Final Settlement of Derivative Claims
Settled as of:
Contracts Reconciled (Percent)
Contracts Valued
(Percent)
Contracts Finally Settled
(Percent)
Estimated Number of Contracts Not Finally Settledc
07/31/2009
45
09/16/2009
53
35
6
5,960
44
11
11/05/2009
5,643
61
50
17
5,262
09/30/2010
95
87
46
3,449
03/31/2011
99
99
59
2,631
12/31/2012
—
—
84
1,014
Panel D: Derivative Claims of Large (“Big Bank”) Counterparties, January 13, 2011
Number of Trades
Initial position, all counterparties
Claims
(Billions of Dollars, Except as Noted)
Number of Contractsd
961,436e
45.31
2,961
Finally settled, all counterparties
69,684
5.04
1,561
Outstanding, all counterparties
891,752
40.37
1,400
Outstanding, thirty largest counterparties
817,221
21.75
148
85
48.00
5
Share of remaining, thirty largest counterparties (percent)
Sources: Debtors’ Disclosure Statement for First Amended Joint Chapter 11 Plan (January 25, 2011); Debtors’ Disclosure Statement for Second Amended
Joint Chapter 11 Plan (June 30, 2011); Debtors’ Disclosure Statement for Third Amended Joint Chapter 11 Plan (August 31, 2011); Lehman Brothers
Holdings Inc.: Debtor’s Motion (November 13, 2008, and January 16, 2009), §341 Meeting (January 29, 2009, and July 8, 2009), State of the Estate (November
18, 2009), Plan Status Report (January 13, 2011), 2013+ Cash Flow Estimates (July 23, 2013); Valukas (2010).
a
Amount equals the value of assets minus liabilities of LBHI-controlled derivative entities.
b
Different numbers were reported for total number of contracts and trades in different reports. Shares are based on the numbers reported in the associated reports.
c
Amount is based on an assumption of 6,340 derivative contracts at the beginning of bankruptcy.
d
Number of contracts excludes the number of guarantee claims (that is, claims based on guarantees by LBHI).
e
Number of trades does not correspond to that reported in Panel B as it comes from a report at a different time, and adjustments were made by the
estate in the interim.
184
The Failure Resolution of Lehman Brothers
organized by major market participants to net their mutually
offsetting positions. However, the netting effort largely failed
as there was little trading during the session.19 LBHI filed
for bankruptcy the following day, but Lehman’s derivatives
entities did so only some days later.20 However, since LBHI
was the credit support party for almost all of Lehman’s
derivatives transactions, its bankruptcy filing constituted a
default event under the ISDA Master Agreement (Appendix D
provides background on the settlement of derivatives in
bankruptcy). More than 6,000 derivatives claims involving
more than 900,000 transactions were filed against Lehman
and its affiliates.21 Counterparties that had terminated their
derivatives contracts or otherwise had claims against the estate
were required, by October 22, 2009, to file a special Derivative
Questionnaire and to provide a valuation statement for any
collateral, specify any unpaid amounts, and supply their
derivatives valuation methodology and supporting quotations.
The settlement of Lehman’s OTC derivatives positions
proceeded along three tracks (Exhibit 3). Most derivatives
contracts were terminated early, under the safe harbor
provisions that provide statutory exceptions to the automatic
stay of debt in bankruptcy (see Appendix D). However, outof-the-money counterparties, which owed money to Lehman,
typically chose not to terminate their contracts. Even after
termination, the parties had to agree to a termination value
of their trades, which proved difficult in illiquid markets and
especially so for large positions; therefore, settlement with
large (“big bank”) counterparties proceeded along a third
track. We describe the settlement of OTC derivatives for each
of these three cases.
option could elect to terminate their transactions by giving
written notice.
The majority of Lehman’s derivatives contracts, by number
(but not by value, as we shall see later), were terminated
shortly after LBHI’s bankruptcy filing. Out of more than
900,000 trades, 733,000 were automatically terminated by
November 13, 2008 (Panel B of Table 2). About 80 percent
of the derivatives counterparties to Lehman Brothers Special
Financing (LBSF) terminated their contracts under the ISDA
Master Agreement within five weeks of the bankruptcy
filing, the largest-ever termination of derivatives transactions
(U.S. Government Accountability Office 2011).
Final settlement of terminated derivatives contracts
required further steps (Appendix D). The Lehman estate had
to 1) reconcile the universe of all trades between Lehman
and a particular counterparty, 2) value each transaction, and
3) negotiate settlement amounts with the counterparty. The
sheer number of derivatives contracts made each of these
steps an arduous process (“Debtors’ Disclosure Statement
for First Amended Joint Chapter 11 Plan,” January 25, 2011).
Accordingly, on November 13, 2008, Lehman asked the court
to approve procedures for entering into settlement agreements
with counterparties that had terminated their contracts with
Lehman, in order to establish termination payments and
the return or liquidation of collateral, without the need for
further action by the bankruptcy court. Lehman asked that
these procedures also apply to counterparties that had not
yet terminated their contracts but were considering doing
so. The court approved these procedures on December 16,
2008. Nevertheless, only 6 percent of ISDA contracts had
been settled by July 2009, with this number rising slowly to
46 percent by September 2010 (Panel C of Table 2).
4.1 OTC Derivatives Contracts That Were
Terminated Early
According to the ISDA Master Agreement, the bankruptcy
filing of LBHI meant that derivatives contracts with
automatic early termination clauses terminated immediately
(Appendix D). In addition, those counterparties of Lehman’s
derivatives entities without the automatic early termination
19
See “Derivatives Market Trades on Sunday to Cut Lehman Risk,” Reuters,
September 14, 2008, available at http://www.reuters.com/article/2008/09/14/
us-lehman-specialsession-idUSN1444498020080914.
4.2 OTC Derivatives Where
Out-of-the-Money Counterparties
Chose Not to Terminate Early
Many nondefaulting counterparties were out-of-themoney and would have owed large termination payments to
Lehman, so they chose not to send a termination notice.22 The
Lehman estate estimated these payments to be of significant
value and feared that market movements would reduce the
amounts owed to it (LBHI, “Debtor’s Motion for an Order
20
For example, Lehman Brothers Special Financing did not file for
bankruptcy until October 3, 2008.
21
The exact total number of Lehman’s derivatives trades and contracts at the
time of bankruptcy remains unclear. Reports by the Lehman estate variously
put the number of trades at 906,000, 930,000, and 1,178,000, and the number
of contracts at 6,120, 6,340, and 6,355.
22
For example, many municipalities and nonprofits had issued floating-rate
bonds and entered into interest rate swaps with Lehman where they paid a
fixed rate and received a floating rate. Some of these swap counterparties were
out-of-the-money to Lehman as the fixed rate was higher than the floating
rate prior to Lehman’s bankruptcy (Braun 2013).
FRBNY Economic Policy Review / December 2014
185
Pursuant to Sections 105 and 365 of the Bankruptcy Code,”
November 13, 2008). Moreover, the counterparties refused to
make required periodic payments to Lehman on out-of-themoney contracts on the grounds of Lehman’s default under the
ISDA Master Agreement.23
Lehman and its counterparties were often unable to agree
on the amount due on contracts when the counterparty
was out-of-the-money, partly because of the prevailing
illiquidity of markets, which made valuing derivatives trades
difficult. Under the Master Agreement, valuation claims are
determined primarily by replacement costs, which diverged
substantially from fair market value owing to the wide bid-
offer spreads at the time. Moreover, Scott (2012) argues that
replacement costs likely did not track actual costs, because
nondefaulting parties had considerable leeway in arriving at
their estimates and also because it was likely difficult to obtain
three dealer quotes as required (see Appendix D).
On November 13, 2008, Lehman asked the court to approve
procedures to realize the value of nonterminated derivatives
contracts either by Lehman assigning them to third parties
in exchange for consideration, or alternatively by mutual
termination. The court gave its approval (LBHI, “Debtor’s
Motion for an Order Approving Consensual Assumption and
Assignment of Prepetition Derivative Contracts,” January 28,
2009), authorizing Lehman to assign nonterminated
derivatives contracts with the consent of unsecured creditors
and the counterparty, but without the need for further court
approval. The effect of the court’s decisions was to strongly
encourage out-of-the-money c ounterparties to comply with
these Alternative Dispute Resolution (ADR) procedures
and to substantively e ngage in settlement and termination
discussions.24 Indeed, by January 2, 2009, just 2,667 contracts
(out of more than 6,000 contracts at the time of bankruptcy)
and 18,000 derivatives trades remained outstanding, and by
June 17, 2009, less than 17 p
ercent of contracts and less than
1 percent of trades were not terminated (Panel B of Table 2).
Assignment of claims moved slowly, partly because
of m
arket illiquidity and the balance sheet constraints of
financial firms, and partly because the positions were less
valuable. For example, some were uncollateralized, had
weak credits, or involved long maturity instruments (LBHI,
“§341 Meeting,” July 8, 2009). Nevertheless, the Lehman estate
23
For example, Metavante Corporation refused to make payments on an
interest rate swap agreement with LBSF (“Debtors’ Disclosure Statement for
First Amended Joint Chapter 11 Plan,” January 25, 2011).
made good progress on collecting derivatives receivables, with
cash collections increasing from less than $1 billion through
November 7, 2008, to about $8 billion through November 6,
2009 (LBHI, “The State of the Estate,” November 18, 2009) and
to about $11.5 billion through June 30, 2010 (LBHI, “The State
of the Estate,” September 22, 2010). As of January 10, 2011,
Lehman had issued notices to counterparties commencing
ADR procedures in connection with 144 derivatives contracts
and resolved fifty-two of these contracts, resulting in receipt of
approximately $356 million (“Debtors’ Disclosure Statement
for First Amended Joint Chapter 11 Plan,” January 25, 2011).
4.3 OTC Derivatives Contracts
with Big Bank Counterparties
The OTC derivatives market was highly concentrated at the
time of LBHI’s bankruptcy (and remains so today), with a
few large banks accounting for a substantial share of market
activity. This fact was reflected in counterparty shares of the
value of derivatives claims against Lehman and, in particular,
the shares of the thirty largest “big bank” counterparties, all of
which were affiliates of thirteen major financial institutions.25
Thus, in January 2011, the Lehman estate reported that, of the
outstanding contracts, the share of the thirty big bank counterparties was 85 percent of the number of trades and 48 percent
of derivatives contracts by dollar value, but only 5 percent of
the number of contracts (Panel D of Table 2).
Settlement of derivatives with big bank counterparties
proved challenging owing to difficult legal and valuation
issues (LBHI, “The State of the Estate,” September 22, 2010).
First, the total amount distributable to derivatives creditors
depended upon the resolution of the basis for the distribution
of creditor claims (that is, whether it should be the assets of
subsidiaries or of Lehman’s consolidated balance sheet—the
“substantive consolidation” issue). As further discussed
in Section 5, after negotiations between Lehman and its
creditors, between 20 and 30 percent of payments owed to
creditors (including derivatives creditors) of affiliates such as
LBSF were reallocated to holding company creditors. Second,
the Lehman estate and the big bank counterparties needed to
negotiate a uniform method for settling the remaining outstanding derivatives contracts.
The Lehman estate argued that big bank counterparties
submitted inflated claims (“Debtors’ Disclosure Statement for
24
The rules of discussions were formalized by the court’s order on
September 17, 2009, approving the ADR and mediation procedures for
nonterminated derivatives trades. The purpose of the order was to promote
“consensual recovery” and to encourage effective communication between
Lehman and its counterparties.
186
The Failure Resolution of Lehman Brothers
25
The thirteen major financial institutions were Bank of America, Barclays,
BNP Paribas, Citigroup, Credit Suisse Group, Deutsche Bank, Goldman
Sachs, JPMorgan Chase, Merrill Lynch, Morgan Stanley, the Royal Bank of
Scotland, Société Générale, and UBS.
First Amended Joint Chapter 11 Plan,” January 25, 2011).26
Their disagreements centered on 1) the time and date of
valuation, 2) the method of valuation (for example, use of the
bid or ask price as opposed to the mid-market price, as well as
the inclusion of additional amounts added to the mid-market
prices), and 3) setoff.27 As previously discussed, the valuation
of claims proved particularly difficult because of the “replacement cost” methodology required by the Master Agreement
and the wide bid-offer spreads at the time.28 Lehman and its
counterparties also disagreed on the discount rate and prices
that were inputs into valuation models (for example, whether
to use end-of-day prices on a particular date).
To avoid the costs and delays of litigating disputes with
the big bank counterparties individually (and a potentially
different outcome in each case), a derivatives claims
settlement framework was included as part of Lehman’s
January 2011 liquidation plan. The framework provided for
rules to settle the half of derivatives claims that remained
outstanding at the time and a commitment to a process and
timeline (LBHI, “The State of the Estate,” September 22, 2010).
The derivatives claims settlement rules offered a standardized
methodology. In particular, these derivatives contracts were
valued at mid-market at the market close of a specified termination date with an “additional charge” based on the maturity
and risk of the contracts (“Debtors’ Disclosure Statement for
Third Amended Joint Chapter 11 Plan,” August 31, 2011).29
Also, the number of maturity “buckets” used for aggregating
and offsetting exposures was reduced. With regard to the
process, the framework was used to determine most unsettled
derivatives claims (all claims except for those already settled,
those not disputed by Lehman, or those previously allowed by
the bankruptcy court).
Confirmation of the Joint Chapter 11 plan by the court on
December 6, 2011, did not completely resolve the settlement
of derivatives with big bank counterparties, as the Lehman
estate had entered into settlement with only eight of thirteen
major financial firms at the time. The slow progress of
negotiations can be gauged by the fact that, in 2012, the
estate settled only about 1,000 of the roughly 2,000 contracts
open at the beginning of the year (LBHI, “2013+ Cash Flow
Estimates,” July 23, 2013). This implies that an estimated
16 percent of contracts remained to be finally settled almost
a year after confirmation of the liquidation plan (Panel C of
Table 2). Nevertheless, sufficient progress was made such that
the Lehman estate was able to make the first distribution to
creditors on April 17, 2012.
Discussion: Settlement of Lehman’s
Derivatives Claims
For a firm, like Lehman, that was planning to liquidate its
assets, the objective of Chapter 11 bankruptcy is to maximize
the present recovery value of the bankruptcy assets of each of
its entities. However, there is a trade-off between obtaining the
highest possible recovery value of assets, which may require
a lengthy bankruptcy process, and minimizing costs (such
as legal and administrative fees) that increase with time.30
Moreover, uncertain and unpredictable resolutions may
destroy value by increasing systemic risk through information
contagion (in other words, bad news about Lehman’s resolution adversely impacting other firms) or fire sales of correlated
assets of entities unrelated to Lehman. Conversely, resolutions
that largely follow case law, and that keep claimants informed
on a regular basis, are likely to mitigate value destruction from
resolution. Accordingly, we assess the efficiency of the claims
settlement process with respect to its duration, predictability,
and transparency.
Promptness of Resolution Varied
across Creditor Claims
26
The disagreements between Lehman and the big bank counterparties stem
from the rights of the debtor and its counterparties under Section 562 of the
Bankruptcy Code.
27
Lehman’s out-of-the-money counterparties attempted to reduce their
payments by “setting off ” the amount they owed to Lehman against money
that (they claimed) Lehman owed to them in a separate transaction.
28
An example of inflated claims resulting from the changed valuation
methodologies occurred with respect to Lehman’s derivatives transactions
with Nomura Holdings (Das 2012). Prior to their termination on September 8,
2008, Nomura appeared to owe Lehman $484 million. Subsequently, however,
Nomura lodged a calculation statement claiming that Lehman owed it
$217 million. The $700 million difference was the result of Nomura changing
from the quotation method to the loss method, according to Lehman.
29
If the big banks could prove that they entered into economically identical
and commercially reasonable replacement trades on the date of LBHI’s filing,
they could use the value of these trades instead of the methodology.
The speed of resolution varied across claimant groups. Retail
OTC derivatives counterparties of Lehman terminated
their contracts within weeks of the bankruptcy filing under
the safe harbor provisions. But despite a perception to the
contrary,31 the final settlement of their claims was a long
30
Covitz, Han, and Wilson (2006) find that firm value initially increases with
time spent in default, but declines thereafter. Earlier research that does not
account for the endogeneity of time in default finds a negative relationship
between value and the time spent in default (see, for example, Acharya,
Bharath, and Srinivasan [2007]).
31
For the contrary perspective, see Liew, Gu, and Noyes (2010), who state that
“counterparties of Lehman Brothers were able to close out their OTC trades
FRBNY Economic Policy Review / December 2014
187
process, proceeding along three separate tracks, requiring
two settlement mechanisms in addition to the one specified
in the ISDA Master Agreement, and involving continuing
litigation and numerous operational problems.32 Thus, about
1,000 derivatives contracts remained “not settled” by the
beginning of 2013, more than four years after the start of
Lehman’s bankruptcy.
The Lehman estate pointed to the need for doing due
diligence on numerous, complex claims on an individual basis
as the chief cause of delay. The Lehman estate had statutory
duties and fiduciary obligations to review and reconcile how
each party reached its early termination amount so that all
creditors would be treated equally (“Debtors’ Disclosure Statement for First Amended Joint Chapter 11 Plan,” January 25,
2011). For example, the Lehman estate had to identify and
object to claims that were inflated in value or were duplicative
of other claims. Claims that involved complex and illiquid
securities were difficult to value. The estate’s determinations
of claims were frequently subject to litigation by creditors.
Indeed, the two new settlement mechanisms approved by the
courts were a means of applying uniform methods to a large
number of claims, and it appears that they proved effective in
facilitating settlement.
Another factor delaying the resolution of claims was the lack
of pre-bankruptcy planning by Lehman, resulting in LBI being
sold to Barclays in haste. The rushed sale caused numerous
problems—uncertainty regarding the number of Lehman
customer accounts transferred to Barclays or left behind, lack of
access to the accounts that were left behind, and litigation with
Barclays, CCPs, and clearing firms regarding the LBI sale—all
of which prolonged the resolution process (see Appendix C).
Finally, the organizational complexity of Lehman
contributed to delays. In many instances, Lehman and its
counterparties were uncertain of the identity of the specific
Lehman entity against which creditors had claims. Moreover,
different Lehman entities had different bankruptcy filing dates
in different international legal jurisdictions, which created
problems in cases where one subsidiary was acting as an agent
of another subsidiary in client transactions. Further, Lehman’s
interconnectedness (in particular, guarantees by the holding
company to affiliates) led to delays as holding company
creditors argued in favor of a greater share of recovery than
expected under strict priority rules.
smoothly under ISDA Master Agreements, despite severely stressed market
conditions.” See also Summe (2012) for a similar viewpoint.
32
Operational problems resulted from market participants that traded with
different Lehman entities having multiple ISDA Master Agreements in
place with different transactions recorded under each contract, according
to Das (2012), who adds that many counterparties’ information systems
inaccurately grouped contracts for determining netting and net exposure.
188
The Failure Resolution of Lehman Brothers
Predictability of Resolution Outcomes
Was Less than Expected
Some legal experts have considered the Chapter 11 process
predictable because it follows a long-standing legal tradition
with an established set of rules for allocating creditor claims
(U.S. Government Accountability Office 2011). This was
only partly true for Lehman’s bankruptcy, as new precedents
were set for many aspects of its resolution. For example, the
allocation of creditor claims did not follow standard priority
rules. While deviations from priority rules are not unusual
in Chapter 11 proceedings, they have declined substantially
over time, dropping from 75 percent of cases before 1990
to only 9 percent during the period 2000-05 (Bharath,
Panchapagesan, and Werner 2010). Moreover, deviations
from absolute priority have typically favored equity holders
(Bharath, Panchapagesan, and Werner 2010), whereas
under Lehman’s Chapter 11 liquidation plan, creditors of
derivatives entities with positive net worth received less than
their strict priority shares, while holding company creditors
received more.
In the Lehman bankruptcy, complex financial structures
were analyzed and adjudicated in the bankruptcy court for the
first time, and consequently the court’s judgments were sometimes controversial and even surprising to many o
bservers
(as acknowledged by Judge Peck in “Lehman Brothers S pecial
Financing Inc. versus BNY Corporate Trustee Services
Limited,” January 25, 2010). Thus, in some cases, Lehman’s
counterparties may have been denied the benefits of certain
safe harbor provisions, such as when the court refused to
enforce “flip clauses” (widely used in c ollaterized debt obligations and other financial structures).33 Since the U.S. court’s
decision contradicted an earlier U.K. court decision, and
the U.S. case was subsequently settled out of court, the legal
validity of flip clauses became uncertain and potentially
affected the credit ratings of financial structures.34 Also, the
33
In the case involving flip clauses, LBSF was a credit default swap
counterparty to a special purpose vehicle that issued credit-linked synthetic
portfolio notes, with LBHI acting as LBSF’s guarantor. The notes were secured
by collateral, which Bank of New York held in trust for the benefit of both
the note holder and LBSF. When LBHI filed for bankruptcy, the swaps were
terminated, and LBSF had priority over the collateral. But Bank of New York
argued that, since LBSF also filed for bankruptcy later, the priority reverted
to the note holder instead because of a “flip clause” specified in the swap
contract. However, the court ruled that the flip clause was unenforceable
under the ipso facto doctrine prohibiting the modification of a debtor’s
contractual rights because of the debtor’s bankruptcy (“Lehman Brothers
Special Financing Inc. versus BNY Corporate Trustee Services Limited,”
January 25, 2010).
34
In other cases, the bankruptcy court was thought to have defined the
rights of nondefaulting parties under safe harbor provisions more narrowly
than previously—for example, by imposing a time limit on a counterparty’s
right to seek relief, as in the Metavante case (LBHI, “Order Pursuant to
settlement of Lehman’s OTC derivatives with large institutional
counterparties followed different rules compared with
those that were terminated early. For example, the v aluation
methodology for calculating termination amounts for big bank
counterparties, as outlined in the derivatives claims settlement
framework, was different from that followed for non-big bank
counterparties.
Transparency of Resolution Was Good,
but Could Have Been Better
The Lehman estate issued numerous reports and created
websites containing archives of court documents and
presentations. Nevertheless, the level and accuracy of detail
provided by the Lehman estate could have been better. For
example, at least three different versions of Lehman’s initial
derivatives positions were provided in different reports.
Moreover, numbers were reported piecemeal rather than in
the aggregate and often without much context. For example,
it is difficult to estimate the total amount paid by the Lehman
estate in consulting and professional fees and administrative
expenses since the inception of the bankruptcy filing. One
report showed the fees and expenses paid since 2011 (the
amount reported in the media), while the fees and expenses
paid prior to 2011 were reported in multiple other documents.
Moreover, the fee and expense categories sometimes
differed between the earlier and later reports. In a similar
vein, information about the number of claims reconciled,
valued, settled, and still open was provided piecemeal and at
different points in time. In some respects, the dribbling out
of i nformation reflected the fact that the Lehman estate was
engaged in settling thousands of complex claims dynamically,
with the relevant information subject to periodic revisions.
Nevertheless, it would be valuable if, in future resolutions, the
bankruptcy estate provided more comprehensive statistics so
that interested parties could obtain a better understanding of
the resolution process.
5. Recovery Estimates for Lehman
Creditors under Chapter 11
At the time of the bankruptcy filing, there were 67,000 claims
against Lehman worth $1.16 trillion (Panel A of Table 3).
Under a plan that Lehman submitted to creditors and
the court on June 29, 2011, initial claims were reduced
to $764 b
illion, after adjusting for duplicate, inflated, and
invalidly filed claims.35 Of this amount, claims totaling about
$214 billion, or 28 percent of the total, were effectively “double
counted” since they were either guarantee claims (claims
based on guarantees by LBHI) or affiliate claims (claims by
Lehman entities against each other).36 After this and other
adjustments, allowed claims to third-party creditors across
twenty-three Lehman entities totaled $362 billion.
Of the total allowed claims, recovered assets were originally
estimated at nearly $84 billion—prior to administrative
expenses of $3.2 billion, amounts due to intercompany entities
or affiliates of nearly $2.9 billion, and operating disbursements
of approximately $3.1 billion—for a net distributable amount
to third-party creditors of $75.4 billion (second column of
Panel A of Table 3). The net amount expected to be distributed
to third-party creditors amounted to a claim payout ratio of
20.9 percent.
As of March 27, 2014, the Lehman estate had made five
distributions to creditors, with total recoveries exceeding the
initial estimates and allowed claims falling below the initial
estimates. Consequently, the recovery ratio for unsecured
creditors has been more than 28 percent (last column
of Panel A, Table 3).37 The amounts distributed include
intercompany claims, so that third-party recovery rates have
been lower than 28 percent. For example, of almost $45 billion
provided in the third, fourth, and fifth distributions,
third-party creditors received about $32 billion. Moreover,
part of the higher recovery rate is owing to a reduction in
claims allowed by the Lehman estate. Nevertheless, recoveries
for third-party creditors appear to have been larger than
35
While the recovery estimates reported in the table were as of May 13, 2011,
the plan was submitted to the court on June 29, 2011.
Sections 105(a), 362, and 365 of the Bankruptcy Code,” September 17, 2009).
However, a counterparty waiting too long to terminate could be deemed
to have waived its right to do so (Charles 2009). Some commentators have
argued that creditor rights under safe harbor provisions were limited when
the court granted Lehman the right to choose the time of termination,
to determine the termination value, and to act as the calculating agent
for valuing derivatives—rights that would normally by exercised by the
nondefaulting party (Ricotta 2011; Das 2012). A counterargument is that
Section 562 of the Bankruptcy Code gives Lehman certain rights to make
some of these determinations. Finally, the Lehman bankruptcy raised new
issues regarding the applicability of safe harbor provisions to setoff rules,
such as whether such provisions may eliminate the requirement that the
obligations are mutual—that is, creditor A and debtor B must owe money to
each other (Smith 2010).
36
For example, a third-party guarantee claim is that of a third party against
LBHI on account of its guarantee of an affiliate and is duplicative of the party’s
direct claim against the affiliate.
37
We estimate the payout ratio for LBHI creditors from the distribution
notices (see LBHI: “Notice Regarding Initial Distributions Pursuant to the
Modified Third Amended Joint Chapter 11 Plan ,” April 11, 2012; “Notice
Regarding Second Distributions Pursuant to the Modified Third Amended
Joint Chapter 11 Plan,” September 25, 2012; “Notice Regarding Third
Distribution Pursuant to the Modified Third Amended Joint Chapter 11
Plan,” March 27, 2013; “Notice Regarding Fourth Distribution Pursuant to the
Modified Third Amended Joint Chapter 11 Plan,” September 26, 2013; “Notice
Regarding Fifth Distribution Pursuant to the Modified Third Amended Joint
Chapter 11 Plan,” March 27, 2014).
FRBNY Economic Policy Review / December 2014
189
Table 3
Estimated Recovery of Creditor Claims under Chapter 11
Panel A: Aggregate Recovery for Lehman and Affiliates, as of March 2014
Billions of Dollars, Except as Noted
Estimated Recovery for Third-Party Creditors
September 15, 2008
Number of claims
Value of claims
Distributions to All Creditors, Unsecured Claims
May 13, 2011
67,000
48,000
1,160
764
March 27, 2014
Reductions related to:
Accounts payable and other
113
Third-party guarantee claims
83
Affiliate guarantee claims
72
Affiliate claims
59
45
Number of claims based on
derivative contracts
Debt
22
Value of claims after reduction
370
Other adjustments
8.5
Estimated allowed claims
361.5
Estimated recovery
303.6
83.7
Administrative expensesa
3.2
Due to intercompany entities
2.9
Operating disbursementsb
3.1
Net amount distributable
75.4
86.0
Payout ratioc (Percent)
20.9
28.3
Panel B: Recovery by Affiliate as of March 27, 2014
Affiliate
Payout Ratio, General
Unsecured Creditorsc
(Percent)
9.7
1,148
49.82
25.23
Shareholder Equity/ Total Assetsd (Percent)
LBHI
Holding company
LOTC
OTC derivatives
13.5
132
1.42
100.00
LBDP
Interest-rate and currency swaps
Interest-rate and FX OTC
derivatives; exchange-traded
derivatives; government bonds
OTC and exchange-traded
foreign currency
51.9
297
0.67
100.00
54.9
7
0.45
100.00
10.3
8
1.58
87.41
LBFP
LBCC
LBCS
LCPI
LBSF
190
Primary Assets
Cash Positiond (Millions of Dollars)
Distributions to
All Creditors,
Unsecured Claims
(Billions of Dollars)
Commodities
Secured and unsecured loans
Interest-rate, currency, credit, and
mortgage derivatives
12.3
30
2.32
67.38
Negative
461
15.41
61.63
7
13.06
30.90
4.3
The Failure Resolution of Lehman Brothers
Table 3 (Continued)
Estimated Recovery of Creditor Claims under Chapter 11
Panel C: Estimated Recovery for Derivative Claims of Large Counterparties as of May 13, 2011
Asserted Claims
(Billions of Dollars)
Allowed Claims
(Billions of Dollars)
Allowed to Asserted Claims
(Percent)
Eight largest counterparties
9.6
6.2
64.6
Thirty largest counterparties
21.8
10.3
47.4
Claimants
Sources: Debtors’ Disclosure Statement for Second Amended Joint Chapter 11 Plan (June 30, 2011); Debtors’ Disclosure Statement for Third Amended Joint
Chapter 11 Plan (August 31, 2011); Lehman Brothers Holdings Inc.: State of the Estate (November 18, 2009, and September 22, 2010), Notice Regarding
Initial Distributions Pursuant to the Modified Third Amended Joint Chapter 11 Plan (April 11, 2012), Notice Regarding Second Distributions Pursuant to
the Modified Third Amended Joint Chapter 11 Plan (September 25, 2012), Notice Regarding Third Distribution Pursuant to the Modified Third Amended
Joint Chapter 11 Plan (March 27, 2013), 2013+ Cash Flow Estimates (July 23, 2013), Notice Regarding Fourth Distribution Pursuant to the Modified Third
Amended Joint Chapter 11 Plan (September 26, 2013), Notice Regarding Fifth Distribution Pursuant to the Modified Third Amended Joint Chapter 11 Plan
(March 27, 2014); Valukas (2010).
Notes: LBHI is Lehman Brothers Holdings Inc.; LBDP is Lehman Brothers Derivative Products; LBFP is Lehman Brothers Financial Products; LCPI is
Lehman Commercial Paper Inc.; LBCS is Lehman Brothers Commodity Services; LBCC is Lehman Brothers Commercial Corporation; LOTC is Lehman
Brothers OTC Derivatives; LBSF is Lehman Brothers Special Financing.
For LBHI, the amount includes $1 billion of incremental liquidation administrative expenses.
From 2011 onwards; the amount includes professional fees and compensation, outsourced services, and information technology activities.
c
Amount equals net amount distributed as percent of estimated allowed claims.
d
Shareholder equity, total assets, and cash position numbers are as of September 14, 2008. LBHI’s cash position includes $509 million seized
post-filing by Bank of America.
a
b
e xpected, helped by settlements with other banks and
Lehman’s foreign subsidiaries.
Based on the cumulated distributions so far, creditors
of the holding company (LBHI) have received 21.3 percent
of their allowed claims in the aggregate. Senior unsecured
creditors of LBHI have received 26.9 percent of their allowed
claims (LBHI, “Notice Regarding Fifth Distribution Pursuant
to the Modified Third Amended Joint Chapter 11 Plan,”
March 27, 2014).38
We examine historical recovery rates to assess whether
LBHI’s recovery rate so far has been significant (as argued
by Scott [2012]) or poor (according to the Federal Deposit
Insurance Corporation [2011]). Average recovery rates for
senior unsecured claims between 1982 and 1999, based on
bonds, loans, and other debt instruments, are estimated at
56 percent for all industries and 59 percent for financial
institutions (Acharya, Bharath, and Srinivasan 2007).
Recovery rates are considerably lower during periods
of d
istress: 19 percentage points lower in recessions
(Schuermann 2004), 15 percentage points lower in periods of
industrial distress (Acharya, Bharath, and Srinivasan 2007),
38
Other creditor groups received considerably less. For example, senior
third-party guarantee claims recovered 16.7 percent and subordinate claims
recovered 0 percent (LBHI, “Notice Regarding Fifth Distribution Pursuant to
the Modified Third Amended Joint Chapter 11 Plan,” March 27, 2014).
and 15 to 22 percentage points lower, depending on the
default event, during credit cycle downturns (Bruche and
Gonzalez-Aguado 2007).39 Thus, even after accounting for
possibly reduced recovery rates owing to adverse credit and
macroeconomic conditions, the recovery rate so far for LBHI
has been low compared with the historical average. With
additional distributions yet to come, the final recovery rate is
expected to be higher, but it remains to be seen whether it will
meet historical norms.
While the average payout ratio for Lehman and affiliates
has been about 28 percent, recovery rates have been higher
for creditors of certain derivatives subsidiaries of LBHI and,
in a few cases, have reached 100 percent (Panel B of Table 3).
The plan had estimated that seven of the twenty-three
Lehman entities would pay all of their claims in full and have
remaining funds for their shareholders. Prior to its bankruptcy filing, Lehman traded derivatives through a number
of wholly owned subsidiaries, both in a trading capacity and
as an end-user, as listed in Panel B of Table 3.40 Lehman’s first
39
In Bruche and Gonzalez-Aguado (2007), the credit cycle is unobservable and
represented by a two-state Markov chain. While the literature does not find a
statistically significant effect of macroeconomic factors on recovery rates (Altman,
Brady, Resti, and Sironi 2005; Acharya, Bharath, and Srinivasan 2007), these
studies have short sample periods that do not include many recession periods.
40
Lehman’s fixed-income derivatives products business was principally
FRBNY Economic Policy Review / December 2014
191
liquidation plan filed in March 201041 had called for maintaining the corporate distinction of each of the twenty-three
Lehman entities that had filed for bankruptcy, implying
that each affiliate would make payments to its creditors on
the basis of its own assets. Derivatives creditors would have
generally benefited from such an approach, given the positive
equity cushions of most Lehman derivatives entities.
General creditors of LBHI argued that parent company
guarantees of affiliates’ debt meant that more debt resided
at the parent level while assets were at the subsidiary level.42
As such, creditors with claims against an affiliate subject to
an LBHI guarantee could recover against both LBHI and the
affiliate. An ad hoc group of ten LBHI creditors submitted
their own liquidation plan on December 15, 2010, proposing
to “substantially consolidate” all affiliates’ assets into one
Lehman entity. In contrast to the existing company structure,
under the consolidated structure, guarantee claims would
be eliminated. Therefore, holders of parent company claims
would receive more with consolidation. Lehman rejected this
plan and, after further negotiations with creditors, submitted
an amended plan on June 29, 2011, that proposed to retain the
corporate formalities of each debtor entity, but to redistribute
the payouts made to certain creditors. After further revisions
to this plan, the Modified Third Amended Plan was finally
confirmed on December 6, 2011, following a creditor vote,
and became effective on March 6, 2012, enabling Lehman to
emerge from bankruptcy and make distributions to creditors.
As a result of the plan, between 20 and 30 percent of
payments owed to creditors of various operating companies
were forfeited and reallocated to the parent company’s
creditors. In particular, distributions due to claim holders
of derivatives entities such as LBSF, Lehman Commercial
Paper Inc., Lehman Brothers Commodity Services, Lehman
Brothers OTC Derivatives Inc., and Lehman Brothers
conducted through LBSF and Lehman’s separately capitalized “AAA”-rated
subsidiaries Lehman Brothers Financial Products and Lehman Brothers
Derivative Products. Lehman’s equity derivatives products business was
conducted through Lehman Brothers Finance, Lehman Brothers OTC
Derivatives Inc., and LBIE, and its commodity and energy derivatives product
business was conducted through Lehman Brothers Commodity Services.
Lehman conducted a significant amount of its spot, forward, and option foreign
exchange business through Lehman Brothers Commercial Corporation.
41
There were four versions of Lehman’s joint proposed Chapter 11 plan
(referred to as the liquidation plan in the text). The original proposal was filed
in March 2010, followed by amended versions on January 25, 2011, June 30,
2011, and August 31, 2011.
42
LBHI was the guarantor to the majority of ISDA derivatives contracts with
about 1.7 million trades and more than 10,000 counterparties (Government
Accountability Office 2011). There were also intercompany claims against
Lehman’s subsidiaries. For example, other Lehman entities filed 630 claims
worth about $19.9 billion against LBI.
192
The Failure Resolution of Lehman Brothers
ommercial Corporation were reallocated to holders of senior
C
unsecured claims and general unsecured claims against LBHI.
Accordingly, while Lehman Brothers OTC Derivatives Inc.,
Lehman Brothers Financial Products, and Lehman Brothers
Derivative Products all had recovery rates of 100 percent,
LBSF (the largest derivatives entity) had recovered about
31 percent, despite having a positive equity cushion (Panel B
of Table 3).
Recovery rates for large derivatives counterparties are
likely to be different from those of other secured creditors.
This is because the Lehman estate followed a different
settlement approach regarding these claims, as discussed in
Section 4. Under the Chapter 11 liquidation plan, the eight
largest financial institutions were allowed about 65 percent
of their asserted claims, while the thirty largest big bank
counterparties were allowed about 47 percent of their asserted
claims (Panel C of Table 3).
Discussion: Recovery Rates of Lehman Creditors
Recovery rates varied across creditor groups. Creditors of
two Lehman derivatives entities received full recovery on
their claims, while customers of centrally cleared securities
were mostly made whole. In contrast, most counterparties of
Lehman’s OTC derivatives suffered substantial losses. What
caused some Lehman creditors to receive better recovery rates
than others?
A crucial factor for LBI customers to receive full recovery
was the availability of Federal Reserve funding for LBI and
Barclays in the first week after bankruptcy, which allowed LBI
to continue operating until it was sold to Barclays. The Fed
also urged LBI’s clearing agents to continue to provide intraday liquidity so that trades could be settled (LBHI, “Debtors
versus JPMorgan Chase Bank, N.A.,” April 19, 2012).
Central clearing allowed Lehman’s positions to be
terminated rapidly and resulted in minimal losses for
Lehman’s customers (Appendix B). However, CCPs and
clearing firms filed numerous suits against the Chapter 11
debtors and the SIPA trustee (Appendix C) that, had these
suits not been decided in favor of Lehman, would have led
to larger losses for Lehman’s customers. Also, despite central
clearing, some of Lehman’s house positions suffered large
losses due to the extreme illiquid market conditions prevailing
during the financial crisis (Appendix B).
The positive net worth of most of Lehman’s derivatives
entities at the time of bankruptcy also helped, although
the largest entity (LBSF) was borderline insolvent with
shareholder equity of only 4 percent of total assets (Panel B
of Table 3). Indeed, derivatives positions were reliable
revenue sources for the Lehman estate during bankruptcy
(Summe 2012). Derivatives creditors could have received
even more if some of their allocations had not been diverted
to larger counterparties of LBHI under the Chapter 11
liquidation plan.
In contrast to centrally cleared derivatives, the settlement
of Lehman’s OTC derivatives claims may have resulted in
significant losses to Lehman (“Debtors’ Disclosure Statement
for First Amended Joint Chapter 11 Plan,” January 25, 2011;
U.S. Government Accountability Office 2011) or to Lehman’s
counterparties. In particular, Lehman’s counterparties used
the safe harbor provisions to terminate contracts when
they stood to gain and to keep alive contracts when they
were out-of-the-money. Further, they refused to make
required periodic payments to Lehman on out-of-the-money
contracts on the grounds of Lehman’s default under the ISDA
Master Agreement.
In other cases, the settlement of Lehman’s OTC derivatives
claims may have resulted in significant losses to Lehman’s
counterparties. Some Lehman counterparties suffered
losses owing to the selection of the termination date for safe
harbor purposes (Ricotta 2011). Although Lehman filed
for bankruptcy protection at about 1:00 a.m. on Monday,
September 15, 2008, the termination date was set as Friday,
September 12, for derivatives subject to automatic termination.
Normally, nondefaulting derivatives counterparties of Lehman
would have attempted to hedge their positions on Monday
to mitigate expected losses on their positions. However, they
could not do so since their positions were deemed to have
terminated two days earlier. Also, in some cases, parties had
sent wire transfers to various Lehman entities on Friday to
satisfy their obligations to make periodic payments, even
though such payments were not required once Lehman had
defaulted (Ricotta 2011). Some of these parties that had elected
automatic early termination tried to revoke their elections ex
post, but such an election is irrevocable.
Scott (2012) argues that twenty-four of Lehman’s top
twenty-five counterparties by number of derivatives
transactions had entered into credit support annexes with
Lehman that required the out-of-the-money party to post
collateral based on mark-to-market liability, greatly mitigating
the effects of a default if counterparties exercised their
rights under these agreements. However, the actual extent of
collateralization is in dispute. For example, it has been alleged
that Lehman did not post sufficient collateral, that it failed to
segregate collateral, and that hypothecated collateral could
not be recovered in a timely fashion (Ricotta 2011). These
problems arose in part because, although counterparties
posted initial margin (or “independent amount”) on their
OTC trades with Lehman, dealers like Lehman generally
do not post initial margin to their buy-side counterparties
(Scott 2012).
Under safe harbor provisions, Lehman’s nondefaulting
counterparties could seize collateral that Lehman posted
to them before default, even if the collateral was posted
just before bankruptcy. Some in-the-money counterparties
suffered losses when, under the credit support annexes
included in their derivatives contracts, Lehman affiliates
either were never required to post collateral or did not post
sufficient collateral (Ricotta 2011).43 As a result, they were
unable to make recovery through the close-out netting process
and became unsecured creditors to the Lehman estate.
Although Lehman typically did not post collateral, it held
collateral posted by its counterparties. Lehman sometimes
commingled its counterparties’ liquid collateral with its own
(less liquid) assets, either because it was allowed to hypothecate
collateral, or because it did not hold counterparty collateral in a
segregated account (Ricotta 2011; Scott 2012). Counterparties
that had allowed Lehman to hypothecate their collateral
to unrelated third parties in connection with securities
transactions that could not be unwound found that their
collateral had become unrecoverable. When Lehman did not
segregate collateral, the collateral became an unsecured claim
in the Chapter 11 cases or subject to Lehman’s SIPA receivership proceedings (Ricotta 2011). It follows that counterparties
did not know when their collateral would be returned to them,
nor did they know how much they would recover given the
deliberateness and unpredictability of the bankruptcy process.
6. Conclusion
The bankruptcy of Lehman Brothers was one of the largest
and most complex in history, encompassing more than
$1 trillion worth of creditor claims, four bodies of applicable
U.S. laws, and insolvency proceedings that involved more
than eighty international legal jurisdictions. The payout ratio
to third-party creditors was initially estimated to be about
21 percent on estimated allowable claims of $362 billion.
While actual distributions appear to have exceeded initial
estimates, some of it has gone to other Lehman entities.
Moreover, recovery rates for Lehman’s senior unsecured
creditors remain below historical averages even after
accounting for possible mitigating factors (such as the state
43
In lieu of posting collateral, LBHI provided credit guarantees for nearly all
the derivatives transactions of its affiliates.
FRBNY Economic Policy Review / December 2014
193
of the economy and the credit cycle). Customers of centrally
cleared securities were generally made whole, and most
customers of Lehman’s broker-dealer were able to transfer
their accounts to other solvent broker-dealers. In contrast,
many counterparties of Lehman’s OTC derivatives suffered
substantial losses.
We argue that some of the losses associated with the failure
of Lehman Brothers may have been avoided in a more orderly
liquidation process. The poor planning of the bankruptcy
process, in particular, stands out as being especially costly. In
contrast, creditor losses would have been more substantial
without the ability of Lehman’s U.S. brokerage subsidiary, and
subsequently of Barclays, to finance positions through the
Federal Reserve’s liquidity facilities.
The size and complexity of Lehman resulted in costly
delays in settling claims. The settlement process was long as
the Lehman estate had a fiduciary duty to do due diligence on
numerous, complex claims on an individual basis. Further, its
determination of claims was frequently litigated, as is typical
for bankruptcies of large firms. Lehman’s organizational
complexity also contributed to delays. For example, in many
instances, Lehman and its counterparties were uncertain of
the identity of the specific Lehman subsidiary against which
creditors had claims. Finally, Lehman’s interconnectedness
led to delays as LBHI creditors argued in court that, since the
194
The Failure Resolution of Lehman Brothers
holding company had guaranteed some of the subsidiaries’
debt, they were entitled to a portion of recovery from
subsidiary assets (the “substantive consolidation” issue).
The predictability of Lehman’s claims settlement
procedures was hindered by the novelty of its business and
financial structure (in the context of bankruptcy cases).
Chapter 11 proceedings are based on the application of case
law relating to the Bankruptcy Court’s prior interpretations
of cases. While existing case law provided a useful starting
point for Lehman’s resolution, the court provided new
interpretations of provisions in the Bankruptcy Code (for
example, regarding some aspects of the safe harbor provisions
for derivatives). In part, this reflected the importance of
complex financial securities that the bankruptcy court had to
analyze for the first time.
In sum, the size and complexity of Lehman, the novelty of
its structure, and the rarity with which such firms go bankrupt contributed to a prolonged and costly resolution. In the
future, because of the Dodd-Frank Act, regulators will have
the option to resolve large, complex financial firms under
the Orderly Liquidation Authority, through the expanded
reach of the FDIC. Details of how such a resolution would
be implemented are still being worked out, making it hard to
evaluate the extent to which the resolution of large nonbank
financial firms will be more efficient going forward.
Appendix A: Glossary
ADR
Alternative Dispute Resolution
LBI
Lehman Brothers Incorporated
CCP
central counterparty
LBIE
Lehman Brothers International (Europe)
CME
Chicago Mercantile Exchange
LBSF
Lehman Brothers Special Financing
DTCC Depository Trust and Clearing Corporation
NSCC
National Securities Clearing Corporation
FDIC
Federal Deposit Insurance Corporation
OCC
Office of the Comptroller of the Currency
FICC
Fixed Income Clearing Corporation
OTC
over-the-counter
GCF
general collateral finance
PDCF
Primary Dealer Credit Facility
ISDA
International Swaps and Derivatives Association
QFC
qualified financial contracts
JPMC
JPMorgan Chase and Company
SIPA
Securities Investor Protection Act
LBHI
Lehman Brothers Holdings Incorporated
TSLF
Term Securities Lending Facility
195
The Failure Resolution of Lehman Brothers
FRBNY Economic Policy Review / December 2014
195
Appendix B: Settlement of Lehman’s Centrally Cleared Positions
The par value of Lehman’s centrally cleared U.S. positions
exceeded $520 billion at the time of bankruptcy (Panel A
of Table B.1). Exchange-traded and some OTC derivatives
contracts (such as futures contracts) were centrally cleared,
and these positions were resolved by central counterparties
acting on behalf of Lehman’s clients (where Lehman acted as
a broker) or on behalf of Lehman (where Lehman traded for
its own accounts), as illustrated in Exhibit 3.44 The resolution
of Lehman’s centrally cleared securities positions by CCPs
proceeded relatively smoothly, as CCPs suspended or imposed
limits on the market access of defaulting Lehman entities
within hours of default (Panel B of Table B.1), with most of
its client and proprietary positions settled with no large losses
to CCPs (CCP12 2009). However, there was controversy
regarding the Chicago Mercantile Exchange’s (CME) handling
of Lehman’s proprietary positions, as described below.
Immediate Response of CCPs to LBHI’s
Bankruptcy Announcement
Lehman traded in almost all developed markets and was a
direct clearing participant on behalf of itself or its clients
in some markets while using third-party clearing arrangements in others. Following the bankruptcy announcement of
LBHI in the United States, there was uncertainty as to which
of Lehman’s international subsidiaries were solvent. Thus,
CCPs with direct clearing relations with Lehman became
unsure about Lehman’s ability to deliver on obligations to
them. After LBHI’s bankruptcy announcement, most of these
CCPs confirmed suspension, declared Lehman in default,
or implemented restricted trading arrangements before
markets opened in the United States. (Panel B of Table B.1).
A few exchanges temporarily allowed trading and settlement
by subsidiaries if they continued to meet CCP obligations
(CCP12 2009). Where Lehman did not have a direct clearing
relationship, the CCPs had no direct exposure to Lehman,
but they worked closely with third-party clearing agents45 to
44
In at least one case, a CCP helped resolve Lehman’s bilaterally cleared
derivative position. Specifically, LCH.Clearnet resolved the default of Lehman’s
interest rate swap portfolio, consisting of 66,000 trades and $9 trillion in
notional value, within three weeks, well within the margin held and without
loss to other market participants. See Managing the Lehman Brothers’ Default,
LCH.Clearnet, available at http://www.lchclearnet.com/swaps/swapclear_for_
clearing_members/managing_the_lehman_brothers_default.asp.
45
Clearing agents are corporations or depositories that act as intermediaries
in the clearing and settlement process. See http://www.sec.gov/divisions/
marketreg/mrclearing.shtml.
196
The Failure Resolution of Lehman Brothers
resolve Lehman’s outstanding positions. Third-party clearers
and trading venues quickly suspended Lehman and prevented
its positions from increasing further (CCP12 2009).
In the United States, the bankruptcy announcement
identified Lehman entities that remained solvent, allowing
U.S. CCPs and clearing agents to continue relationships with
solvent Lehman entities (although the relationship with LBI
would prove to be contentious, as discussed in Appendix C).
The CCPs of the Depository Trust and Clearing C
orporation,
namely the Fixed Income Clearing Corporation (FICC)
and the National Securities Clearing Corporation (NSCC),
confirmed on September 15, 2008, that Lehman’s s ubsidiaries
remained solvent participants of the CCP (CCP12 2009).
ICE Clear U.S. and the CME also announced that Lehman
continued to meet commitments to the clearinghouse.
Default Management and Risk Reduction by CCPs
CCPs, by taking on the obligations of their clearing members,
are exposed to risk, which they manage through a variety of
strategies (for example, through margins and other member
contributions, and capital and insurance for use in the event
of default). In Lehman’s case, CCPs used similar approaches to
limit their exposure, with some exceptions influenced by local
regulation (CCP12 2009).
In many markets, Lehman acted as a broker, making and
receiving payments on behalf of its clients. Insolvency of a
broker typically results in clients facing restricted access to
their accounts. In response to Lehman’s insolvency, CCPs
acted quickly to transfer (or facilitate transfer under the
client’s direction) Lehman’s client accounts to other
nondefaulting clearing participants. In the United States,
LBI’s client accounts were mostly transferred to Barclays
Capital or Ridge Clearing and Outsourcing Solutions, Inc.
(a clearing services provider), as further discussed in
Appendix C. Overall, the vast majority of Lehman’s clients
obtained access to their accounts within weeks (and sometimes days) of Lehman’s bankruptcy (CCP12 2009).
Lehman’s house positions were the outcome of
proprietary trading on behalf of itself. With limited third-
party interest, most CCPs closed out these positions. In the
United States, following the appointment of the SIPA trustee
on S eptember 19, 2008, the DTCC announced on October 30,
2008, that it had wound down LBI’s outstanding obligations.
FICC netted and liquidated $329 billion in par value of outstanding forward trades in mortgage-backed securities and
$190 billion in gross government bond positions (CCP12 2009).
Appendix B:
(Continued)
Settlement of Lehman’s Centrally Cleared Positions (Continued)
Table B.1
Resolution of Lehman’s Centrally Cleared Positions
Panel A: Lehman’s Centrally Cleared Positions at Time of Bankruptcy
Central Counterparty
Par Value of Positions
(Billions of dollars)
Asset Type
CME
Derivatives
FICC
MBS forwards, government bonds
NSCC
Equity, municipal and corporate bonds
Netted and
Liquidated by
4.00*
09/19/2008
519.00
10/30/2008
5.85
10/30/2008
Panel B: CCP Actions Following LBHI Bankruptcy Filing
Date
Actions of Global Central Counterparties with Respect to Lehman Entities
09/15/2008
• Six CCPs confirm no clearing relationship with Lehman
• Six CCPs confirm Lehman continues to meet obligations
• Eight CCPs announce default or suspension of Lehman
• One CCP announces restricted trading/clearing for Lehman
09/16/2008
• Four CCPs announce default or suspension of Lehman
• LCH.Clearnet and two CCPs commence transfer of client accounts
• Three CCPs complete close-out of positions
09/19/2008
• Two CCPs close positions without loss
• CME closes out Lehman house positions
• FICC and NSCC begin close-out of house positions
• LCH.Clearnet announces 90 percent risk reduction of positions
• LCH.Clearnet and another CCP largely complete transfer of client positions and close out house positions
09/26/2008
• One CCP completes transfer of client accounts
• Two CCPs close out positions
10/03/2008
• FICC, NSCC, and another CCP close out house positions without loss
Sources: CCP12 (2009); “Debtors’ Disclosure Statement for Second Amended Joint Chapter 11 Plan” (June 30, 2011); Valukas (2010).
Notes: CME is Chicago Mercantile Exchange; FICC is Fixed Income Clearing Corporation; NSCC is National Securities Clearing Corporation.
*Aggregate margin requirements on Lehman’s customer and house positions.
NSCC inherited a $5.85 billion portfolio of equities,
municipal bonds, and corporate bonds, used $1.9 billion in
pledged securities to settle outstanding equity obligations,
and liquidated or hedged remaining positions (CCP12 2009).
NSCC’s portfolio included $3.8 billion in options exercises
and assignments from the Options Clearing Corporation for
the quarterly expiration on September 19, 2008, which was
liquidated with no losses to other NSCC members.46
46
See “DTCC Successfully Closes Out Lehman Brothers Bankruptcy,” http://
www.bloomberg.com/apps/news?pid=newsarchive&sid=aojt5wVkz_EM.
LBI had large derivatives positions at the CME, where it
was a clearing member. At the time of its bankruptcy, LBI’s
margin requirements at the CME that were related to its
proprietary and public customer positions totaled roughly
$4 billion, accounting for more than 4 percent of the margin
requirements of all CME clearing members (Panel A of
Table B.1). Despite the size of LBI’s positions, they were
unwound in four days. Nonetheless, there were difficulties
with the settlement, as discussed below.
On September 12, 2008, the CME was informed by
federal regulators of LBHI’s expected bankruptcy or sale
and began preparing for a possible liquidation or transfer
FRBNY Economic Policy Review / December 2014
197
Appendix B:
(Continued)
Settlement of Lehman’s Centrally Cleared Positions (Continued)
of LBI positions (Valukas 2010, p. 1844). Owing to the large
size and complexity of Lehman’s exchange positions, the
CME judged that an open market sale would not be prudent
(Valukas 2010, p. 1845). Instead, on September 14, the CME
selected six firms and disclosed LBI’s house positions to
them in order to solicit contingent bids on these positions
(Valukas 2010, p. 1846). The bids, received from five of the
six firms, implied substantial losses to LBI as it would lose
the majority (or, in some cases, all) of its posted margins on
these positions. On September 15, the CME instructed LBI
to liquidate its proprietary position in bulk, the first time that
it had conducted a forced transfer/liquidation of a clearing
member’s position.47 The CME took this action, even though
LBI was not in default of its margin requirements, because it
felt that LBI would be liquidated before too long.
Between September 15 and September 17, LBI attempted
to find buyers for its house positions, but was unable to do so
except for its natural gas positions (Valukas 2010, p. 1849).
On September 17, the CME learned that Barclays would not
assume all of LBI’s customer positions and that LBI was likely
to file for liquidation on September 19. Consequently, that
same evening, the CME decided to re-solicit bids from the
five firms that had previously submitted bids. On the morning
of Thursday, September 18, the CME transferred LBI’s
proprietary positions to three firms.
The bulk sale resulted in a loss to LBI on its proprietary
position that exceeded $1.2 billion and an additional loss
of $100 million over margin requirements (Valukas 2010,
p. 1854). LBI’s portfolio at the CME, largely intended to hedge
Lehman’s OTC swaps contracts that were guaranteed by LBHI,
became outright positions after the bankruptcy filing.48 The
inability to offer both legs of the hedged positions meant that
LBI could not liquidate the outright positions on favorable
terms, because counterparties would require substantial
additional collateral and margins (“Trustee’s Preliminary
Investigation Report and Recommendations,” August 25, 2010).
47
However, amid the confusion, LBI modestly added to its position over the
next two days as Lehman traders either did not show up for work or received
inadequate direction from management.
198
The Failure Resolution of Lehman Brothers
48
This is because the swaps contracts terminated when the guarantor, LBHI,
defaulted. Therefore, LBI’s hedge position stood on its own.
Appendix C: Settlement of Lehman’s Customer Positions under SIPA
The insolvency proceedings involving LBHI on September 15,
2008, severely limited the daily funding sources of LBI, and
it was able to continue operations only by borrowing from
the Fed, as detailed in Section 3.49 On September 19, 2008,
the court appointed a trustee under the Securities Investor
Protection Act of 1970 to “maximize the return of customer
property to customers of LBI as defined by the law, while
at the same time maximizing the estate for all creditors.”
Different from Chapter 11, SIPA was a liquidation proceeding,
with an emphasis on returning customer property wherever
possible (Giddens 2008).
The LBI resolution was the largest and most complex in
SIPA history. Almost 125,000 customer claims worth almost
$190 billion were filed (Panel A of Table C.1). Even prior to
his formal appointment, the SIPA trustee assisted in the
transfer of LBI’s customer accounts to Ridge Clearing and
Outsourcing Solutions Inc. on behalf of NeubergerBerman,
resulting in the transfer of more than 38,000 customer
accounts worth over $45 billion (Panel A of Table C.1).50 On
September 19, 2008, Barclays acquired select, but not all,
broker-dealer assets and customer accounts of LBI.51 Originally,
it was believed that Barclays would leave behind few significant
customer accounts; accordingly, the SIPA proceedings would
largely be a vehicle for effectuating customer account transfers
to Barclays (“Trustee’s Preliminary Investigation Report and
Recommendations,” August 25, 2010).
Beginning September 23, 2008, the SIPA trustee supervised
and authorized the transfer of more than 72,500 private
investment management accounts amounting to more than
$43 billion to Barclays (Panel A of Table C.1). Effectively, these
LBI account holders became Barclays account holders, and
their account assets appeared on their Barclays account statements
(“Trustee’s First Interim Report,” 2009).
In contrast to these (mostly retail) customer accounts that
were transferred within weeks of LBI’s liquidation filing, the
resolution of institutional customer claims through the SIPA
claims process remains ongoing. The resolution of institutional
claims occurred through account transfers and the SIPA
claims process. After Barclays unexpectedly refused to assume
LBI’s prime brokerage accounts, a majority of these accounts
were transferred by the SIPA trustee to other broker-dealers,
using an innovative protocol that expedited the transfer
process (“Trustee’s First Interim Report,” 2009). Almost
300 accounts worth close to $3.50 billion were transferred
through the SIPA trustee’s Prime Brokerage Protocol (Panel A
of Table C.1). However, owing to the complexity of the process,
most account transfers were only partial (“Trustee’s First
Interim Report,” 2009).
Numerous claims remained pending after the account
transfers, including thousands of customer accounts that
Barclays left behind, claims of Lehman’s European broker-dealer
LBIE, and intercompany claims of LBHI and other Lehman
affiliates.52 These claims included both customer and general
creditor claims and were determined through the SIPA claims
process starting on December 1, 2008 (Giddens 2008). The
process proved challenging because of complex issues of
statutory interpretation and the need for extensive reconciliation and analysis. Nearly 10,000 claims were investigated,
denied customer status, and closed. Nevertheless, by March 29,
2013, more than 14,000 claims had been resolved, and
customers and general creditors received a distribution of about
$13.5 billion (Panel A of Table C.1), the bulk of which went to
satisfy LBIE’s intercompany claims (Panel B of Table C.1).
A relatively small number of claims remain contested
(Panel A of Table C.1) and, in order to streamline the
resolution of general creditor claim disputes, the SIPA trustee
recently sought and received a court order establishing
ADR procedures (“Trustee’s Tenth Interim Report,” 2014).
49
LBHI’s rushed Chapter 11 filing also forced Lehman’s European brokerdealer LBIE into administration in the United Kingdom on the morning
(local time) of September 15, 2008. LBI assets that had been traded
in overseas markets through LBIE (which acted as LBI’s clearing and
settlement agent for certain LBI overseas trades) became tied up in the
LBIE administration process. At the same time, LBIE demanded more than
$8 billion from LBI related to transactions allegedly made just before LBIE
entered administration.
Discussion: Resolution of Lehman’s Customer
Accounts under SIPA
The resolution process has resulted in 100 percent recovery for
50
customers, a significant achievement for SIPA. Nevertheless,
in his investigative report, the SIPA trustee noted many legal
51
52
Shortly after LBHI’s bankruptcy filing, Neuberger Berman (which had used
LBI as its clearing broker) transferred its clearing services to Ridge Clearing
and Outsourcing Solutions Inc.
Barclays also did not acquire LBI house positions, the resolution of which is
discussed in Appendix B.
LBIE’s claims included those on its own behalf and those on behalf of LBIE
customers, for which LBI acted as custodian and clearing broker.
FRBNY Economic Policy Review / December 2014
199
Appendix C: Settlement of Lehman’s Customer Positions under SIPA (Continued)
Table C.1
Estimated Recovery of Customer Claims under SIPA
Panel A: Summary of Customer Claims Resolutions as of March 29, 2013
Number of Claims
Amount (Billions of Dollars)
124,989
188.57
—
105.78
110,920
92.30
To Barclays
72,527
43.25
To Neuberger Berman
38,106
45.57
Total claims
Less: Total claims resolved by transfers or claims process
Less: Claims distributed by accounts transfers
287
Through Trustee’s Prime Brokerage Protocol
3.49
14.23
Remaining claims
Claims distributed through SIPA claims process
Claims unresolved
14,069
13.48
—
0.75
Panel B: Customer Claims Distributed through SIPA Claims Process, by Group, as of March 29, 2013
Market Value of Securities and Cash
(Billions of Dollars)
Share of Total (Percent)
Non-affiliate
1.62
12.0
LBIE
9.23
68.5
LBHI
2.37
17.6
Other affiliates
Total
0.26
1.9
13.48
100.0
Sources: Trustee’s Fifth Interim Report (2011) and Trustee’s Ninth Interim Report (2013).
Notes: SIPA is Securities Investor Protection Act; LBHI is Lehman Brothers Holdings Inc.; LBIE is Lehman Brothers International (Europe).
and systemic difficulties in the liquidation process (albeit
unnoticed by customers whose accounts were treated as
intact despite the difficulties) and made recommendations for
improvements (“Trustee’s Preliminary Investigation Report
and Recommendations,” 2010). Retail customer accounts were
transferred quickly, although reconciliation of accounts and
delivery of property held in custodial banks around the world
took more than a year (Giddens 2010). In contrast, resolution
of institutional customer claims through the SIPA claims
process remains ongoing.
The rushed liquidation of customer accounts left behind by
Barclays resulted in a disorderly process of unwinding LBI’s
customer and intercompany balances (“Trustee’s Preliminary
Investigation Report and Recommendations,” 2010). There
was inadequate understanding as to how the interests of
customers whose accounts Barclays rejected would be affected,
leading to prolonged disputes with Chapter 11 creditors and
Barclays. For example, it was initially believed that only a
200
The Failure Resolution of Lehman Brothers
few customer accounts not transferred to Barclays would be
liquidated under SIPA, but a substantial number of customer
accounts were actually left behind.
In addition, CCPs and clearing agents took unilateral
adversarial actions that made it difficult for the SIPA trustee
to obtain access to customer property and records. Thus, at
the time of the bankruptcy filing, JPMC unilaterally shut off
access to information systems, thereby preventing LBI and
the SIPA trustee from identifying and protecting customer
accounts (“Trustee’s Preliminary Investigation Report and
Recommendations,” 2010). JPMC also did not honor customer
segregation requirements. These issues were ultimately
resolved through formal agreements between JPMC and the
SIPA trustee, but in the meantime, the ability of the SIPA
trustee to transfer customer property was impaired.
Moreover, the Depository Trust and Clearing Corporation
and the Office of the Comptroller of the Currency (OCC)
threatened emergency actions that harmed the account
Appendix C:
(Continued)
Settlement of Lehman’s Customer Positions under SIPA (Continued)
t ransfer process (“Trustee’s Preliminary Investigation Report
and Recommendations,” 2010). The OCC threatened to
liquidate all LBI positions unless Barclays stepped into
LBI’s shoes by having LBI’s accounts at OCC transferred to
Barclays. Although Barclays agreed, customers of LBI who
did not transfer to Barclays had difficulty accessing their
OCC positions and margins. Similarly, DTCC was unwilling
to provide settlement services if Barclays did not take over
LBI positions. The issue was settled when Barclays agreed
to deposit the purchase price for LBI (due to the estate) to
DTCC, but there was less cash available to settle customer
claims in the interim.
The transparency of the SIPA liquidation process was
good. The SIPA trustee has issued ten interim reports so far,
in addition to a detailed preliminary investigative report on
various aspects of LBI’s resolution.
FRBNY Economic Policy Review / December 2014
201
Appendix D: The Settlement of OTC Derivatives Contracts in Bankruptcy
Derivatives settlement procedures, as documented under the
ISDA Master Agreement, attempt to enable the nondefaulting
party to assert a claim for an amount that, if fully recovered,
would place it in the same position absent the default (Scott 2012).53
To do so involves four steps: 1) terminate contracts and
unwind all open transactions, 2) determine the value of each
transaction, 3) perform close-out netting, and 4) pay out net
amounts. The amount owed to or from a nondefaulting party
on account of default is equal to the net value of the
derivatives, as determined according to the selected valuation
methodology plus any unpaid amounts offset by the value of
the collateral. If the amount due to the nondefaulting party is
positive, then it becomes an unsecured creditor to the estate.
In a bankruptcy, derivatives and other qualified financial
contracts are awarded special legal treatment exempting them
from several provisions of the Bankruptcy Code, thereby
creating a safe harbor.54 First, derivatives creditors can net
offsetting positions with the debtor, seize and liquidate
collateral, and choose whether to close out and terminate
positions right after bankruptcy without being subject to the
automatic stay. Relatedly, creditors have broad rights to set off
debts owed to the debtor against debts due from the debtor if
a setoff provision has been included in the ISDA Master
Agreement. Second, they are exempt from certain creditor
liabilities related to pre-bankruptcy agreements such as
fraudulent conveyance liability (arising from the debtor selling
its own assets prior to bankruptcy for less than fair value) and
preference rules (the need to return preferential payments
received just before bankruptcy or to give back preferential
collateral calls). The remainder of this section focuses on the
first exemption relating to the procedures for termination,
liquidation of collateral, netting, and setoff.
The termination procedure for creditors is described by an
ISDA Master Agreement that lists the default events triggering
termination. Specifically, contracts terminate automatically
if the derivatives contract has an automatic early termination
53
ISDA is an industry trade association that has developed two documents
that are fundamental to any OTC derivatives transaction: the 1992 ISDA
Master Agreement and the 2002 ISDA Master Agreement. Multiple
derivatives transactions may be documented under a single Master
Agreement that contains alternative provisions to be selected by the
two signatories.
54
More formally, “safe harbor provisions” are provisions in the
U.S. Bankruptcy Code ensuring that derivatives contracts and other QFCs
are enforced according to their terms by creditors even after the debtor
files for bankruptcy, subject to certain exceptions under the code. Bliss and
Kaufman (2006) and Roe (2011) discuss the desirability and rationale of safe
harbor provisions.
202
The Failure Resolution of Lehman Brothers
clause or, alternatively, the nondefaulting party has the choice
(but not the obligation) to terminate by giving written notice
to the defaulting party. Naturally, the nondefaulting party
has an incentive not to terminate the contract when it is
out-of-the-money; moreover, in such cases, it has the right to
suspend periodic payments to the defaulting party under the
Master Agreement. Termination of a Master Agreement terminates all derivatives transactions under that agreement. The
Master Agreement is supplemented or amended by a schedule
that (among other things) states whether or not the derivatives transactions are supported by a guarantor or other credit
support provider. If so, a default by a credit support provider
will constitute a default event under the Master Agreement.
With early termination under a Master Agreement, parties
can seize any collateral posted pursuant to the agreement. A
derivatives transaction may include a credit support annex,
which is a security agreement that describes any collateral
pledged in the derivatives transactions. Typically, liquid
collateral (such as U.S. Treasury securities or agency securities)
is posted (Ricotta 2011). Collateral is “marked to market” and
the amount due to or from a party (its “exposure”) is calculated
periodically. Either one side or both sides to a transaction may
post collateral. The credit support annex may also permit a
party to hypothecate collateral posted and delivered by the
other party.
The valuation framework implicitly envisions a liquid
market such that the nondefaulting party closes out its open
positions at market rates and then establishes replacement
hedges to offset expected price changes (Das 2012).
Accordingly, v aluation of contracts requires determining the
exact timing of valuation, the method used, and the calculation
agent carrying out the valuation. Under the 1992 Master Agreement, parties can choose between the market quotation method
and the loss method. The market quotation method allows nondefaulting parties any unpaid amount plus replacement transactions valued based on quotes from at least three reference
market-makers; the loss method entitles the nondefaulting
party to “an amount that party reasonably determines in good
faith to be its total losses” from the terminated transactions. The
2002 Master Agreement uses the close-out amount approach,
which combines elements of the quotation and loss methods.55
55
Similar to the quotation method, the close-out amount approach entitles
the nondefaulting party to any unpaid amounts. Similar to the loss method,
it also allows a “close-out amount” equal to the replacement costs of the
terminated trades where the determining party may “use [any] commercially
reasonable procedures in order to produce a commercially reasonable result.”
Appendix D: The Settlement of OTC Derivatives Contracts in Bankruptcy
Appendix (Continued)
(Continued)
Once contract values are established, close-out netting is
used to determine the net settlement amount. Close-out
netting involves the calculation of gains or losses for each
party upon termination of a derivatives contract, repeating the
calculation for all of the derivatives transactions involving
the two parties and then offsetting the resulting amounts.
After applying any setoff rights and the value of collateral
posted, the procedure yields a single payment from one party
to the other. If a party has multiple derivatives transactions
with different affiliates of a firm, then netting requires written
agreements with the affiliates. If no such agreements exist,
then the ability to net depends on the local law of the jurisdiction (in particular, the applicable insolvency law), which often
prohibits multilateral setoffs (for example, derivatives counterparty A sets off an amount it owes to Lehman affiliateB
against an amount Lehman affiliate C owes to derivatives
counterparty A).
FRBNY Economic Policy Review / December 2014
203
References
Acharya, V., S. Bharath, and A. Srinivasan. 2007. “Does IndustryWide Distress Affect Defaulted Firms? Evidence from Creditor
Recoveries.” Journal of Financial Economics 85, no. 3
(September): 787-821.
Debtors’ Disclosure Statement for First Amended Joint Chapter 11
Plan of Lehman Brothers Holdings Inc. and Its Affiliated Debtors.
January 25, 2011. Available at http://www.wilmingtontrust.com/
lehman/pdf/Disclosure_Statement_First_Amended_01_25_11.pdf.
Altman, E., B. Brady, A. Resti, and A. Sironi. 2005. “The Link between
Default and Recovery Rates: Theory, Empirical Evidence, and
Implications.” Journal of Business 78, no. 6 (November): 2203-28.
Debtors’ Disclosure Statement for Second Amended Joint Chapter 11
Plan of Lehman Brothers Holdings Inc. and Its Affiliated
Debtors. June 30, 2011. Available at http://www.wilmingtontrust
.com/lehman/pdf/Second_Amended_Plan_filed_by_Debtors_
dated_06_30_11.pdf.
Azarchs, T., and S. Sprinzen. 2008. “How Is the Lehman Brothers
Holding Inc. Bankruptcy Progressing?” Standard and Poor’s
Global Credit Portal, RatingsDirect.
Bharath, S., V. Panchapagesan, and I. Werner. 2010. “The Changing
Nature of Chapter 11.” Available at SSRN: http://ssrn.com/
abstract=1102366 or http://dx.doi.org/10.2139/ssrn.1102366.
Bliss, R., and G. Kaufman. 2006. “Derivatives and Systemic Risk:
Netting, Collateral, and Closeout.” Journal of Financial
Stability 2, no. 1 (April): 55-70.
Braun, M. 2013. “Lehman Reaches beyond Grave Seeking
Millions from Nonprofits.” Bloomberg, May 14. Available at
http://www.bloomberg.com/news/2013-05-14/lehman-reachesbeyond-grave-to-grab-millions-from-nonprofits.html.
Bruche, M., and C. Gonzalez-Aguado. 2007. “Recovery Rates, Default
Probabilities, and the Credit Cycle.” CEMFI working paper
no. 0612.
CCP12–The Global Association of Central Counterparties. 2009.
“Central Counterparty Default Management and the Collapse of
Lehman Brothers.” April.
Charles, G. 2009. “OTC Derivative Contracts in Bankruptcy: The
Lehman Experience.” New York State Bar Association NY Business
Law Journal 13, no. 1 (Spring): 14-17.
Covitz, D., S. Han, and B. A. Wilson. 2006. “Are Longer Bankruptcies
Really More Costly?” Board of Governors of the Federal Reserve
System Finance and Economics Discussion Series, no. 2006-27, June.
Das, S. 2012. “In the Matter of Lehman Brothers.” Wilmott 2012,
no. 59 (May): 20-29.
Debtors’ Disclosure Statement for Third Amended Joint Chapter 11
Plan of Lehman Brothers Holdings Inc. and Its Affiliated Debtors.
August 31, 2011. Available at http://dm.epiq11.com/LBH/
Document/GetDocument/1430484.
Duffie, D., A. Li, and T. Lubke. 2010. “Policy Perspectives on OTC
Derivatives Market Infrastructure.” University of Chicago. MFI
Working Paper Series, no. 2010-002, January.
Durham, S. 2010. Terminating Derivative Transactions: Risk
Mitigation and Close-Out Netting. Practising Law Institute.
Federal Deposit Insurance Corporation. 2011. “The Orderly
Liquidation of Lehman Brothers Holdings Inc. under the DoddFrank Act.” FDIC Quarterly 5, no. 2: 31-49.
Fleming, M., and A. Sarkar. 2014. “Where Was Value Destroyed in
the Lehman Bankruptcy?” Unpublished paper, Federal Reserve
Bank of New York.
Giddens, J. 2008. “In re Lehman Brothers Inc.: A Securities Investor
Protection Act Liquidation.” Case No. 08-01420 (JMP) SIPA,
December 17. Available at http://dm.epiq11.com/LBI/document/
GetDocument.aspx?DocumentId=1324539.
———. 2010. “SIPA Liquidation Overview.” American Bankruptcy
Institute, New York City Bankruptcy Conference, April. Available
at http://www.abiworld.org/committees/newsletters/busreorg/
vol9num5/sipa.pdf.
Jackson, T. H., 2012. “Bankruptcy Code Chapter 14: A Proposal.” In
K. E. Scott and J. B. Taylor, eds., Bankruptcy Not Bailout: A
Special Chapter 14. Hoover Institution Press.
Lehman Brothers Holdings Inc. (LBHI). Debtor’s Motion for an
Order Pursuant to Sections 105 and 365 of the Bankruptcy Code
204
The Failure Resolution of Lehman Brothers
References (Continued)
to Establish Procedures for the Settlement or Assumption and
Assignment for Prepetition Derivative Contracts. November 13,
2008. Available at http://www.swapfinancial.com/documents/
document/19.
———. Debtor’s Motion for an Order Approving Consensual
Assumption and Assignment of Prepetition Derivative Contracts.
January 28, 2009. Available at http://dm.epiq11.com/LBH/
Document/GetDocument/1096464.
———. §341 Meeting. January 29, 2009. Available at http://
dm.epiq11.com/LBH/Document/GetDocument/1130920.
———. §341 Meeting. July 8, 2009. Available at http://dm.epiq11
. com/ LBH/Document/GetDocument/1130931.
———. Order Pursuant to Sections 105(a), 362, and 365 of the
Bankruptcy Code to Compel Performance of Contract and to
Enforce the Automatic Stay. September 17, 2009. Available at
http://dm.epiq11.com/LBH/Docket#Debtors=1906&Docket
Number=5209&RelatedDocketId=&ds=true&maxPerPage=
25&page=1.
———. Notice Regarding Third Distribution Pursuant to the
Modified Third Amended Joint Chapter 11 Plan of LBHI and
Its Affiliated Debtors. March 27, 2013. Available at http://
dm.epiq11. com/LBH/Document/GetDocument/2452746.
———. 2013+ Cash Flow Estimates. July 23, 2013. Available at http://
dm.epiq11.com/LBH/Document/GetDocument/2453379.
———. Notice Regarding Fourth Distribution Pursuant to the
Modified Third Amended Joint Chapter 11 Plan of LBHI and Its
Affiliated Debtors. September 26, 2013. Available at http://
dm .epiq11. com/LBH/Document/GetDocument/2453025.
———. Notice Regarding Fifth Distribution Pursuant to the Modified
Third Amended Joint Chapter 11 Plan of LBHI and Its Affiliated
Debtors. March 27, 2014. Available at http://dm.epiq11.com/
LBH/Document/GetDocument/2473191.
Lehman Brothers Special Financing Inc. versus BNY Corporate
Trustee Services Limited. Chapter 11 Case No. 08-13555 (JMP),
Adversary Proceeding No. 09-01242 (JMP). January 25, 2010.
———. The State of the Estate. November 18, 2009. Available at http://
dm.epiq11.com/LBH/Document/GetDocument/1130926.
Available at http://www.gpo.gov/fdsys/pkg/USCOURTS-nysb1_09-ap-01242/pdf/USCOURTS-nysb-1_09-ap-01242-0.pdf
———. The State of the Estate. September 22, 2010. Available at
http://dm.epiq11.com/LBH/Document/GetDocument/1265151.
Liew, C., J. Gu, and K. Noyes. 2010. “Derivatives Netting Not Fully in
———. Plan Status Report. January 13, 2011. Available at http://
dm.epiq11.com/LBH/Document/GetDocument/1309170.
Ricotta, P. 2011. “Some Lessons Learned from the Lehman Brothers
———. Notice Regarding Initial Distributions Pursuant to the
Modified Third Amended Joint Chapter 11 Plan of LBHI
and Its Affiliated Debtors. April 11, 2012. Available at http://
dm.epiq11. com/LBH/Document/GetDocument/1730908.
———. Debtors versus JPMorgan Chase Bank, N.A. Chapter 11
Case No. 08-13555 (JMP), Adversary Proceeding No. 10-03266
(JMP). April 19, 2012. Available at http://www.gpo.gov/fdsys/pkg/
USCOURTS-nysb-1_10-ap-03266/pdf/USCOURTS-nysb-1_10ap-03266-0.pdf.
———. Notice Regarding Second Distributions Pursuant to the
Modified Third Amended Joint Chapter 11 Plan of LBHI and
Its Affiliated Debtors. September 25, 2012. Available at http://
dm.epiq11.com/LBH/Document/GetDocument/2452720.
Place in Asia.” Asia Risk. August 4.
Bankruptcy Proceedings.” In Financial Derivatives in
Bankruptcy, ACFA Meeting, January 12. Available at http://
acfalaw.org/system/docs/79/original/Derivatives_In_Bankruptcy.
pdf?1301233552.
Roe, M. 2011. “The Derivatives Market’s Payment Priorities as
Financial Crisis Accelerator.” Stanford Law Review 63, no. 3
(March): 539-90.
Schuermann, T. 2004. “What Do We Know about Loss Given
Default?” In D. Shimko, ed., Credit Risk: Models and
Management. 2nd ed. London: Risk Books.
Scott, H. 2012. “Interconnectedness and Contagion.” Discussion
Paper, Committee on Capital Markets, November 20.
FRBNY Economic Policy Review / Forthcoming
205
References (Continued)
Smith, E. 2010. “Setoff Issues in Derivative Transactions: The
Lehman Experience.” Paper presented at the 2010 Annual
Meeting and Education Conference, American College of
Investment Council. Available at http://aciclaw.org/sites/default/
files/sessions/attachments/2013/ACICAM10_Session4b_
SetoffIssuesinDerivativeTransactions.pdf.
Summe, K. A. 2012. “An Examination of Lehman Brothers’
Derivatives Portfolio Postbankruptcy: Would Dodd-Frank Have
Made a Difference?” In K. Scott and J. Taylor, eds., Bankruptcy
Not Bailout: A Special Chapter 14. Hoover Institution Press.
Trustee’s Preliminary Investigation Report and Recommendations.
Case No. 08-01420 (JMP) SIPA. November 19, 2009. Available at
http://dm.epiq11.com/LBI/Document/GetDocument/1320923.
Trustee’s First Interim Report for the Period September 19, 2008
through May 29, 2009. Case No. 08-01420 (JMP) SIPA. Available
at http://dm.epiq11.com/LBI/document/GetDocument. aspx?
DocumentId=1320941.
Trustee’s Fifth Interim Report for the Period October 27,
2010 through April 22, 2011. Case No. 08-01420 (JMP)
SIPA. Available at http://dm.epiq11.com/LBI/document/
GetDocument. aspx?DocumentId=1355722.
Trustee’s Ninth Interim Report for the Period October 25,
2012 through June 14, 2013. Case No. 08-01420 (JMP)
SIPA. Available at http://dm.epiq11.com/LBI/document/
GetDocument. aspx?DocumentId=2385679.
Trustee’s Tenth Interim Report for the Period June 15, 2013 through
February 14, 2014. Case No. 08-01420 (SCC) SIPA. Available
at http://dm.epiq11.com/LBI/document/GetDocument.aspx?
DocumentId=2462769.
U.S. Government Accountability Office. 2011. “Complex Financial
Institutions and International Coordination Pose Challenges.”
GAO-11-707, July 19.
———. 2013. “Financial Company Bankruptcies: Need to Further
Consider Proposals’ Impact on Systemic Risk.” GAO-13-622,
July 18.
Valukas, A. 2010. “Report of the Examiner in the Chapter 11
Proceedings of Lehman Brothers Holdings Inc.” March 11.
Available at http://jenner.com/lehman/.
Wasserman, R. 2006. “A Step-by-Step Look at Chapter 11.” In
Inside the Minds: Best Practices for Corporate
Restructuring: Leading Lawyers on Communicating with
Creditors, Analyzing Debt, and Filing for Bankruptcy.
Aspatore Books.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
206
The Failure Resolution of Lehman Brothers
Joseph H. Sommer
Why Bail-In? And How!
• Financial firm insolvency is very different in
kind than the insolvency of ordinary firms.
• The key distinction is that only financial firms
are dominated by “financial liabilities”—
liabilities whose value is greater than
the net present value of their associated
income streams.
• The “bail-in” insolvency process respects
this distinction, converting the claims of
the parent company’s creditors to equity
while paying the subsidiaries’ creditors on
time and in full.
• Bail-in impairs only the nonfinancial liabilities
in the parent and preserves the financial
liabilities in the subsidiaries. It therefore
preserves the firm’s liquidity and riskshifting abilities.
• For systemically important financial firms,
bail-in averts systemic risk.
Joseph H. Sommer is an assistant vice president and counsel at the
Federal Reserve Bank of New York.
joseph.sommer@ny.frb.org
1. Introduction
B
ank resolution is a big topic these days.1 (“Resolution” is a
term of art, meaning something like “insolvency process.”)
This is especially true for megabanks—large international
financial conglomerates.2 Most bank regulators are unhappy
with standard insolvency law, such as the Bankruptcy Code
(Code). They often favor a novel process. The generic term is
“bail-in.”3 The Federal Deposit Insurance Corporation (FDIC)
has its own version, called “single point of entry.”
This raises two questions. Why should bank regulators
dislike standard insolvency law? And why should bail-in make
them happy? This article answers these questions.
1
E.g., citations contained infra notes 3, 4 (second paragraph), 8, 28, 34,
48, 56, 61, 65.
2
Throughout this discussion, I shall use the terms “megabank,” “large
financial firm,” or “large international financial conglomerate” as if they
all meant the same thing. I do not use the jargon term “SIFI,” the acronym
for “systemically important financial institution.” This article does not need
a systemic risk boogeyman, although it helps.
3
D. Wilson Ervin, a banker at Credit Suisse, invented the concept.
Paul Calello & Wilson Ervin, From Bail-Out to Bail-In,
The Economist 95 (Jan. 28, 2010). Mr. Ervin told me that he
conceived the idea around September 2008.
The author thanks Barry Adler, Thomas Baxter, Wilson Ervin, Mark Flannery,
Charles Gray, Joyce Hansen, HaeRan Kim, Lisa Kraidin, James McAndrews,
Hamid Mehran, Donald Morgan, Brian Peters, Michael Schussler, and
David Skeel. The views expressed in this article are those of the author
and do not necessarily reflect the position of the Federal Reserve Bank of
New York or the Federal Reserve System.
FRBNY Economic Policy Review / December 2014
207
My answers are not quick ones. The theory for a quick
answer is not there. Few bankruptcy scholars or practitioners
know about financial firm insolvency. Unlike bankruptcy,
the secondary legal literature on financial firm insolvency is
sparse.4 Not everybody has read it, and besides, I have a few
notions of my own.
The first section of this article therefore discusses
megabank insolvency. Ordinary bankruptcy law makes many
tacit assumptions as to what a generic firm should be. Many
of these assumptions are invalid—or even inverted—for
financial firms. The second section defines and discusses
bail-in. The third section defines, discusses, and dismisses the
alternatives to bail-in.
With this teaser, let us get started.
4
I know of one excellent, if ancient, monograph. Hirsch Braver,
Liquidation of Financial Institutions (1936). It remains useful for a
few technical issues; cf. Office of the Comptroller of the Currency,
Instructions to National Bank Receivers (1932). There are three
useful modern monographs: David A. Skeel, The New Financial
Deal: Understanding the Dodd-Frank Act and Its (Unintended)
Consequences (2011); Bankruptcy Not Bailout: A Special
Chapter 14 (Kenneth E. Scott & John B. Taylor, eds., 2012) (“Hoover
Institution”); Eva Hüpkes, The Legal Aspects of Bank Insolvency:
A Comparative Analysis of Western Europe, the United States
and Canada (Kluwer 2000). The law review literature is sparse between
the Depression and the 2008 crisis. Robert R. Bliss & George G. Kaufman,
U.S. Corporate and Bank Insolvency Regimes: A Comparison and Evaluation,
2 Va. L. & Bus. Rev. 143 (2007); Thomas C. Baxter, Joyce M. Hansen
& Joseph H. Sommer, Two Cheers for Territoriality, 78 Am. Bankr.
L.J. 57 (2004); David A. Skeel, The Law and Finance of Bank and
Insurance Insolvency, 76 Tex. L. Rev. 723 (1998); Peter B. Swire, Bank
Insolvency Law Now that It Matters Again, 42 Duke L.J. 469 (1992);
William R. Buck, Jr., Comment, Bank Insolvency and Depositor
Setoff, 51 U. Chi. L. Rev. 188 (1984). See also, e.g., Group of Thirty,
International Insolvencies in the Financial Sector 84 (1998);
G-10 Contact Group on the Legal and Institutional Underpinnings of the
International Financial System: Insolvency Arrangements and Contract
Enforceability (September 2002) at http://www.bis.org/publ/gten06.htm.
Recently, the topic has become more trendy: e.g., Skeel, supra; Hoover
Institute, supra; Douglas G. Baird & Edward R. Morrison, Dodd-Frank for
Bankruptcy Lawyers, 19 Am. Bankr. Inst. L. Rev. 287 (2011); David A. Skeel
& Thomas H. Jackson, Transaction Consistency and the New Finance in
Bankruptcy, 112 Colum. L. Rev. 152 (2012); Peter Conti-Brown, Elective
Shareholder Liability, 64 Stan. L. Rev. 409 (2012); Randall D. Guynn,
Are Bailouts Inevitable?, 29 Yale J. Reg. 121 (2012); Adam J. Levitin, In
Defense of Bailouts, 99 Geo L.J. 435 (2011); John R. Bovenzi, Randall
D. Guynn & Thomas H. Jackson, Too Big to Fail: The Path to a
Solution (Bipartisan Policy Center 2013); Thomas F. Huertas, Safe to Fail,
28 Butterworth’s J. of Int’l Banking & Fin. L. 407 (2013); Paul Tucker,
Resolution and Future of Finance (May 20, 2013) (available at http://www.bis.
org/review/r130606a.pdf?frames=0
(last visited June 12, 2013); High-Level Expert Group on
Reforming the Structure of the E.U. Banking Sector (available at
http://ec.europa.eu/internal_market/bank/docs/high-level_expert_group/
report_en.pdf) (last visited September 26, 2013).
208
Why Bail-In? And How!
2. Why Megabanks Are Different
Megabanks are more than big banks. They are more complex;
they are more interlocked; they are more global. They also
have a peculiar corporate structure and—most importantly—
peculiar liabilities. They are atypical firms. As we shall see,
they need atypical insolvency law.
This section does four things. It starts with the balance sheet
of megabanks. All financial firms have peculiar balance sheets.
Megabanks are more peculiar yet. Megabanks are highly
leveraged. Their liabilities are generally financial products,
often very liquid. Finally, megabanks are conglomerates, with
close connections among component entities.
After these balance-sheet concerns, we then look at the
business of megabanking. Several traits stand out: personnel,
interconnections, and a global reach. These all affect
megabank insolvency.
We then look at the nature of bank supervision—
including megabank supervision. Supervision is tightly tied
to insolvency, much as corporate finance is tied to corporate
insolvency. The incentives of local supervisors are a powerful
force in cross-border megabank insolvency.
Add all these together, and we have our conclusion:
Megabank insolvency is different from that of
other businesses.
2.1 The Peculiar Balance Sheet of Megabanks
We begin by looking at the balance sheet of large financial
firms. A “financial firm,” for our purposes, has a high degree
of leverage. It also has something I call “financial liabilities.”
Megabanks also have a conglomerate structure.
We begin with the two traits that megabanks share with
other financial firms: leverage and financial liabilities.
Leverage
Almost all financial firms are highly levered. Debt-equity
ratios of 1:1 are typical for ordinary firms: the widget
maker of the textbook. Financial firms’ debt-equity ratios
are much higher: about 15:1-30:1 for banks and securities
firms,5 and somewhat less for insurers. This leverage has
some implications.
5
Sebnem Kalemli-Ozcan, Bent Sorensen & Sevcan Yesiltas, Leverage Across
Firms, Banks, and Countries, 88 J. Int’l Econ. 284 (2012).
First, financial firms maintain good credit despite their
high leverage. Before the recent crisis, banks typically got
ratings around A- to AA.6 This need do no violence to the
basic tenets of corporate finance. Financial firm assets are
typically much safer (i.e., lower in variance) than the assets of
general business firms. They must be. The classical bank asset
is somebody else’s debt. Debt is paid before profit. A firm that
specializes in holding debt will have less risk than firms that
must pay the debt.
However, megabanks hold many assets other than simple
debt. Some of these other assets are risky indeed. To reduce
the variance of these assets, megabanks use diversification
and hedging schemes. Diversification and imperfect hedging,
of course, rely on historical behavior. Beyond a few standard
deviations, history is bunk—so-called “tail risk.” Megabanks
are inherently sensitive to tail risk in their models, more
so than less sophisticated firms. We have seen this several
times in the past few decades: Askin Capital, Long-Term
Capital Management, and AIG come to mind. None of these
firms had bank charters, but our definition of “megabank”
needs no charter.
Second, high leverage is hard to measure. In a highly
levered firm, the value of assets is close to the value of
liabilities. A small error in measurement can lead to a large
error in reported leverage. Measurement is harder in the
insurance industry, where the (contingent) liabilities of claims
are probably more difficult to measure than asset values. But
it is bad enough in banking, and amplified by the greater
leverage in the banking industry. This is especially true for
megabanks with substantial contingent liabilities in the form
of derivatives contracts. Similarly, a small change in the
variance of bank assets can lead to a large change in bank risk.
This has some unpleasant implications for governance. As
Jensen and Meckling have told us, overleveraged firms may
gamble with their creditors’ money.7 But the leverage and risk
of banks is difficult to measure. A small increase in either is
hard to verify, and banks traditionally had weak creditors,
anyway. This is both a good argument for capital regulation
(smaller risk substitutions are more effective in leveraged
firms), and for supervision (a creditors’ agent argument
independent of deposit insurance).
Third, leverage is hard to define, even if it is measurable.
The notion of leverage distinguishes between some liabilities
called “capital” and other liabilities. This distinction may
still exist in corporate law, but corporate finance theory
views it as mere superstition.8 There are only classes of risk,
with different classes of control appropriate to the risk class.
Nonetheless, financial regulators think that it is important to
sharply distinguish some kinds of liabilities from others.
They do not draw the line at equity. They will also consider
some kinds of subordinated debt and preferred stock to
be capital. Other kinds of debt—even long-term debt—do
not so qualify.
The regulators are right, and the theory is wrong. Not all
liabilities are created equal, like beads on a linear string of
risk. As we shall see, some liabilities—financial liabilities—
are indeed different in kind.
Financial Liabilities
Financial liabilities are a key concept in this article—the key
concept of this article. Since induction trumps deduction,
a list should precede my definition. Financial liabilities are
those that only financial firms are in the business of incurring.
They include things like bank deposits, derivative contracts,
insurance policies, and repos. Corporate debt or trade credits
are not financial liabilities. This is also true of the corporate
debt or trade credit of financial firms.
With our intuition set, we should now define the term.
We need a definition that makes a difference in insolvency.
Two such definitions come to mind.
Perhaps the best definition of a financial liability is one
whose value is impaired by the insolvency process. Yes,
insolvency does nothing if it does not affect liabilities. But
I mean this in a certain special sense, one evoked by an old
bankruptcy lawyers’ joke: “not only does the food taste awful,
but the portions are too small.” This joke is funny because
it is true. The portions must be too small: insolvency must
impair liabilities. Somebody will not get what they bargained
for. And the food tastes awful: bankruptcy destroys value. But
value destruction is not an inherent trait. An ideal insolvency
process can impair liabilities quickly, smoothly, and with no
collateral damage. But the Bankruptcy Code, although pretty
good, is not an ideal process.
We usually think in terms of destroyed asset value—
deadweight losses, such as administrative costs, lost goingconcern value, and so on. Value is destroyed on the asset side,
and modern insolvency law mitigates this destruction. But
6
Frank Packer & Nikola Tarashev, Rating Methodologies for Banks, BIS
Quarterly Review 39, 41, 49 (June 2011) (citing Fitch and Moody’s ratings).
A notch or two of these ratings consisted of government support, but they are
still investment grade. Id. at 50.
7
Michael Jensen & William Meckling, Theory of the Firm: Managerial
Behavior, Agency Costs, and Capital Structure, 3 J. Fin. Econ. 305 (1976).
8
To be fair, some corporate finance theorists are beginning to see the wisdom
to this superstition, at least where systemic risk is concerned. Oliver Hart
& Luigi Zingales, A New Capital Regulation for Large Financial Institutions,
13 Am. L. & Econ. Rev. 453, 455-54 (2011). But as we shall see, the issue is
a more general one than systemic risk.
FRBNY Economic Policy Review / December 2014
209
some special liabilities—financial liabilities—may also lose
value in the insolvency process, beyond any value measured
by the net present value of the claim. They, too, suffer a
deadweight loss.
Financial liabilities are more than claims to a future stream
of income. Yes, they are that, but they are something else,
too. Some of them, such as bank deposits, are also sources
of liquidity. Others—derivatives or insurance policies—shift
risk. These liabilities are also credit instruments. But unlike,
say, corporate bonds, their credit nature is incidental to—
but inherent in—their liquidity or risk-shifting functions.
Liquidity and risk shifting are valuable in themselves—
valuable beyond the face value of the financial liability. This is
not a conjecture. It is a revealed preference. Insurance policies
cost more than their net present value. Liquid debts pay less
well than illiquid ones.
This has potent implications in insolvency. If financial
liabilities have value that goes beyond their face value,
the extra cost of impairing them in insolvency is itself a
deadweight loss, exceeding the cost of impairing other
liabilities. This cost includes impaired liquidity or risk
shifting, as well as an impaired payment stream. This cost
is substantial, and it can be enormous if it takes the form of
systemic risk.9
Good insolvency law would protect these liabilities:
preserve their liquidity or risk-shifting functions. This would
be at the cost of other liabilities: the familiar ModiglianiMiller seesaw. However, this seesaw has no fixed pivot:
priority creates value. The other liabilities are mere streams
of income, with no other function. The cost of impairing
them is less than the cost of impairing financial liabilities.
This leads to the central policy implication of this article:
Financial liabilities deserve priority treatment in insolvency
law. Such priority exists today, ordinal and often temporal.
Financial liabilities are often paid first in line and often first
in time, before any payments to other creditors. We return
to this point below, after a detour through a few of these
financial liabilities.
A checking account is a financial liability. People hold
these accounts because they are liquid. A delayed insolvency
distribution is an illiquid distribution. An illiquid distribution
is costly. For an individual, these costs may include bad credit
reports, eviction, or delayed medical care. For a firm, these
costs may include strained relations with creditors, or even
its own insolvency. For a megabank, the costs can extend to
chain-reaction illiquidity and insolvency: systemic risk. In a
typical bank insolvency, checking accounts are typically paid
in full with no delay.
9
See infra text accompanying note 23.
210
Why Bail-In? And How!
Bank deposits are not the only financial liability. Insurance
policies are another. Consider a term life policy as an
example, with an insolvent insurer. To healthy policyholders,
the policy is worth little, because it is easy to replace. But
consider a person who purchased the policy at a low cost
when healthy, has subsequently developed cancer, and
has a right to renew. What does it mean for an insolvency
distribution to treat all policyholders equally? A quick pro rata
distribution would be a disaster for the sick policyholder,
who will not be able to replace the policy on the market. A
future claims estimation process could be a nightmare of
litigation. Insurance insolvencies, therefore, are often the
opposite of bank insolvencies: very slow, so that all claims may
come to fruition. Policyholders are first in line, even if the
line moves slowly.
Brokerage accounts are financial liabilities, much like bank
deposits. We usually think of them as direct property rights
of the customer, with the broker acting as a kind of bailee.
But the property law of securities, contained in Uniform
Commercial Code (UCC) Article 8, shows that they are
generally relationships between the broker and customer,
rather than a direct property right of the customer against the
issuer.10 Again, the liquidity of these securities is important.
Issuers structure debt as a security rather than a direct loan
to enhance its liquidity, either in the distribution process or
the secondary market. The Security Investors Protection Act
(SIPA) insolvency, as well as Subchapter III of Chapter 7 of the
Code, further enhance this liquidity. They rapidly transfer the
customer securities of insolvent brokers to another party that
can comply with customer orders.
The protected contracts of the Federal Deposit Insurance
Act (FDI Act) and Bankruptcy Code can also be financial
liabilities. (These liabilities are characteristic of megabanks.)
Derivative contracts are intended to shift risk. Insolvency
impairs this function. Many securities contracts have a similar
function. Many of these transactions hedge portfolios of
market risk. Repo creates liquidity.
In other words, financial contracts are contracts in which
credit risk is incidental. The creditor in such contracts is not
primarily an investor: paying money now to get more later.
Instead, it wants liquidity, or insurance, or other kind of risk
shifting. It is the creditor that is different, not the debtor. To a
financial firm, the proceeds of a financial liability are a form
of credit, used like any other credit. This suggests a second
definition of a financial liability: a product of a firm, sold to
10
Some of the more important pieces of secondary literature:
James Steven Rogers, Policy Perspectives on Revised U.C.C. Article
8, 43 U.C.L.A. L. Rev. 1432 (1996); Kenneth C. Kettering, Repledge
Deconstructed, 61 U. Pitt. L. Rev. 45 (1999); Steven L. Schwarcz,
Intermediary Risk in a Global Economy, 50 Duke L.J. 1541 (2001).
customers. This definition pretty much overlaps with the first.
It is less analytically pleasing, because it does odd things like
exclude interdealer derivatives contracts. But it explains much,
works well, and is easy on the intuition.
This notion of a financial liability as a product has
implications for insolvency law, apart from priorities.
Insolvency law assumes that firms often need a breathing
spell from their creditors, so that they can pick themselves up,
continue operating, and start reorganizing. It therefore places
all claims in a collective procedure and places a moratorium
on efforts to collect assets. However, financial products are
operations of the financial firm. Freezing performance on a
financial product, whether by automatic stay or treatment as
a claim, is akin to prohibiting a carmaker in Chapter 11 from
making and selling cars, or an airline from selling tickets,
buying jet fuel, and flying planes.11
Since a financial liability is a product, it contains some
goodwill. It is worth more to the issuer than the mere
proceeds of other liabilities. Nobody thinks of customer
loyalty in the bond market, but it is quite common in the
insurance market, or the retail market for bank deposits. In
other words, some financial liabilities—e.g., bank deposits
or insurance policies—can be firm-specific. They are more
valuable if kept with the firm’s business than paid off in an
insolvency distribution. Therefore, purchasers will assume
these liabilities for a discount. They need less than one dollar
of assets to assume a dollar’s worth of these liabilities.
This turns a standard bankruptcy argument inside out.
Bankruptcy scholars argue that reorganizations are generally
more efficient than liquidations because reorganizations
preserve the value of firm-specific assets.12 Financial firms
may have fewer firm-specific assets than other firms. But
they have firm-specific liabilities. These liabilities also
require reorganization of the firm—but no alteration of these
liabilities. Other liabilities would bear the brunt.
This article uses the concept of financial liabilities to
define financial firms. This excludes many firms that have
financial assets: e.g., leasing, factoring, lending, or mortgage
companies. These firms raise their funds from banks and on
the bond market, like any other ordinary business firm. They
do not issue financial liabilities. They are not financial firms,
for our purposes.
Our definition of financial firms might be narrow, but it
accords with U.S. law. The law defines banks by their unique
power to issue one financial liability: the deposit. (This power
is legally necessary, if not always sufficient.13) The Bankruptcy
Code treats leasing, factoring, lending, or mortgage
companies as ordinary industrial firms. The Code can succeed
with these firms, as it did with CIT Group. It only excludes
those firms defined by financial liabilities: banks and insurers.
When the Code sees something like a financial liability, it
typically feels protective. Financial contracts such as swaps
and repos are exempt from the stay and most avoidance
provisions. Subchapter III of Chapter 7 transfers customer
security positions rapidly, again free of the stay and with
limited avoidances. Trade credit is a special case. It is not a
financial liability—it is worth no more to the creditor than
its net present value. However, paying the trade credit in
full is often worth more to the estate than any loss to other
creditors, since an angry trade creditor can refuse to deal. So
the Code creates a twenty-day priority and a forty-five-day
quasi-priority in the form of a right of reclamation.14
12
A “liquidation” sells the assets of the firm and distributes the sales proceeds
down the priority ladder of creditors. A “reorganization” leaves the assets alone,
and readjusts the liabilities by eliminating or reducing junior claims, converting
senior claims to junior ones, and lengthening some surviving senior claims.
Chapter 11 of the Bankruptcy Code employs both liquidation (usually of
bulk business) and reorganizations. See infra text accompanying note 68.
11
To express this in more abstract language, firms have both operations and
financing. The task of reorganization is to rearrange the claims of financiers,
without disturbing operations. However, operations require continuing
financing. Insolvency law must separate the two somehow. The Code does
this with the automatic stay and post-petition lending. The automatic stay
keeps the erstwhile financiers away from the operations; post-petition lending
funds ongoing operations. This decomposition is impossible for a financial
firm, if there is no distinction between financing and products. Bail-in works
by segregating the financial products (and thus the operations) from the
nonproduct financing. This segregation requires an insolvency priority for
the financial products. The segregation only needs to be good enough, not
perfect. Even the Dodd-Frank Act (see Section 3.3 below) has a one-day
stay on financial contracts.
Liquidation and reorganization are not the only insolvency law techniques.
Bank and insurance insolvency law allows for a transfer of liabilities
to a solvent party, compensating the transferee with assets. This is the
“bridge bank” and “purchase and assumption” transaction of the FDI Act
or the “bridge company” of the Dodd-Frank legislation. As we shall see
below, a bridge company can be tantamount to a reorganization. See infra
text accompanying notes 55-56.
13
Necessary and sufficient: 12 U.S.C. § 378(a)(1). Necessary but not
sufficient: 12 U.S.C. § 1841(c)(1)(B) (also needs commercial lending);
12 U.S.C. § 1813(a)(2) (also needs incorporation).
14
Financial contract provisions: 11 U.S.C. §§ 362(b)(6), (7), (17), (27),
546(e)-(g), 555-56, 559-61; Subchapter III: 11 U.S.C. §§ 741-52; Sales
priority: 11 U.S.C. § 503(b)(9); Right of reclamation: 11 U.S.C. § 546(c).
FRBNY Economic Policy Review / December 2014
211
Bankruptcy practice expands on these statutory hints with
so-called “critical vendor orders.” These orders, awarded at
the beginning of the process, grant ordinal and temporal
priority to select liabilities. These orders originally protected
suppliers, for the reasons discussed above. They now often
cover consumers who have paid but have not received value:
e.g., warranties or airline tickets.15 (Since modern critical
vendor orders often support non-vendors, I henceforth use an
acronym: CVO.) Consumer CVOs preserve the reputation of
the firm to its customers: an apparent requisite of successful
reorganization.
CVO treatment also applies to true financial liabilities, at
least those few liabilities of Code entities that are not already
exempted as repo or derivative contracts. Customer securities
and commodities positions are financial liabilities. The Code
demands their rapid transfer, and regulatory segregation
principles usually give them priority. Casinos are financial
firms; casino chips are financial liabilities; they pass free of the
stay with administrative priority.16 I know of only one other
Code entity with financial liabilities: money transmitters. The
liabilities are the payments in transit: funded, but not paid.
The Code does not recognize these liabilities as special. State
law works around the Code, by pairing these liabilities to
segregated assets:17 a statutory trust.
However, most megabanks have an enormous volume of
financial liabilities: far greater than their other liabilities. They
do not pair these liabilities to specified assets. The argument
for CVOs only works if the favored liabilities are paired with
segregated assets, or are few.18 Therefore, the CVO approach
will not work for megabanks.
The difference between financial firms and others, then,
is one of degree, rather than kind. Financial liabilities
dominate the balance sheet of financial firms. Other firms
may have some peculiar liabilities. But they do not have
enough of them to interfere with formal Code doctrine. CVOs
are nonstatutory and limited in scope. Therefore, financial
firms require an explicit priority for financial liabilities: one
absent from the Code.
Conglomeration
Megabanks are seldom—if ever—single entities. Instead, they
are typically conglomerates. The parent is typically a bank
or a holding company (in U.S. law, the latter.) Some of the
subsidiaries have financial liabilities: banks, insurers, securities
dealers, derivatives dealers, or the like. Other subsidiaries
do not: mortgage banks, venture capital firms, asset holding
companies, various kinds of middlemen, or the like. There
are also special purpose vehicles (SPVs), which purport to
be bankruptcy-remote, but often operate with megabank
resources. Except for the bank and perhaps a reinsurer, there
is little cross-border branching: each country (or at least each
major country) gets its own set of subsidiaries.
Most of these affiliates are centrally controlled, sharing risk
management, personnel, business, reputation, and operations.
They also lend to each other. Typically, the parent and bank
are the main sources of interaffiliate credit, because they are
the most creditworthy entities. The bank is creditworthy
because regulators limit its interaffiliate credit exposures.19
Some other affiliates have their own credit, as standalone
business units or SPVs. But most do not. The credit of most
megabank entities depends on that of the organization. And
conversely. Except for the insurance industry, parent financial
firms seldom let their affiliates become insolvent, even when
there is no question of legal exposure. This is at least as old
as the salad oil swindle of 1963, in which American Express
rescued its warehouse company. And we saw it in 2007-08,
when parent firms rescued their shadow banks, despite a clear
legal separation between them.
The net result is an organization that is hard to
decompose in insolvency, even if interaffiliate books were
perfect. (“[I]mperfection in intercompany accounting is
assuredly not atypical in large, complex company structures.”20)
Insolvency law treats the legal entity as the basic unit upon
which it operates. Insolvency law acknowledges that affiliation
usually calls for unified administration, but otherwise treats
the separate entities with great respect.
15
The oddest CVO may have been In re Marvel Entertainment Group,
209 B.R. 832 (D.Del. 1997) (comic books paid for but not delivered
to children). The leading case on CVOs is clearly In re Kmart,
359 F.3d 866 (7th Cir. 2004). Judge Easterbrook suggested a statutory
basis for this and proposed an economic rationale. Id. at 872-73. He
argued that the CVO priority is justified when it increases the value of the
estate to the other creditors.
16
In re TCI 2 Holdings LLC, 428 B.R. 117, 180 (Bkrtcy. D.N.J. 2010).
17
Uniform Money Services Act § 701(c) (2004). Cf. Ronald Mann, The Rise
of State Bankruptcy-Directed Legislation, 25 Cardozo L. Rev. 1805 (2004).
18
See supra note 15 and accompanying text. A sufficiently large volume
of CVO priorities cannot make the other creditors better off.
212
Why Bail-In? And How!
19
See 12 U.S.C. §§ 371c & 371c-1; E.U. Conglomerate Directive:
Directive 2002/87/EC of the European Parliament and of the Council
of 16 December 2002 on the supplementary supervision of credit
institutions, insurance undertakings, and investment firms in a
financial conglomerate and amending Council Directives 73/239/EEC,
79/267/EEC, 92/49/EEC, 92/96/EEC, 93/6/EEC and 93/22/EEC, and
Directives 98/78/EC and 2000/12/EC of the European Parliament and of the
Council 2003 O.J. (L 35/1).
20
In re Owens-Corning, 419 F.3d 195, 215 (3d Cir. 2005) (Ambro, J.)
Yet megabanks continue to use these affiliated structures,
for several reasons.21 First, regulators sometimes force them
to. A good example of this is the separation between banking
and securities underwriting/dealing. This is near mandatory
in U.S. law,22 but rare in Europe. Second (although first in the
hearts of corporate lawyers) is tax avoidance. Tax avoidance
does not increase welfare; credit impairment decreases
welfare. Third, securitization relies on separate entities.
Fourth, insurance insolvency law is incompatible with other
insolvency law. This requires that the insurance business of a
firm be in a separate subsidiary. Fifth, insurance companies
do not care as much about credit as they care about tail
risk. For them, separate subsidiaries reduce tail risk at a
reasonable cost to credit. Other reasons doubtless exist—some
good and some bad.
Entity proliferation certainly complicates insolvency
law. But good or bad, it is a fact. Megabanks are complex
and highly interconnected conglomerates. Any megabank
resolution scheme must deal with this.
traders, quantitative analysts. The work that they do is directly
linked to the profitability of the firm, so the profit of a business
unit or subunit is a reasonable proxy for their performance—
and thus their pay. Because these personnel face outward, the
top performers have a reputation throughout the industry.
This reputation adheres to them (or their team), more
than it does the bank that employs them. In other words,
they have very little firm-specific human capital: much like
superstar athletes or scholars.
Since these personnel have weak ties to their firms, they
can easily leave for another megabank. Megabanks are
aware of this, and seek to hold their stars with deferred pay
packages. However, these packages contain credit or market
risk, and are less credible if a megabank appears weak. Hence,
if a megabank appears weak, its successful high-paid teams
tend to go elsewhere. This run on human capital can parallel
a run on more conventional parts of the balance sheet.
Interconnectedness and Systemic Risk
2.2 The Peculiar Ecology of Megabanks
We now turn to a few attributes of megabanks that do not
show on their balance sheets, yet do affect their insolvency.
First, megabanks have little specific human capital. Highpaid individuals and teams can run from the megabank
almost as quickly as deposits can. Second, megabanks are
interconnected. Third, most megabanks are international,
spread across many legal regimes.
The Human Factor
All banks are subject to a run on their liabilities. Megabanks
are subject to a run on their personnel.
Most megabanks contain many high-paid sales and trading
personnel who are not management: investment bankers,
21
Thomas C. Baxter, Jr. & Joseph H. Sommer, Breaking Up Is Hard to Do:
An Essay on Cross-Border Challenges in Resolving Financial Groups, in
Systemic Financial Crises: Resolving Large Bank Insolvencies
175 (Federal Reserve Bank of Chicago 2005); Richard Herring & Jacopo
Marcassi, The Corporate Structure of International Financial Conglomerates:
Complexity and Its Implications for Safety and Soundness, in The Oxford
Handbook of Banking (Allen N. Berg, Philip Molyneux & John O. S.
Wilson, eds.) (Oxford 2012).
Megabanks are highly interconnected. This implies that a
weakness in one megabank can become a weakness in all.
The mechanism is unimportant. It could be a pure panic
attack, with bad news for one bank imputed to all. Or
perhaps a markdown of an asset class by one bank triggers
markdowns by all. Or an industry-wide hedging model goes
south. Or perhaps one megabank’s liabilities are others’ assets.
Or perhaps a clearinghouse goes bad, blocking liquidity.
Leverage and liquidity stress seem to be important.
To make matters worse, the asset side also becomes
illiquid in times of general stress. (I make no claims of causal
direction here.) Therefore, asset liquidity dries up precisely
when a megabank most needs this liquidity. Tradable assets
are not naturally liquid; they are only liquid because of legal
rules and market conventions. Market liquidity is at best
factitious. In times of stress, it may become fictitious.
Megabanks, like any other bank, are subject to runs on
their liabilities. Bank transaction deposits are liquid by design:
always susceptible to a run.23 Some megabank liabilities, such
as commercial paper or repo, expire very quickly and are
also liquid. Bank derivative liabilities are ordinarily illiquid,
because derivative contracts commonly remain outstanding
for years at a time. However, this illiquidity is illusory; most
derivative contracts have hair-trigger closeout provisions, and
also demand constant collateral calls. When the bank is under
22
The few remaining shards of the Glass-Steagall Act still restrict the equities
activities of national banks. But perhaps more significant these days are the
Securities and Exchange Commission’s capital requirements, which would be
prohibitive if applied to a bank’s balance sheet.
23
Douglas W. Diamond & Philip H. Dybvig, Bank Runs, Deposit Insurance,
and Liquidity, 91 J. Pol. Econ. 401 (1983).
FRBNY Economic Policy Review / December 2014
213
stress, the closeout provisions loom large, and the collateral
calls generally create greater demands on liquidity.
The mechanism is not important: only the results. Banks
are interconnected. Chain-reaction illiquidity or insolvency
is possible: the systemic risk boogeyman. Systemic risk events
are not common, and seldom trace back to a single cause.
But they are frightening. We need not know the etiology of
systemic risk—its consequences are enough.
With all this being said, systemic risk events are not the
norm, even for megabank insolvency. Megabanks can usually
ride out times of financial stress: for example, the 1997 Asian
financial crisis, or the 1987 stock market break. Conversely,
financial firms often collapse in isolation, even large firms.
They still go down quickly, but they go down smoothly. Enron
is one example; others are Barings Bank, Drexel Burnham
Lambert, Refco, MF Global, and Amaranth Advisors. Some of
these firms’ failures created stress (Drexel, Barings); others did
not (Refco, Enron, Amaranth, MF Global.) These firms mostly
went down in relatively good times, often from some kind of
fraud. Systemic risk is reserved for times of extreme market
stress. But systemic risk inheres in the balance sheets and
business practices of megabanks.
Internationality
As a stylized fact, most megabanks are international.There are,
of course, some exceptions, but internationality is the norm.
The insolvency of international firms is more complex
than that of domestic firms. Cross-border insolvency may
entail multiple and competing insolvency administrations
of a firm. Each administrator uses its own law to conduct
the proceeding (lex concursus), although it usually defers to
the relevant local law governing assets and liabilities. Lex
concursus includes process, avoidances, priorities, conflicts of
law, and any stays. Setoff and netting may be lex concursus, or
may be local law.
The basic unit of responsibility is the entity. The emerging
norm is that of a central administrator, with other juris
dictions in a supporting role. These other jurisdictions
conduct “ancillary proceedings” that assist the main
proceeding. In a liquidation, the ancillary proceeding
collects assets and distributes them to the central receiver for
distribution. In a reorganization, the ancillary jurisdiction
enforces the stay and does whatever asset collections
are necessary. This cooperation requires a consensus on
roles. Who runs the central proceeding? Who assumes the
214
Why Bail-In? And How!
ancillary role? This consensus is codified in an international
instrument: a model law.24
But this norm is limited to entity insolvency. Cooperation
on conglomerates is more informal, and not nearly as
effective, since the entity is the basic unit of insolvency law.
Jurisdictions are not likely to cede their primacy on their
local entities. However, most cross-border conglomerate
insolvencies still work themselves out, albeit awkwardly
and inefficiently. There are some incentives for cooperation.
The assets of ordinary firms are typically firm-specific:
the justification for the automatic stay. Local liquidations
will destroy value, and the automatic stay buys time for
cooperation. Industrial insolvencies are common, and large
firms are spread over the globe. This supports a norm of
reciprocity. A jurisdiction may agree to a subordinate role
now, in return for a central role later.
The incentives for cooperation are far weaker in financial
firm insolvencies. Financial insolvencies (especially major
ones) are more rare, and asset specificity less intense. As we
shall see below, parochial regulators weaken these incentives
even more. Not only are incentives for cooperation weaker:
cooperation is more difficult. To preserve liquidity and
confidence, megabank resolutions must be very fast. Certain
parts of them are over with almost before they start. And
furthermore, megabank insolvencies give very little warning.
Cooperation, then, must be ex ante. For sovereigns, this is
much more difficult than ex post cooperation. It involves
ceding sovereignty, rather than extending comity.
There is a final problem. International banks, unlike
most international firms, tend to use branches rather than
subsidiaries. (The bank is typically branched; the rest of
the megabank is typically compartmentalized by national
subsidiary.) This tends to complicate insolvency law. There is
no international consensus on the insolvency of cross-border
bank entities.25 Indeed, the Model Law has an express
carve-out for bank insolvency. Some jurisdictions (such as
the United States) treat branches as if they were separate local
juridical entities. Most claim to subject them to conventional
ancillary proceedings—at least in theory.
24
UNCITRAL Model Law on Cross-Border Insolvency, 36 I.L.M. 1386
(1997). The United States adopted this model law in 2005 as
Chapter 15 of the Bankruptcy Code.
25
See infra notes 31-35 and accompanying text.
2.3 Prudential Supervision
It is impossible to understand financial insolvency law
without understanding something about bank supervision.
Here we talk about two things: the role of supervisors
in ordinary times, and their role and behavior in the
neighborhood of insolvency.
What Do Supervisors Do?
Most office workers have a difficult time explaining their jobs
to their children. Financial supervisors have it worse—they
have a difficult time explaining their job to adults. Here I will
try to explain it to you—at least enough to understand their
role in financial insolvency.
Any discussion of financial supervision must begin with
the distinction between supervision and financial regulation.
This distinction is often a fine one, because the same agencies
often do both. So let me clarify my terms. Henceforth, I will
use “supervision” and “regulation” as if they were distinct
categories. A “supervisor” is a person or agency who performs
supervision. “Regulator” is an ambiguous term, at least in
natural language. I shall respect nature, and use it to refer to
the persons or agencies that supervise and/or regulate, with
the function left to context.
Financial regulation applies conventional regulatory
techniques to financial firms. These techniques are part of the
rule-of-law enterprise. A regulator promulgates and enforces
rules, and often issues licenses. Rule promulgation is a mixture
of policy, prudence, and authority: a mini-legislative function.
The product is a rule that has force of law. Licensing can be
a more discretionary activity. Nevertheless, it involves some
kind of process defined by legal rules and generally subject to
judicial review.
Supervision, in contrast, fills a gap somewhere between
business and law. Supervisors are invested with legal
powers: the power to collect information and typically
some enforcement power.26 But law enforcement is only
secondary to what they do. Instead, they try to ensure that
a financial firm is well run: with good operations, good risk
management, good compliance, good management, and a
good business plan. This task resembles many other roles,
but not the same as any of them. One could view supervisors
as agents of the financial firm’s creditors, with a bias toward
safe over profitable operation. (Remember that the creditors
of financial firms are typically disperse and weak.) Or maybe
supervisors resemble agents of top management, ensuring
compliance with internal policies. Or perhaps a kind of
auditor: an independent line of reporting to the board of
directors, assessing the effectiveness of top management. Or
perhaps supervisors are the bankers’ answer to the agricultural
extension service: ensuring that the best practices of good
banks spread to the others.
We have two take-home lessons here. The first is that
supervisors know quite a bit about banks, which suggests
a role in insolvency law. The second is that supervisors
traditionally felt protective of the creditors of the firms they
supervised. Nowadays, they also seek to protect the financial
system as a whole.
How Do Supervisors Behave in the
Zone of Insolvency?
As a general rule of corporate finance, the creditors of a
firm become increasingly powerful as a firm approaches
insolvency.27 However, financial firms are different, because
they do not have powerful creditors. The creditors of banks
and insurers are widely dispersed, and not in the business
of lending money. (This is inherent in the definition of
“financial liability.”) They have no covenants that can enable
a governance role. Derivatives creditors are less widely
dispersed and are professionals. However, derivatives credit
exposures are typically collateralized and enjoy favorable
insolvency treatment. Derivatives creditors therefore have
little incentive to govern.28
The firm’s supervisor typically steps into the gap and
is the contingent control party. It becomes more active as
the firm’s prospects decline: something formalized in the
“Prompt Corrective Action” system, linked to the bank’s
capital position.29 In the United States, the supervisor is
not the creditor of the bank, unless it is the FDIC, which
is subrogated to depositors’ claims. But the supervisor, in
whatever jurisdiction, plays the traditional strong creditor
role: exerting increasing control over weaker banks. If not a
strong creditor itself, the supervisor might be the strong agent
of the weak creditors.
Supervisors’ active role with weak financial firms extends
to the insolvency of financial firms. Almost invariably, the
supervisor has the right to initiate the insolvency procedure.
27
Oliver Hart, Firms, Contracts, and Financial Structure (1995).
28
26
For the legal status of the examination function, see Cuomo v. Clearing
House Ass’n, 557 U.S. 519 (2009).
Cf. Mark J. Roe, The Derivatives Players’ Payment Priorities as Financial
Crisis Accelerator, 63 Stan. L. Rev. 539 (2011).
29
12 U.S.C. §§ 1831o, 5366.
FRBNY Economic Policy Review / December 2014
215
This pattern is an international one, at least for banks.30 The
right of initiation tends to be exclusive in the United States,
but is superimposed on ordinary process elsewhere.
In the United States, the role of supervisors goes even
further, to the administration of insolvency. They are logical
choices for this role. As supervisors, they know about the
firm and know the firm’s business. However, supervisory
administration is not universal. In U.S. banking law, the FDIC,
uninsured bank regulators, and state insurance supervisors
all administer insolvencies. Securities firm insolvencies are
typically administered by the Securities Investor Protection
Corporation: an agency with no supervisory role. Overseas,
supervisory administration is even less common.
This strong creditor role creates problems for international
supervision. Supervisors are national actors, with national
incentives. As Ernest Patrikis once put it:31
When faced with the prospect of bankruptcy at
a multinational bank, it is the solemn duty of each
bank supervisor to do all that can possibly be done
to ensure that the adverse financial effects fall on no
customer or counterparty of the bank. But failing
that, they should fall in another jurisdiction.
Because supervisors are national, they take a partial view:
particularly host-country supervisors. They want to protect
customers of the host entity and are not inclined to sacrifice
their wards to global interests. This affects their supervision
of troubled firms and insolvency law. Effective supervisors
prefer to supervise and liquidate their local branches as if
these branches were separate juridical entities. The branches
of effective supervisors will pay their creditors (defined by
territory, not nationality) in full. This preference may be
inefficient (at least ex post), and even unfair, according to the
equal treatment norm. But strong local supervisors are more
interested in the welfare of the creditors of their firms than
they are interested in a globally efficient proceeding.
This policy is codified into U.S. banking law, which treats
the insolvent local branches of foreign banks as if they were
separate juridical entities.32 International insolvency standards
acknowledge this approach. The UNCITRAL Model Law—
which favors integrated proceedings—exempts financial
entities.33 Even where not codified in law, it is often part of
practice. The United Kingdom is ordinarily a devout believer
in cooperative insolvency. This belief did not apply to the
insolvency of the Icelandic banks (which had U.K. branches)
or the administration of the London brokerage of Lehman
Brothers. The E.U. Recovery and Resolution Directive aspires
to collective action within the European Union. However,
it retains a local resolution option for local supervisors who
dislike collective action.34
This policy interferes with megabank insolvency. As stated
above, megabanks—despite their complex structure and
international operations—are tightly integrated firms. But
when megabanks get weak, supervisors work for the creditors
of the entities they supervise. Aggressive local supervisors
will move assets to their jurisdictions or liabilities from their
jurisdiction, thus making home country resolution more
difficult.35 Home and host-country supervisors become less
cooperative, precisely when cooperation is most needed. If
it comes down to insolvency, they may become territorial.
These foreign jurisdictions, if successful in grabbing enough
assets, have no incentive to do anything but liquidate,
destroying value.
2.4 Summary
Things look grim for megabank insolvency. As complex
international conglomerate firms, megabanks seem doomed
to piecemeal competitive procedures: the worst kind. Their
key personnel are likely to run, along with their liquid
liabilities. Both assets and financial liabilities are likely to be
impaired by any insolvency process—assuming that they can
be sorted out across affiliate lines. Small asset impairments in
these highly leveraged organizations translate to large equity
impairments. All this without even bothering to invoke the
boogeymen of interconnectedness and systemic risk!
Fortunately, the cavalry is coming.
33
30
Hüpkes, supra note 4, at 80. Hüpkes’ study was limited to the United States,
Canada, and E.U. jurisdictions, but I believe that the conclusion applies
almost anywhere.
31
Group of Thirty, International Insolvencies in the Financial
Sector 84 (1998).
32
See 12 U.S.C. § 3106(j); N.Y. Banking L. § 606. The federal and other state
statutes are modeled after the New York statute.
216
Why Bail-In? And How!
36 I.L.M., supra note 24, at 1389 art. 1(2).
34
See Article 83(6), 83a(4) (local E.U. authorities may go their own way if
they articulate a reasoned dissent to their fellow E.U. authorities.) This right
is even more explicit with non-E.U. insolvency authorities. See Articles 86-87.
This document is still in preparation. A recent draft can be found at
http://register.consilium.europa.eu/pdf/en/13/st11/st11148-re01.en13.pdf
(last visited Aug. 26, 2013.)
35
Baxter et al., supra note 4, at 77.
3. Bail-In
It is time to answer the two questions I posed at the
beginning. Why do bank regulators like bail-in? And why do
they dislike other modes of resolution? We start with bail-in:
the topic of this section.
To analyze bail-in, we must describe it. But which bail-in
to describe? Bail-in is a work in progress, not tested law like
the Code. The FDIC is working on its “single point of entry”
concept. But the FDIC has not released all the details yet, so I
will take the do-it-yourself approach. The first subsection is an
idealized description of one possible bail-in scheme, inspired
by the theory of Section 2 and some imagination. The
second subsection puts this description in a broader context
and engenders a bit more theory. Finally, I will look at the
FDIC’s current plans.
3.1 The Mechanics of Bail-In
Bail-in is a stripped-down form of reorganization36 working
at warp speed. As we have seen, financial liabilities lose
additional value if reorganized to other debt. But other
liabilities only lose net present value: let us call them “bonded
debt.” Bail-in subordinates bonded debt and reorganizes
only it into equity—mostly overnight. If there is enough
bonded debt, the financial liabilities are untouched: ordinal
and temporal priority. The bail-in process should create a
well-capitalized firm the next morning, before the financial
liabilities have had a chance to run. The hope is that this
process works as smoothly as a recapitalization with
government money—with no government money at risk.
The bonded debt bails out the financial liabilities: hence the
sobriquet “bail-in.”
I have just been a bit too glib. Debt subordination is an
old trick in financial insolvency law,37 but it is not enough.
“Subordination” is an entity concept. Megabanks are
conglomerates, not unitary entities. How, then, to instantly
reorganize the nonfinancial debt of conglomerates, without
touching the financial debt? Such debt might exist in many
entities, in many jurisdictions. The creditors might be
affiliates. Overnight reorganization of such debt requires
a tremendous amount of information and jurisdictional
coordination. This task is an impossible one.
Fortunately, this is not the task of bail-in. By a stroke of
luck, the Bank Holding Company Act encourages the parent
entity of a financial firm to be pretty much a pure holding
company.38 This means that the parent entity does not rely on
financial liabilities. Furthermore, the parent is the cheapest
source of funding in the organization. Therefore, the parent
can downstream this cheap debt to the subsidiaries.39 This
means that the third-party liabilities of the subsidiaries are
mostly financial. Because of limited liability, the debt of
the parent is “structurally subordinated” to the debt of its
subsidiaries. The creditors of a solvent subsidiary are paid
in full, even if the parent is insolvent. And finally, a parent
reorganization involves only one entity. Only one jurisdiction
is responsible for the parent’s insolvency: one set of rules, one
set of acts, and one set of incentives. International insolvency
law works better for single entities than conglomerates.
This all looks a bit too pat. The problem of megabank
insolvency is too hard, and bail-in seems too easy. Let
us slow down and look at the details of bail-in, starting
with its sequence.
The Sequence
Bail-in begins before it begins. The regulator must prepare
for the bail-in well in advance. There are two pre-initiation
processes: fast and slow. The slow process is one of discourse
with clearinghouses and foreign regulators. The goal is not
agreement, but the formation of reciprocal expectations.
The clearinghouses and foreign regulators expect the parent
regulator to rescue all relevant subsidiaries, at the expense
of the parent. The parent’s regulator expects cooperation in
return. These expectations are not mutual obligations. There
are no obligations until the parent regulator decides to rescue
the subsidiaries. Only then do the reciprocal expectations
crystallize into reciprocal obligations.
38
36
37
See supra note 12.
Insurance policies are priority debts in insurance law. Bank insolvency
law has inconsistent priority rules. Compare Jennings v. U.S. Fidel. &
Guar. Co., 294 U.S. 216 (1935) (priority debts abhorrent to the National Bank
Act); 12 U.S.C § 5390(b) (priorities in Dodd-Frank Act which do not
privilege financial liabilities) with 12 U.S.C. § 1821(d)(11) (depositor
priority); UCC § 4-216 (priority for checks in collection); Merrill Lynch
Mortgage Capital, Inc. v. Federal Deposit Insurance Corp., 293 F.Supp.2d 98
(D.D.C. 2003) (priority for special deposits).
12 U.S.C. § 1841 et seq. The holding company can do anything that its
nonbank subsidiaries can do, so financial liabilities are possible. However, there
is no need to keep these liabilities in the parent. They are not very common in
practice and can be moved out of the parent without much cost. This trick—
limiting parents to nonfinancial liabilities—does not work for all jurisdictions,
some of which have the bank as a top-tier parent. Such jurisdictions must rely
on explicit subordination, whether by priority or contract.
39
The bank subsidiary is an exception to this, since it, too, is a cheap source of
funding. However, Sections 23A and 23B of the Federal Reserve Act restrict
the bank’s ability to fund its affiliates. 12 U.S.C. §§ 371c, 371c-1.
FRBNY Economic Policy Review / December 2014
217
The fast process is fast indeed: days or even hours, if
necessary. At this point, we shall call the official actor the
“receiver,” likely the regulator in another guise. The receiver
must assess the situation as best it can and make two key
decisions: whether to support the subsidiaries, and the
amount of the debt haircut at the parent level. The receiver
will probably want to recapitalize all subsidiaries. (There are
complexities to this, discussed below.)
Initiation has four immediate consequences. First, the
receiver’s second decision becomes action. The receiver
selects parent debt. In doing so, it climbs up the liability
stack as far as it needs: certainly equity and preferred stock,
then subordinated debt on up to senior debt, if needed. Each
class but one is either untouched or fully selected. One class
may be partially selected: the one sandwiched between the
untouched and fully converted classes. Any unselected debt
is paid according to contract. The selected debt becomes
new equity, to be distributed later. The result is a parent with
much less debt, and substantial equity. The proper debt-equity
conversion is the most difficult decision that the regulator
must make. Too small a conversion, and the reorganization
will not be credible. An overlarge conversion unnecessarily
disrupts creditor expectations. (The risks are asymmetric.)
The second consequence of initiation is that the receiver
can (but need not) exercise classical receiver powers for some
time. It may replace management (if necessary), do some early
transactions, and possibly alter the governance of the firm.
The active part of the receivership could be over as soon as
reliable private governance is in place: a few weeks. Or it could
persist longer.
The third consequence of initiation is an automatic antiipso facto provision that invalidates cross-default clauses keyed
to the parent. Since the subsidiaries will likely be solvent
in bail-in (see below), any invocation of these cross-default
clauses would be opportunistic behavior of counterparties.
Enforcement of this stay requires some measure of crossborder cooperation: either through harmonized insolvency
law, or changes in industry-standard documentation, or
changes in regulation.
The fourth consequence of initiation is the recapitalization
of the subsidiaries, probably by debt relief. The parent can
afford to relieve its subsidiaries’ debt, because it has very little
debt service itself. Presto! All affiliates that relied on parental
funding are now reasonably capitalized. There will probably
not be much private liquidity in times of stress, even though
the bail-in creates an extremely well capitalized megabank.
However, government liquidity to the parent will serve
temporarily and can also recapitalize any subsidiaries that did
not rely on parental credit. There is not much risk in lending
to the now-well-capitalized parent.
218
Why Bail-In? And How!
The fast work is all done. Only one operation on the
balance sheet remains, but this takes some time. Who gets the
new equity? It is operationally easy to relegate debt to equity—
just name a number. This number can be arbitrary, if the
process respects debt priority. True, the relegated debt holders
lose their old debt. But they are compensated in new equity.
As long as debt priority is respected, the aggregate value of
the new equity should have precisely the same value as the old
debt aggregate. This is true notwithstanding the amount of
relegated debt.
The problem, of course, is that only the aggregate value is
preserved. There will likely be several classes of claimant, each
insisting that it is entitled to plenty of new equity. These claims
are harder to resolve. Any resolution would require some rules
and a few months of time. They may require the imprimatur
of an Article III court, or at least plenary review by an
Article III court.40 The entire process of equity distribution
would have to do the following:
Handle claims. Most of this is fast and mechanical, as
the claims will typically be those of bondholders. Therefore,
the process is simply that of identifying bondholders and
their assignees.
Compute new equity to the claimants. This is not
mechanical. Even the simplest form is complex: valuation
of the firm, and equity distribution following the priority
ladder of claims. The Code follows another path: a negotiation
process that culminates in a plan. Such a process gives more
voice to claimants and thus may be more legitimate than a
judicial valuation. However, it encourages strategic claimant
behavior, runs the risk of delay, and requires judicial review.41
Note that a bail-in negotiation process would be more limited
than the one in the Code. The Code negotiation process
chooses which debt to impair, as well as the conversion of
impaired debt to equity. In contrast, bail-in impairs parent
debt at the very beginning of the process, to create confidence
at the subsidiary level.
Distribute new equity to the claimants. This is
relatively mechanical, but time-consuming. The distribution
itself is fast enough, but the antecedent securities law
disclosures take time.
This process could be compressed into a few months,
with appropriate procedures. Time is significant, but not
of the essence. This process has a limited role: who gets
40
41
United States v. Raddatz, 447 U.S. 667 (1980).
There is a third way: giving junior classes an option to buy out the
senior classes at face value. See, e.g., Lucian Bebchuk, A New Approach to
Corporate Reorganizations, 101 Harv. L. Rev. 775 (1988); Philippe Aghion,
Oliver Hart & John Moore, The Economics of Bankruptcy Reform,
8 J.L. Econ. & Org. 523 (1992). However, this approach presumes a working
capital market, which is unlikely during a financial crisis.
Table 1
Timeline of the Integrated Bail-In Process
Prior Steps
Bail-In
Years
Talk to:
Foreign regulators,
clearinghouses
Days
Overnight
Days to Weeks
Value firm
Debt to equity
Liquidity support
Deal with:
Foreign regulators,
clearinghouses
Subsidiary debt swap
New governance
Restructuring
Months
Restructure business
Access to public
liquidity
Receiver takes over
New securities
registration
Ipso facto relief
New securities
distribution
Foreign regulator
approval
Receivership ends
Long Term
Restructure business
Add parent debt
or shrink bank
Pay dividends
Source: Tabular summary of text.
how many shares in the reborn enterprise. This role is
especially limited because the market for corporate control
of financial firms is a tightly regulated one that favors widely
distributed shareholdings.
This is the only slow part of bail-in. The rest is fast.
Financial liabilities are unaffected, and the firm’s
operations are unscathed.
The distribution of equity is the end of the legal process,
but not the end. The reorganized bank will have a strong
balance sheet, but may not have a strong business. The
megabank probably became troubled in the first place
because its operations were insufficiently profitable, or
perhaps too risky. The megabank will have to shed its
bad operations. This is not a primary task for the bail-in,
which apart from installing new management and maybe
governance, works mostly on the balance sheet. Rather, it is a
task for the restructuring stage, although there will be some
overlap with the earlier process. A timeline of the integrated
bail-in process is shown in Table 1.
Bail-in requires two things to succeed in full. First,
there must be enough debt at the parent to credibly fill the
consolidated capital shortfall, and the receiver must be
willing to haircut it accordingly. This requires regulation,
as discussed below. Second, bail-in must inspire confidence.
For this, adequate capital is necessary, but not sufficient.
A sufficient liquidity backstop is also necessary, as is the
cooperation of foreign regulators. But even these are not
sufficient. An adequately capitalized firm might still not
engender enough confidence to survive as a going concern.
However, even such a failure would be a success. The firm
will still survive as an orderly liquidating organization, if not
as a business unit. The liquidity backstop assures that it will
not need to dump its assets on the market. Such a failure will
internalize credit risk on the bondholders, and not destroy
asset or liability values.
The Guarantee Problem
Guarantees pose a technical problem for bail-in. Guarantees,
for our purposes, include anything that pierces the
corporate veil of affiliates: straight guarantees, collateral,
or keep-well agreements, for instance. The guarantee can
run from parent to subsidiary, or cross-stream, or even
upstream. This definition does not include guarantees of
unaffiliated organizations.
Guarantees preclude the receiver’s option to abandon
a subsidiary. A guaranteed subsidiary is welded to its
guarantor. This sounds like little loss; bail-in will usually
recapitalize all the subsidiaries. However, this is the ex post
fallacy: ignoring incentives. The option of abandoning a
subsidiary is credit risk for its creditors and regulators. This
risk is an incentive to monitor subsidiaries.42 An ideal bail-in
would be time-inconsistent: ex ante putting the subsidiaries
at risk and ex post bailing them out. Time-inconsistent
42
Cf. Baxter et al., supra note 4 (context of branching: the rationale for
territorial branch liquidations); Roe, supra note 28 (derivatives).
FRBNY Economic Policy Review / December 2014
219
policies do not work in a frictionless world. But frictions exist
aplenty: notably uncertainty about regulatory action and the
credit risk aversion of financial product counterparties.
If the receiver does not have the ex post option to abandon
a subsidiary, the parent regulator has no ex ante bargaining
position with foreign regulators. In other words, foreign
regulators will have less incentive to cooperate with parent
regulators, because they know that their local subsidiary
will leave no creditor behind. This is not an insolvency
problem. Instead, it is a regulatory problem, encouraging
local regulators—who have local knowledge and power—
to free ride off the parent regulator, who does not.
This analysis of guarantees is incomplete, and at most
establishes a prima facie case. But it is enough to serve my
purpose. Guarantees are a significant issue in bail-in, and
one without an easy solution.
If regulation of guarantees is useful, the bail-in process
itself can regulate, by subordinating parental guarantees.
A subordinated guarantee remains fully effective against a
healthy parent. It therefore assures subsidiary creditors that
the healthy parent will not walk away from the subsidiary. But
subordinated guarantees do not protect the subsidiaries of an
insolvent parent. Therefore, subordinated guarantees preserve
the receiver’s freedom of action in the event of insolvency.
The Regulations
Bail-in assumes a bank supervisory process, e.g., monitoring
a weakening business and restructuring the bailed-in firm.
Bail-in also requires some adjunct regulation. Fortunately, this
regulation is neither extensive in scope nor difficult to draft:
Mandatory debt. Bail-in requires an adequate level of
parental debt: enough to recapitalize the largest foreseeable
shortfall. Market forces may not provide enough of such
debt, since firms may prefer to issue liabilities through the
subsidiaries, as profitable financial products. This argues
for minimum mandatory debt at the parental level.43
A mandatory debt regulation is easy to draft and comply
with. The amount of debt could key off Basel risk-weighted
methodology or the value of the financial liabilities, held by
third parties with the subsidiaries.
Cross-affiliate guarantee. The insolvency process
can subordinate parental guarantees, but it cannot affect
cross-affiliate guarantees, because bail-in will put few, if
any, subsidiaries into insolvency. A holding company might
be tempted to use these guarantees to deny the receiver
the ability to abandon a particular subsidiary.44 Similar to
guarantees are other close relationships, such as service
agreements, cross-stream debt, common names, and the like.
This problem is not a fatal one, but it is not easy to fix in
insolvency law. It suggests a regulatory approach.
Claims trading. Valuation is one of the slower parts of
bail-in. During its pendency, the ultimate value of the claims
will be uncertain. This valuation uncertainty is likely to create
an active market in claims, along with the invariable portfolio
repositioning of debt-holders who may not (want to) hold
equity. There is nothing wrong with this; it is part of every
modern Code reorganization.
However, this trading is likely to concentrate the
claims, which will concentrate the ultimate equity
holdings. U.S. bank regulation is chary of concentrated
equity holdings. A concentrated equity holder might itself
become a bank holding company, which is illegal without a
license.45 There will probably be some need to reconcile the
claims trading process with the ownership limitations of the
Bank Holding Company Act.
Parent liabilities. Bail-in works best when the parent has
no financial liabilities. This might imply some reinterpretation
of the Bank Holding Company Act, to prohibit the few
financial liabilities that a modern holding company parent
might have. It might go a bit further. Some nonfinancial
liabilities are typically subject to a CVO (e.g., trade credit)
or a bankruptcy priority, such as employee compensation.
From the perspective of bail-in, the best holding company is
a pure shell, without any operations or even a building lease or
telephone bill.
3.2 The Meaning of Bail-In
Now that we have discussed the mechanics of bail-in, it is
time to put this technique into context, with three brief essays.
I shall first discuss why this technique works. I then discuss
the implications of bail-in for the notion of bank capital.
I conclude with a few words on the limits of this technique.
43
Such mandatory parent debt is current regulatory policy, although not
yet implemented. http://federalreserve.gov/newsevents/speech/
tarullo20131018a.htm (last visited November 12, 2013.) For a more
sophisticated argument, see James McAndrews, Donald P. Morgan,
João Santos & Tanju Yorulmazer, What Makes Large Bank Failures So Messy
and What to Do about It?, 20 FRBNY Econ. Pol. Rev., 229 (2014).
220
Why Bail-In? And How!
44
See supra text accompanying note 42.
45
12 U.S.C. §§ 1844(a), 1847.
Table 2
Balance Sheet Data of Selected Large Banks in 2006
Consolidated Liabilities (L)
(Trillions of dollars)
Equity (E)
(Trillions of dollars)
Long Liabilities (LL)
(Trillions of dollars)
E/L
(Percent)
LL/E
(Percent)
LL/L
(Percent)
(LL+E)/L
(Percent)
JPMorgan Chase
1.24
116
145
9.4
1.25
11.7
21.0
Lehman Brothers
0.53
18
82
3.4
4.55
15.5
18.8
Citibank
1.88
119
290
6.3
2.44
10.1
16.4
Goldman Sachs
0.80
34
126
4.2
3.71
15.8
20.0
Source: Securities and Exchange Commission Form 10-K consolidated balance sheets filed in 2007.
Notes: The long-term debt is consolidated, and thus may count long-term third-party debt at the subsidiaries. However, most of this subsidiary debt (if it exists)
can be cheaply moved to the parent, so it is useful for bail-in. I would dearly love to argue that Table 2 proves that the 2006 levels of long-term debt were sufficient
to avert the disaster of 2008, if only bail-in had been around. I am not certain that this is true.
Why Bail-In Works
Parent-level bail-in is quick and simple, compared with the
alternatives. Since everything happens at the parent level, the
complexity of the conglomerate matters little, if subsidiaries
are safe and everybody cooperates. Parent-level bail-in is
strongest at the crisis stage—the beginning. Compared with
the alternatives, it economizes on information, planning, and
implementation when time is short and the stakes are high.
The early stages of bail-in are operationally tractable, even
with the time constraints. The debt haircut may be a difficult
judgment call, but is operationally easy. The subsidiary debt
forgiveness and liquidity provision are conceptually simple,
and operationally straightforward. With some luck, they can
restore confidence in the firm, preserving its going-concern
value. At worst, bail-in creates an orderly liquidation.
International cooperation is the most complex of the
early steps. But fortunately, the scope of the cooperation is
limited. The foreign regulator must keep its subsidiaries out
of local insolvency proceedings, perhaps provide liquidity,
and discourage declarations of default. Clearinghouses must
not close their members out. Fortunately, bail-in aligns the
cross-border incentives, at least if all the subsidiaries are safe.
For the foreign regulators, bail-in shifts all the pain to the
home country, at least if we assume that foreign regulators
care no more about their bondholders than they do about
domestic bondholders in nonfinancial firms. The home
country also wants bail-in, because it is likeliest to preserve
the financial firm.
The process creates few perverse incentives, because the
parent creditors cannot expect the public to assume their
credit risk. It does not concentrate the industry further. It
may encourage a shift of liabilities to the subsidiary. But
the fix for this is easy: mandatory debt at the parent level. It
may encourage inappropriate downstream and cross-stream
guarantees, but there are fixes to this, too.46 There will be some
tail risk. But this is not a significant problem. There is plenty
of bail-in ammunition in most large banks’ balance sheets,
and the banks can afford it. As Table 2 shows, there is nothing
unnatural about the kind of balance sheet that supports bailin. Most large banks in 2006 had substantially more long-term
liabilities than equity capital. And these liabilities understate
the bail-in-able debt, because the parent also had substantial
short-term liabilities.
The New Meaning of Capital
Our core insight is that only financial firms have financial
liabilities. Bail-in succeeds because it subordinates and
separates the nonfinancial liabilities from the financial
liabilities. This transforms our understanding of bank capital.
Capital regulation presupposes that junior liabilities should
protect senior liabilities. This makes no sense in ordinary
corporate finance theory, because nobody needs protection.
Every voluntary investor assumes its risk, compensated
by the pricing and contractual terms it bargained for.
46
See supra text accompanying notes 42-44.
FRBNY Economic Policy Review / December 2014
221
Why protect it from its bargain? (We ignore nonadjusting
creditors and strategic behavior.) But this article does not
use ordinary corporate finance theory. This article extends
corporate finance theory to include financial products as well
as ordinary debt, held for investment. Holders of financial
products lose more in insolvency than the net present value of
the difference between their claim and their share. An efficient
contract gives them priority regardless of their bargain,
averting deadweight loss.
Bail-in transforms the meaning of capital. In bail-in,
parental debt does exactly the same thing as equity: it protects
financial liabilities from a degradation of value. If this is
the function of capital, we may conclude that with bail-in,
all nonfinancial liabilities are capital! It also means that in a
bail-in regime, megabanks currently hold much more capital
than we thought they did.47 But with poor insolvency law,
there is no access to it.
Not all capital is created equal. But it is hard to say
which forms of capital are better. Debt might provide better
protection than equity. It is easier to measure. It disciplines
management,48 especially if continuously issued. From a
supervisor’s perspective, it provides superior information
to equity. The price of debt reflects only downside risk: the
supervisor’s main concern. Finally, a debt-heavy structure
ensures plenty of bail-in ammunition.49
Proponents of equity structures have their argument,
too. Inadequate equity encourages excessive risk-taking.50
Also, low-equity structures enter insolvency more
often than high-equity structures. Insolvency is costly.
The cost of insolvency argues for more equity—a lower
probability of default.
The term structure of parent debt also makes a difference.
Short tenors are more sensitive than long tenors, because
the primary market constantly assesses them. Alternatively,
long-term debt protects a firm from transient market
sentiment. Banks arguably need such protection more than
commercial firms, because they do not have commercial
47
See Table 2: compare E/L column to (LL+E)/L column. This is directly
contrary to the Admati-Hellwig hypothesis: that the low levels of Basel Tier I
capital imply that banks are severely undercapitalized. Anat Admati &
Martin Hellwig, The Bankers’ New Clothes: What’s Wrong with
Banking and What to Do about It (2013). If bail-in works as I expect it
to, their apparently radical recommendation of 20-30 percent equity is pretty
close to a plea for the status quo. The Admati-Hellwig thesis tacitly assumes
that these debt liabilities are irrelevant: i.e., bail-in does not work.
paper backstops. Then again, banks have a fair amount of
liquid assets; hence less need for something like a commercial
paper backstop.
I do not seek to optimize parental debt and equity. It is
enough to say that they both serve as capital in a workable
bail-in regime.
Bail-In and Systemic Risk
I cannot stress the point enough: the case for bail-in does
not need the systemic risk boogeyman. The boogeyman is
real and scary enough, but also a rare beast. Bail-in works
well for isolated megabank insolvencies, which are far more
common.51 If there is enough debt in the parent, the worst
result is pretty good: an orderly liquidation that does not
impair financial liabilities, dissipate asset values, or put public
funds at risk. And bail-in has a good chance of preserving the
firm as a going concern.
Bail-in should also mitigate systemic risk. I have been
agnostic on the causes and mechanisms of systemic risk,52
but liquidity and leverage have a lot to do with it. Bail-in
eliminates the leverage problem: the bonded debt of the
parent protects the subsidiaries’ creditors. Liquidity support
is credible. If the government can print money and does not
assume substantial credit risk (bonded debt again), public
liquidity has no real cost, even before public benefits are
considered. Furthermore, bail-in can work as quickly as
systemic risk can materialize. Since the early stages of bail-in
are administratively simple, it also scales well. It can work on
many firms at the same time, if necessary.
Bail-in will probably create its own stresses. A bailed-in
firm will likely mark many of its assets down. These asset
markdowns might force other firms to do the same, adding
to the systemic risk of multiple bail-ins. However, I believe
that this particular risk may be a chimera. Bail-in is scalable.
It is a reorganization, needing no outside resources, apart
from liquidity and regulatory attention. These resources are
not scarce, at least in the United States, as we have seen in
2008. A contrarian could even argue that multiple bail-ins are
less stressful than single ones. Bail-in may be stigmatizing,
but multiple bail-in stigmatizes an industry, not a firm. This
may decrease the risk of soft failure. Counterparties can
avoid a stigmatized firm, but have a harder time avoiding a
stigmatized industry.
48
Michael C. Jensen, Agency Cost of Free Cash Flow, Corporate Finance, and
Takeovers, 76 Am. Econ. Rev. 323 (1986).
49
Some of these arguments are made more analytically in McAndrews et al.,
supra note 43.
51
For a short list, see supra Section 2.2.
50
52
See supra text accompanying note 23.
See supra note 7.
222
Why Bail-In? And How!
Limits on Bail-In
Bail-in has at least five limits: maybe a sixth if you are
worried about multiple bail-ins. Here we discuss the five.
First, bail-ins can only marshal limited resources: the
nonfinancial liabilities of the parent. This would have been
plenty for the crisis of 2008. But it is not enough for any
imaginable crisis. Since megabanks are in the business of
financial liabilities, they can afford to issue only so many
nonfinancial liabilities. For a sufficiently large shock, systemic
risk will remain. Nor will breaking up megabanks eliminate
the systemic risk problem, even with bail-in. The failure of
a small bank may not endanger the system. But systemic
dangers—such as asset collapses—will systemically endanger
even small banks.
This is reminiscent of the catastrophic risk problem of
the insurance industry. Capital markets, no matter how
ingeniously organized, can only handle so much risk.
Leviathan must always lurk at the far end of the risk tail. All
we can do is stretch the tail a little longer, further away from
our workaday world.
Bail-in must also muster another resource: governmental
liquidity. With enough parental liabilities to bear the risk,
governmental liquidity is a free good, at least in principlebut maybe not always in practice. Liquidity is only free if
the government debt market is deep enough. This is almost
certainly true in jurisdictions like the United States, where
financial panics increase government liquidity, as investors
rush to public debt. But, as Iceland has shown, it is possible for
a jurisdiction to be smaller than its banks. Bail-in might have
operational problems in such a jurisdiction.
Third, as discussed above, bail-in has a soft failure mode.
Counterparties may not have enough confidence in the firm
to stick with it, even if they know they will be repaid in full.
If so, bail-in ceases to be a reorganization, and becomes a
kind of controlled wind-down. Such a failure is a successful
one: this bug is really a feature. Financial creditors get paid in
full, at the expense of nonfinancial creditors. This both averts
systemic risk and imposes market discipline on nonfinancial
creditors. But it does destroy the business.
The fourth limit does not exist in principle, but may be a
significant problem in practice. Megabanks are international
firms. Bail-in requires a fair degree of ex ante legal
harmonization and ex post cooperation. This is no problem in
principle: both the ex ante and ex post incentives are strong,
as argued above. But legal harmonization derogates from
sovereignty. The history of insolvency treaties has not been
a good one. Ex post cooperation has had some success, but
cooperation is hardest in times of panic. As a political matter,
can a home-country receiver promise to make good on a
massive hole in a foreign subsidiary? As an economic matter,
can it afford not to? Can a host-country official (be seen to)
rely on the kindness of strangers?
There are, however, some grounds for optimism. Since
most of the action takes place at the parent, the necessary
harmonization is narrow in scope. I can only think of two
major issues (there may be more.) Creditors of the parent may
seek to enforce their claims against parental assets overseas—
the stock and upstream debt of the subsidiaries. And we have
already mentioned, in Section 3.1, that bail-in requires that
jurisdictions not enforce ipso facto cross-default clauses. The
first problem was solved—or at least addressed—over a decade
ago by recognition of main and ancillary proceedings.53
If a bail-in follows the established rules of the road, the
parent creditors will have no recourse outside the main
proceeding. The ipso facto problem might also be tractable.
Legal harmonization might require super-sovereignty, but
banking law contains super-sovereign forces. The Basel
process, for instance, encouraged enforcement of ipso facto
clauses in derivatives contracts. The ISDA model agreement
could remove or modify these clauses, and ISDA seems
to have done so.
There is one other limit to the bail-in concept. It is limited
to financial firms. Bail-in cannot replace the Code. Bail-in
buys speed at the cost of flexibility. This speed is needed for
the financial liabilities that define financial firms, but other
liabilities can survive the automatic stay. Bail-in presupposes
a certain corporate structure. It also presupposes prudentially
regulated firms, and requires a capital regulatory scheme. It is
a specialist: good for the peculiar world of financial firms, but
not exportable elsewhere.
3.3 The FDIC and Bail-In
Title II of the Dodd-Frank Act empowers the FDIC to resolve
financial conglomerates.54 The FDIC formally adopted
the single point of entry (SPOE) approach to implement
Title II in a December 2013 release that is currently out for
comment.55 This release does not contain all the details, such
as: the details of the valuation and equity distribution, or the
criteria for recapitalizing subsidiaries. But the outline is good
enough. SPOE is a form of bail-in at the parent. Instead of
53
See supra text accompanying note 24. But see supra notes 32-34.
54
12 U.S.C. § 5381 et seq.
55
“Resolution of Systemically Important Financial Institutions: The Single
Point of Entry Strategy,” 78 Fed. Reg. 76614 (Dec. 18, 2013).
FRBNY Economic Policy Review / December 2014
223
working directly on the parent entity, it uses an intermediate
“bridge company.” 56 The FDIC will transfer all or most of the
assets of the holding company parent to a bridge company,
retaining many or all of the parent liabilities in the estate.
It will then issue the stock of the bridge company to estate
claimants in satisfaction of their claims. This liquidating
distribution in kind is almost identical to a classical
reorganization, although it entails a de novo entity.
Their approach should work, if there is enough debt in the
parent. (This task is the Federal Reserve’s.) Bail-in requires
liquidity support, but the Dodd-Frank Act provides it,
through the FDIC and the Treasury.
SPOE relies on Title II, and Title II is drafted as an
insolvency process of last resort. The entity must be on the
eve of insolvency (defined broadly), the insolvency must have
systemic consequences, and there must be no good alternative
to Title II resolution. The procedural barriers are high, as well:
a recommendation by the Board of Governors of the Federal
Reserve System and another agency; a determination by the
Secretary of the Treasury (in consultation with the President),
and either approval by a district court or the acquiescence of
the firm’s board of directors.
This hard trigger has its downside. Bail-in works well, even
when systemic risk is not on the table. With the appropriate
parental capital structure, bail-in improves the balance
sheet, preserves going-concern value, does not result in
concentration, and displaces poor management. Bail-in is not
an inherently desperate measure. It should not be reserved for
desperate times.
The hard trigger is not only too narrow; it also harms the
bail-in process. The hard trigger means that there will be no
SPOE practice emerging from experience in low-stakes cases.
This is troublesome. “The life of the law has not been logic;
it has been experience.”57 (Chapter 11 is a good illustration
of this maxim, as is the administration of the FDI Act.58) The
Title II hard trigger does not allow for experience. The FDIC
will have to get it right the first time, with high stakes and no
latitude for error. Fortunately, however, bail-in is a simple and
robust idea. We may never get the experience with SPOE that
we have with Chapter 11. But we will not need as much of it.
There is a second problem, complementary to the first.
People are more confident with well-tested procedures. Bail-in
may not require confidence to provide an orderly liquidation
or avert systemic risk: enough capital and liquidity support
should do the trick. However, it does require confidence to
56
See supra note 12.
57
Oliver Wendell Holmes, Jr., The Common Law 1 (1881).
58
David A. Skeel, Jr. Debt’s Dominion: A History of Bankruptcy Law in
America (2003). A similar history of bank insolvency has not been published.
224
Why Bail-In? And How!
preserve the bailed-in entity as a business concern. Such
confidence relies on practice and custom: “an instinctive
confidence based on use and years.”59 This is the same
confidence by which customers buy airline tickets from
airlines in Chapter 11.
4. Alternatives to Bail-In
This section answers the other question posed at the
beginning: why are regulators unhappy with the alternatives
to bail-in? It examines three alternatives: fast asset sales,
Chapter 11, and private law. It concludes that they are all
worse than bail-in. Some may not work at all.
None of these comparative arguments requires systemic
risk. Bail-in is better for any megabank failure—even
localized failures.
4.1 Fast Asset Sales
The asset sale proposal of Melaschenko and Reynolds60 looks
attractive. It takes place at the parent level. All the assets
and some of the liabilities of the parent go into a temporary
holding company, which operates for a few months, until a
buyer emerges. The proceeds of the sale pay off the creditors,
much like the sale of a business in bankruptcy.
The proposal looks much like bail-in, and should be about
as quick. It has the further virtue of placing a market value on
the firm. It may even work. But even if it works, it will work
worse than bail-in. It assumes too much: a competitive market
for corporate control, and no antitrust problems. Bail-in
suffers from neither problem. Let us review the bidding,
starting with the market for corporate control.
Market for Corporate Control
In a perfect market for corporate control, the sale price would
be the best measure of firm value. Real-world markets for
corporate control are not perfect, but they are far better than
the forced asset sale of a megabank.
59
Walter Bagehot, Lombard Street: A Description of the Money
Market 33 (Richard D. Irwin 1972) (1873).
60
Paul Melaschenko & Noel Reynolds, A Template for Recapitalising Too-Bigto-Fail Banks, BIS Quarterly Rev. 25 (June 2013).
To begin with, the market would be thin. Thanks to the
Bank Holding Company Act, bidders would be few. Only
megabanks—or organizations willing to become megabanks—
can buy megabanks. Not all of them would necessarily be
bidding. In times of financial panic, many banks might prefer
to avoid the action. And the best offer would not necessarily
win. It may come from the weakest bidder.
The market would be thinner because it would likely be
lumpy: of the whole bank. Units could be sold, in principle,
just the same way that any functioning megabank could sell a
business unit. But in practice, things are a bit more difficult,
because the buyer is not certain that the seller would remain
as a going concern. (Sales of business units usually feature
warranties and service contracts.) And the more pieces that
are sold, the more skittish the financial liability holders of
the unsold pieces would be, and the less the unsold pieces
would be worth.
The assets would also be a bit lemony, as well as lumpy.
It takes a lot of due diligence to purchase a very large bank.
This is especially true for a megabank known to have weak
business units—especially in a time of financial panic when
asset prices are unmoored from asset values.
Not only are the assets lemony and lumpy, they are also
volatile. There will be some time between bidding and closing.
This is time that the subsidiary’s creditors could decide that
they do not like the bidder, and disappear, taking the value
of the firm with it. The same is true for others. A temporary
entity with a “for-sale” sign around its neck might have more
problems retaining customers and key employees.
An impaired market for corporate control is still a market
for corporate control. If the regulator is determined to sell the
megabank cheaply, some buyer will probably emerge—maybe
even in a financial panic. The buyer will probably get a very
good deal. This is precisely the problem. The creditors would
do better in bail-in, which does not require a functioning
market for corporate control.
Competition
Despite their problems, we know that fast asset sales or
mergers of megabanks can work if the acquiring megabank
is strong and/or if the acquisition is assisted. We have seen
them work in 2008. JPMorgan Chase Bank acquired Bear
Stearns and Washington Mutual Bank. Wells Fargo Bank
acquired Wachovia Bank. Bank of America acquired Merrill
Lynch. These are successes; they averted target insolvency.
Transaction flow was smooth; financial liabilities were
unimpaired. The Washington Mutual transaction created
some angry creditors of the parent (which was not sold), but
the parent had no financial liabilities.
But they are only partial successes. Each sale replaced a sick
megabank with a bigger megabank. This is poor competition
policy.61 There are not that many megabanks: the industry
is concentrated. Only a megabank can acquire another
megabank, so mergers concentrate the industry further.
Bail-in is clearly superior in this regard. A successful
bail-in has only a marginal effect on competition; an
unsuccessful bail-in only eliminates a competitor, without
creating a bigger one.
4.2 The Bankruptcy Code
Bail-in is a form of reorganization. The Chapter 11 reorganization
is the jewel in the crown of the Code. Why can’t megabanks
just use Chapter 11, on a parent-only basis? A parent-only
Chapter 11 would be similar to bail-in: protecting the
financial liabilities at the expense of the parent’s bonded debt.
This question has a consensus answer: “Chapter 11 will
not work.” This is true even though Chapter 11 is better
than SPOE in some respects. It has a much lower initiation
trigger than Title II,62 and it does not require a separate
bridge entity.63 Despite some early support for unvarnished
Chapter 11,64 most Code proponents now say that Chapter 11
needs some improvements.65 This is a good place to examine
the weaknesses of Chapter 11 in financial insolvency. We start
with a very brief introduction to Chapter 11. We then discuss
the flaws of Chapter 11.
61
This is so at least in the eyes of Congress. See 12 U.S.C. §§ 1852, 5363
(limiting acquisitions of large financial firms.)
62
See supra text accompanying notes 56-59.
63
See supra text accompanying notes 55-56.
64
E.g., Kenneth Ayotte & David A. Skeel, Bankruptcy or Bailouts?,
35 J. Corp. L. 469 (2009); Stephen J. Lubben, Systemic Risk and Chapter 11,
82 Temp. L. Rev. 433 (2009); Skeel, supra note 4; Baird & Morrison, supra
note 4 (more-or-less equating Code to FDI Act process).
65
See Hoover Institution, supra note 4 (“Chapter 14”); Bovenzi, Guynn &
Jackson, supra note 4.
FRBNY Economic Policy Review / December 2014
225
A Quick Tour of Chapter 11
Problems of Chapter 11: A Checklist
I implore readers with any knowledge of the Chapter 11 process
to skip this short section. I wrote it only for a hypothetical
reader who is new to the topic and does not know how
grossly I simplify.
The Chapter 11 process begins with a petition filed in
court, generally by the debtor. No judicial action is needed;
filing alone is effective and creates an “estate” in the entity
that filed. After filing, the incumbent management typically
continues to operate the estate, although a court may select
other management. Management serves as a fiduciary for the
estate, supervised by the courts. As fiduciary for the estate,
it is responsible only for the estate, not third parties such as
employees or financial counterparties of subsidiaries.
Filing creates an immediate stay on all debt-collection
efforts. Nevertheless, debtors typically need liquidity for their
continuing operations, and financial firms definitely need
liquidity. The Code lets a post-petition debtor borrow on a
priority basis: so-called “DIP financing.”66 At this stage, the
Chapter 11 process bifurcates. In one path, the court—at the
behest of management or the creditors—sells the bulk of the
business as an operating concern. This so-called “Section 363”
path is favored these days, because it is much faster (weeks
to months) than the alternative path: a true reorganization.67
The reorganization seeks the same end-state as a bail-in: a new
capital structure. However, its process is completely different.
A Chapter 11 reorganization is a negotiating process. To
oversimplify, creditors form committees of similar claims. The
debtor and committees negotiate among themselves and come
up with a “plan” that reorganizes the liabilities of the firm in a
more sustainable fashion: transforming senior debt into junior
classes (or even equity), and short-term debt into long-term
debt. This often takes a year or more. Dissenters complain to
the court. If the court deems the plan fair, it “confirms” the
plan, over the dissenters’ objections. If not, the negotiation
cycles again. Upon confirmation, the firm is reorganized, with
a more sustainable capital structure.
Note that this negotiation process conflates two processes
distinct in bail-in: the relegation of old debt, and the
distribution of new instruments. In bail-in, the first process
occurs at the beginning. The second process occurs at the end,
with the valuation of the firm. In Chapter 11, both processes
occur synchronously, with the plan confirmation at the end.
66
“DIP” is an acronym for “debtor in possession”: i.e., the incumbent
management, which usually continues operating the firm, subject to judicial
supervision. “DIP financing” is a term of art; it applies to any post-petition
financing, regardless of who is running the firm.
67
See supra note 12 for more on reorganizations.
226
Why Bail-In? And How!
This section is a list of Chapter 11 elements that may
impede a megabank resolution. Again, we assume a
parent-only Chapter 11, much like bail-in, but using
the Code. Chapter 11 reform advocates have noted and
addressed some of these elements, but not all. Some of
these elements are easy to fix, at least conceptually. But
some go to the very structure of Chapter 11.
This list could be a useful checklist for Chapter 11
reform advocates.
Adjudication and compensation. Chapter 11 has
no concept of ex post compensation; it relies on ex ante
adjudication. This must be so—a court has no fund with
which to compensate claimants for its errors. But this also
limits the speed of the process. If compensation is impossible,
any significant decision must be adjudicated, which entails
due process. Bankruptcy courts can be very quick—for courts.
But the faster they act, the less legitimate their process—a
point that emerged from the Chrysler and General Motors
(GM) reorganizations. And the faster they must act, the more
meaningless the appellate review, in which the appellate court
is asked to unscramble an omelet prepared by the bankruptcy
court. In contrast, Dodd-Frank contemplates errors and
provides for their ex post compensation.68 This places far less
strain on due process.
Bank Holding Company Act. Claims trading is common
in modern bankruptcy practice. Active investors seek a stake
in the firm that will give them the best possible position in
negotiations. Typically, this position translates to a controlling
equity stake. However, the Bank Holding Company Act limits
control. Claims trading without regulatory approval, then,
may lead to illegal control. This problem exists in bail-in
as well as Chapter 11.
Capital regulation. Chapter 11 does not regulate the
ex ante capital structure of an enterprise. It reorganizes the
capital structure it is given. As we have seen, this is not
enough: the parent needs enough bonded debt to bail out all
the financial liabilities. A regulatory fix is necessary.
Capital structure (shape). Chapter 11 tends to produce a
thin capital structure ex post: an outcome of the negotiation
process. The capital structure may also have some optionality.
Neither is reassuring to creditors at the subsidiary level who
can run during the pendency of the process. And both may be
the subject of regulatory displeasure.
Capital structure (timing). Chapter 11 produces a capital
structure at the end of the process, not the beginning. Even if
this capital is adequate, it may be too late. Financial creditors
68
12 U.S.C. § 5390(d)(2)(B). For the quantum of compensation, see infra text
accompanying notes 70-72.
need the most assurance at the beginning of the process. They
are more likely to run if they do not immediately see a hefty
capital cushion. There is a bit of time-inconsistency here;
more ordinary amounts of capital may placate these creditors
at the end of the process. But this time-inconsistency does
not connote a logical inconsistency. Asset values are more
questionable at the beginning: imperfect information.
It is worth noting that the successful bankruptcy
reorganization of CIT did not suffer from this timing
problem. CIT was an insolvent lending company with a very
solvent bank subsidiary (segregated and insured, to boot.)
The lending company was funded like an ordinary industrial
corporation, with bonded debt. The only protected financial
liabilities at the bank subsidiary did not run; the nonfinancial
liabilities were trapped in the process and could not run.
This provided enough time for Chapter 11 to work on the
nonfinancial liabilities.
Derivatives and repo closeout. Chapter 11 permits
unrestricted closeout of derivatives and repo transactions.
All Chapter 11 advocates have recognized this problem and
have proposed fixes. In the context of a parent-level bail-in,
this problem is in the cross-default clause of the subsidiaries’
contracts, discussed above. The solution for Chapter 11 would
be the same as that for bail-in: a stay on derivative closeouts
triggered by parental filing.
Governance. Chapter 11 has a complex governance
structure, not suited to fast resolution. For consequential
decisions, everybody has a say, with the court’s word as the
final one. This works because the automatic stay buys the
necessary time. But the automatic stay buys no time in a
financial insolvency, and the most consequential decisions
are on the first day. The first day is the day to flash the most
money: “shock-and-awe” DIP financing must be in place.
The financial creditors will run on the first day, unless they
are assured by subsidiary recapitalization and a thick layer
of reorganized parent equity. Clearinghouses and foreign
regulators also need assurance on the first day—preferably
earlier. There is no way the court can do this all on the first
day and provide ordinary bankruptcy due process.
Initiation. As a practical matter, the debtor initiates
Chapter 11. Since unsecured creditors traditionally receive
low recoveries, it is hard to avoid concluding that Chapter 11
begins later than it should. Most Chapter 11 reform advocates
have proposed a regulatory role in initiation, to supplement
the debtor’s role.
Liquidity. Chapter 11 has no public liquidity provider.
Normally, the private sector suffices; DIP loans are profitable.
However, megabanks need far more liquidity than most
industrial firms. Also, megabank insolvency often occurs
during a financial panic, when liquidity lending dries up.
Most Chapter 11 reform advocates have discussed a public
liquidity provider, generally assigning it the role of a DIP
lender. But DIP lenders generally play a very strong role in
a Chapter 11 process, which contradicts the general belief
among these advocates that the executive branch should
have limited discretionary power in Code bankruptcies.
The governmental DIP role was controversial in the GM
and Chrysler bankruptcies and should be more so in a
megabank bankruptcy.
Planning. Bankruptcy planning is part of modern
Chapter 11 practice: creditors need notice, and the court
must approve the ordinary operations and DIP financing
of the debtor. Much of this planning is not necessary for a
financial reorganization at the parent level. The operations
are contained in the subsidiaries, unaffected by the parent’s
filing. However, financial firms require one unique form of
planning that does not lend itself well to Chapter 11. They
must extend and obtain many pairs of credible conditional
reciprocal promises: that the parent will recapitalize its
subsidiaries and in consideration, that the subsidiaries’
regulators and clearinghouses will let them live. Bankruptcy
judges are generally realistic and businesslike, but no judge
can engage in extended secret ex parte communications in
advance of a filing.
Limits of structural subordination. Bail-in only operates
on the parent. Bail-in protects the subsidiaries’ creditors with
structural subordination: only the parent liability holders bear
losses. This is true regardless of the implementation of bail-in:
the FDIC’s SPOE approach, or Chapter 11. This protection
is complete if only the parent is insolvent. Structural
subordination does not necessarily protect the creditors of
insolvent subsidiaries. It can only work if the insolvent parent
recapitalizes the insolvent subsidiary.
This is more difficult to do in Chapter 11 than in something
like SPOE, even if there is adequate parental debt. The
problem lies in two key elements of Code ideology. First, the
Code looks no further than the welfare of the insolvent entity.
This ignores externalities—including those related to affiliates
or systemic risk. Second, the Code views the welfare of the
entity as consisting solely of maximum recovery for creditors,
consistent with the Code’s priority scheme.69 Therefore, Code
ideology demands that any recapitalization of a subsidiary by
an insolvent parent must benefit the creditors of the parent.
This is easy if the subsidiary is solvent. Every dollar that
flows from the parent’s creditors to the solvent subsidiary will
increase the value of the subsidiary by at least a dollar.
69
This is the logic of Judge Easterbrook’s defense of CVO priority. See supra
note 15. He argues that CVO priority is justified only if it enhances the
aggregate recovery of the creditors not given the CVO priority.
FRBNY Economic Policy Review / December 2014
227
But justification is harder if the subsidiary is insolvent. Any
value that the parent injects into the subsidiary will first go to
the creditors of the subsidiary, not the equity of the subsidiary.
Such a capital injection will only benefit the parent if we make
some special assumptions.
These special assumptions can be plausible. But they are
contestable, and they must convince a bankruptcy judge. Judges
may be willing to fudge close cases, mumbling “going-concern
value,” or the like. But many judges fully buy into the ideology
and might not want to fudge. And there are limits on what even
heterodox judges can do. The chief of these limits is time. The
subsidiaries must be recapitalized on the first day of the process.
Can a judge do this with any semblance of due process? And
putting due process aside, can the judge possibly have enough
information or time to make a sound decision?70
Consider, for example, a megabank with one deeply
indebted, but highly systemic, subsidiary and a number of
other subsidiaries that are doing well. If the parent has not
guaranteed this subsidiary, a bankruptcy judge would have a
hard time recapitalizing it. Or consider the contrary case:
a foreign non-systemic subsidiary with a very uncooperative
supervisor. The home country supervisor might want to let
it go: to encourage cooperation in the future. A bankruptcy
court might not. Or as a final case, consider the Section 363
sale of a major subsidiary to another large financial firm. This
may be in the best interest of the creditors, but might create an
excessively large firm.
No bail-in scheme can ignore the plight of the parent’s
creditors. The Constitution requires some solicitude: creditors
must do at least as well as they would in liquidation.71
Dodd-Frank and the FDI Act meet this standard, if not the
higher standard of the Code. This guarantee is denominated
in monetary terms, so the FDIC can act now and compensate
later, if necessary.
4.3 Private Law
The Dodd-Frank Act mandated consideration of
contingent convertible debt as a source of bank capital:
perhaps a substitute for equity. Can this idea be
extended? Is contingent convertible debt a plausible
private-law substitute for bail-in? This would resemble
Professor Adler’s “chameleon equity” proposal for corporate
restructuring, which converts old debt to new equity
upon a trigger event.72
A private law insolvency process would require a
contractual formula for debt conversion. In a megabank,
there is no time to wait for arbitration or adjudication. It is
difficult to imagine a trigger that does not contain either basis
or manipulation risk: possibly both. Management can control
reported capital levels; market participants can affect bond
prices. Both indexes also contain basis risk. Capital levels vary
with the macroeconomy: bond prices with the term structure
of interest rates.
Furthermore, the amount of convertible debt would have
to be large: as large as the amount of bail-in debt. And this
debt would have to convert to something. The something
could be equity. If so, the dilution will be very large: probably
exceeding standard shareholder protections entrenched in the
corporate certificate of the parent. This problem goes away if
the something is nothing: a fixed-income instrument junior to
equity. But what creditor would trust a debtor with this kind
of incentive to default?
Finally, contractual bail-in is still bail-in. It is hard
to see how it would work without public law: a stay on
ipso facto provisions, official-sector liquidity support,
international negotiations among public officials and
clearinghouses, and the like.
70
Judge Peck, who presided over the Lehman case, was very frank on the
difficulties of adjudicating the first few days. In re Lehman Brothers Holdings,
445 B.R. 143 (Bkrtcy S.D.N.Y. 2011).
71
Neblett v. Carpenter, 305 U.S. 297 (1938); cf. Doty v. Love, 295 U.S. 64 (1935).
The Code has a similar standard for distributions among classes of creditors.
11 U.S.C. § 1129(a)(7)(A)(ii). However, the Code maximizes the value
of the entity as a whole.
72
Cf. Barry Adler, Financial and Political Theories of American Corporate
Bankruptcy, 45 Stan. L. Rev. 311 (1993).
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
228
Why Bail-In? And How!
James McAndrews, Donald P. Morgan, João A. C. Santos, and Tanju Yorulmazer
What Makes Large Bank
Failures So Messy and What
Should Be Done about It?
• The failures of large banks are not only
costly—they destroy asset value and consume
legal resources—but also destabilizing, in that
they spill over to other financial institutions
and cause more widespread instability.
1. Introduction
T
• The stabilizing effects of an at-risk debt
requirement cannot be achieved by simply
requiring more equity; bail-in-able debt and
equity are not perfect substitutes in providing
financial stability if the resolution authority
is slow to close the bank.
his article uses “messy” repeatedly, so we should be clear
at the outset what we mean by this term. Simply put, we
mean that the failures of large banks are costly—in terms of
destruction of asset value arising from fire sales—and also
destabilizing—meaning their failure can threaten the operation
of financial markets generally. We maintain that messy failures,
so defined, are unique to large, complex, and interconnected
banking firms. A small bank failure is costly, in terms of
lost local output (Ashcraft 2005), but it does not threaten
the smooth functioning of the financial system at large.
Thus, small bank failures are costly, but not destabilizing.
The failure of a large nonfinancial firm can also be costly,
but it is not usually considered destabilizing; when the
bankruptcy of General Motors Company was considered,
most of the discussion was about lost jobs, not the stability
of the automobile sector.
We contend that the reliance of large banks on uninsured
financial liabilities is a key reason why their failures are
so messy. We define uninsured financial liabilities (UFL)
according to Sommer (2014) as liabilities that are issued
specifically by financial firms, that is, uninsured foreign
and domestic deposits, repurchase agreements (repos),
James McAndrews is an executive vice president and the director of research,
Donald P. Morgan an assistant vice president, João A. C. Santos a vice
president, and Tanju Yorulmazer a former research officer at the Federal
Reserve Bank of New York.
The authors thank Beverly Hirtle and Joseph Sommer for helpful comments.
Minh Phan provided excellent research assistance. The views expressed in
this article are those of the authors and do not necessarily reflect the position
of the Federal Reserve Bank of New York or the Federal Reserve System.
• The messiness of these failures can be traced
in part to large banks’ reliance on uninsured
financial liabilities (UFLs). UFLs include
uninsured foreign and domestic deposits,
repurchase agreements (repos), commercial
paper, and trading derivative liabilities.
• To ease the problem of large banks’ disorderly
failures, regulators might require the banks
to issue a certain amount of long-term
“bail-in-able” debt, or “at-risk” debt that
converts to equity in resolution.
Correspondence: don.morgan@ny.frb.org
FRBNY Economic Policy Review / December 2014
229
commercial paper, and trading derivative liabilities.1 These
liabilities are special for two reasons. First, unlike a regular
debt liability of a nonfinancial firm, uninsured financial
liabilities confer money-like or liquidity services that may
be impaired or destroyed in bankruptcy. This is one reason
why the failure of financial firms is especially costly or messy.
Another reason is that uninsured financial liabilities are
runnable. Runs on the large firms relying heavily on UFL
(or financial liabilities that are not fully collateralized) trigger
fire sales that inflict losses not just on the firm in question,
but also on other firms with similar portfolios of assets. That
is what we mean by destabilizing—it is the threat of systemic
consequences associated with the failure of a very large bank.
Our claim that the liabilities of financial firms are
the defining feature that makes failures messy is not
incompatible with the view that illiquid asset holdings or
organizational/global complexity contributes to messy
failures. While illiquid assets and organizational complexity
are undoubtedly important, we suggest that large banks’
liability structure is the defining feature that leads to messy
failures. Simplifying a bit, uninsured financial liabilities
are those liabilities that are runnable. When a financial
firm experiences a run or fears a run in some part of its
organization, it can trigger a fire sale of its assets as well as
runs by holders of runnable liabilities in other parts of the
firm or in other firms. So, in our view, the risk of a run is
the element that catalyzes the fire sales and other rapid and
destabilizing effects of a failure. The run creates a messy
situation because as the holders of runnable liabilities run,
they steal time from all other decisionmakers to respond
in an orderly manner. When the firm fails, those holders
of UFL that have not run lose twice, in the sense that they
may ultimately receive a pro rata share of the asset values,
which typically involves a loss, but they also will have lost
the services they had counted on—for example, having a
deposit that they would normally use to provide liquidity
at a moment’s notice to make purchases or investments.
We present some direct evidence in support of our
hypothesis that uninsured financial liabilities contribute
to messy failures. Using data on all failed banks and thrifts
(herein “banks”) resolved by the Federal Deposit Insurance
Corporation (FDIC) from 1985 to 2011, we first show that
banks more reliant on UFL in the year before their failure
experience larger contractions in UFL in the ensuing year.
This simple fact is consistent with the notion that UFL holders
1
Commercial paper issued by large bank holding companies (BHCs) is
distinguished from nonfinancial company commercial paper in that the large
BHCs tend to “make markets” in their own commercial paper, standing ready
to buy it back under most circumstances. This feature makes commercial
paper effectively demandable debt.
230
What Makes Large Bank Failures So Messy?
are prone to run. We then show that the estimated cost of
failures to the FDIC is increasing in the amount of UFL on a
bank’s balance sheet in the year before failure. We take that as
evidence for our premise that greater reliance on UFL leads
to runs and fire sales of assets, which make failure costlier.
Having discussed what we think makes large bank
failures so messy, we then turn to the question of what
to do about it. Following Calello and Ervin (2010),
European Commission (2012), Tarullo (2013), and others,
we advocate that BHCs be required to issue a certain
amount of long-term “bail-in-able” debt or, as we prefer,
“at-risk” debt that converts to equity in resolution (we call
it “at-risk” because the debt is at risk of being converted to
equity). If issued in sufficient quantities, the at-risk debt
requirement immunizes UFL holders from losses and thus
reduces their incentive to run.2 An at-risk debt requirement
would also have helpful incentive effects as it would tend to
discourage the over-issuance of UFL (although not so bluntly
as an outright ceiling) that Stein (2012) highlights in the
context of short-term debt.
One of the central contributions of this article is to
counter the argument that the stabilizing effects of an
at-risk debt requirement could be achieved by simply
requiring more equity, thus obviating the need to impose
a new requirement for this class of liabilities. According to
that view, requiring x units of equity and x units of at-risk
debt is no different, in stability terms, than requiring 2x in
equity. To investigate the claim requires one to consider
how the resolution authority behaves—that is, when it will
shut down the firm. Using a simple model, we show that
at-risk debt and equity are not strictly substitutes, assuming
(plausibly, we think) that the resolution authority is slow
to close a failing institution. The resolution authority in
our model is “slow” in the sense that it will shut down
and resolve a firm only once its (book) equity capital is
exhausted. Granting that assumption, we show that holders
of uninsured financial liabilities are less likely to run on a
bank that has x in long-term debt and x in equity than a
bank that has 2x in equity; resolution turns out to be more
frequent under the at-risk debt requirement, but also more
orderly. The at-risk debt functions as “capital in resolution”
that serves to stall runs by holders of uninsured liabilities.
2
We envision that the bail-in would happen in resolution under the FDIC’s
proposed single point of entry (SPOE) receivership. Under SPOE, the
FDIC would take over the holding company and transfer its assets to a
bridge financial holding company. The bridge bank would be capitalized by
bailing in the subordinated and unsecured term debt held in the receivership.
By taking over at the holding company level, the operating subsidiaries
(for example, the bank) could continue with business as usual. Since
the bridge bank would be well capitalized (and have adequate liquidity
provided by the Orderly Liquidation Fund housed in the U.S. Treasury),
uninsured liability holders should have little incentive to run.
Where we may differ from other proponents of an
at-risk debt requirement is that we advocate scaling the
requirement by the amount of uninsured financial liabilities
held by the (consolidated) entity. The logic for this scaling
is derived directly from our model of messy failures. First,
the at-risk debt, scaled to the amount of UFL, will provide a
buffer in resolution to protect holders of financial liabilities,
forestalling runs by them. Forestalling those runs will
reduce the messiness of the firm’s failure. Consequently,
designing the requirement to stop runs by the holders of
UFL is as important to a successful requirement as is the
buffering role of providing capital in resolution. Finally,
by imposing such a requirement scaled to the amount of
uninsured financial liabilities, and because issuing at-risk
debt is expected to be costly to the firm, the requirement
can provide the firms with incentives to reduce their
reliance on UFL, which would improve the overall stability
of funding by targeting the weak link in the large banks’
funding models: uninsured financial liabilities. Stein (2012)
argues that banks produce externalities when they issue
short-term, money-like liabilities, which can consist of both
insured liabilities and the uninsured financial liabilities
that we are focused on. Tying an at-risk debt requirement
to those liabilities would force firms to internalize those
externalities to some extent.
In contrast with those who, in seeking to end the
too-big-to-fail problem, suggest “breaking up the banks”
or reimposing more stringent separation of commercial
and investment banking as mandated in the Glass-Steagall
Act of 1933, we offer a seemingly less radical but equally
consequential change. We suggest that it is the liability side
of today’s large financial firms that should be restructured:
The uninsured financial liabilities should be separated
from the equity capital by an amount of long-term (at-risk)
debt. To issue more UFL, the firm would be required in
time to issue additional long-term (at-risk) debt. This
structure of the liabilities of a large financial firm would
assist in protecting the firm against runs, provide capital in
resolution, and produce incentives for those firms to avoid
excessive reliance on runnable liabilities. These benefits are
not without costs, nor would they fully ensure against messy
failures (topics we discuss later), but they would improve the
chances that failures would be avoided in the first place and,
if encountered, be of a more manageable scale.
The next section makes some preliminary points about the
problem of “messy” bank failures. Section 3 presents evidence
that UFL holders at failing banks are prone to run and that
those runs add to the cost of resolving those failures. Section 4
advocates and provides analytics in support of a long-term
(at-risk) debt requirement as a way to deal with the problem
of messy bank failures. Section 5 provides a general discussion
of our results. Section 6 summarizes our findings.
2. Preliminaries
Why are bank failures more disruptive than those of
nonfinancial firms? As Sommer (2014) explains, bank failures
are different because banks issue money as liabilities.3 One can
think of “money production” as one of the most important
services provided by banks. While textbooks often define
banks as intermediaries that gather the savings of households
and lend to productive enterprises, most economic models
of banks emphasize the point that banks issue deposits, or
other money-like liabilities (Diamond and Dybvig 1983
and Gorton and Pennacchi 1990), and that the demandable
deposits issued by banks are the source of messy failures of
banks when the depositors run.
More recently, banks have expanded their organizational
forms and activities (see Avraham, Selvaggi, and Vickery
[2012]). As reviewed by Gorton and Metrick (2010), the
rise of “shadow banking” has led to innovative forms of
liabilities, such as repos, that are the functional equivalent
of what used to be provided only by deposits.4 Gorton and
Metrick (2010) argue that repos are therefore a type of
money because they are liquid, functionally demandable
at par due to their largely overnight tenor, and able to
function as an overnight store of value. Similarly, other
forms of uninsured financial liabilities, such as commercial
paper issued by banks, are also demandable at par for large
customers that request the financial firm to “buy back” its
paper. As a result, a large amount of big financial firms’
funding is made up of uninsured financial liabilities, which
provide the monetary services of demandability at par and
apparent safety. They are consequently runnable.
It is important to note that U.S. and much international
law recognizes the unique characteristics of some uninsured
financial liabilities and specifically excludes them from the
stay that bankruptcy imposes on creditors. For many repo
contracts, and for most derivative contracts, the creditors can
exercise their right of close-out and sell collateral immediately.
This carve-out specifically recognizes that those claims
3
Versions of this point have been made before. Friedman and Schwartz
(1963) famously argued that the Great Depression was aggravated by
bank failures that contracted the supply of bank liabilities—that is, money.
Corrigan (1982) made a similar point, although more narrowly, in his
famous paper “Are Banks Special?”
4
To be sure, repo finance has been around for decades, but its use
has grown exponentially.
FRBNY Economic Policy Review / December 2014
231
on the firm are “special” and that the law in many cases
allows holders of those claims to exit their claim (by selling
collateral) rather than having to petition the bankruptcy
court for it. In addition, the special resolution regime for
banks and deposit insurance also recognizes the social value
of preserving the main financial liabilities of a bank—its
deposits—even in the event of the bank’s failure.
Most bank deposits in the United States are insured by the
FDIC.5 Because insured depositors are relatively unaffected
by the failure, a bank has the capacity to issue additional
deposits even if it is economically insolvent, in the sense
that the market value of its liabilities exceeds that of its
assets. Consequently, a bank is typically put into resolution
by its supervisor. In the United States, the FDIC resolves
failed U.S. banks. For most of these failed banks, the capital
structure is relatively simple, consisting primarily of insured
deposits along with equity, but often with an additional
portion of deposits that are uninsured. The firm is resolved
in one of several ways, often by transferring deposits and
an equivalent amount of assets to another bank in such a
way that depositors maintain full access to their deposit
accounts without interruption.
Our thesis is that bank failures are messy because holders
of uninsured financial liabilities can and do run to avoid
the consequences of failure. Financial liabilities are often
redeemable on demand at par, or subject to frequent rollover.
As financial liability holders run, the bank must borrow to
replace the funding it loses to the run, or sell assets quickly.
The asset sales can lead to deeply discounted prices (that
is, fire sales), (further) imperiling the solvency of the bank
and imposing costs on unaffiliated parties. In addition,
because other financial institutions demand uninsured
financial liabilities from banks because of their money-like
properties, the failure of the issuing bank can bankrupt the
institutions holding their liabilities (apart from fire sales).
The leading example, of course, is the money market fund
Reserve Primary Fund; after Lehman Brothers filed for
bankruptcy, that fund “broke the buck” after Lehman Brothers
filed for bankruptcy because it was holding $535 million of
Lehman’s commercial paper.6
To be clear, we are not saying that reliance on UFL is
the only feature that makes bank failures costly. We know
from Ashcraft (2005) that even small bank failures are costly
in terms of forgone output. His findings could reflect that
5
According to Federal Financial Institutions Examination Council,
Consolidated Reports of Condition and Income, insured deposits made
up 61 percent of all domestic deposits in the fourth quarter of 2012.
bank failures destroy the private information that banks
develop about their borrowers so that erstwhile borrowers
become credit constrained after the failure. Our position is
that larger banks’ reliance on uninsured financial liabilities
is what makes their failures messy—that is, both costly and
destabilizing to other banks and the financial system. In
other words, small bank failures are “merely” costly, but large,
UFL-dependent bank failures are messy.
3. Testing Our Thesis
Recall our thesis that uninsured financial liabilities
contribute to messy (costly and destabilizing) large bank
failures for two reasons. First, the money-like services
provided by those liabilities are destroyed in the event
of failure. Second, UFL are runnable, which can lead
to fire sales of assets that not only destroy value at the
failing institution, but can also have spillover costs on
other institutions with similar asset holdings. This
section provides some evidence on both points. First we
show that UFL holders at failed banks are prone to run.
Then we provide evidence that greater reliance on such
liabilities leads to messier—that is, costlier—failures.7
Chart 1 plots the various components of UFL—uninsured
domestic deposits, foreign deposits, repos, commercial
paper, and derivative liabilities—by BHC asset decile. In
general, UFL increases with BHC size, primarily because of
increasing reliance on uninsured deposits. For BHCs in the
90th percentile, the class comprising megabanks, there is
a sharp increase in the share of liabilities accounted for by
UFL. The jump reflects increased reliance on virtually every
component of UFL except uninsured domestic deposits. This
chart neatly makes the point that if, as we maintain, reliance
on UFL makes for messy bank failures, then we would expect
large bank failures to be especially messy.
To test the hypothesis that UFL holders are prone to run
when a bank is in distress, we turned to the FDIC database
on failed banks. The data include 1,619 instances of failed
banks or thrifts (“banks”) between 1985 and 2011. Summary
statistics for the banks, including those for a number of
variables we use in a subsequent regression, are reported in
Table 1. The statistics are measured at the quarter of failure,
unless otherwise indicated. The average assets of the failed
banks over this period (at the quarter of failure) totaled
only about $275 million, so these are not the large banks
that most interest us. Nevertheless, the data represent a
useful laboratory to test our ideas.
6
The Reserve Primary Fund was also holding $250 million of
medium-term notes. See http://www.reuters.com/article/2010/04/14/
reservefund-lehman-idUSN1416157520100414.
232
What Makes Large Bank Failures So Messy?
7
Since we are studying smaller bank failures here, we do not test for
evidence that UFL is associated with more financial instability.
To test the run hypothesis, we estimated the
following regression:
Chart 1
Uninsured Financial Liabilities,
Sorted by Asset Decile
UFLit - UFLit − 4
βUFL
____________
Assets
= a + _______
Assetsit − 4 + λlog(Assetsit − 4) + εit − 4 .
it − 4
it − 4
Our hypothesis is that β < 0, that is, failing banks or thrifts
experience larger runoffs of UFL over the year before their
failure, the larger their UFL holding the year before failure.
Despite the t subscript, this is not a panel regression; we are
simply regressing the scaled, four-quarter change in UFL on
the UFL four quarters earlier for the set of 1,619 failed banks
and thrifts. The regressions include fixed effects for the state
in which the failure occurred and the type of insurance fund.8
50
40
Percentage of total liabililities
Repos
Commercial paper
Foreign deposits
Derivative liabilities
Uninsured domestic deposits
30
20
10
8
Before 1989, there were two federal deposit insurance funds, one administered
by the FDIC, which insured deposits in commercial banks and state-chartered
savings banks, and another administered by the Federal Savings and Loan
Insurance Corporation (FSLIC), which insured deposits in savings associations
with state or federal charters. In 1989, the Financial Institutions Reform, Recovery,
and Enforcement Act (FIRREA) specified that thereafter the FDIC would be
the federal deposit insurer of all banks and savings associations and would
administer both the FDIC fund, which was renamed the Bank Insurance Fund
(BIF), and the replacement for the insolvent FSLIC fund, renamed the Savings
Association Insurance Fund (SAIF). Although it was created in 1989, the SAIF
was not responsible for savings association failures until 1996. From 1989 through
1995, savings association failures were the responsibility of the Resolution Trust
Corporation (RTC). In February 2006, the Federal Deposit Insurance Reform
Act of 2005 provided for the merger of the BIF and the SAIF into a single Deposit
Insurance Fund (DIF). Necessary technical and conforming changes to the law were
made under the Federal Deposit Insurance Reform Conforming Amendments Act
of 2005. The merger of the funds was effective on March 31, 2006.
0
1st
2nd
3rd
4th
5th
6th
7th
8th
9th
10th
Sources: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data); Federal Financial Institutions Examination
Council, Consolidated Reports of Condition and Income.
Note: The chart plots the UFL components of U.S. bank holding
companies (BHCs) as a percentage of total liabilities at different asset
sizes. To construct this chart, we split the set of BHCs in 2012:Q4 into
deciles, according to total asset size. We proxy BHC-level uninsured
domestic deposits for a particular asset decile with bank-level
uninsured domestic deposits for the same decile. All other line
items are obtained from the Federal Reserve’s Y-9C Form.
Table 1
Summary Statistics Calculated for Failed Banks and Thrifts from 1985 to 2011
Variables
Observation
Mean
Log [estimated loss to FDIC]
1,619
9.15
Uninsured financial liabilities (thousands of dollars) — lag 4Q
1,619
71,467.16
Uninsured financial liabilities / assets — lag 4Q
1,619
0.11
Log [uninsured financial liabilities / assets] — lag 4Q
1,619
GDP growth
1,619
Log [assets]
Median
9.11
Standard
Deviation
Minimum
Maximum
2.06
0.00
41,3971.59
0.00
8,233,800.00
0.08
0.11
0.00
0.84
0.10
0.07
0.10
0.00
0.61
0.62
0.64
0.64
-2.30
1.95
1,619
10.98
10.73
1.53
7.46
17.05
Assets (thousands of dollars)
1,619
274,726.91
45,573.00
1,141,216.61
1,731.00
25,455,112.00
Commercial real estate loans / assets
1,619
0.21
0.15
0.18
0.00
0.78
Real estate owned / assets
1,619
0.05
0.04
0.05
0.00
0.53
Loans past ninety days / assets
1,619
0.02
0.01
0.03
0.00
0.28
Total equity capital / assets
1,619
-0.01
0.00
0.06
-0.48
0.52
Asset growth
1,619
-12.35
-14.84
21.80
-63.43
359.58
3,877.00
15.35
Source: Federal Deposit Insurance Corporation.
Notes: All balance sheet variables are measured at the date of failure. Asset growth (yearly rate) is measured at the quarter of failure.
FRBNY Economic Policy Review / December 2014
233
Table 2
Is Higher UFL Associated with More UFL Runoff at Failed Banks?
Failed Banks
(1)
UFL / assets — lag 4Q
Log assets — lag 4Q
(3)
(4)
-0.502***
-0.507***
-0.268***
-0.287***
[0.037]
[0.059]
[0.061]
[0.092]
0.002
Constant
Healthy Banks
(2)
0.008***
0.007**
0.006
[0.002]
[0.003]
[0.003]
[0.006]
-0.005
-0.073***
-0.045
-0.037
[0.018]
[0.025]
[0.029]
[0.046]
Observations
1,619
1,619
1,619
1,619
Adjusted R-squared
0.334
0.361
0.084
0.083
Fund FE
YES
NA
State FE
YES
NO
YES
Source: Authors’ calculations.
Notes: The table reports ordinary least squares (OLS) regression and robust standard errors (clustered by time and state) in parentheses. The dependent variable
is the change in UFL over the previous year, scaled by assets four quarters before failure. For a placebo test, we tested whether the relationship between lagged
UFL and the change in UFL holds for a matched sampled of healthy banks. Healthy banks were matched by state, entity type, assets (within 25 percent of
matching failed banks), and date. Robust standard errors are presented in brackets.
*** p < 0.01
** p < 0.05
* p < 0.1
The results are reported in Table 2, models 1 and 2. Consistent
with the hypothesis, we observe β < 0, with the estimate
significant at the 1 percent level. The point estimate in
model 2 (with all the fixed effects) implies that a failing
bank or thrift with the mean ratio of UFLt − 4 ∕ Assetst − 4
(11 percent) experiences a runoff of 5.5 percent of assets.
We can express the run in dollar terms if we assume that
the bank with mean UFLt − 4 ∕ Assetst − 4 also has mean assets
($275 million). In that case, the bank would experience a
run of 0.055 × $275 = $15 million. Note from the summary
statistics (Table 1) that failing banks did experience substantial
asset contractions in the year before their failure.9
To see if our run regressions were simply picking up
regression toward the mean, we also estimated placebo
regressions for a set of matched nonfailing (healthy) banks.
The healthy banks were matched by state, entity type (bank
or thrift), asset size, and date.10 In fact, we do observe a
9
Our premise is that a run on UFL triggered a contraction. However, we
cannot rule out the opposite causality—that is, that assets were contracting
so the UFL was allowed to run off.
significant relationship between the lagged level of UFL and
the change in UFL, suggesting that some regression toward
the mean may explain some of the link between lagged UFL
and UFL runoff observed for models 1 and 2. Note, however,
that the coefficient on lagged UFL in models 1 and 2 is
substantially larger for failed banks—almost twice as large, in
fact. Using a Chow test, we can reject at below the 1 percent
level that the coefficient on lagged UFL for failed banks in
model 1 equals the corresponding coefficient for healthy
banks in model 3.11 We take the extra sensitivity of the change
in UFL to lagged UFL for failed banks as evidence that failing
banks do experience runs by holders of UFL.
The greater tendency for UFL to run off from failed banks
is apparent in the histograms plotted in Chart 2 and Chart 3.
The histogram for the failed banks is skewed negative while the
histogram for the healthy, matched banks is more symmetrically
distributed around zero. The skewness statistic for failed banks
is -0.939. The statistic for healthy banks is -0.004.
Now we present some regression evidence consistent with
the hypothesis that higher UFL is associated with costlier
10
The healthy banks were considered a match by assets if their assets were
within 25 percent of the failed bank.
234
What Makes Large Bank Failures So Messy?
11
We cannot do a Chow test for models 2 and 4 because the fixed effects differ.
Chart 2
Chart 3
Failed Banks
Healthy Banks
Frequency
Frequency
800
300
600
200
400
100
200
0
0
-40,000
-20,000
0
20,000
40,000
Change in UFL from previous year (thousands of dollars)
-10,000
-5,000
0
5,000
10,000
Change in UFL from previous year (thousands of dollars)
Source: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data); Federal Financial Institutions Examination Council,
Consolidated Reports of Condition and Income.
Source: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data); Federal Financial Institutions Examination Council,
Consolidated Reports of Condition and Income.
Note: The chart plots the UFL components of U.S. bank holding
companies (BHCs) as a percentage of total liabilities at different asset
sizes. To construct this chart, we split the set of BHCs in 2012:Q4 into
deciles, according to total asset size. We proxy BHC-level uninsured
domestic deposits for a particular asset decile with bank-level
uninsured domestic deposits for the same decile. All other line
items are obtained from the FR Y-9C.
Note: The chart plots the UFL components of U.S. bank holding
companies (BHCs) as a percentage of total liabilities at different asset
sizes. To construct this chart, we split the set of BHCs in 2012:Q4 into
deciles, according to total asset size. We proxy BHC-level uninsured
domestic deposits for a particular asset decile with bank-level
uninsured domestic deposits for the same decile. All other line
items are obtained from the FR Y-9C.
failures. As before, we use the FDIC’s data on bank failures,
except now we focus on estimated losses (to the FDIC)
associated with bank and thrift failures; the estimated loss is the
difference between the amount disbursed from the insurance
fund and the amount estimated to be ultimately recovered
from liquidation of the receivership estate.12 According to our
hypothesis, failing banks with more UFL in the period leading
up to their failure are more likely to have to “fire sale” assets,
and the attendant liquidation costs should be expected to
increase the costs of the failure to the deposit insurer.
Our regression model is
we use the same set of variables shown by Schaeck (2008)
to influence FDIC losses on failures. We also include fixed
effects for the state where the failure occurred, the transaction
type (failure or assistance), and the type of insurance fund.
We report ordinary least squares (OLS) estimates and
Tobit estimates (since the dependent variable is truncated
at zero). We predict β > 0.
Columns 1 and 2 of Table 3 reveal a positive and significant
(at the 5 percent level) relationship between the costs of
failure and the level of UFL four quarters earlier, that is,
the failures of banks with more UFL are costlier. Given that
the distribution of UFL is so heavily skewed toward larger
institutions, we tried splitting the sample and estimating the
model separately for failed institutions with assets below
the median for the sample ($45.6 million) and institutions
with assets above the median. Splitting the sample reveals
an interesting difference: The positive relationship between
the cost of failure and the amount of UFL holds only for the
larger failed banks in the sample; for the smaller banks, there
is also a positive relationship, but it is not significant. The
OLS coefficient estimate in model 3 implies that a 10 percent
(roughly one standard deviation) increase in the ratio of
log (_____
Assets it ) = α + βlog( _______
Assetsit − 4 ) + λ'Controlsit + εit .
it − 4
it
Losses
UFL
On the right-hand side, we lag UFL by four quarters for
consistency with the run regression results.13 For controls
12
See the FDIC’s data on failed banks at http://www2.fdic.gov/hsob/
SelectRpt.asp?EntryTyp=30.
13
Assets on the left-hand side are measured at the quarter of failure.
FRBNY Economic Policy Review / December 2014
235
Table 3
Is Higher UFL Associated with Costlier Banks?
Assets > Median
All Banks
Log UFL / assets – lag 4Q
(1)
(2)
(3)
(4)
(5)
(6)
OLS
TOBIT
OLS
TOBIT
OLS
TOBIT
1.010**
GDP growth
Log assets
0.663
0.650
[0.429]
[0.428]
[0.491]
[0.481]
[0.834]
[0.817]
-0.118**
-0.119**
-0.065
-0.065
-0.114
-0.116
[0.050]
[0.050]
[0.060]
[0.059]
[0.082]
[0.081]
0.755***
[0.056]
Commercial real estate loans / assets
0.949***
[0.328]
Real estate owned / assets
4.929***
[0.587]
Loans past ninety days / assets
6.256***
[1.018]
Total equity capital / assets
3.386***
[0.659]
Asset growth
0.004**
Constant
Assets < Median
1.005**
0.755***
[0.056]
0.955***
[0.327]
4.952***
[0.585]
6.291***
[1.008]
3.398***
[0.652]
0.004**
1.449***
0.748***
[0.085]
1.516***
[0.440]
6.710***
[1.104]
7.297***
1.455***
0.747***
[0.083]
1.533***
[0.433]
6.768***
[1.088]
7.352***
0.990***
0.992***
[0.084]
[0.083]
0.214
0.216
[0.410]
[0.401]
3.089***
[0.646]
4.846***
3.097***
[0.632]
4.858***
[2.066]
[2.003]
[0.832]
-4.816***
-4.831***
-2.221***
[1.352]
[1.322]
[0.692]
[0.673]
0.002
0.002
0.006**
0.006**
[0.812]
2.231***
[0.002]
[0.002]
[0.003]
[0.003]
[0.002]
[0.002]
-0.009
-0.021
-0.673
-0.684
-2.052**
-2.078**
[0.543]
[0.540]
[0.942]
[0.917]
[0.846]
[0.833]
1,619
809
809
810
810
Observations
1,619
Adjusted R-squared
0.611
Fund FE
YES
YES
YES
YES
YES
YES
State FE
YES
YES
YES
YES
YES
YES
0.512
0.379
Source: Authors' calculations.
Notes: The table reports regression estimates and robust standard errors (in parentheses). The dependent variable is the estimated cost of failure to the FDIC
per assets. Coefficients are estimated over the indicated number of failures over the period 1985 to 2011.
UFL to assets is associated with a 15 percent increase in the
ratio of estimated costs to assets. This should be viewed as a
lower bound of the costs associated with UFL because our
dependent variable does not capture the effect of fire sales
on the solvency of other banks. Note also that the cost of
failure is significantly increasing in the log of assets; failures
of larger banks are messier.
236
What Makes Large Bank Failures So Messy?
4. What to Do about the
Problem of Messy Failures?
Having argued and provided some evidence that reliance on
uninsured financial liabilities is one reason why large bank
failures are so messy, we now turn to the question of what
to do about it. We cannot simply argue that banks should
eschew the use of such liabilities because the liquidity they
create is socially valuable. Instead, we join the chorus of those
calling for a long-term debt requirement, where the debt
is bail-in-able—that is, it converts to equity in resolution.
Given that the debt is at risk of being converted to equity, we
prefer the term at-risk debt. We have three points to make
regarding the potential benefits of an at-risk (or subordinated)
debt requirement based on the amount of a bank holding
company’s financial liabilities.
The first point, which we spend some time on, is to counter
what is perhaps the most important possible objection to an
at-risk debt requirement. Stated simply, the objection is that
equity and at-risk debt are substitutes in terms of providing
financial stability. For example, suppose that the BHC has
$1 trillion in risk-weighted assets and a $75 billion Tier 1
common equity requirement; furthermore, consider an at-risk
debt requirement of an additional $75 billion. Then one might
object, why not make the Tier 1 common equity requirement
equal to $150 billion? In that case, the bank’s UFL will be
roughly equally protected against shocks to asset values,
and the BHC will not be put into resolution as frequently.14
Therefore, the at-risk debt requirement is superfluous relative
to an equity requirement that is higher by the exact amount of
the at-risk debt requirement.
Treating equity and at-risk debt as equally costly (that is,
not granting any benefits to the tax deductibility of interest
expense on debt), one still has to consider three issues before
concluding that the protection achieved by an at-risk debt
requirement can be duplicated by a larger equity requirement.
One has to specify 1) the rule by which the resolution
authority puts the BHC into resolution, 2) the process by
which losses accrue, and 3) the incentives of the bank to issue
uninsured financial liabilities.
First, because long-term debt and equity are generally more
expensive forms of funding for a financial firm, we assume
that, without the requirement to issue at-risk debt, the BHC
would issue UFL to the extent feasible, up to its required
equity.15 Second, we assume—and this is critical—that the
resolution authority puts the BHC into resolution only after
it has experienced losses in excess of its equity.16 Finally,
14
We are ignoring the fact that if the protection takes the form of equity,
the bank will pay higher taxes out of cash flow. This may reduce the
retained wealth available for UFL protection.
15
Equity is more expensive than debt generally because interest is tax
deductible. Long-term debt is usually considered more expensive than
short-term debt because of the greater uncertainty associated with the longer
maturity. In addition, the higher cost of long-term debt may not be offset by
lower costs of other liabilities of the firm, in violation of the Modigliani-Miller
framework; if there are agency problems (conflicts of interest between
shareholders and creditors), creditors may prefer lending with a “short
leash”—that is, short-term. Pushing them away from their natural habitat will
require a maturity premium that makes long-term debt more expensive.
16
This assumption is not implausible; in the bank failure data we studied
earlier, only two out of 1,619 failures did not entail losses to the FDIC.
Prompt corrective action implies in principle that the FDIC should close
we assume that the loss-generating process is a relatively
“smooth” one, so that there are no large jumps to default;
instead, the BHC transits through relatively small losses to
larger losses (this process could be a random walk, but the size
of incremental losses, if not continuous, is small; alternatively,
and more realistically, it could be a process with significant
serial correlation). With these three assumptions, we now
demonstrate that a larger equity requirement is not equivalent
to an equity requirement plus an at-risk debt requirement.
Consider a BHC with a large equity requirement
($150 billion in our example) versus one with both equity and
at-risk debt requirements (a $75 billion equity requirement
and a $75 billion at-risk debt requirement). We assume, for
this exercise, that both BHCs have issued the same amount
of UFL and they both have the same asset composition.
Now, when the firm has the high equity requirement, all of
its remaining liabilities are in the form of UFL. As the firm
experiences losses that grow from 13 to 14 to 15 percent of
its risk-weighted assets, the holders of the UFL realize that
they have no further “buffer” that would limit their exposure
if losses grow from those levels.17 Knowing, furthermore,
that the resolution authority will not put the BHC into
resolution until losses exceed 15 percent of risk-weighted
assets, the holders of the UFL will likely run on the BHC.
As the run creates fire sales by the BHC, imposing losses on
other parties, the resolution of the firm will be messy, and
the government may feel the need to bail out the BHC’s UFL
holders to forestall the run.
In contrast, consider the BHC with both the equity and
the debt requirement. In this case, losses of half the previous
size will exhaust the BHC’s equity. When losses rise from
5 to 6 to 7.5 percent of risk-weighted assets, the holders of
the UFL realize that the firm has losses that equal its equity
and that it will likely be put into resolution. However, they
also recognize that the $75 billion of at-risk debt provides
a source of “capital in resolution” that, in the event of the
firm’s resolution, provides a buffer against further losses
from eroding the value of the firm’s UFL. Consequently,
the UFL holders have little reason to run. As a result, the
resolution authority could put the BHC into resolution
without triggering a run, allowing a greater chance for an
banks before capital is depleted and the FDIC is exposed to losses. However,
as just noted, losses to the FDIC are the rule in FDIC failures. Nonetheless,
our assumption can be weakened. What is required is that 1) there are
dead-weight costs to resolution that will deplete assets available to pay out
to holders of UFL, and 2) the timing of the resolution is uncertain, so that by
the time it occurs there is a sufficient probability applied to the outcome that
UFL holders will not be made whole in the course of the resolution or that
payouts to them will be delayed.
17
The losses and equity values discussed in this section are all in book terms.
FRBNY Economic Policy Review / December 2014
237
orderly resolution (holding fixed the potential signaling effects
on other firms). So if society were to substitute long-term
at-risk debt for equity, one would expect more frequent
failures of firms, but these failures would be less likely to be
accompanied by runs on the firm—that is, they would be less
likely to be messy. By contrast, if long-term at-risk debt were
deployed in addition to the minimum regulatory equity capital
requirement, then, all else equal, losses that deplete capital
would be no more frequent but would be less messy.
In summary, the difference between “loss bearing” capacity
in which one is expressed solely as an equity requirement and
the other is split between an equity requirement and an at-risk
debt requirement is this: An at-risk debt requirement results
in more frequent resolutions of BHCs, but these resolutions
are more orderly. Essentially, under our assumptions, a
requirement consisting solely of equity results in little
expected protection for the holders of UFL in those extreme
events in which equity is exhausted, resulting in runs on the
firm. This, in turn, reduces the chances that resolution can be
accomplished in an orderly way, putting greater pressure on
the government to bail out the UFL of the firm.
We can make the same point about the benefits of
an at-risk debt requirement more generally using some
algebra. Consider a model with three dates, t = 0,1,2,
and a representative bank with the following balance sheet:
Assets
A
Liabilities
UFL
LD
E
The bank has assets worth A, which it funds with UFL,
long-term debt, LD, and equity, E.18 UFL can be redeemed
at t = 1. Long-term debt can be redeemed only at the last date
t = 2. All liability-side variables are valued as of date t = 2.
LD is at risk, or bail-in-able, because it is junior to UFL. That
is, in the event of default, long-term debtholders are paid only
after UFL debtholders have been reimbursed in full.
We assume that the return on the bank’s assets is random
and that the bank can suffer losses at dates 1 and 2. In
particular, we assume two states of the world: The good
state occurs with probability 1 - a, and the bad state occurs
with probability a. If the good state of the world occurs, the
bank does not suffer any losses, and the value of its assets
18
Note the absence of insured deposits; we show in the appendix that the
case for an at-risk debt requirement is even stronger when the bank has
insured deposits because insured depositors are senior to UFL creditors
and therefore the latter are more likely to run.
238
What Makes Large Bank Failures So Messy?
is A at t = 2. If the bad state of the world occurs, the bank
suffers losses L1 at t = 1. Further, if the bad state of the
world occurs, with probability 1 - β, the bank does not
suffer any further losses at t = 2, in which case the value of
its assets is A - L1, but with probability β, the bank suffers
additional losses L2 at t = 2, in which case the value of its
assets is A - L1 - L2 at t = 2.
We now consider two alternative funding structures
for the bank in our model:
Case I (all equity): The bank holds no long-term debt, only
equity. The bank’s balance sheet thus has the following form:
Assets
A
Liabilities
UFL
LD1 = 0
E1
Case II (equity and long-term debt): The bank holds some
long-term debt and some equity, where the sum of the two
is equal to the equity the bank holds in Case I (all equity).
Hence, the bank’s balance sheet has the following form:
Assets
A
Liabilities
UFL
LD2 = E1 - E2
E2 < E1
We assume that the bank makes the following promises
to its UFL creditors: If they withdraw their funds at t = 1,
they will receive 1 unit; and if they choose to roll over their
claims and withdraw their debt at t = 2, they will receive
the return of rs > 1 at t = 2. In order to see UFL creditors’
rollover incentives, consider the following scenario: Suppose
that A - L1 - L2 < UFL < A - L1. Under these conditions,
in the bad state of the world, if the bank experiences further
losses, it does not have enough funds to pay UFL creditors in
full at t = 2, whereas the bank can pay them in full at t = 2 if
it does not experience any further losses. In Case I (all equity)
the bank has positive equity E1 - L1 > 0 at t = 1. Suppose
that E2 < L1 so that, under Case II (equity and long-term debt),
the bank has negative equity at t = 1 in the bad state. Note that,
if the probability of the bank experiencing additional losses at
t = 2 (β) is sufficiently high, UFL creditors will be concerned
about the solvency of the bank and decide not to roll over their
claims, resulting in a run on the bank.
We model UFL creditors’ rollover decision at t = 1 as
follows: If a UFL creditor withdraws, he receives 1 unit. If he
rolls it over, he expects to receive
A - L1 - L2
β ( ___________
) + (1 - β)rs ,
UFL
since with probability 1 - β, the bank does not experience
additional losses and an UFL creditor receives the promised
amount rs, and with probability β, the bank experiences
additional losses and the creditor receives a pro rata share of
the bank’s return at t = 2 with other UFL creditors. Long-term
creditors receive nothing because, by assumption, they hold a
junior claim. Hence, the UFL holders will withdraw as long as
A - L1 - L2
β ( ___________
) + (1 - β)rs < 1,
UFL
(1)
that is, when β is sufficiently high:
rs - 1
β > ____________
= β*.19
A-L -L
1
2
rs - __________
UFL
Hence, for β > β*, it is optimal for UFL creditors not
to roll over their claims, and, consequently, in the bad state,
there will be a run on the bank at t = 1 unless the regulator
intervenes. Note that in the benchmark case, where there is
no intervention by a regulator, long-term at-risk debt and
equity provide the same level of buffer for losses; they are
substitutes.20 Next, we modify the benchmark case to show
how long-term debt and equity can have different effects once
regulatory intervention is possible.
Suppose that a regulator intervenes if, and only if, the bank
has negative equity.21 We assume the regulator can make this
decision before UFL creditors decide whether they will roll
over their debt (say, at t = 1/2).
Then, at t = 1/2 in Case I, where the bank has all equity,
the bank has a positive equity of E1 - L1 > 0, so that the
regulator leaves the bank open. However, for β > β*, the
probability of further losses is large enough that UFL creditors
do not roll over, resulting in a run on the bank.22
19
Note that we are assuming depositors are risk neutral. If they were
risk averse, the threshold for running would differ.
20
To see that explicitly, substitute the balance sheet identity
A = UFL2 + LD2 + E into (1).
21
22
To be clear, the meaning here is book value of equity, not market value.
One can argue that the regulator can intervene if it anticipates a run, even
though the bank may have positive equity at the moment. We can extend the
model and allow the value of β to be uncertain, either high or low, and in
expectation the bank can pay all wholesale creditors (or has positive equity) so
that the regulator does not intervene. But once the high value of β is realized,
the run starts, and it is too late for the regulator to intervene to prevent it.
To contrast, consider Case II, where the bank has some
equity and some long-term debt. Since in the bad state of the
world the bank’s equity is already wiped out (E2 - L1 < 0),
the regulator has to intervene. The long-term debt (by
providing, in the event of resolution, a loss absorber in front
of the uninsured financial liabilities) allows the regulator to
take the “right” action (when it follows a rule of intervening
when the capital has been wiped out).
The analysis above suggests that an at-risk debt requirement
can add to the stability of a BHC by preventing runs by UFL
creditors. It should be noted that more frequent (but more
orderly) resolutions would be expected only if an at-risk debt
requirement were put in place at the expense of a lower equity
requirement. However, if the at-risk debt requirement were
met by substituting UFL with long-term debt, then there would
be no expectation of more frequent resolutions.
Next, we show that the amount of long-term debt should
be increasing in the amount of UFL the bank uses for the
same level of the threshold value
β*. To perform the analysis,
_
we fix the equity of the bank at Eand change UFL and LD. In
particular, for the same level of β*, we obtain
∂β*
∂UFL
∂β*
∂LD
dβ* = _____
dUFL + ___
dLD = 0.
rs - 1
Using β* = ____________
A - L - L and the balance sheet identity
1
2
rs - __________
UFL
_
A = UFL + LD + E
, we can show that
_
∂β*
∂UFL
sign(_____) = sign(-(rs - 1)(LD + E
- L1 - L2)),
which is negative, and
∂β*
∂UFL
sign(_____
) = sign(UFL(rs - 1)), which is positive.
∂β*/∂UFL
dLD
Hence, we have _____
= _______
> 0.
dUFL
∂β*/∂LD
If the bank wants to increase UFL, it needs to hold more
long-term debt for the same level of bank stability (as
measured by the likelihood of runs by UFL). We can
perform the same analysis where we keep the equity of the
bank fixed as a fraction of the bank’s assets. In that case, we
obtain similar results, but the required increase in long-term
debt is less compared with the previous case. This is because
when the bank’s balance sheet expands due to an increase
in UFL, its equity increases (while keeping the capital ratio
constant). The increase in the bank’s equity provides some
cover for the holders of UFL, and the required increase in
long-term debt can be less compared with the previous case.
FRBNY Economic Policy Review / December 2014
239
Recall that we had three points to make about the benefits
of an at-risk debt requirement. We now turn to the second:
the internalization of an externality. While banks produce a
socially valuable, money-like service when they issue UFL,
they may create too much of a good thing. As Stein (2012)
and others have noted, there are externalities associated with
the production of short-term debt; banks capture the social
benefit of the production of short-term debt, but they do not
always internalize its costs—namely, fire sales.23 In the event
of, or anticipation of, a crisis, banks are forced to “fire sale”
assets to meet their short-term obligations, a move that can
exacerbate the crisis by weakening the solvency of banks
with similar assets. As Stein (2012, p. 2) explains, “banks
may engage in excessive money creation, and may leave the
financial system overly vulnerable to costly crisis.”
Requiring banks to issue long-term at-risk debt in
proportion to their financial liabilities can force banks to
internalize the external costs associated with UFL issuance.
The at-risk debt requirement forces banks to deviate from
their privately optimal liability structure (because long-term
debt is costlier than short-term debt), and, under our
proposal, the required deviation is increasing in the amount
of UFL. Thus, banks are inclined to be less reliant on UFL in
their balance sheet choices.
The third potential benefit of an at-risk debt requirement is
that debt can provide a useful signal of risk to supervisors. As
Gropp, Vesala, and Vulpes (2006) point out, market indicators,
such as spreads on debt, have the advantage of being more
frequently observed and more forward-looking than accounting
data. Bond spreads, in particular, have the advantage over
equity prices in that spreads are not increasing in volatility as an
institution nears default; bond spreads represent the downside
perspective of supervisors and the FDIC. Gropp, Vesala, and
Vulpes (2006) show that both subordinated bond spreads and
equity prices help predict bank downgrades, but at different
horizons. Both have marginal predictive power compared
with bank accounting data.
5. Discussion
To reiterate, we have said that at-risk debt plays the role of
capital in the resolution of a firm. We have said also that basing
the at-risk debt requirement on the amount of UFL issued
by a firm serves the purpose of providing additional capital
23
The short-term debt emphasized in Stein (2012) seems very close to our
concept of UFL. See also Gorton and Metrick (2010), Gorton and Pennacchi
(1990), and Holmström and Tirole (1998, 2011).
240
What Makes Large Bank Failures So Messy?
in resolution for those firms whose failure would likely be
the messiest (because of the high level of UFL among their
liabilities). Given that long-term debt is costlier than short-term
debt, the at-risk debt requirement would also provide an
incentive for firms to reduce their reliance on UFLs.
Consider a large financial firm whose liabilities consist
solely of insured deposits, with a large amount of equity. In
our suggested rule for at-risk debt shown above, the firm
would have a zero requirement of long-term debt. Is that
reasonable? We would argue that it is reasonable because the
liability structure of the firm would resemble a small bank
whose failures are not typically messy (recall that all the
firms we are discussing are subject to prudential regulation
and supervision). Since insured deposit holders are not
prone to run, the failure itself is unlikely to be extremely
messy. In this case, the deposit insurer would provide the
“capital” in resolution of the firm.
What would our proposal for basing an at-risk debt
requirement on the amount of UFL issued by a firm imply
about the amount of long-term debt large banks would have
to issue? Calibrating the requirement is beyond the scope of
this article, but conceptually we are proposing a rule of the form
LTDi = aUFLi .
Chart 4 plots the amount of UFL per assets as of the
fourth quarter of 2012 for the set of twenty-two BHCs
with more than $100 billion in assets. The chart shows
considerable variation in reliance on UFL, so the amount
of at-risk debt required, per dollar of assets, would
vary accordingly across BHCs.
In practice, given the complexity of the large financial
firms, it is difficult to measure precisely how much UFL a firm
has issued, because for some liabilities it is not perfectly clear
whether they are “financial” liabilities or exactly how runnable
they are (for example, it may be unclear what proportion of
its commercial paper a firm would buy back). Consequently,
it may be preferable to base an at-risk debt requirement
on the size of the firm as measured by either total assets or
risk-weighted assets combined with the amount of UFL they
issue, or make the requirement the greater of the two, such as
a requirement expressed as
LDTi = min{aUFLi, bTotal Assets}.
The parameters a and b can be chosen to make sure
that, in the event of a firm’s failure, the resolution authority
could forgive the long-term debt of the separate subsidiaries,
as needed, to provide them with additional capital.
Chart 4
UFL to Assets at the Largest BHCs
AIG
AMEX
GMAC
GS
MS
COF
BMO
BBT
FITB
STI
PNC
RBS
REG
WFC
TD
BOA
USB
JPM
HSBC
BNY
CTI
STT
6. Conclusion
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
UFL/assets
Source: Board of Governors of the Federal Reserve System,
Consolidated Financial Statements of Bank Holding Companies
(FR Y-9C data).
Note: The largest BHCs have total assets more than
U.S.$100 billion, according to the FR Y-9C in 2014:Q4.
would have sufficient long-term at-risk debt on hand to
provide capital that would cover a variety of scenarios
regarding the firm’s asset values.
The large firms we discuss are most often organized as a
holding company with many subsidiaries. How would the
at-risk debt requirement apply to a bank holding company?
One possibility would be to measure, at each subsidiary, the
amount of UFL that the subsidiary has issued to third parties.
The holding company would then be required to issue at-risk
debt in the amount of a multiplied by the total UFL issued
by all of the firm’s subsidiaries. In turn, the subsidiaries could
borrow from the holding company an amount of long-term
debt equal to a multiplied by the UFL issued by the subsidiary.
This arrangement would be consistent with the single point of
entry receivership approach to resolution that the FDIC has
proposed.24 Under that approach, the FDIC would take only
the holding company into resolution, with the intention of
maintaining the operating subsidiaries as going concerns. The
at-risk debt of the holding company would be converted into
equity of the bridge company. The bridge holding company
24
See http://www.fdic.gov/about/srac/2012/2012-01-25_resolution-strategy.pdf.
Accessed August 22, 2013.
If the Lehman Brothers bankruptcy proved anything, it was that
large bank failures are messy; they destroy value, they consume
legal resources, and, not least, they spill over to other financial
institutions and cause more widespread instability. This article
has suggested a unifying framework for understanding why
large bank failures are so messy. The reason for the messy
failures, we have argued, is banks’ heavy reliance on uninsured,
money-like financial liabilities, such as uninsured deposits,
repos, trading liabilities, commercial paper, and the like. The
liquidity services of those liabilities get destroyed in failure, and
the holders of those uninsured liabilities are prone to run as
the bank approaches failure, which can cause fire sales. Both of
these consequences make large bank failures messy.
We provide simple, direct evidence for our thesis. First, we
show that failed banks that relied more on uninsured financial
liabilities in the year prior to their failure experienced greater
contractions in uninsured financial liabilities over the ensuing
year. This finding is consistent with the hypothesis that
holders of uninsured financial liabilities are prone to run.
Second, we show that the cost of bank failures to the FDIC
was increasing in the amount of uninsured financial liabilities
in the year before the crisis. We take that finding as consistent
with the premise that distressed banks’ heavy reliance on
uninsured financial liabilities subjects them to runs and fire
sales, which increases the cost of the failure. That is, it makes
the failure messier (although our regression does not capture
the spillover to other institutions).
We join Calello and Ervin (2010), the European
Commission (2012), Tarullo (2013), and others in
recommending a long-term “at-risk” debt requirement
as an additional measure to help cope with the problem of
large banks’ messy failures. Having such debt convertible to
equity at failure provides a form of capital in resolution that
can, in principle, stall runs by uninsured liability holders.
Furthermore, sizing the requirement by the amount of
uninsured financial liabilities, as we recommend, helps
internalize the external costs (the risk of fire sales) of issuing
money-like uninsured financial liabilities.
While we recommend an at-risk debt requirement as a
way to deal with messy bank failures, we realize that such a
requirement is not a panacea. First, it is not entirely clear how
thick the market would be for at-risk, or “bail-in-able”, debt;
the peculiarities of pricing such an instrument could hamper
FRBNY Economic Policy Review / December 2014
241
its development. Second, there is the potential for unstable
market dynamics associated with an at-risk debt requirement.
Even a small rumor about losses at a large bank could cause
issuers’ debt prices to collapse and make it difficult for the
bank to issue new debt, which would potentially create a
crisis for the firm. So the issuance dynamics must be carefully
considered when requiring periodic issuance by a firm.
Firms should not be put into resolution solely because of
temporary disruptions in the market for their long-term debt.
Finally, this proposal, like many others, does not prevent the
buildup of systemic risk and the experience of contagion and
contagious defaults among firms. Consequently, we think
that this single approach, like all other approaches, cannot
by itself eliminate the too-big-to-fail problem. Instead, we
think this approach is an effective step in the right direction
to limit the most damaging feature of too-big-to-fail
242
What Makes Large Bank Failures So Messy?
financial firms: the fragility inherent in their reliance on
uninsured financial liabilities.
To be clear, we are recommending an at-risk debt
requirement as a supplement—not a substitute—for other
macroprudential regulations, including equity capital
requirements. In our discussion and argumentation, we
needed to consider the argument of whether 2x in equity
was as effective in limiting the messiness of large financial
firms’ failures as x in equity and x in long-term at-risk
debt. However, we conclude that at-risk debt and equity
are not substitutable. In particular, we do not suggest that
at-risk long-term debt should serve to fulfill equity capital
requirements, nor do we suggest that equity be allowed to
fulfill the at-risk long-term debt requirement. Our view is
that at-risk long-term debt should substitute for uninsured
financial liabilities, not equity capital.
Appendix
A - L1 - L2 - ID
β (______________
) + (1 - β)rs < 1,
UFL
Insured Deposits
Suppose now that the bank funds a portion of its assets
with insured deposits, ID. In this case, the bank balance
sheet has the following form:
Assets
A
Liabilities
ID
UFL
LD
E
Further assume that ID is senior to all other creditors in
bankruptcy. Suppose that A - L1 - L2 < ID + UFL and
ID + UFL < A - L1. Hence, in the bad state of the world, if
the bank experiences additional losses, it will not have enough
funds to pay all insured depositors and the UFL creditors in
full at t = 2, whereas it can pay them in full at t = 2 if it does
not experience additional losses.
Assuming that UFL holders follow a rollover decision
at t = 1 similar to that adopted in the benchmark
case, they will withdraw if
that is, if β is sufficiently high:
rs - 1
β > ________________
= β' < β*
A - L - L - ID
1
2
rs - ______________
UFL
Hence, for β > β', it is optimal for UFL holders to
withdraw at t = 1 in the bad state of the world, triggering
a run on the bank (unless the regulator intervenes). Note
that β' is decreasing in ID.
In the presence of insured deposits, the run threshold
for the probability of further losses is lower compared with
the benchmark case, that is, β > β'. Hence, UFL creditors
are more likely to run at t = 1 when the bank suffers losses
of L1. The reason is that the UFL creditors are junior to
insured depositors in bankruptcy so that, compared with the
benchmark case, UFL recover less in bankruptcy (and even
less so when the bank has more insured deposits). As a result,
early intervention by the regulator is even more important,
and the conclusions of the benchmark case about the
desirability of long-term debt are strengthened.
FRBNY Economic Policy Review / December 2014
243
References
Ashcraft, A. 2005. “Are Banks Really Special? New Evidence from the
FDIC-Induced Failure of Healthy Banks.” American Economic
Review 95, no. 5 (December): 1712-30.
Avraham, D., P. Selvaggi, and J. Vickery. 2012. “A Structural View
of U.S. Bank Holding Companies,” Federal Reserve Bank of
New York Economic Policy Review 18, no. 2 (July): 65-81.
Basel Committee on Banking Supervision. 2011. “Global Systemically
Important Banks: Assessment Methodology and the Additional Loss
Absorbency Requirement.” Consultative Document, July. Available
at www.bis.org/publ/bcbs201.pdf. Accessed August 22, 2013.
Brewer, E., and J. Jagtiani. 2013. “How Much Did Banks Pay to Become
Too-Big-to-Fail and to Become Systemically Important?” Journal
of Financial Services Research 43, no. 1 (February): 1-35.
Calello, P., and W. Ervin. 2010. “From Bail-Out to Bail-In.”
The Economist, January 28.
Corrigan, E. G. 1982. “Are Banks Special?” Annual Report of the
Federal Reserve Bank of Minneapolis.
Diamond, D., and P. Dybvig. 1983. “Bank Runs, Deposit Insurance,
and Liquidity.” Journal of Political Economy 91, no. 3
(June): 401-19.
European Commission. 2012. “High-Level Expert Group on
Reforming the Structure of the EU Banking System: Final Reports”
(E. Liikanen, chairman), October 2. Available at ec.europa.eu/
internal_market/bank/docs/high-level_expert_group/report_en.pdf.
Federal Deposit Insurance Corporation Advisory Committee on
Systemic Resolution. 2012. “Resolution Strategy Overview,”
January 25. Available at fdic.gov/about/srac/2012/2012-01-25_
resolution-strategy.pdf. Accessed August 22, 2013.
Federal Deposit Insurance Corporation and Bank of England.
2012. “Resolving Globally Active, Systemically Important
Financial Institutions.” Mimeo.
Gorton, G., and A. Metrick. 2010. “Regulating the Shadow Banking
System.” Brookings Papers on Economic Activity, Fall: 261-312.
Gorton, G., and G. Pennacchi. 1990. “Financial Intermediaries
and Liquidity Creation.” Journal of Finance 45, no. 1
(March): 49-71.
Gropp, R., H. Hakenes, and I. Schnabel. 2011. “Competition,
Risk-Shifting, and Public Bail-out Policies.” Review of
Financial Studies 24, no. 6 (June): 2084-120.
Gropp, R., J. Vesala, and G. Vulpes. 2006. “Equity and Bond Market
Signals as Leading Indicators of Bank Fragility.” Journal of
Money, Credit, and Banking 38, no. 2 (March): 399-428.
Holmström, B., and J. Tirole. 1998. “Private and Public Supply
of Liquidity.” Journal of Political Economy 106, no. 1
(February): 1-40.
———. 2011. Inside and Outside Liquidity. Cambridge,
Massachusetts: MIT Press.
Schaeck, K. 2008. “Bank Liability Structure, FDIC Loss, and Time
to Failure: A Quantile Regression Approach.” Journal of
Financial Services Research 33, no. 3 (June): 163-79.
Sommer, J. 2014. “Why Bail-In? And How!” Federal Reserve
Bank of New York Economic Policy Review 20, no. 2
(December): 207-28.
Stein, J. 2012. “Monetary Policy as Financial Stability Regulation.”
Quarterly Journal of Economics 127, no. 1 (February): 57-95.
Strahan, P. 2013. “Too Big to Fail: Causes, Consequences,
and Policy Responses.” Annual Review of Financial
Economics 5 (December): 43-61.
Tarullo, D. 2013. “Toward Building a More Effective Resolution
Regime: Progress and Challenges” available at
www.federalreserve.gov/newsevents/speech/
tarullo20131018a.htm#f8.
Friedman, M., and A. Schwartz. 1963. A Monetary History of the
United States, 1867-1960. Princeton: Princeton University Press.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the accuracy,
timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in documents
produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
244
What Makes Large Bank Failures So Messy?
@FedFRASER