The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Can We Explain Banks' Capital Structures?*
BY MITCHELL BERLIN
B
ank capital has been much in the news
during the recent financial crisis. In 2008
and 2009 the U.S. government injected $235
billion of capital into the banking system as
part of the Troubled Asset Relief Program (TARP). In
2009, bank regulators carried out a full-scale evaluation
of the capital adequacy of 19 large banking organizations,
ultimately requiring 10 of these organizations to increase
their capital levels. While most commentators agree that
regulatory capital levels are too low for large organizations
— especially large organizations that create systemic risks
— financial economists have only recently been paying
attention to what factors actually govern banks’ capital
choices. In this article, Mitchell Berlin discusses how
understanding bank capital decisions over the 20-year
period prior to the recent crisis can provide insights that
may help us to evaluate reform proposals.
After posing the question, “Why
are banks so averse to raising equity?”
a recent column in The Economist an-
Mitchell Berlin
is a vice president
and economist
in the Research
Department of
the Philadelphia
Fed. He is the
head of the
department’s
Banking and
Financial Markets section. This article
is available free of charge at www.
philadelphiafed.org/research-and-data/
publications/.
www.philadelphiafed.org
swers, “The usual laws of corporate finance do not seem to apply to banks.”
The reason the column suggests is
that deposits are insured; so uninsured
sources of funding (such as equity) are
relatively expensive. This view is fairly
widespread, and not just among business columnists. Indeed, most theoretical models of the banking firm assume
that banks hold the minimum amount
of equity required by regulation.1
*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
While this view appears plausible,
it actually contradicts the evidence
of the last 20 years, which shows
that banks do not appear to hold the
minimum amount of equity required
by regulators. Furthermore, while
banks are typically highly leveraged
compared with most nonfinancial
firms, this doesn’t mean that similar
forces are not at work when banks and
nonfinancial firms choose their capital
levels. To the contrary, empirical work
by banking scholars supports the view
that market forces have been an important determinant of banks’ capital
decisions since the early 1990s.
Bank capital has been much in the
news during the recent financial crisis.
In 2008 and 2009 the U.S. government
injected $235 billion of capital into the
banking system as part of the Troubled
Asset Relief Program (TARP).2 And
in 2009, bank regulators performed
a full-scale evaluation of the capital
adequacy of 19 large banking organizations, ultimately requiring 10 of these
organizations to increase their capital
levels.3 While most commentators
agree that regulatory capital levels are
too low for large organizations — especially large organizations that create
systemic risks — financial economists
1
To be fair, theorists often assume that banks
hold the minimum capital level mainly as a matter of convenience when they are not primarily
concerned about the bank’s choice between
debt and equity.
2
This total includes capital injected into a range
of financial institutions, not all of which were
commercial banking organizations. See the
report from the Government Accountability
Office for more details about TARP.
3
See the Board of Governors’ two accounts of
the Supervisory Capital Assessment Program
(SCAP).
Business Review Q2 2011 1
have only recently been paying attention to what factors actually govern
banks’ capital choices. Understanding
bank capital decisions over the 20-year
period prior to the recent crisis can
provide insights that may help us to
evaluate reform proposals. (See Some
Bank Capital Reform Proposals.)
CAPITAL STRUCTURE IN
NONFINANCIAL FIRMS
While banks may be special along
a number of dimensions, in the first
instance, banks are firms. So to understand bank capital, a sensible starting
point is to take stock of our current
knowledge about capital structure
decisions by firms in general. First,
some terminology: We can think about
capital structure in a few equivalent
ways. Sometimes it is easiest to talk of
the firm’s leverage ratio, the value of the
firm’s debt divided by the value of its
total assets. Alternatively, we sometimes talk of its capital ratio, the value
of the firm’s equity (or, often in the
case of banks, some broader measure of
regulatory capital) divided by the value
of its assets.4
The Dynamic Tradeoff Model.
Capital structure has been an active
area of research in financial economics
for the last 50 years.5 Despite inevitable
differences of opinion among researchers, the current consensus is that the
4
Regulators use the term “leverage ratio” to
refer to the value of a bank’s tier 1 capital over
total assets. (See Bank Capital Regulation for a
definition of tier 1 capital and other regulatory
terminology.) Throughout the text, I will use
the term “capital ratio” to refer to common
equity divided by assets and I will specify whenever I use some regulatory measure of capital
or assets.
5
Most accounts of the modern theory of capital
structure begin with the capital structure irrelevance theorem of Nobel laureates Franco
Modigliani and Merton Miller, who showed
conditions under which a firm’s capital structure
does not affect its value. Subsequent researchers have systematically examined the effects of
relaxing these conditions.
2 Q2 2011 Business Review
Some Bank Capital Reform Proposals
I
n addition to the widespread view that banks should be
required to maintain higher capital levels than under Basel I,
banking researchers and policymakers have made a number of
proposals to reform bank capital regulation.
A number of researchers have proposed that banks be
required to maintain a layer of contingent convertible debt. The element
common to all versions of this proposal is that when bank capital falls below
some level, the debt converts to equity, thereby reducing the bank’s leverage
automatically. Proposals differ in the details of how conversion is triggered. For
example, in Mark Flannery’s proposal, conversion is triggered when the market
value of equity falls below a predetermined level. Alternatively, the Squam Lake
Working Group for Financial Institutions proposes that conversion should be
triggered only when both the book value of equity falls below a predetermined
level and bank regulators announce that there is a systemic crisis.*
Other researchers have proposed that banks be assessed a higher capital
charge based on some measure of their contribution to systemic risk. This
approach seeks to address the issue that banks will not take into account
the costs they impose on other institutions, and ultimately taxpayers, when
they take risks that increase systemic risk. For example, Viral Acharya, Lasse
Pedersen, Thomas Philippon, and Matthew Richardson have proposed that
bank capital requirements (or a systemic risk insurance fee) be partially based
on a financial institution’s contributions to episodes of severe stock market
declines. Other researchers have proposed other measures of an institution’s
contribution to systemic risk; for example, Tobias Adrian and Markus
Brunnermeier propose that capital charges be based on the covariance between
an institution’s stock price and those of other large financial institutions.
It is important to note that contingent capital schemes and schemes
that impose capital charges for systemic risk are potentially complementary
approaches.
* The various proposals contain extended discussions of the main issues in dispute. Flannery
views his scheme more as a means of mobilizing market discipline and early regulatory intervention than as a mechanism for recapitalizing a financial system already in serious crisis. The
Squam Lake group worries that conversion triggers based on the market price of equity will lead
to market manipulation that would increase instability. It views conversion primarily as a means
of recapitalizing institutions once the system is already in crisis.
empirical evidence is consistent with a
dynamic tradeoff model in which firms
choose a target leverage ratio to which
they actively adjust over some period
of time. Furthermore, alternative views
in which firm managers make financing decisions with little or no thought
to hitting a target leverage ratio have
received little empirical support to
date. But even its proponents recognize
that the standard model has limited
power to explain firm capital structure
decisions.6
In the standard model, a firm
6
See, for example, two recent reviews of the
capital structure literature by Christopher Parsons and Sheridan Titman and by Murray Frank
and Vidhan Goyal.
www.philadelphiafed.org
chooses its target leverage to balance
the benefits and costs of increasing
its debt level. Much of the literature
has focused on the deductibility of
interest payments as the primary
benefit of higher debt: A firm’s interest
payments to bondholders and other
lenders are treated by the firm as an
expense and, thus, lower the firm’s tax
bill. In contrast, dividend payments
to the firm’s stockholders are not
deductible. If this were the whole
story, firms would choose to be fully
debt financed. But debt also generates
costs. A highly levered firm with a
lot of interest payments can get into
trouble in difficult financial times. At
the minimum, a firm may be forced
to postpone investment projects and
use all incoming cash to meet interest
payments. At the worst, a firm might
actually face default and bankruptcy if
it can’t pay its creditors. (In contrast,
postponing or cutting dividend
payments do not lead to default.)
These costs are usually grouped under
the term costs of financial distress.
Factors That Reliably Affect
Leverage. Empirical studies that
cover different time periods, samples
of firms, and countries indicate that
a firm’s leverage tends to be higher
when a firm is larger, when it has
more tangible assets, and when its
market-to-book ratio — the value of
the firm’s stock divided by the book
value of its assets — is lower. Most
researchers interpret these factors as
evidence that concerns about financial
distress play an important role in the
firm’s capital structure choice. Large
firms have more diversified sources
of cash, and thus, they are less likely
to face a sudden cash shortfall. A
firm’s tangible assets include machines
and inventories, assets that could
potentially be sold much more easily
than a firm’s intangible assets: its
trademarks, its reputation for quality,
brand recognition, or the accumulated
www.philadelphiafed.org
knowledge of its workforce. In the
event of a decline in cash flows, a firm
may be able to avoid default by selling
some of its tangible assets. The marketto-book ratio is often interpreted
as a measure of the firm’s growth
opportunities, for example, future
investment activities that investors
see as valuable — and, thus, raise the
firm’s stock price — but which are
not yet embodied in assets in place.
When a firm has valuable growth
valuable growth
opportunities, it may be particularly
costly when declines in cash flow force
it to delay new investments.7
In addition, researchers have
found that a firm’s leverage depends
importantly on its industry and that its
leverage is high when the firm’s own
profitability is low. These factors don’t
fit as comfortably into the tradeoff
model. The importance of industry
effects simply shifts the inquiry one
step further back: What is it about
7
The reader may note that none of the enumerated factors are clearly related to the tax
benefits of debt. Until John Graham’s work,
the consensus view was that taxes had limited
ability to explain firms’ leverage decisions.
Recent dynamic models have uncovered more
evidence for the importance of taxes, but
research continues to suggest that firms do not
take on as much debt as models would predict.
See Graham’s article, as well as the literature
reviews cited in footnote 6 for fuller discussions
of taxes and capital structure.
an industry that explains high or low
leverage? And while the negative
relationship between profits and
leverage can be squared with some
versions of the tradeoff model, the
effect is probably best viewed as an
unexplained empirical regularity.8
Firms Actively Adjust Toward a
Target. While firms may have a target
leverage ratio, factors often shift a firm
away from its target; for example, a
sudden increase in sales might increase
retained earnings, thereby reducing the
firm’s actual leverage ratio. Since new
debt issuance is costly, the firm may
take some time to get back to its target.
In surveys of chief executive officers
and chief financial officers, over 70
percent of the firms report that they
have either a strict target or a target
range for their leverage ratio.9 This
survey evidence is supported by formal
empirical studies, but researchers
report widely disparate estimates of
the speed with which firms adjust,
with estimates ranging from very slow
(Eugene Fama and Kenneth French
estimate that firms adjust at a rate of
7 to 10 percent per year) to very fast
(Mark Flannery and Kasturi Rangan
(2006) estimate an adjustment rate of
34 percent per year), and researchers
are far from achieving consensus.
Furthermore, studies disagree as to
whether the target is fixed or whether
it may vary over time in a systematic
way.
Most economists would agree
with the statement, “It takes a
model to beat a model.” This means
that to evaluate a particular model,
researchers compare it to alternative
models, mainly by asking how well
8
This negative relationship is consistent with
Stewart Myers’s pecking order model, examined
in the next section.
9
See John Graham and Campbell Harvey’s
article.
Business Review Q2 2011 3
each explains the facts, in this case,
firms’ capital structure choices.10 To
date, no alternative to the dynamic
tradeoff model has found strong
empirical support. In particular,
researchers have found only limited
support for alternative models that
predict no target leverage ratio. The
most influential of these is Stewart
Myers’s pecking-order model, in which
firms finance investments out of cash
whenever possible, sell debt only if cash
flows are too low, and sell new equity
only as a last resort. According to this
view, a firm’s leverage ratio increases
when its cash flows drop and it is
compelled to sell new debt to finance
expenditures, and its leverage ratio
declines when cash flows increase and
internal funds build up. In contrast to
the assumption of tradeoff models, a
firm manager in a pecking-order type
world will make no attempt to actively
adjust toward some target.11
Limits of the Dynamic Tradeoff
Model. The empirical importance of
industry effects and of other variables
that might be interpreted in ways
that have little to do with a tradeoff
between tax savings and the costs of
financial distress, for example, firm
size, firm profitability, or market-tobook value, limits our confidence
in the dynamic tradeoff model.
Furthermore, in an important recent
paper, Michael Lemmon, Michael
Roberts, and Jaime Zender highlight
the limited explanatory power of the
10
Of course, it is possible for different models
to help explain different aspects of a firm’s
decision-making or for one model to explain
decision-making by some types of firms, for
example, large firms, but not others.
11
See Frank and Goyal’s review article for
further discussion of the empirical evidence for
and against the pecking-order model and other
models that predict no target, for example,
Malcolm Baker and Jeffrey Wurgler’s view that
managers’ decisions to issue securities are driven
by attempts to time the market.
4 Q2 2011 Business Review
model. They find that, even including
industry effects, the traditional model
explains at most 30 percent of the
variation in firms’ capital structures;
an economist would say that the model
has limited power to explain the data.
Perhaps more important, Lemmon
and his co-authors find that firm fixed
effects have a lot more explanatory
power than all of the traditional
factors put together. A fixed effect is
a persistent factor associated with a
particular firm: We know it’s there,
and we know that it helps explain
the firm’s choice of capital structure;
we just don’t know what it is. This
useful as a guide to understanding or
prediction.12
BANK CAPITAL STRUCTURE
Bank Capital Levels Over Time.
Banks are highly levered firms. In
Reint Gropp and Florian Heider’s
international sample of large banks
in 2004, median leverage was nearly
93 percent in book value terms and
just over 87 percent when measured
in market value terms. Compare this
with the median book and market
leverage of Frank and Goyal’s sample
of nonfinancial firms in 2004 of 24
percent and 23 percent, respectively.13
! " # $
$
$ $
finding is a challenge for the tradeoff
theory because it suggests that much
of the variation in firms’ leverage is
potentially explicable by some model of
firm decision-making, just not the one
we have.
The controversy over the speed of
adjustment toward the target and the
stability of the target presents further
challenges for the theory. The model
is less persuasive when the speed of
adjustment is slow; a firm that adjusts
to its target over a period of 10 or 15
years begins to look more and more
like a firm with no target at all. And
the problem with time-varying targets
is much like the problem with firm
fixed effects and industry effects. A
theory that depends on factors (firm,
industry, time) that help “explain” a
firm’s leverage ratio in the statistical
sense, but without any underlying
economic intuition, may not be very
Bank capital levels have not
always been so low. In the U.S., commercial banks had equity-to-asset
ratios (measured at book value) of over
50 percent in 1840.14 This ratio fell
continuously until 1945, at which point
it remained roughly stable in the 6 to 8
12
Other recent challenges to the dynamic
tradeoff model are even more fundamental. For
example, Xin Chang and Sudipto Dasgupta
show that simulations with random stock and
bond selling can generate dynamic capital
structures that look a lot like a firm moving
toward a target.
13
That said, banks are not unique in maintaining high leverage ratios. For example, in Ivo
Welch’s listing of the 30 most highly levered
firms in February 2006, only 11 were financial
firms and none were commercial banks.
14
The numbers prior to 1980 come from the
article by Allen Berger, Richard Herring, and
Giorgio Szegö. Note that the numbers are not
strictly comparable over time and so should be
viewed as an indicator of trends.
www.philadelphiafed.org
percent range until the 1970s. Examining the figure at the bottom of the
page, we see that the weighted average
book value equity ratios for bank holding companies (BHCs) had declined
to the 4 to 6 percent range by 1980
and then rose to 6 percent in the latter
half of the 1980s, mainly in response
to the imposition of uniform capital
guidelines in 1985.15 (See Bank Capital
Regulations for a summary description
of U.S. bank capital regulation and for
definitions of all terms.)
Bank capital ratios increased
dramatically after 1990, when Basel I
capital requirements were first imposed. Book equity-to-asset ratios for
large BHCs rose from approximately 6
percent in the late 1980s to over 8 percent in the 1990s and 9 percent until
the financial crisis of 2008. The rising
trend since 1990 is even more striking
in market value terms. The average
market value of bank equity to the
market value of assets for the largest
100 BHCs rose from 6 percent in 1990
to over 15 percent from 1996 through
the second half of 2007.16
Banks Hold More Capital Than
the Regulatory Minimum. The rise in
bank capital ratios since 1990 also corresponded to an increase in regulatory
capital ratios. For their sample of large
BHCs, Flannery and Rangan (2008)
find that risk-weighted tier 1 capital
ratios rose from under 8 percent in
1986 to over 10 percent by 1995. This
ratio showed a declining trend through
2006 but remained above 8 percent
throughout the period, comfortably
above the 6 percent level required for a
bank to be considered well-capitalized
for regulatory purposes and well above
the regulatory minimum of 4 percent.17
Examining the entire distribution
of large BHCs’ regulatory capital ratios,
Flannery and Rangan (2008) show
that by 1992 more than 95 percent of
large BHCs had tier 1 capital ratios
at least 1.5 percentage points higher
17
Interestingly, this decline coincided with an
increase in tier 2 capital. Two trends appear to
be at work: first, a shift toward riskier assets,
and second, a shift toward nonequity sources
of regulatory capital. This raises a range of
important (and difficult) issues about the appropriate way to measure capital adequacy. To
the extent that the risk weights on off-balancesheet assets (or other assets) were too low — for
example, regulation may have underestimated
BHCs’ commitment to support off-balance-sheet
vehicles — BHCs may not have been as well
capitalized as they appeared in the early 2000s.
than the regulatory minimum. This
percentage rises to 100 percent for
most years through 2001. Berger and
his co-authors (2008) examine a larger
sample of BHCs and show that this
trend continued through 2006.18 They
show that 99 percent of large BHCs
had tier 1 capital ratios that qualified
them as well capitalized in 2006. The
lion’s share of these firms had tier 1
risk-weighted capital ratios between 10
and 12 percent.
Banks Actively Manage Toward
a Target. It is clear that throughout
the 1990s and into the 2000s, banks
overwhelmingly held capital levels
greater than the regulatory minimum,
but this raises a question: What factors
determine banks’ capital levels? One
possibility is that the bank capital
18
Flannery and Rangan’s (2008) sample includes
the largest 100 BHCs in each year, while Berger
and his co-authors (2008) use a larger sample of
all BHCs with assets in excess of $150 million.
FIGURE
BHC Capital Ratios*
Percent
14
12
Unweighted Mean of All BHCs
10
15
A bank holding company is any company
that controls one or more commercial banks.
The figure displays bank capital ratios both for
the largest 100 BHCs and for a larger group of
BHCs. The figure also displays unweighted average capital ratios to show that the main trends
are not driven by a small number of very large
banks.
8
Weighted Mean of All BHCs
6
Weighted Mean of Top 100 BHCs
4
2
16
Flannery and Rangan (2008) show that the
increase in the average capital ratio corresponds
to a rightward shift in the entire distribution
of market values of equity from the 1986-1989
period to the 1998-2001 period. The 2008
article by Berger and co-authors suggests that
this distribution continued to shift rightward
through 2006, although they focus on regulatory capital.
www.philadelphiafed.org
Unweighted Mean of
Top 100 BHCs
0
1980
Q1
1983
Q1
1986
Q1
1989
Q1
1992
Q1
1995
Q1
1998
Q1
2001
Q1
2004
Q1
2007
Q1
*
The capital ratio is measured as the book value of common equity over total assets.
Source: Bank Reports of Condition. “All BHCs” refers to BHCs with assets greater than $150 million.
Business Review Q2 2011 5
Bank Capital Regulation
P
rior to the 1980s, bank regulators had no formal uniform capital requirements, although regulators
evaluated banks’ capital levels as a part of their regulatory review. In 1985, U.S. bank regulators imposed
uniform requirements, largely in response to concerns about the secular decline in bank capital. Banks
were required to maintain at least a 5.5 percent primary capital ratio — equity plus loan-loss reserves/
total assets — and a 6.0 percent secondary capital ratio — primary capital plus various subordinated debt
instruments/total assets.
The Basel Accord of 1988 first imposed binding capital requirements in 1990, although these were phased in
over the next two years. The goals of the Basel Accord were to: (i) raise capital levels for most banks; (ii) increase
international uniformity in regulatory capital standards; (iii) adjust capital requirements to better reflect actual credit
risk; and (iv) impose capital requirements for some off-balance-sheet exposures. The following provides the basic
elements of Basel I capital requirements. (See, for example, Anthony Saunders and Marcia Millon Cornett’s textbook
for a more complete treatment.) European (but not U.S.) banks have been subject to Basel II capital requirements since
2008.
Tier 1 capital = Common equity + Preferred noncumulative stock + Minority interests in consolidated
subsidiaries.a
Tier 2 capital = Tier 1 capital + Allowances for loan losses + Perpetual preferred stock + Subordinated debt +
Various hybrid capital instruments.b
Note: The amounts of some of the components of tier 1 and tier 2 capital are limited to some maximum value. For
example, preferred noncumulative stock can be no more than 25 percent of tier 1 capital.
Risk-weighted assets: Each asset has a risk weight, reflecting the risk of default. For example, a Treasury security
carries a zero risk weight, while a commercial loan carries a 100 percent risk weight. In addition, off-balance-sheet
assets, such as commitments to lend, are assigned a conversion factor. For example, an unused two-year loan commitment
increases on-balance-sheet assets 50 cents for each dollar of the commitment; that is, the conversion factor is 0.5. Total
risk-weighted assets are the sum of all assets, with each asset weighted by its risk weight.
Each BHC, each bank within a BHC, or any stand-alone bank is subject to three basic capital requirements:
Leverage requirement: Tier 1 capital/Total assets must exceed 4 percent.
Tier 1 capital requirement: Tier 1 capital/Total risk-weighted assets must exceed 4 percent.
Total capital requirement: Tier 2 capital/Total risk-weighted assets must exceed 8 percent.
BHCs that wish to engage in international activities and pay lower deposit insurance premiums, among other
benefits, must be well capitalized. To be well capitalized, the BHC must maintain a tier 1 capital ratio no less than 5
percent, a tier 1 risk-based capital ratio no less than 6 percent, and a tier 2, risk-based capital ratio no less than 8 percent.
a
Preferred stock confers no voting rights and pays a fixed dividend. Dividend payments on preferred stock must be paid before common stockholders
are paid any dividends, but contractual payments to debt holders have priority over preferred dividends. Unlike a missed interest payment, a missed
dividend payment is not an event of default. Unlike for cumulative preferred stock, missed dividend payments on noncumulative preferred stock are
not added to future dividend payments. When a BHC owns a majority of the shares of a subsidiary, the subsidiary is consolidated into the balance
sheet of the parent BHC. If the BHC owns less than 100 percent of the shares, the equity share is considered a minority interest in its consolidated
subsidiary.
b
Perpetual preferred stock has no fixed maturity and any missed dividend payments are added to future dividend payments. The interest payments
on subordinated debt instruments are contractual payments that must be paid before any stockholders receive dividend payments. Failure to make
interest payments leads to default. Subordinated debt has lower priority than deposits or senior debt, so depositors (or the FDIC standing in for
depositors) or senior debt holders must be fully paid off before subordinated debt holders receive any payments. Hybrid capital instruments included
in tier 2 capital refer to a range of securities, including deeply subordinated debt instruments. These have lower priority than ordinary subordinated
debt and make interest payments only under specified contractual conditions.
buildup reflected pecking-order behavior and that the capital buildup was
an accidental byproduct of the strong
6 Q2 2011 Business Review
revenue growth for banks during this
period. This behavior might have
been reinforced by regulators’ prefer-
ence to see better capitalized banks.
The evidence strongly suggests
that this is not the case. Beginning
www.philadelphiafed.org
with a study of the 1980s, Berger and
co-authors, in their 2008 paper, find
that banks sold new equity when
their earnings increased, a finding
at odds with pecking-order behavior.
Examining the capital buildup of the
1990s, Berger and co-authors find that
BHCs systematically offset new equity
issues carried out to finance mergers by redeeming existing shares, also
consistent with active management of
their capital ratios. Furthermore, Flannery and Rangan (2008) estimate an
empirical model of bank market capital
ratios for the 1990s and conclude that
the mechanical effect of increases in
earnings accounts for only 3 percent of
the capital buildup in the 1990s. So, as
in the literature for nonfinancial firms,
researchers do not find much support for pecking-order models of bank
capital.
What Do We Know About
Banks’ Target Leverage? It is important to note that the literature on
what determines banks’ target leverage
ratios is relatively small, the samples
and model specifications are different,
and not all findings are consistent;
so all results should be regarded as
preliminary.19 I focus primarily on
those results that are consistent across
studies and that pertain to leverage
ratios or capital ratios (common equity/
assets) measured at market prices.20
Consistent with the literature on
nonfinancial firms — and also with
many other studies in the banking
literature — all researchers find a
positive relationship between banks’
asset size and target leverage. That is,
larger banks are less well capitalized.
This finding is consistent with the
view that larger banks are better
diversified and less likely to breach
their target leverage.21 Also in line
with the previous capital structure
literature, researchers find that most
of their models’ explanatory power
comes from a firm-specific fixed effect,
again, a reflection of our limited
understanding of the cross-sectional
%
&
'$ &
' # ( $ $
variation in bank capital structure
choices. Existing research also agrees
that banks adjust quickly toward
their target; indeed, the adjustment
speeds exceed the top end of the range
previously estimated for nonfinancial
firms.22
Finally, the studies by Gropp
and Heider and by Flannery and
Rangan (2008) document a negative
19
I focus on the results of Berger and co-authors
(2008), Flannery and Rangan (2008), and
Gropp and Heider, all of which cover sample
periods through at least 2000. These articles
contain references to earlier contributions
that address similar questions for earlier time
periods.
20
The leverage ratio is comparable to the measure typically used in studies of nonfinancial
firms. Furthermore, regulatory definitions of
capital pose difficult questions about the quality
of the capital, for example, whether the instruments included in capital should be thought of
as equity or debt. And risk-weighted measures
of assets raise a host of questions about whether
the risk weightings are reasonable.
www.philadelphiafed.org
relationship between bank leverage
ratios and a measure of bank asset risk,
although Flannery and Rangan (2008)
find this result only for the second
half of their sample period, 1994-2000.
This result is consistent with the view
that bank leverage decisions are driven
by market pressures; that is, investors
or other bank counterparties demand
that a bank with more portfolio risk
be better capitalized.23 The view that
market pressures increased in the late
1990s is in tune with other empirical
research showing that the costs of
uninsured funding sources became
more risk sensitive in the 1990s.
Interestingly, Flannery and Rangan
(2008) find no such relationships for
the 20 largest U.S. banks. They argue
that market participants view the
largest banks as too big to fail and
that this suppresses the relationship
between risk and leverage.
Researchers have tried to
distinguish between two possible
types of explanations to explain
variations in capital levels over time
and across banks. The first possibility
21
It is also consistent with the view that larger
banks were undercapitalized, in particular, that
their capital provisions were too low given the
probability of very bad economic outcomes,
so-called tail risk.
22
Interestingly, Flannery and Rangan (2008)
find that adjustment speeds are faster for banks
nearer their minimum capital requirement and
that banks with poor regulatory ratings adjust
relatively slowly. They interpret the latter result
as evidence of the difficulties such banks face in
selling new equity.
23
Flannery and Rangan (2008) do not find a
negative relationship between leverage and
risk for the first half of their sample period,
1987-1994. Their interpretation is that market
forces became more important in determining
bank capital structure decisions throughout the
sample period. We should be cautious in our interpretation of a negative relationship between
asset risk and bank leverage. Better capitalized
banks may simply choose to take fewer risks,
perhaps reflecting the risk preferences of the
owners or managers.
Business Review Q2 2011 7
is that regulatory capital requirements
actually determine bank capital but
that banks hold some cushion above
the required capital level to reduce the
likelihood of a regulatory intervention
or the need to raise capital or reduce
assets at short notice. The second
possibility is that bank capital levels
are determined in the market,
perhaps according to some tradeoff
model similar to the model in the
standard capital structure literature.
(Indeed, Gropp and Heider estimate
a canonical tradeoff model, with only
small alterations to account for certain
distinctive characteristics of banking
firms.)
To this point, researchers have not
found a way to persuasively distinguish
these hypotheses in the data, although,
in my view, Flannery and Rangan
(2008) present the most convincing
evidence against the equity cushion
view. They show that bank asset
volatility is not positively related to the
excess of book capital over required
capital (the cushion), inconsistent with
the view that the cushion is chosen
to protect the bank against the risk of
poor outcomes that would breach the
regulatory capital requirement.24
THEORIES OF BANK CAPITAL
STRUCTURE
Although there is a large
theoretical literature on what makes
banks special, a surprisingly small
number of banking theorists have
addressed banks’ capital structure
decisions. While the empirical
evidence doesn’t yet firmly reject the
view that banks hold the regulatory
minimum plus some cushion, the high
capital levels of the last 20 years have
24
This is consistent with the results of Berger
and co-authors (2008), who do not find any
relationship between earnings volatility and
book leverage or any other measures of regulatory capital.
8 Q2 2011 Business Review
led some theorists to explore optimal
capital decisions driven by market
pressures, in the context of the modern
theory of the banking firm.25
Banks Hold Illiquid Assets and
Provide Liquid Liabilities. The high
leverage we observe for banks is closely
related to what makes banks special.
First, unlike those of nonfinancial
firms, banks’ liabilities are used as
money (for example, demand deposits)
and as a safe store of savings that
can be called on at short notice (for
example, certificates of deposit). More
recently, other types of bank liabilities,
for example, asset-backed securities,
While a diversified portfolio of loans is
less risky than any single loan, a bank
must monitor its loans to ensure that
portfolio returns are adequate to pay
off the bank’s depositors and other
creditors. Besides the view that bank
capital is determined by regulatory
requirements, there are (broadly) two
different views of the role of bank capital, both of which revolve around the
view of banks as specialists in monitoring borrowers. But the underlying
mechanisms are quite different.
Bank Capital Promotes Monitoring. In a number of models, the
banker’s incentive to monitor borrow-
' $ '
)
'#
have served as collateral for a host of
financial transactions.26 Since liquid
liabilities are a primary output of
the banking firm, we should expect
banks to be highly levered. At the
same time, to be useful in exchange
or as a source of liquid savings, banks’
liabilities need to have little risk of
default and, even more important,
should not require customers to carry
out a careful evaluation of the bank’s
assets. (Imagine having to examine
a bank’s annual report each time you
accept a check drawn on that bank.)
Meanwhile, bank assets are risky.
25
Samu Puera and Jussi Keppo’s article presents
a formal model in which a bank holds an equity
cushion above its regulatory capital requirement. The size of the cushion reflects the bank’s
costs of securing funds from outside investors in
the event that it suffers losses.
26
See Gary Gorton’s account of securitization
and the repo market.
ers depends on stockholders’ equity investment. In particular, recent articles
by Franklin Allen, Elena Carletti, and
Robert Marquez, and by Hamid Mehran and Anjan Thakor use this idea to
explain why banks would hold capital
in excess of regulatory requirements.
In these models the banker acts in the
interests of the bank’s stockholders,
perhaps because he or she has substantial stockholdings or because his or her
pay is tied to the bank’s stock price.
Although the models differ in many
significant ways, they share a similar
basic intuition: Stockholders gain
only when profits are positive, that is,
when enough loans are repaid to cover
the bank’s debt payments. The more
equity invested by stockholders, and
thus the lower the bank’s leverage, the
smaller the share of the loan revenues
that must be paid out to debt holders
when revenues exceed debt payments.
Thus, the gains from increasing the
www.philadelphiafed.org
likelihood of successful loans through
monitoring are greater when the equity
investment is greater.
This is only half of the story
because it doesn’t explain the limits
on the bank’s equity. In both models,
the authors simply assume that equity
is a relatively costly means of funding
loans, mainly to focus attention on
the relationship between monitoring
and leverage. Among other factors, a
higher relative cost for equity funding
might arise if (i) deposits are insured;
(ii) insiders have more information about the quality of the bank’s
portfolio than potential outside equity
investors; or (iii) we take into account
the value of producing bank liabilities
that facilitate exchange.
Each of the models contains
interesting empirical predictions. In
particular, Allen and his co-authors
show theoretically that banks will hold
more capital when they lend in more
competitive markets. This prediction
illustrates an important feature of their
model: market discipline is imposed by
borrowers, rather than capital markets.
Intuitively, borrowers gain when they
are monitored more closely by banks,
and banks’ incentives to monitor are
stronger both when bank capital is
higher and when borrowers pay higher
loan rates. Everything else equal,
borrowers prefer that banks charge
lower loan rates; so when loan market
competition is strong, banks compete
for borrowers by lowering rates and
holding more capital. When competition is weak, banks can charge higher
loan rates and hold less capital without
undermining their commitment to
monitor. This prediction has yet to be
tested empirically.
Mehran and Thakor’s paper has
a host of empirical predictions, most
notably the prediction that bank
equity capital and bank value will be
positively related in the cross-section.
Intuitively, a bank with a low cost of
www.philadelphiafed.org
capital has a comparative advantage in
monitoring borrowers, and a bank that
monitors more will have a higher value. In the cross-section, Mehran and
Thakor argue that we should observe
that banks with more equity capital
will also be more valuable. They find
support for this hypothesis in their
empirical analysis of merger deals in
the U.S. between 1989 and 2007.27
Deposits Promote Monitoring; Bank Capital Reduces Bank
Failures. Douglas Diamond and
Raghuram Rajan present a model
in which bankers are hired by the
bank’s suppliers of funds, for example,
depositors or stockholders, to monitor
borrowers. In their model, bankers seek
to grab as large a share of borrowers’
payments as they can; that is, bankers
don’t automatically share a common
interest with any of the bank’s claimants, either its borrowers or its suppliers of funds. If there were a single
banker and a single depositor, the
banker would threaten to withdraw
his expertise and knowledge about the
borrower, that is, to stop monitoring
and force the depositor to accept lower
interest payments. Since the loan is
much less valuable without the banker,
the banker can use his or her threat
to walk away to capture a significant
share of the firm’s loan payments at the
expense of the depositor.28
But things are different if there
27
Using goodwill — the difference between the
purchase price of a bank and its book value of
assets — as a proxy for the value of the bank’s
continuing relationships with its borrowers,
Mehran and Thakor also predict (and find
empirical support for) a positive relationship
between equity capital and goodwill. This result
is consistent with Flannery and Rangan’s (2008)
and Gropp and Heider’s finding that a BHC’s
leverage is lower when its market-to-book ratio
is higher.
28
The reader who finds this story too melodramatic should view it as a metaphorical way of
modeling the very realistic conflicts between
managers and claimants that can’t be easily
resolved through incentive contracts.
are lots of depositors. Diamond and
Rajan argue that, in this case, the
deposit contract has a strong disciplinary effect if, when multiple depositors
withdraw funds at once, a run on
the bank develops.29 Faced with the
threat of a depositor run, the banker
will choose to monitor borrowers (or
else the loan will not pay off) and will
make promised payments to depositors.
Deposits are hard claims that impose
discipline on bankers.
If a hard-working banker could always pay off his or her depositors, Diamond and Rajan’s model would predict
that banks could be fully funded by
deposits. The threat of a run would
impose discipline, but the threat would
never actually be carried out. But bank
loans can go bad for reasons other
than poor monitoring or an attempt
by the banker to keep loan revenues,
for example, an economic downturn.
In this case, the banker may be unable
to pay off depositors, depositors will
run, and many loans will have to be
liquidated inefficiently.
This is where bank capital comes
in. Bank capital serves as a buffer in
the event of a decline in loan revenues.
Equity is a soft claim. In the event
that depositors withdraw their funds,
stockholders take a loss to ensure that
all depositors can be paid off and fewer
loans have to be liquidated.30 But this
creates a tradeoff: The better capitalized the bank, that is, the more heavily
29
For a run to develop, the deposit contract
must require the bank to pay off depositors who
want to withdraw their funds on a first-come,
first-served basis. In the banking literature this
is called a sequential service constraint.
30
Actually, in the model, bank capital might
also take the form of long-term subordinated
debt. It is important that the depositors have
priority, but in the model, there is no real
distinction between equity and long-term subordinated debt. Thus, in Diamond and Rajan’s
model, market forces would affect regulatory
capital, not just equity.
Business Review Q2 2011 9
the bank is financed by soft claims,
the weaker the discipline imposed on
the banker. Since it reduces the threat
of a run, bank capital ensures that the
banker captures a larger share of the
bank’s profits.
While Diamond and Rajan’s
model has been quite influential —
increasingly so, since the financial
crisis reminded banking scholars that
banks might actually fail — there has
been no systematic attempt to test
whether it helps explain variations in
bank capital over time or in the crosssection.31
CONCLUSION
While the experience of the 1990s
and 2000s is inconsistent with the
view that banks hold only the minimum required amount of equity, it is
31
Mehran and Thakor argue that Diamond and
Rajan’s model counterfactually predicts a negative relationship between a bank’s value and its
capital level in the cross-section.
10 Q2 2011 Business Review
difficult to address The Economist’s
claim that the usual laws of corporate
finance do not apply to banks. Over
50 years of theoretical and empirical
research into nonfinancial firms’ leverage decisions has identified factors that
are consistently related to leverage,
but one would be hard pressed to say
that we have a firm understanding of
the usual laws of corporate finance.
Empirically, too much of the variation
in nonfinancial firms’ capital structures is explained by dummy variables
representing the firm’s industry and
the firm itself. While this is better
than no explanation at all, it is more
an invitation to further research than
a settled set of laws.
Furthermore, while banking researchers have rejected the simple view
that capital requirements are binding,
they have only begun to explore the
determinants of bank leverage decisions empirically or theoretically. For
example, the banking literature has yet
to establish convincingly whether bank
capital decisions are determined by
market pressures — perhaps including pressures from borrowers as well as
investors — or whether they are best
explained as banks meeting regulatory
requirements while holding an extra
equity cushion.
While these issues do not directly
answer the pressing question of how
much capital banks should hold, they
are directly relevant to the inquiry. In
particular, capital requirements are
much more difficult to enforce when
they are binding; if banks wish to hold
less than the regulatory minimum (or
the minimum plus a cushion), they
have a strong incentive to evade these
requirements through a variety of
strategies. This incentive increases as
the difference between the regulatory
requirement and the desired level of
capital increases. Understanding the
extent to which market forces are
working with or against a new capital
regulation should help policymakers
understand the costs of enforcement. BR
www.philadelphiafed.org
REFERENCES
Acharya, Viral, Lasse Pedersen, Thomas
Philippon, and Matthew Richardson.
“Regulating Systemic Risk,” Chapter 13
in Viral Acharya and Matthew Robinson,
eds., Repairing the U.S. Financial Architecture. Hoboken: John Wiley and Sons, 2009.
Adrian, Tobias, and Markus Brunnermeier.
“CoVaR,” Working Paper, Federal Reserve
Bank of NY Staff Reports 348 (September
2008).
Chang, Xin, and Sudipto Dasgupta.
“Target Behavior and Financing: How
Conclusive Is the Evidence?” Journal of
Finance, 64 (2009), pp. 1767-96.
Diamond, Douglas, and Raghuram Rajan.
“A Theory of Bank Capital,” Journal of
Finance, 55 (2000), pp. 2431-65.
Economist, The. “Economics Focus: Buffer
Warren,” October 31, 2009, p. 88.
Graham, John, and Campbell Harvey.
“The Theory and Practice of Corporate
Finance: Evidence from the Field,” Journal
of Financial Economics, 61 (2001), pp. 1-52.
Gropp, Reint, and Florian Heider. “The
Determinants of Bank Capital Structure,”
European Central Bank Working Paper
(September 2009).
Fama, Eugene, and Kenneth French. “Testing Trade-off and Pecking Order Predictions About Dividends and Debt,” Review
of Financial Studies, 15 (2002), pp. 1-33.
Lemmon, Michael, Michael Roberts,
and Jaime Zender. “Back to the Beginning: Persistence and the Cross-Section
of Corporate Capital Structure,” Journal of
Finance, 63 (2008), pp. 1575-1608.
Baker, Malcolm, and Jeffrey Wurgler.
“Market Timing and Capital Structure,”
Journal of Finance, 57, (2002) pp. 1-30.
Flannery, Mark. “Stabilizing Large Financial Institutions with Contingent Capital
Certificates,” Working Paper, University of
Florida (October 2009).
Mehran, Hamid, and Anjan Thakor.
“Bank Capital and Value in the CrossSection,” Review of Financial Studies, 24:4
(2011), pp. 1019-67.
Berger, Allen. “The Relationship Between
Capital and Earnings in Banking,” Journal
of Money, Credit and Banking, 27 (1995),
pp. 432-56.
Flannery, Mark, and Kasturi Rangan.
“Partial Adjustment and Target Capital
Structures,” Journal of Financial Economics,
79 (2006), pp. 469-506.
Modigliani, Franco, and Merton Miller.
“The Cost of Capital, Corporation
Finance, and the Theory of Investment,”
American Economic Review, 48 (1958), pp.
261-97.
Berger, Allen, Richard Herring, and
Giorgio Szegö. “The Role of Capital in
Financial Institutions,” Journal of Banking
and Finance, 19 (1995), pp. 393-430.
Flannery, Mark, and Kasturi Rangan.
“What Caused the Bank Capital Buildup of the 1990s?” Review of Finance, 12
(2008), pp. 391-429.
Berger, Allen, Robert DeYoung, Mark
Flannery, David Lee, and Ozde Oztekin.
“How Do Large Banking Organizations
Manage Their Capital Ratios?” Journal of
Financial Services Research, 34 (2008), pp.
123-49.
Frank, Murray, and Vidhan Goyal. “Capital Structure Decisions: Which Factors
Are Reliably Important?” Financial Management, forthcoming.
Allen, Franklin, Elena Carletti, and Robert
Marquez. “Credit Market Competition and
Capital Regulation,” Review of Financial
Studies, 24:4 (2011), pp. 983-1018.
Board of Governors of the Federal Reserve
System. “The Supervisory Capital Assessment Program: Design and Implementation,” April 2009.
Board of Governors of the Federal Reserve
System. “The Supervisory Capital Assessment Program: Overview of Results,” May
2009.
Brealey, Richard, and Stewart Myers. Principles of Corporate Finance, 7th ed., Chapter
18. Burr Ridge, IL: McGraw-Hill.
www.philadelphiafed.org
Frank, Murray, and Vidhan Goyal. “Tradeoff and Pecking Order Theories of Debt,”
in Espen Eckbo, ed., The Handbook of
Empirical Corporate Finance. Elsevier, 2008.
Gorton, Gary. “Slapped in the Face by
the Invisible Hand,” Working Paper, Yale
School of Management (2009).
Government Accountability Office.
“Troubled Asset Relief Program: Status of
Efforts to Address Transparency and Accountability Issues,” January 2009.
Parsons, Christopher, and Sheridan Titman. “Empirical Capital Structure: A Review,” Foundations and Trends in Finance, 3
(2008), pp. 1-93.
Puera, Samu, and Jussi Keppo. “Optimal
Bank Capital with Costly Recapitalization,” Journal of Business, 79 (2006), pp.
2163-2201.
Saunders, Anthony, and Marcia Millon
Cornett. Financial Institutions Management:
A Risk Management Approach, 6th edition.
McGraw Hill, 2007.
Squam Lake Working Group on Capital
Regulation. “Reforming Capital Requirements for Financial Institutions,” Working
Paper (March 2009).
Welch, Ivo. “Common Flaws in Empirical Capital Structure Research,” Working
Paper (2007).
Graham, John. “Taxes and Corporate
Finance,” in Espen Eckbo, ed., The
Handbook of Empirical Corporate Finance.
Elsevier, 2008.
Business Review Q2 2011 11
Why Do Markets Freeze?*
BY YARON LEITNER
I
n normal times, investors buy and sell
financial assets because there are gains
from trade. However, markets do not always
function properly — they sometimes “freeze.”
An example is the collapse of trading in mortgagebacked securities during the recent financial crisis. Why
does trade break down despite the potential gains from
trade? Can the government intervene to restore the
normal functioning of markets? In this article, Yaron
Leitner explains what a market freeze is and some of the
theories as to why these freezes occur.
A puzzling feature of the recent
financial crisis is the collapse of trading volume and the lack of transactions in many financial markets that
were historically quite liquid. This is
strange because we expect demand
and supply forces to generate a price at
which trade will occur. However, like
everything else in life, markets are not
perfect, and they may not always function properly. Why do markets seize
up, even when there are potential gains
from trade? Can the government intervene to restore the normal functioning of markets? We begin by explaining
what we mean by a market freeze.
Yaron Leitner is a
senior economist
in the Research
Department of
the Philadelphia
Fed. This article
is available free
of charge at www.
philadelphiafed.
org/research-anddata/publications/.
12 Q2 2011 Business Review
WHAT IS A MARKET FREEZE?
In normal times, investors buy and
sell financial assets for various reasons.
First, they may have different opinions
as to what assets are worth. Those who
think an asset is worth more than its
current price will buy, and those who
think the asset is worth less will sell.
Second, investors may have different
needs. For example, one investor is
about to retire and would like to hold
relatively safe assets; another investor
is young and may prefer to hold risky
assets, taking the chance of getting
a higher return. The first investor
can reduce the risk in his portfolio by
selling shares of stocks he owns to the
second investor. Another example is
the sale of mortgage-backed securities:
A bank originates a mortgage and then
sells it to other investors. In this way,
*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
the bank replenishes its funds and can
use the sale’s proceeds to originate
another loan.
One way to think of the examples
above is to say that each investor attaches a different “value” to the asset,
where this value incorporates his own
assessment about the asset’s cash flows
(for example, stock dividends) as well
as his own preferences (for example,
attitude toward risk). If one investor
attaches a high value and another
investor attaches a low value, there
are potential gains from trade. As long
as trade takes place at a price that is
between the two values, both investors are better off. If there are many
buyers and sellers, trade will take place
until the market “clears.” The marketclearing price is the price at which
demand equals supply. That is, no one
wants to sell below the price, and no
one wants to buy above the price. In
normal times, market-clearing prices
represent “fair values,” which reflect
expected cash flows and individuals’
attitudes toward risk. You can think
of fair values as the price that would
be agreed on between a willing buyer
and a willing seller, with neither being required to act, and both having
reasonable knowledge of the relevant
facts.1
A market freeze refers to a situation
in which trade does not occur despite
1
This is the IRS definition (Publication 561).
The Financial Accounting Standards Board
(FASB) defines fair value as the “price that
would be received to sell an asset or paid to
transfer a liability in an orderly transaction between market participants at the measurement
date.” FASB then explains each term in the
definition above in more detail. (See Statement
of Financial Accounting Standards No. 157,
September 2006.)
www.philadelphiafed.org
the potential gains from trade. An
example is the collapse of trading in
mortgage-backed securities during the
recent financial crisis. (See the figure.)
One of the challenges during
a market freeze is the lack of market prices from which we can infer
fair values. While in practice some
transactions may occur, these transaction prices may not represent fair
values because only a limited number
of transactions take place and/or some
investors trade only because they must.
For example, to avoid bankruptcy, a
firm might be forced to sell its assets
at a very low price, one that does not
represent the fair value. While a lack
of market prices is a symptom of one
problem, it can also cause additional
problems, since potential buyers may
not know how much to bid for the
assets. For example, when you buy a
house, you look at the prices at which
similar houses in the area were sold.
However, if no houses were sold recently, it may be hard to come up with
a price.2 The lack of market prices also
led BNP Paribas, France’s largest bank,
to halt withdrawals from three of its
investment funds in 2007. In a statement, BNP Paribas said that “the complete evaporation of liquidity in certain
market segments of the U.S. securitization market has made it impossible to
value certain assets fairly regardless of
their quality of credit rating.” Alain
Papiasse, head of BNP Paribas’s asset
management and services division,
said in an interview, “For some of the
securities there are just no prices…As
there are no prices, we can’t calculate
the value of the funds.”3
In practice, it is hard to tell
precisely whether gains from trade
exist because we do not observe
investors’ needs and we do not know
the valuations they have in mind.
Thus, a simple explanation for the lack
of trade might be that investors do not
trade because they do not need to; for
example, they have exhausted all the
gains from trade and have reached the
desired outcome. While possible in
theory, it is difficult to imagine that
changes in preferences or portfolio
objectives could explain the dramatic
collapse in trading we observe in the
figure.
Another simple explanation for
a market freeze is that assets have
become more risky, so investors are
reluctant to hold them. However,
there is no simple relationship between
changes in asset risk and the volume
of trading. Increased risk may actually
result in more trade, since those who
already hold the assets may rush to sell
them. And if the price is low enough,
other investors may be willing to
buy, as in our example in which the
investor who wants a safe portfolio
sells shares of stock to the investor
who wants to take more risk. Thus,
when assets become more risky, we
may see a market crash in which prices
drop significantly but not necessarily a
market freeze.
Other explanations involve a more
puzzling situation in which investors
do not trade even though it may seem
that trade can make them better off. In
particular, we explain why Investor A,
who owns an asset, may not sell it to
Investor B, even though both investors
know that Investor B attaches a higher
value to the asset.4
4
In a related Business Review article, Ronel Elul
discusses some other features of a liquidity crisis, such as a large decline in prices, a flight to
quality, and a liquidity spiral, wherein an initial
drop in prices propagates to a large decline.
FIGURE
Nonagency MBS Issuance
Billions of $USD
350
300
250
200
150
100
2
William Lang and Leonard Nakamura provide
a formal model for this. They show that a lack
(or a low amount) of recent home sales reduces the precision of appraisals. This, in turn,
forces lenders to require larger down payments,
thereby affecting current sales.
50
0
Q1 Q3
02 02
Q1 Q3
03 03
Q1 Q3
04 04
Q1 Q3
05 05
Q1 Q3
06 06
Q1 Q3
07 07
Q1 Q3
08 08
Q1 Q3
09 09
3
See the article, “BNP Paribas Freezes Funds as
Loan Losses Roil Markets,” Bloomberg, August
9, 2007.
www.philadelphiafed.org
Source: Inside Mortgage Finance MBS Database (2002-2007), Inside MBS & ABS (2008-2009)
Business Review Q2 2011 13
INVESTORS MAY NOT KNOW
HOW TO QUANTIFY RISK
Introductory finance classes
usually teach students how to calculate
present values, that is, how to answer
the question, “How much is an asset
worth?” An important ingredient
in this process is an estimate of the
asset’s expected cash flows. Another
important ingredient is the asset’s risk
or, in other words, how likely you are
to obtain each potential outcome.
Financial companies use various
models to quantify assets’ risk. In
normal times, these models seemed
to have worked pretty well. However,
when house prices started falling,
and homeowners defaulted on their
mortgages, investors realized that
the models did not work; that is, the
assumptions behind the models were
incorrect. Investors knew that there
was risk, but they did not know how to
quantify it.
What should investors do in this
case? One response might be to avoid
buying the asset and try to sell it if you
already own it. As mentioned earlier,
in this case, the outcome could be a
huge volume of sell orders and a price
collapse, but not a market freeze.
Another response, discussed in a
paper by David Easley and Maureen
O’Hara, is to stay with the status quo,
that is, do nothing. This response leads
to a market freeze. Investors avoid
buying because they do not want to
pay too much, and those who already
own the asset avoid selling because
they do not want to sell at too low a
price. For example, suppose you own
an asset and someone offers to buy it
for $50. You may think this price is
too low because the asset might be
worth $70. Now suppose someone
offers to sell you additional units of
the asset at $50 per unit. (So then
you would have the units you already
own plus additional ones.) If you are
sure that the assets are worth $70,
14 Q2 2011 Business Review
buying additional units at $50 would
be a good thing. However, if you are
afraid that the price might fall to $30,
you will not buy additional units at
$50. Thus, you may simply sit on the
fence and do nothing. If everyone else
behaves like you, the market freezes.
The underlying assumption is
that if an investor thinks there is more
than one plausible way to value the
asset, he trades only if he is better
off given every plausible scenario.5 In
the example above, you (the investor)
thought it was plausible that the asset
might be worth $70, but it was also
plausible that it could be worth only
$30. If you had bought the asset at $50,
you would be better off if the asset is
actually worth $70, but worse off if
the asset is worth only $30. If, instead,
you sold the asset at $50, you would
be better off if it is worth only $30,
but worse off if the asset is worth $70.
Therefore, you did nothing.6
Quoted Prices May Be Biased
Relative to Fair Values. Easley and
O’Hara’s model has some interesting
implications for the debate on how to
establish fair values when markets are
frozen. For example, FASB suggests
using quoted prices.7 In normal times,
quoted prices reflect fair values, since
transactions occur at or close to these
5
Different investors may have different scenarios in mind, depending on whether they are
optimistic or pessimistic.
6
Economists have used the words “uncertainty
aversion” and “inertia” to describe the behavior
of the investor above. Uncertainty aversion
means that investors behave as if the worst-case
scenario will happen. This type of behavior
and its implications for a liquidity crisis are
discussed in Ronel Elul’s Business Review article.
Inertia means that investors act (buy or sell)
only if they expect to be happy with their
decision, given any plausible model for valuing
the assets. If they can come up with even one
plausible model under which they expect to lose
money, they do nothing.
7
See Statement of Financial Accounting Standards No. 157.
prices, and it doesn’t really matter if we
use bid or ask prices, since for highly
liquid assets the two prices are roughly
the same.8
Suppose now that the markets
are frozen. Easley and O’Hara show
that while investors may continue to
quote bid and ask prices, these prices
may be biased relative to fair values,
since no trades occur at these prices.
Consider, for example, the bid price.
You might guess that the fair value is
above the bid price because investors
who do not know how to quantify
the risk may play it safe by offering to
buy at a low price. While this may be
true in some cases, Easley and O’Hara
obtain the surprising result that, in
other cases, the fair value may actually
be below the bid price. The intuition
is that the bid price reflects the beliefs
of only one investor — the one who
is the most optimistic; however, fair
values should reflect the beliefs of all
investors, including those who are very
pessimistic and bid very low prices.
Similarly, Easley and O’Hara show that
the ask price may overestimate the fair
value, but it may also underestimate it.
Easley and O’Hara suggest that using
the average of the bid and ask spread
might be better than using just one of
these two quotes. However, they also
point out that this measure may be
biased relative to the fair value; it may
overestimate the fair value, but it may
underestimate it.9
ASYMMETRIC INFORMATION
Another explanation suggested
by economists for the recent market
freeze is an increase in asymmetric
information, that is, a situation in
which some investors have better
8
The bid price is the highest price someone is
willing to pay for an asset. The ask price is the
lowest price a seller will accept when selling the
asset.
www.philadelphiafed.org
information than other investors. For
example, the seller of a used car may
know whether the car is a lemon, but
the buyer would have no knowledge
of that. This is different from the
situation in the previous section in
which two investors had different
opinions as to what the true value
was, but none of them had better
information than the other.10,11
In a well-known article, George
Akerlof, who won a Nobel Prize in
9
The discussion above explains why we may
observe large bid and ask spreads during a
financial crisis. A small bid and ask spread
exists during normal times, even if markets
are competitive. The small spread reflects the
risk of trading with investors who have better
information or the risk that the value of dealers’
inventories (that is, the units of assets they may
need to hold temporarily when buying and selling) will fall. I discussed these two components
of the spread in more detail in an earlier Business Review article. During a financial crisis, a
spread exists not only for the two reasons above
but also because investors do not know how to
value the assets.
10
Many market observers emphasized the problems of asymmetric information in the markets
for mortgage-backed securities during the
financial crisis. For example, the original TARP
proposed that the Treasury Department purchase mortgage-backed securities from banks in
an auction. Commenting on this plan, finance
professors Glenn Hubbard, Hal Scott, and Luigi
Zingales note that “such an approach raises significant problems – most significant is the risk
posed by asymmetric information regarding the
value of these securities. Because the holders
of complex and incomparable mortgage-related
securities have more information regarding their
worth than does Treasury, Treasury is at a huge
disadvantage and will likely overpay.” See the
article, “From Awful to Merely Bad: Reviewing
the Bank Rescue Options,” Wall Street Journal,
February 7, 2009.
11
Why would asymmetric information increase
suddenly? Gary Gorton has suggested that
initially investors thought that mortgage-backed
securities were “safe,” so the fact that the seller
might have had more information was not an
issue. However, when indexes of subprime risk
began to fall, investors realized that mortgagebacked securities were not safe; that is, investors
realized that some market participants were
willing to pay a premium to protect themselves
against subprime loan default. At this point, the
fact that the seller may know more about the
likelihood of default became an issue.
www.philadelphiafed.org
economics in 2001, has shown that
an information asymmetry, such as
that in the example above, can lead
to a market breakdown. The idea is
this: If you think someone has more
information than you do, you will be
afraid to trade with him for fear of
being exploited.
The following example illustrates
this. Bank A (the seller) originates a
loan, which it wants to sell to Bank
B (the buyer). The value of the loan
example, if Bank A agrees to sell at
$60, Bank B can conclude that the
value of the loan to Bank A is between
$0 and $60. Since Bank B values the
loan at $20 more than Bank A, this
means that the value of the loan to
Bank B is between $20 and $80, or
$50, on average. Thus, if Bank B buys
the loan at $60, it expects to lose
money, on average. If Bank B offered
a higher price, it would expect to lose
even more. However, since Bank A
depends on the borrower’s likelihood of
default; the value is $60 to Bank A and
$80 to Bank B. Both banks know that
Bank B has a higher valuation because
it can do a better job of monitoring
the borrower and collecting the loan.
Everyone is better off if Bank A sells
the loan to Bank B. Any price between
$60 and $80 could work.
The sale above may not go
through, however. The problem is that
since Bank A originated the loan,
it has a better idea of whether the
borrower is likely to default. That is,
Bank A has more information than
Bank B. For example, suppose Bank
A knows that the loan is worth $60 to
itself and $80 to Bank B. In contrast,
Bank B knows only that it values the
loan at $20 more than Bank A, and
that the value of the loan to Bank A is
somewhere between $0 and $100, with
each value equally likely. Then Bank
A will not sell the loan to Bank B,
despite the fact that both banks know
that the loan is more valuable to Bank
B than to Bank A.
Why won’t the sale go through?
Whenever Bank A agrees to sell, Bank
B can conclude that Bank A values
the loan at the sale price or less. For
knows that its own valuation of the
loan is $60, it will agree to sell only if
the price is at least $60.
CAPITAL CONSTRAINTS
The effects of asymmetric information are magnified when the seller
has large inventories of assets but can
sell only a fraction of them. In particular, regulators may use the new sale
price to reassess the value of the seller’s
remaining assets. If the value drops,
regulators may require more capital, or
even worse, if the seller is a bank, the
regulator may shut it down. Potential
lenders may also use market prices to
decide how much to lend and whether
to roll over loans, even if regulators
don’t require banks to use mark-tomarket accounting.12 Whether the
constraints are imposed by a regulator or by market participants, we can
simply say that the seller is subject to
12
Under mark-to-market accounting, assets
are valued based on the recent market price of
identical or similar assets. For example, if you
bought a share of stock for $50 and the stock
now trades for $20, the “mark-to-market” value
of your stock is only $20, even though the “book
value” is $50.
Business Review Q2 2011 15
capital constraints, meaning that the
seller must ensure that the market
value of his inventories is high enough
relative to the value of his liabilities.
The cost of violating the constraint is
assumed to be very high.
Depending on Leverage, We
May Observe Increased Trade or a
Market Freeze. Concerns about the
market value of his remaining assets
may induce the seller to reject offers
that he would accept if he were not
subject to a capital constraint. Thus,
we may observe less trade compared
with the situation in which the only
problem was asymmetric information. However, in some cases, we may
actually see more trade. The reason is
that the buyer may understand that a
profitable trade would be scuttled by
the seller’s capital constraint and may
offer to buy the asset at a higher price.
Since a higher price increases the
chances that the seller will accept the
offer, trade is more likely to happen.
In a recent working paper, Philip
Bond and I show that whether we see
more trade or less trade depends on
the seller’s “leverage,” meaning that
it depends on the size of his liabilities
relative to the market value of his assets or, alternatively, on how tight his
capital constraint is. When leverage
is low, inventories have no effect on
trade. When leverage is moderate,
inventories increase the likelihood of
trade. Finally, when leverage is high,
the market freezes. We also show that
a market freeze may be preceded by
increased trade and an increase in leverage. This pattern is consistent with
what we have seen in the recent crisis.
The reason is as follows: If the
seller has only moderate leverage, the
buyer can ensure that the seller’s capital constraint is satisfied by increasing
the bid. This reduces the buyer’s expected profits from the transaction but
still allows him to profit, on average.
However, if the seller’s leverage is too
16 Q2 2011 Business Review
high, such that the value of his assets
is just a little bit above the value of his
liabilities, the buyer must increase the
bid by a lot to ensure that the seller’s
capital constraint remains satisfied
after the transaction. However, with
such a high price, the buyer expects to
lose money and may prefer not to bid
at all.
The reasoning above also explains
why we may see increased trade before
the market freezes. Like the seller, the
buyer may also have inventoried assets, and the buyer may be concerned
about their market value. Under
some circumstances, when the buyer
purchases new assets, the market value
of his existing assets falls.13 In turn, he
becomes more leveraged and his capital constraint tightens. This forces him
to bid a higher price in the next trade,
which increases the chances that the
next seller will accept the offer — so
we may see more trade. However, at
some point, when the buyer continues
to accumulate assets, he becomes overleveraged and he can no longer bid for
the asset because whatever he does, he
will either expect to lose money or he
will violate his capital constraint. This
is when the market freezes.14
Policy Implications. Our model
suggests a caveat to proposals that
would require sellers of asset-backed
securities to retain a stake on their
own books. In particular, regulatory
13
This might happen, for example, if the buyer
already has some assets similar to the one he
purchases, and if the fact that the seller was
willing to sell indicates that the value of these
assets is lower than initially thought.
14
For our results, we do not need to assume
mark-to-market accounting, where inventoried
assets are being evaluated “technically” based
on the price of the last transaction. We could
assume instead that regulators or potential lenders make inferences from the sale price, just as
the buyer did in the previous section, and that
they use these inferences to assess the value of
inventoried assets.
interventions to buy up assets may
need to be large enough to buy all
or most of a seller’s assets. Selling
assets helps the seller raise cash —
which strengthens his balance sheet.
However, selling assets also reduces the
value of the assets that remain on the
balance sheet — which weakens the
balance sheet. Buying all of the seller’s
assets eliminates this second effect.
Thus, requiring sellers to retain some
stake in the assets they sell may lead to
a market freeze.15
Another implication is that
piecemeal government interventions
to facilitate asset sales may not be
feasible. When potential buyers are
highly leveraged, they are reluctant to
buy assets for fear of creating a new
price that will reduce the value of their
inventoried assets. The government
could then unfreeze the market by
buying the assets, rather than having
the highly leveraged buyers buy the
assets. Since the government may
have less information than the seller,
it must offer a low enough price so
that it can break even, on average.16
However, by creating this lower price,
the government may harm other
potential buyers who previously chose
not to trade, since the new price can
be used to reevaluate their inventoried
assets. Alternatively, if the government
does not want to hurt potential buyers,
it could offer them a subsidy or could
increase the price it pays to the seller.
However, these options impose a cost
on taxpayers.
15
This possibility must be weighed against the
possible benefits of requiring the seller to retain
a stake in his own assets. Such a requirement
may discourage loan originators from making
bad loans in the first place. See, for example,
Senate bill S. 3217 - Restoring American Financial Stability Act of 2010 (April 15, 2010).
16
This assumes that the government has a
higher valuation for the asset. Otherwise, the
government can never break even, on average.
www.philadelphiafed.org
FEAR OF FIRE SALES CAN LEAD
TO A MARKET FREEZE
During the recent financial crisis,
we observed not only a market freeze
but also a contemporaneous credit
crunch, during which banks were
reluctant to make loans. Douglas
Diamond and Raghuram Rajan suggest
that both problems may have a common root: the fear of a fire sale, that
is, the fear that banks will be forced to
sell their assets at prices that are well
below fair values.
Why Do Fire Sales Occur?
Suppose a firm runs into liquidity
problems and needs to raise cash.
Ideally, the firm would sell its assets
to the buyer who values them the
most, such as another firm in the same
industry. However, this buyer may be
experiencing financial difficulties at
the same time as the firm and may be
unable to raise the money to buy the
assets at a fair value. The firm may
then attempt to sell its assets to a firm
outside its industry, but this other firm
might place a lower value on the assets.
For example, if all airlines are losing
money, an airline that runs into bankruptcy might need to sell its assets to a
financial firm with an airplane leasing
subsidiary. This financial firm may
not value the assets as much because
it may take time for it to find a lessee
and put the aircraft in service, especially during a recession. In this case,
the sale price might be well below the
price that firms in the airline industry
would pay if they had the money.17
17
The discussion above is based on the paper
by Andrei Shleifer and Robert Vishny. Todd
Pulvino provides empirical evidence consistent
with Shleifer and Vishny’s model. Using aircraft
sale transactions that occurred from 1978 to
1991, he shows that during a recession, an airline that is more financially constrained is more
likely to sell to a financial institution (rather
than to another airline) and that financial institutions pay, on average, 30 percent less than the
market price.
www.philadelphiafed.org
Similarly, when a bank runs into
financial problems, it may need to
sell its assets at fire sale prices simply
because other banks that value its
assets don’t have enough cash to pay
fair prices. (Or, alternatively, if there
is only one bank with cash, that bank
may use its monopoly power to lower
During the recent
its bid.) In our context, different
valuations may arise because of
different expertise. For example, some
financial firms specialize in mortgagebacked securities (they know how
to value and how to market these
securities), while other firms don’t.
These less knowledgeable firms may
be willing to buy the assets only if they
get a large enough discount, which
may also reflect the fact that they have
less information about the assets. Note,
however, that once conditions in the
financial sector improve and the banks
that value the assets the most are no
longer cash constrained, the price of
the asset is expected to return to its
fair value.
Viral Acharya, Douglas Gale, and
Tanju Yorulmazer expand the intuition
above to explain why a bank may not
be able to roll over short-term loans,
even though the bank posts collateral
whose value is expected to be high
in the long term. In their paper, the
problem is that if the bank defaults,
the lender must sell the collateral in a
fire sale to another bank, which also
borrows short term and which can
also default on its loan. If this second
bank defaults, its lender must also sell
the collateral in a fire sale to a third
bank, which can borrow short term
and default, and so on. Anticipating
this, the initial lender may not be
willing to lend against the full value of
the collateral. (In the language of the
finance profession, the initial lender
may require a large “haircut.”18)
The Prospect of a Fire Sale May
Cause a Market Freeze. The prospect
of a fire sale will be reflected in the
price today because, instead of buying
today, a potential buyer can wait and
buy later. For example, if investors
think that there is a 50 percent chance
that the price next month will be $100
and a 50 percent chance that the price
will be only $20 (because of a fire sale),
the most they will be willing to pay
today is the average price of $60.19
Douglas Diamond and Raghuram
Rajan show that the possibility of a
fire sale can lead to what financial
economists call “debt overhang” and,
in turn, a market freeze. In their
model, a bank is reluctant to sell its
assets today, even though this could
save it from potential bankruptcy in
the future, because the gains from
selling at today’s low price are captured
by the firm’s creditors rather than its
shareholders. To see that, let’s continue
with the example above. Suppose
the bank owes $60 to its creditors,
to be paid next month. If the bank
does not have any financial problems,
18
For example, if the face value of a bond used
as collateral is $100, but the lender is willing to
lend only $80 against it, we say that the haircut
is 20 percent.
19
If the investor does not care about risk (that
is, he is risk neutral) and if the interest rate is
very low, say 0 percent, the price today will be
$60. Otherwise, the price will be lower.
Business Review Q2 2011 17
it can sell its asset next month for
$100, pay its creditors, and distribute
the rest ($40) to its shareholders.
However, if the bank runs into a
financial problem and is forced to
sell its assets at a price of only $20, it
cannot fully pay its creditors and its
shareholders get nothing. On average,
the bank’s shareholders expect to
obtain $20 (0.5*40+0.5*0) and the
bank’s creditors expect to obtain $40
(0.5*60+0.5*20). Now suppose instead
that the bank can sell its assets today
at $60. Then the bank can pay back
its creditors, but nothing is left for
its shareholders. Hence, the bank’s
shareholders will prefer not to sell,
despite the financial risk. And if the
bank’s manager acts on behalf of its
shareholders, he will not sell, and the
market will freeze.20
The Prospect of a Fire Sale
Can Also Make Banks Reluctant
to Lend. Diamond and Rajan’s model
explains not only a freeze in asset
markets but also a contemporaneous
credit freeze, which is consistent with
what we saw in the recent recession.
Banks may be induced to hoard cash
rather than to lend because if there is
a fire sale, cash on hand could make
them a fortune, since they would pay
less for assets than what they are truly
worth.21
Diamond and Rajan discuss
various interventions through which
the government can reduce the
prospects of fire sales and unfreeze the
market. For example, the government
can induce banks to sell their assets
by offering to pay more than other
potential buyers offer. However, as
in the previous section, this does not
necessarily imply that the government
is expected to lose money. In
particular, if the government can hold
the assets until the price comes back
to fair value, the government could
potentially make money. However, this
argument ignores the potential costs
involved in managing those assets.
CONCLUSION
Economists have suggested a few
explanations for the recent freeze in
asset markets, such as: (1) investors
did not know how to quantify risk; (2)
asymmetric information has increased;
21
20
Note that if the bank’s creditors were in control, they would decide to sell the asset today.
The conflict of interest between shareholders
and creditors (debt holders) described above is
a common problem in corporate finance. See,
for example, the well-known paper by Michael
Jensen and William Meckling and the wellknown paper by Stewart Myers.
18 Q2 2011 Business Review
In a recent working paper, Lucian Bebchuk
and Itay Goldstein suggest a different explanation for the recent credit freeze. In their paper,
a bank is reluctant to lend to a firm with a good
investment opportunity because the bank is
afraid that other banks won’t lend and the firm
will fail. In another paper, Ricardo Caballero
and Arvind Krishnamurthy show that banks
that are worried about worst-case scenarios
may hoard liquidity instead of lending to one
another.
(3) banks were concerned about the
effect of transactions on the value of
their inventories; or (4) banks did not
want to sell their assets at low prices
that reflected the possibility of a future
fire sale.
While it is unlikely that a single
model will explain everything — after
all a model is not reality — each model
sheds light on some aspect of the crisis.
For example, one model explains the
large bid-and-ask spreads and the
relationship to fair values, another
explains the increased trade before the
market froze, and yet another explains
the contemporaneous freeze in credit
markets.
The models also help us think
about the effects of government interventions. For example, if banks are
worried about the effects of transactions on their inventoried assets, the
government may need to buy all or
most of the assets on the seller’s balance sheet in order to unfreeze the
market; however, creating a lower
market price may impose a cost on
other market participants. If banks
are worried about future fire sales, the
government may help by reducing the
chance of fire sales, for example, by
closing weak banks, infusing capital
into banks that face liquidity problems,
buying assets, or injecting capital into
potential buyers.22 BR
22
These government interventions are discussed
in Diamond and Rajan’s paper.
www.philadelphiafed.org
REFERENCES
Diamond, Douglas W., and Raghuram G.
Rajan. “Fear of Fire Sales and the Credit
Freeze,” Quarterly Journal of Economics
(forthcoming).
Lang, William W., and Leonard I.
Nakamura. “A Model of Redlining,”
Journal of Urban Economics, 33 (1993), pp.
223-34.
Easley, David, and Maureen O’Hara.
“Liquidity and Valuation in an Uncertain
World,” Journal of Financial Economics, 97
(2010), pp. 1-11.
Leitner, Yaron. “Liquidity and Exchanges,
or Contracting with the Producers,”
Federal Reserve Bank of Philadelphia
Business Review (First Quarter 2004).
Elul, Ronel. “Liquidity Crises,” Federal
Reserve Bank of Philadelphia Business
Review (Second Quarter 2008).
Myers, Stewart C. “Determinants of
Corporate Borrowing,” Journal of Financial
Economics, 5 (1977), pp. 147-75.
Bond, Philip, and Yaron Leitner. “Market
Run-Ups, Market Freezes, and Leverage,”
Federal Reserve Bank of Philadelphia,
Working Paper 10-36 (November 2010).
Gorton, Gary B. “The Panic of 2007,”
NBER Working Paper 14358 (September
2008).
Pulvino, Todd. “Do Asset Fire Sales Exist?
An Empirical Investigation of Commercial
Aircraft Transactions,” Journal of Finance,
53 (1998), pp. 939-78.
Caballero, Ricardo J., and Arvind
Krishnamurthy. “Collective Risk
Management in a Flight to Quality
Episode,” Journal of Finance, 63 (2008), pp.
2195-2230.
Jensen, Michael C., and William
H. Meckling. “Theory of the Firm:
Managerial Behavior, Agency Costs and
Ownership Structure,” Journal of Financial
Economics, 3 (1976), pp. 305-60.
Acharya, Viral, Douglas Gale, and Tanju
Yorulmazer. “Rollover Risk and Market
Freezes,” Journal of Finance (forthcoming).
Akerlof, George A. “The Market for
‘Lemons’: Quality Uncertainty and the
Market Mechanism,” Quarterly Journal of
Economics, 84 (1970), pp. 488-500.
Bebchuk, Lucian, and Itay Goldstein. “SelfFulfilling Credit Market Freezes,” Working
Paper (2009).
www.philadelphiafed.org
Shleifer, Andrei, and Robert W. Vishny.
“Liquidation Values and Debt Capacity: A
Market Equilibrium Approach,” Journal of
Finance, 47 (1992), pp. 1343-66.
Business Review Q2 2011 19
Understanding House-Price Dynamics*
BY MAKOTO NAKAJIMA
F
or most homeowners, housing is the
single most important component of their
nonpension wealth. Therefore, a change in
house prices greatly affects the total wealth
of many households. Furthermore, movements in house
prices can affect people’s lives indirectly. For example,
the surge in the number of mortgage defaults and
foreclosures during the recent recession was triggered in
part by a drop in house prices, and this surge damaged
the health of the financial institutions that either
directly or indirectly owned mortgage loans. In turn, the
deteriorating health of the financial sector was one of the
factors contributing to the recession. Naturally, for both
policymakers and for people who want to make sound
financial decisions, it is important to understand how and
why house prices move. In this article, Makoto Nakajima
explains a simple theory that helps us better understand
house-price dynamics. The theory — called the user
cost-rent equivalence — is based on the close relationship
between user costs, which are the costs of owning a house
for a year, and rents.
The ups and downs of house
prices affect our lives substantially.
Makoto Nakajima
is an economist
in the Research
Department of
the Philadelphia
Fed. This article
is available free
of charge at www.
philadelphiafed.
org/research-anddata/publications/.
20 Q2 2011 Business Review
About two-thirds of U.S. households
own a house, and for most
homeowners, housing is the single
most important component of their
nonpension wealth. Therefore, a
change in house prices greatly affects
the total wealth of many households.
*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.
For example, if there is a large
drop in the price of a house, the
homeowner is more likely to receive
less money when selling his house in
the future. Under this circumstance,
it is probably a sound decision to cut
back on household expenditures.
House prices are also important for
the one-third of households who are
not homeowners, since many of them
are young households that are saving
money to buy their first house. Higher
house prices could force many of
them to delay or give up their plans
to buy a house. Lower house prices
help young households while hurting
homeowners.1
Moreover, the recent recession
seems to suggest that movements in
house prices also affect people’s lives
indirectly. The surge in the number
of mortgage defaults and foreclosures
was triggered in part by a drop in
house prices. Furthermore, this surge
damaged the health of the financial
institutions that either directly or
indirectly owned mortgage loans,
and the deteriorating health of the
financial sector was one of the factors
contributing to the recession.
Naturally, for both policymakers
and for people who want to make
sound financial decisions, it is
important to understand how and
why house prices move. This article
presents a simple theory that helps
us better understand house-price
dynamics. The theory — called the
user cost-rent equivalence — is based
1
See the Business Review article by Wenli Li
and Rui Yao for a more detailed analysis of how
house-price changes affect the consumption and
well-being of American households.
www.philadelphiafed.org
on the close relationship between user
costs, which are the costs of owning a
house for a year, and rents.
We’ll start with some observations
about the housing market, then
review recent economic research that
analyzes house-price dynamics. Since
economists are still trying to improve
their understanding of how house
prices move, there are many theories
that explain house-price dynamics
other than the one presented in this
article. We will take a brief look at
some of the other theories. Then we’ll
discuss the theory that we focus on in
this article and examine how elements
that affect house prices, according to
our theory, change over time and the
implications of such changes for house
prices. Finally, we’ll carry out a simple
numerical exercise to see what fraction
of the recent rise in house prices
can be accounted for by the theory
presented here and by the data.
Interested readers are encouraged
to look at Wenli Li and Fang Yang's
related Business Review article, which
analyzes the economic benefits and
costs of homeownership.
SOME OBSERVATIONS ABOUT
HOUSE PRICES
The trend of the average house
price between 1975 and 2009 is shown
in Figure 1. This is a real index in the
sense that the house prices shown in
the figure are relative to the prices
of nonhousing goods and services. A
constant real house price doesn't mean
that the nominal house price (the ones
we see in newspaper ads) is constant;
rather, it most likely means that house
prices are, on average, rising at the
same pace as other goods we regularly
purchase. The average house price
rose about 1.5 percent faster than
other prices per year over this period.
What is striking about the figure is
that the trend is relatively flat until the
mid-1990s. Since then, there has been
www.philadelphiafed.org
a substantial increase (until the end
of 2006) and a substantial drop (since
2007). Around the end of 2006, when
the average house price peaked, house
prices were about 60 percent higher
than their level in the mid-1990s.
The recent increase and decline
in the average house price have been
accompanied by similar changes in the
homeownership rate (Figure 2). The
figure plots both the homeownership
rate (left scale) and the average
real house-price index (right scale),
which was shown in Figure 1. Until
the mid-1990s, about 64 percent of
U.S. households lived in housing
that they owned. But in 2005, the
homeownership rate went up to 69
percent and then came down to 67
percent. Matthew Chambers, Carlos
Garriga, and Don Schlagenhauf
find that the increase is an outcome
of demographic changes as well as
developments in the mortgage loan
market, in particular, the proliferation
of new types of mortgage loans with
low down-payment requirements and
low introductory rates.
Although this article focuses on
how and why the national average
house price moves, it is important to
keep in mind that behind the average
house-price dynamics, there are
substantial differences across regions
of the U.S. (Figure 3). The Pacific,
New England, and Middle Atlantic
regions exhibit the most volatile
movements. On the other hand, the
average house price in the West-South
Central region changed very little
between 1975 and 2009. The houseprice bubble and subsequent burst that
we often hear about does not apply
equally to all regions of the U.S. In
general, the regions that experienced a
larger increase in house prices are also
experiencing a larger drop in house
prices. The level of average house
prices in regions with volatile houseprice movements is still high compared
with that in the mid-1990s.
House-price dispersion across U.S.
FIGURE 1
Real House Price Index for the U.S. (1975 = 100)
Index
200
180
160
140
120
100
80
60
40
20
0
1975 1978
1981 1984 1987
1990 1993 1996 1999 2002 2005 2008
Data source: Federal Housing Finance Agency
Business Review Q2 2011 21
FIGURE 2
Homeownership Rate and House Price
Percent
Percent
70
200
69
Homeownership Rate
180
68
160
67
140
66
House Price Index
120
65
100
64
80
63
60
62
40
61
20
0
60
1975 1978 1981 1984 1987 1990
1993 1996 1999
2002 2005 2008
Data source: U.S. Census Bureau and Federal Housing Finance Agency
Note: Homeownership rate is computed by dividing the number of households living in owneroccupied housing units by the total number of households.
FIGURE 3
Real House Price Index for U.S. Regions
(1975 = 100)
Index
450
400
350
300
East North Central
Pacific
East South Central
South Atlantic
Middle Atlantic
West North Central
Mountain
West South Central
New England
USA
250
200
150
100
50
0
1975 1978
1981 1984 1987
1990 1993 1996 1999 2002 2005 2008
Data source: Federal Housing Finance Agency
22 Q2 2011 Business Review
cities also increased, and the dispersion across cities is even larger than
the dispersion across regions. A study
by Stijn Van Nieuwerburgh and PierreOlivier Weill focuses on this increasing
dispersion of house prices across U.S.
cities. They show that house-price dispersion across U.S. cities can increase
when the dispersion of wages across
cities increases. For example, higher
house prices in San Francisco reflect
the higher wages earned by people living in San Francisco.
Finally, let's look at the trend of
average rents. It is important to know
the dynamics of rents because, as
mentioned above, the theory presented in this article suggests a strong
link between house prices and rents.
Figure 4 shows the trend of average
real rents for primary residences since
the 1970s. Like average house prices,
average rents have gone up since the
mid-1990s. However, the fluctuations
are much less pronounced. The average annual growth rate of rents is 0.5
percent, compared with a 1.5 percent
average annual growth rate of house
prices. However, we need to be aware
that rents have some measurement
issues. In their study, Theodore Crone,
Leonard Nakamura, and Richard
Voith argue that the growth rate of
rents has been higher than the official
data suggest.
RECENT ATTEMPTS TO
UNDERSTAND HOUSE-PRICE
DYNAMICS
Because of their obvious importance, particularly in recent years,
house-price dynamics have been an active area of research. Perhaps the most
important question is, why did house
prices go up substantially? Theories
that attempt to explain rising housing
prices can be placed into three groups.
The first group of studies deals
with the inflexible nature of housing supply; it takes time to build a
www.philadelphiafed.org
FIGURE 4
Real Rent of Primary Residence (1975 = 100)
Index
140
120
100
80
60
40
20
0
1975 1978
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008
Data source: Bureau of Labor Statistics
Note: Real rent is computed by dividing the "rent of primary residences" (notice this does not include
the imputed rents of owner-occupied housing) by the consumer price index (CPI) less shelter.
house, and land is not always available,
especially in a city’s center (think of
Manhattan). Motivated by the observation that house prices went up more
in metropolitan areas, where space
is tighter, Edward Glaeser, Joseph
Gyourko, and Raven Saks investigate
the role of supply-side restrictions,
such as land-use restrictions, in the
recent house-price boom. They find
that tightened housing-supply regulations played some role in generating an
upward trend in house prices. Morris
Davis and Jonathan Heathcote, in
their study, break down the changes in
house prices into changes in land prices and changes in the price of building
materials and find that changes in land
prices drive house-price dynamics. If
the prices of building materials are
volatile, it could explain at least a part
of house-price dynamics, but they show
that that is not the case. In another
study, Nobuhiro Kiyotaki, Alexander
Michaelides, and Kalin Nikolov look
www.philadelphiafed.org
at the role that a limited supply of land
plays in shaping house-price dynamics. Their model indicates that in an
economy in which the total value of
land is large relative to the total value
of real estate (consisting of land and
the structures built on it), house prices
react more strongly to changes in the
economy's economic growth or interest rate. We can interpret their result
as confirming the importance of the
limited availability of housing supply in
shaping house-price dynamics.
The second group of theories
investigates why demand for housing
increased over time. In interesting but
controversial work, Gregory Mankiw
and David Weil argue that house
prices are driven by demographic
changes. When baby boomers were in
their prime (30s and 40s), a time when
people tend to buy bigger houses, total
housing demand was larger. A natural
consequence is that as baby boomers
age and retire, housing demand and
house prices decline. Whether and
to what extent Mankiw and Weil’s
theory is true remains to be seen. In
my 2005 study, I argued that demand
for housing, especially owner-occupied
housing, increases when income is
more volatile. This is because housing
is a big part of people’s total wealth
(housing made up about 40 percent of
total wealth in 2004, according to the
Survey of Consumer Finances), and it
is natural for people to save more and
prepare for bad times when income is
more volatile. I find that a part of the
rise in house prices can be attributed
to the fact that individual wages have
become more volatile since the 1970s.
The third group of theories focuses on the role of expectations in shaping house-price dynamics. The role of
expectations might be important because house prices seem more volatile
than factors that are naturally thought
to affect house prices (often called
fundamentals), such as income and
mortgage rates. I will briefly describe
three studies in this group. The irrational exuberance theory of Robert Shiller
is the most well known.2 If everybody
thinks that house prices will go up,
house prices could go up only because
more people try to buy now, expecting capital gains from owning a house.
When house prices are increasing only
because people expect prices to go up,
and not because the fundamental drivers of house prices are changing, the
increase is commonly called a bubble.
When increases in house prices are a
bubble, there is no reason for prices to
stay at a higher level.3 If people suddenly start thinking that house prices
will drop, house prices could actually
2
See Robert Shiller’s 2005 book. Shiller
analyzes the U.S. housing market in his 2007
article.
3
The Business Review article by Timothy Schiller analyzes the bubble hypothesis.
Business Review Q2 2011 23
drop. Shiller discusses a variety of factors that contribute to bringing about
such irrational exuberance, including
cultural and psychological factors.
Monika Piazzesi and Martin
Schneider look at survey evidence
to analyze expectations. They use
the Michigan Survey of Consumers,
which is a useful data set for this
purpose because it asks respondents
about current and future house prices.
According to Piazzesi and Schneider’s
study, the proportion of households
that are optimistic about future house
prices is about 9 percent, on average.
However, what is more interesting is
that they also find that the proportion
of such optimistic households
increased from 10 percent to 16
percent during the recent house-price
boom. Motivated by this evidence,
Piazzesi and Schneider propose a
theory whereby some households'
expectations are driven by momentum.
When house prices are increasing
for a while for some reason, these
momentum households can keep house
prices going up for a bit longer, because
they believe that house prices will
keep increasing, based on their recent
experience, and they behave like
households with irrational exuberance.
James Kahn proposes an alternative theory as to how house prices
are linked to expectations. When
the economy is growing faster, as in
the 1990s, people’s income increases
faster, and thus, future rents rise
faster. Notice that house prices today
reflect future rents because if you buy
a house today, you don't need to pay
higher rents in the future. Therefore,
if income, and thus rents, are expected
to grow faster, people try to buy rather
than rent a house today. Consequently,
house prices go up today just because
of a positive change in expectations
about future income growth. According to Kahn's theory, expectations for
sustained high income growth were the
24 Q2 2011 Business Review
driving force for the recent increase in
house prices.
How are the various studies
presented above related to the usercost theory of house prices that I will
present? In what follows, a rising trend
in rents, which is consistent with the
combination of inflexible supply and
growing demand for housing, and
expectations for future house-price
growth will be important in generating
house-price growth. I will use these
factors similar to the way they’re used
in a study by James Poterba and another by Charles Himmelberg, Christopher Mayer, and Todd Sinai. The
latter study, using the same approach
employed in this article, concludes that
"as of the end of 2004, our analysis
reveals little evidence of a housing
bubble." Himmelberg and co-authors
also look at differences in house-price
dynamics across U.S. cities, while
this article focuses on movements in
average house prices nationally. This
article also emphasizes the importance
of expectations in driving house prices.
THEORY OF THE USER COSTS
OF HOUSING AND RENTS
The user-cost theory is based on
two elements: how user costs are determined, and the equivalence between
user costs and rents. Let's look at these
elements one at a time.
User costs are the costs of owning
a house for a year instead of renting
it. What are the components of user
costs? As explained by Poterba and
by Himmelberg and co-authors, there
are five major components of the user
costs of housing. First, there is the
interest cost, which can be interpreted
in two ways. If a person buys a house
with a mortgage loan, he has to pay
interest on the mortgage every year.
The total mortgage interest payment
is approximately the annual mortgage
interest rate multiplied by the house’s
value (house price). However, some
people buy houses without mortgage
loans. Even if a person buys a house
without taking out a mortgage, there is
an opportunity cost, which is the profit
missed by taking one action over another. In the current context, he loses
the interest income that he would have
earned if he had saved and invested
the money instead of using the money
to buy a house. The forgone interest income can be expressed as the
interest rate multiplied by the house’s
value (house price). In either case, the
interest cost can be represented as the
house price times the annual interest
rate.
Second, homeowners are required
to pay property taxes. Since property taxes also depend on the house’s
value, property tax payments can be
computed as the house price times the
property tax rate.
Third, in the U.S., homeowners
can deduct mortgage interest payments
and property tax payments from their
taxable income, up to some limit.4 This
deduction indirectly reduces the cost
of ownership. The benefit derived from
the deduction can be represented as
the sum of mortgage interest payments
and property tax payments multiplied
by the deduction rate.
Fourth, homeowners have to pay
4
The amount of mortgage interest payment
deduction is capped at the interest on the first
$1 million in mortgages.
www.philadelphiafed.org
for maintenance and repairs. It is also
natural to assume that the cost is approximately proportional to the house’s
value. A bigger or more valuable house
requires more money for maintenance
and repairs.
Finally, expectations about future
changes in house prices affect user
costs today, even before the changes
are realized. For example, suppose
you expect that the value of your
house will drop by 5 percent in the
near future. That means that you will
lose 5 percent of the house’s value by
keeping the house. The expected cost
of owning would be higher, taking this
future 5 percent loss into account. On
the other hand, if you expect that the
house’s price will go up by 10 percent, this makes owning profitable by
exactly 10 percent of the house’s value.
Thus, the cost of owning a house, taking into account the expected gain just
by holding on to it, will decrease by
the same amount. In sum, a change in
the expected future value of the house
has the effect of indirectly changing
user costs.
How can we use these components of user costs to understand
house-price dynamics? This is where
the other important element of our
theory — the close relationship
between user cost and rent — comes
into play. If there is a house that can
be either rented or purchased, the cost
of renting the house must be close to
the user costs if the house is owned.
Why? If the rent is much higher than
the user costs, somebody can buy
the house, rent it out, and make a
profit because the costs of owning and
maintaining the house (user costs) are
lower than the income from renting
the house (rent). Under this circumstance, demand for housing will rise as
people try to buy houses and exploit
the opportunity, and this pushes up
house prices. On the other hand, if
the rent is much lower than the user
www.philadelphiafed.org
costs, the opposite is likely to happen: A homeowner can save money
by selling his house and renting one
instead. If there are a lot of people trying to sell their houses and rent, house
prices would fall, reflecting the weak
demand for housing. In the end, we
should expect that user costs and rents
will end up close to each other when
houses are both rented and purchased.5
We will use this (approximate) closeness between user costs and rents to
examine how house prices are affected
by changes in interest rates, rents, and
so forth.
From the discussion above, we
know how user costs are determined,
and we also know that user costs
should be close to rents. In addition,
all of the major components of user
costs — interest cost, property taxes,
deduction of mortgage interest payments, maintenance and repair costs,
and expectations about future changes
in house prices — are approximately
proportional to house prices. In other
words, in general, all of the components will be larger if the house price
is higher. Now, consider a normal
situation in which user costs are equal
to rents. Suppose the interest rate goes
up. Since the interest cost is a part
of the user cost, when the interest
rate goes up, user costs should go up
if nothing else changes. However, as
discussed above, when the user cost
exceeds rent, it is beneficial for homeowners to sell their house and rent
instead. This decline in demand for
housing would put downward pressure
on house prices, bringing house prices
5
In the language of finance and economics, this
condition, which indicates that prices of substitutable things should be close to each other,
is called an arbitrage-free condition: Nobody
can make a profit by taking advantage of the
price difference between two assets (arbitrage),
because the prices will adjust to eliminate the
arbitrage opportunity.
and thus user costs back to their initial
level. As a result, user costs and rents
will be equalized, with a higher interest
rate and lower house prices.
Let’s look at another example.
What happens if rents turn out to be
higher than in the normal situation,
but other things remain unchanged?
When rents are higher than user costs,
renters would become homeowners
and save money. This would push up
housing prices. House prices would rise
until user costs and rents are balanced
again. In summary, here’s how each
element in user costs and rents is related to house prices: House prices are
higher when rents are higher, interest
rates are lower, property tax rates are
lower, the tax deduction rate is higher,
maintenance and repair costs are
lower, and house prices are expected to
rise in the future. (See User Cost-Rent
Equivalence for a more formal representation of the theory.)
However, these relationships are
valid only when all other things do not
change. For example, suppose the government decides to raise the property
tax rate. Higher property taxes mean
higher user costs and lower house
prices, if nothing else has changed.
However, a landlord’s natural response
might be to increase rents so that (at
least a part of) the additional property
tax is passed on to the tenants. When
rents and the property tax rate both
increase, it is hard to say what should
happen to house prices, according to
our theory.
THEORY MEETS DATA
Now let’s look at how three of
the six elements that affect user costs
— rents, interest rates, and expected
changes in house prices — have
changed over time and discuss whether
and to what extent such changes help
us rationalize the changes in house
prices. I do not discuss the other three
— property tax rate, tax deductions,
Business Review Q2 2011 25
and maintenance and repair costs —
since it is hard to capture changes in
the trends of these factors.6
Rents. As we saw in Figure 4,
house prices and rents tend to move
together (for example, look at the early
1980s). These synchronized dynamics are exactly what the equivalence
between user costs and rents would
suggest. Although rents have been less
volatile than house prices, real rents
and real house prices, on average, have
been steadily increasing over time. To
understand why, we will look at both
the supply side and the demand side.
On the supply side, a natural answer
is the limited supply of land, especially
in and around metropolitan areas.
House prices and rents are increasing because the land on which houses
and apartments are built has become
more and more scarce. The two studies mentioned earlier — the one by
Edward Glaeser and his co-authors
and the one by Nobuhiro Kiyotaki and
his co-authors — find evidence that
supports the importance of the limited
availability of land for the rising trend
in house prices. On the demand side,
it is natural that house prices and
rents increase when the supply cannot
adjust flexibly and the population —
and therefore demand — is growing.
That is the implication of the work by
Mankiw and Weil reviewed earlier. My
own research, cited earlier, supports
the notion that demand for housing
increases when individual income is
becoming more volatile. When the
availability of land, and thus housing,
6
Property tax rates differ state by state, and
thus, it is hard to capture the trend of the
average property tax rate. The effect of tax
deductions on house prices is difficult to
measure because the federal income tax features
a progressive structure and various kinds of deductions and exemptions. Moreover, there has
been no clear trend in terms of different levels
of the income tax rate since 1975. Finally, there
has been no substantial change in maintenance
and repair costs.
26 Q2 2011 Business Review
User Cost-Rent Equivalence
F
ormally, the equivalence between user costs and rents can be
written in the following way:
Rent = User cost = (Interest rate + Property tax rate –
(Mortgage interest rate + Property tax Rate) * Tax deduction
rate + Maintenance cost rate – Expected rate of capital gain) * House price
For the simple exercise on page 27, I set the parameters as follows. The
property tax rate is set at 1.5 percent per year. Maintenance and repair costs
are set at 2.5 percent of house value per year. The tax deduction rate is set at
25 percent. These are the numbers used in the study by Charles Himmelberg,
Christopher Mayer, and Todd Sinai. The expected nominal house-price growth
rate is set at 3.7 percent per year, which is the average between 1975 and 2004.
Finally, I add 2 percent as the risk premium of owning instead of renting,
following Himmelberg and co-authors. Rents are 102 in 1997 (normalized such
that the 1975 level is 100) and 115 in 2007. The interest rate is 6.6 percent in
1997 and 4.7 percent in 2007.
is limited, such an increase in demand
pushes up house prices.
Interest Rates. Let's look at the
interest rate, which is the second
element that affects user costs. Figure 5
shows two types of interest rates: a 30year fixed-rate conventional mortgage
interest rate and the interest rate on
10-year Treasury securities. Thirty-year
fixed-rate conventional mortgage loans
are the type of mortgage loans the
majority of homeowners obtain when
purchasing a house. According to the
American Housing Survey, in 2005
90 percent of U.S. primary mortgages
were fixed-rate mortgage loans.
As easily seen in Figure 5, both
interest rates have been dropping
steadily since the early 1980s.
According to the theory of user costs
and rents, when the interest rate is
declining, so is the user cost of owning
a house, and house prices will increase.
Moreover, the effect of changes in
interest rates on house prices becomes
stronger when the interest rate is lower.
For example, suppose the mortgage
interest rate declines from 2 percent
to 1 percent. This 1 percentage point
decline in the interest rate halves the
interest rate and, thus, the interest
cost. On the other hand, suppose the
interest rate drops from 10 percent to
9 percent. Although the interest rate
drops by 1 percentage point again,
this reduces the interest cost by only
one-tenth.
Expected Changes in House
Prices. The third element that
determines user costs is expectations.
Although expectations about future
changes in house prices are difficult
to measure precisely, the literature
discussed earlier supports the idea that
people might have expected possibly
rapid increases in house prices to
continue in the future, especially from
the mid 1990s through 2006. These
expectations lowered the user cost of
housing and resulted in an increase in
house prices.
In summary, there is evidence to
suggest that rents gained consistently,
interest rates fell steadily, and people
www.philadelphiafed.org
FIGURE 5
Mortgage and Treasury Interest Rates
Percent
20
18
30-year Mortgage
16
14
12
10
8
6
10-year Treasury
4
2
0
1975 1978
1981 1984 1987
1990 1993 1996 1999 2002 2005 2008
Data source: Federal Home Loan Mortgage Corporation and Board of Governors of the Federal
Reserve System
expected strong growth in house
prices between the mid-1990s and
the mid-2000s. According to the
theory presented in this article, these
elements are consistent with rising
house prices during the same period.
Moreover, if we were to observe the
opposite — that is, rents falling,
interest rates rising, and expected
house prices falling — the user-cost
theory would suggest that it would
not be surprising to see house prices
decrease.
A NUMERICAL EXAMPLE
By combining the user-cost theory
and the actual data on rents and interest rates described above, we can
generate house-price dynamics implied
by the theory and the data. By comparing the actual house-price data and
the data implied by the theory, we can
learn to what extent the theory helps
us understand house-price dynamics.
As an example, let’s look at the
www.philadelphiafed.org
question of how much of the observed
rapid increase in house prices between
1997 and 2007 can be explained by the
user-cost theory. As explained above,
the combination of steadily increasing
rents and declining interest rates is
consistent with an upward trend in
house prices. We basically follow the
strategy of Himmelberg and co-authors
in setting numbers for this exercise.
(More details can be found in User
Cost-Rent Equivalence.) An important
assumption is the expected growth
rate of house prices. Let’s assume that
people expect that a nominal houseprice growth rate of 3.7 percent per
year will continue. This is the average
house-price growth rate between 1975
and 2005. Notice that this is a rather
conservative assumption because this
growth rate is lower than the growth
rate we observed between the 1990s
and early 2000s.
The user-cost theory, combined
with the observed changes in rents and
interest rates and a moderate assumption about expectations, implies that
house prices went up by an average
of 3.3 percent per year (39 percent
between 1997 and 2007). This increase
is smaller than 4.2 percent, which is
the actual annual growth rate in average house prices from 1997-2007 (51
percent during the entire period). The
simple theory of user costs accounts
for about 80 percent of the growth
rate of house prices during the period.
The unexplained part might be due to
changes in expectations or innovations
in the mortgage market, such as the
introduction of new types of mortgage
instruments.
How sensitive is the result to
a different assumption about the
expected growth rate of house prices?
For example, if we assume a low expected house-price growth rate of 1.85
percent per year (which is half of 3.7
percent), our theory implies that house
prices went up by 2.8 percent per year
between 1997 and 2007 — lower than
3.3 percent but still a large proportion
of the observed 4.2 percent annual
growth rate during the same period.
Finally, let me briefly discuss the
recent sudden reversal of the trend in
house prices. The numerical example
generates a sudden reversal of the
trend when there is a sudden reversal
in expectations about the future trend
of house prices. For example, suppose, in 2007, the expected annual
growth rate of house prices suddenly
dropped from 3.7 percent to zero, but
everything else remained the same.
Then the house price suggested by the
model becomes 12 percent lower. The
size of the drop is exactly the same as
the drop in the national average house
price index between 2007 and the
third quarter of 2009. The change in
expectations can be related to changes
in fundamentals (for example, prospects for future income growth may
have suddenly become bleak with the
Business Review Q2 2011 27
economy slowing down) or it could be
unrelated to changes in fundamentals
(for example, the bursting of a bubble).
We need a dynamic model that
incorporates expectations to systematically analyze the sudden reversal in
the trend, and many attempts, some
of which are mentioned in this article,
are being made to improve our understanding in this area.
CONCLUSION
This article presents a simple
theory of house prices based on
the equivalence between user costs
and rents. Although it is a simple
relationship, the theory tells how
different types of housing market data
are related to each other. For example,
we use the theory to show that the
observed increase in house prices since
the mid-1990s is consistent with the
increase in rents, declining interest
rates, and reasonable expectations
about future house-price growth.
The theory indicates that the sudden
reversal of the trend in house-price
growth is related to changes in
expectations. BR
Kiyotaki, Nobuhiro, Alexander
Michaelides, and Kalin Nikolov. "Winners
and Losers in Housing Markets," Princeton
University, unpublished paper (2008).
Poterba, James M. "Tax Subsidies to
Owner-Occupied Housing: An AssetMarket Approach," Quarterly Journal of
Economics, 99:4 (1984), pp. 729-52.
Li, Wenli, and Rui Yao. "Your House Just
Doubled in Value? Don't Uncork the
Champagne Yet!" Federal Reserve Bank of
Philadelphia Business Review (First Quarter
2006), pp. 25-34.
Schiller, Timothy. "Housing: Boom
or Bubble?" Federal Reserve Bank of
Philadelphia Business Review (Fourth
Quarter 2006), pp. 9-18.
REFERENCES
Chambers, Matthew, Carlos Garriga, and
Don E. Schlagenhauf. "Accounting for
Changes in the Homeownership Rate,"
International Economic Review, 50:3 (2009),
pp. 677-726.
Crone, Theodore, Leonard I. Nakamura,
and Richard Voith. "Rents Have Been
Rising, Not Falling, in the Postwar Period,"
Review of Economics and Statistics, 92:3
(August 2010), pp. 628-42.
Davis, Morris A., and Jonathan Heathcote.
"The Price and Quantity of Residential
Land in the United States," Journal of
Monetary Economics, 54:8 (2007), pp.
2595-20.
Li, Wenli, and Fang Yang. "American
Dream or American Obsession?
The Economic Benefits and Costs of
Homeownership," Federal Reserve Bank
of Philadelphia Business Review (Third
Quarter 2010).
Glaeser, Edward L., Joseph Gyourko, and
Raven E. Saks. "Why Have Housing Prices
Gone Up?," American Economic Review,
95:2 (2005), pp. 329-33.
Mankiw, N. Gregory, and David N. Weil.
"The Baby Boom, the Baby Bust, and the
Housing Market," Regional Science and
Urban Economics, 19 (1989), pp. 235-58.
Himmelberg, Charles, Christopher
Mayer, and Todd Sinai. “Assessing High
House Prices: Bubbles, Fundamentals
and Misperceptions,” Journal of Economic
Perspectives, 19:4 (2005), pp. 67-92.
Nakajima, Makoto. "Rising Earnings
Instability, Portfolio Choice, and Housing
Prices," University of Illinois, UrbanaChampaign, unpublished paper (2005).
Kahn, James A. "What Drives Housing
Prices?," Federal Reserve Bank of New
York Staff Report, No. 345 (2008).
28 Q2 2011 Business Review
Shiller, Robert J. Irrational Exuberance, 2nd
edition. Princeton: Princeton University
Press, 2005.
Shiller, Robert J. "Understanding Recent
Trends in House Prices and Home
Ownership," National Bureau of Economic
Research Working Paper 13553 (2007).
Van Nieuwerburgh, Stijn, and PierreOlivier Weill. "Why Has House Price
Dispersion Gone Up?," Review of Economic
Studies, 77 (2010), pp. 1567-1606.
Piazzesi, Monika, and Martin Schneider.
"Momentum Traders in the Housing
Market: Survey Evidence and a Search
Model," American Economic Review, 99:2
(2009), pp. 406-11.
www.philadelphiafed.org
RESEARCH RAP
Abstracts of
research papers
produced by the
economists at
the Philadelphia
Fed
You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/research-and-data/
publications/research-rap/. Or view our working papers at: www.philadelphiafed.org/research-and-data/
publications/.
TOOLS FOR ASSESSING
MULTIVARIATE ASPECTS OF
BAYESIAN DSGE MODELS
This paper develops and applies tools
to assess multivariate aspects of Bayesian
dynamic stochastic general equilibrium
(DSGE) model forecasts and their ability
to predict co-movements among key
macroeconomic variables. The authors
construct posterior predictive checks to
evaluate the calibration of conditional and
unconditional density forecasts, in addition
to checks for root-mean-squared errors and
event probabilities associated with these
forecasts. The checks are implemented
on a three-equation DSGE model as well
as the Smets and Wouters (2007) model
using real-time data. They find that the
additional features incorporated into the
Smets-Wouters model do not lead to a
uniform improvement in the quality of
density forecasts and prediction of comovements of output, inflation, and interest
rates.
Working Paper 11-5, “Evaluating DSGE
Model Forecasts of Co-movements,” Edward
Herbst, University of Pennsylvania, and Frank
Schorfheide, University of Pennsylvania, and
Visiting Scholar, Federal Reserve Bank of
Philadelphia
EXPLAINING FLUCTUATIONS IN
TRADE DURING THE RECENT
RECESSION
The authors examine the source of
the large fall and rebound in U.S. trade
www.philadelphiafed.org
in the recent recession. While trade fell
and rebounded more than expenditures
or production of traded goods, they find
that relative to the magnitude of the
downturn, these trade fluctuations were
in line with those in previous business
cycle fluctuations. The authors argue that
the high volatility of trade is attributed
to more severe inventory management
considerations of firms involved in
international trade. They present empirical
evidence for autos as well as at the aggregate
level that the adjustment of inventory
holdings helps explain these fluctuations in
trade.
Working Paper 11-6, “U.S. Trade and
Inventory Dynamics,” George Alessandria,
Federal Reserve Bank of Philadelphia; Joseph
P. Kaboski, University of Notre Dame; and
Virgiliu Midrigan, New York University
RECENT ADVANCES IN THE
ESTIMATION AND EVALUATION
OF DSGE MODELS
Estimated dynamic stochastic
equilibrium (DSGE) models are now
widely used for empirical research in
macroeconomics as well as for quantitative
policy analysis and forecasting at central
banks around the world. This paper reviews
recent advances in the estimation and
evaluation of DSGE models, discusses
current challenges, and provides avenues for
future research.
Working Paper 11-7, “Estimation and
Evaluation of DSGE Models: Progress
Business Review Q2 2011 29
and Challenges,” by Frank Schorfheide, University of
Pennsylvania, and Visiting Scholar, Federal Reserve Bank
of Philadelphia
MEASURING THE EFFECT OF EXTENSIONS
OF UNEMPLOYMENT INSURANCE ON THE
UNEMPLOYMENT RATE
This paper measures the effect of extensions
of unemployment insurance (UI) benefits on the
unemployment rate using a calibrated structural
model that features job search and consumptionsaving decision, skill depreciation, UI eligibility, and
UI benefit extensions that capture what has happened
during the current downturn. The author finds that the
extensions of UI benefits contributed to an increase in
the unemployment rate by 1.2 percentage points, which
is about a quarter of an observed increase during the
current downturn (a 5.1 percentage point increase from
4.8 percent at the end of 2007 to 9.9 percent in the fall
of 2009). Among the remaining 3.9 percentage points,
2.4 percentage points are due to the large increase in
the separation rate, while the staggering job-finding
probability contributes 1.4 percentage points. The last
extension in December 2010 moderately slows down
the recovery of the unemployment rate. Specifically,
the model indicates that the last extension keeps the
unemployment rate higher by up to 0.4 percentage
point during 2011.
Working Paper 11-8, “A Quantitative Analysis of
Unemployment Benefit Extensions,” Makoto Nakajima,
Federal Reserve Bank of Philadelphia
FIXED VS. FLOATING EXCHANGE
RATES: A RECONSIDERATION OF THE
CONVENTIONAL WISDOM
According to conventional wisdom, fiscal policy
is more effective under a fixed than under a flexible
exchange rate regime. In this paper the authors
reconsider the transmission of shocks to government
spending across these regimes within a standard New
Keynesian model of a small open economy. Because of
the stronger emphasis on intertemporal optimization,
the New Keynesian framework requires a precise
specification of fiscal and monetary policies, and their
interaction, at both short and long horizons. The
authors derive an analytical characterization of the
transmission mechanism of expansionary spending
policies under a peg, showing that the long-term real
30 Q2 2011 Business Review
interest rate always rises in response to an increase in
government spending if inflation rises initially. This
response drives down private demand even though
short-term real rates fall. As this need not be the case
under floating exchange rates, the conventional wisdom
needs to be qualified. Under plausible medium-term
fiscal policies, government spending is not necessarily
less expansionary under floating exchange rates.
Working Paper 11-9, “Floats, Pegs, and the
Transmission of Fiscal Policy,” Giancarlo Corsetti,
Cambridge University; Keith Kuester, Federal Reserve
Bank of Philadelphia; and Gernot J. Müller, University of
Bonn
USING NEW TIME SERIES TO STUDY THE UK
ECONOMY DURING WORLD WAR I AND THE
INTERWAR PERIOD
This article contributes new time series for studying
the UK economy during World War I and the interwar
period. The time series are per capita hours worked and
average capital income, labor income, and consumption
tax rates. Uninterrupted time series of these variables
are provided for an annual sample that runs from 1913
to 1938. The authors highlight the usefulness of these
time series with several empirical applications. The
per capita hours worked data are used in a growth
accounting exercise to measure the contributions of
capital, labor, and productivity to output growth. The
average tax rates are employed in a Bayesian model
averaging experiment to reevaluate the Benjamin and
Kochin (1979) regression.
Working Paper 11-10, “UK World War I and Interwar
Data for Business Cycle and Growth Analysis,” James M.
Nason, Federal Reserve Bank of Philadelphia, and Shaun
P. Vahey, Australian National University
COMPARING BORROWER OUTCOMES
AFTER DIFFERENT TYPES OF CREDIT
COUNSELING
This paper compares outcomes for borrowers who
received face-to-face credit counseling with similarly
situated consumers who opted for counseling via
the telephone or Internet. Counseling outcomes are
measured using consumer credit report attributes
one or more years following the original counseling.
The primary analysis uses data from a sample of
26,000 consumers who received credit counseling
either in-person or via the telephone during 2003. A
www.philadelphiafed.org
second sample of 12,000 clients counseled in 2005
and 2006 was provided by one of the agencies to
examine Internet delivery. Technology-assisted delivery
was found to generate outcomes no worse — and at
some margins better — than face-to-face delivery of
counseling services.
Working Paper 11-11, “Is Technology-Enhanced Credit
Counseling as Effective as In-Person Delivery?,” John
M. Barron, Purdue University, and Michael E. Staten,
University of Arizona, and Visiting Scholar, Federal
Reserve Bank of Philadelphia
EXPLORING THE LINK BETWEEN LENDERS’
HOUSE-PRICE EXPECTATIONS AND
SUBPRIME LENDING
This paper explores the link between the houseprice expectations of mortgage lenders and the extent
of subprime lending. It argues that bubble conditions
in the housing market are likely to spur subprime
lending, with favorable price expectations easing the
default concerns of lenders and thus increasing their
willingness to extend loans to risky borrowers. Since
the demand created by subprime lending feeds back
onto house prices, such lending also helps to fuel an
emerging housing bubble. The paper, however, focuses
on the reverse causal linkage, where subprime lending is
a consequence rather than a cause of bubble conditions.
These ideas are illustrated in a theoretical model, and
empirical work tests for a connection between price
expectations and the extent of subprime lending.
Working Paper 11-12, “Subprime Mortgages and
the Housing Bubble,” Jan K. Brueckner, University of
California—Irvine; Paul S. Calem, Board of Governors
of the Federal Reserve System; and Leonard I. Nakamura,
Federal Reserve Bank of Philadelphia
HOW IS THE RISKINESS OF THE POOL OF
HELOC ORIGINATIONS AFFECTED OVER
THE CREDIT CYCLE?
The authors empirically study how the underlying
riskiness of the pool of home equity line of credit
originations is affected over the credit cycle. Drawing
from the largest existing database of U.S. home
equity lines of credit, they use county-level aggregates
of these loans to estimate panel regressions on the
characteristics of the borrowers and their loans, and
competing risk hazard regressions on the outcomes of
the loans. The authors show that when the expected
www.philadelphiafed.org
unemployment risk of households increases, riskier
households tend to borrow more. As a consequence, the
pool of households that borrow on home equity lines of
credit worsens along both observable and unobservable
dimensions. This is an interesting example of a type
of dynamic adverse selection that can worsen the risk
characteristics of new lending, and suggests another
avenue by which the precautionary demand for liquidity
may affect borrowing.
Working Paper 11-13, “Credit Cycle and Adverse
Selection Effects in Consumer Credit Markets: Evidence
from the HELOC Market,” Paul Calem, Board of
Governors of the Federal Reserve System; Matthew
Cannon, CoreLogic; and Leonard Nakamura, Federal
Reserve Bank of Philadelphia
EFFECTS OF LENDERS’ ACCESS TO
INFORMATION ABOUT BORROWERS’ PAST
DEFAULTS
In many countries, lenders are restricted in their
access to information about borrowers’ past defaults.
The authors study this provision in a model of repeated
borrowing and lending with moral hazard and adverse
selection. They analyze its effects on borrowers’
incentives and access to credit and identify conditions
under which it is optimal. The authors argue that
“forgetting” must be the outcome of a regulatory
intervention by the government. Their model's
predictions are consistent with the cross-country
relationship between credit bureau regulations and
the provision of credit, as well as the evidence on the
impact of these regulations on borrowers’ and lenders’
behavior.
Working Paper 11-14, “Bankruptcy: Is It Enough to
Forgive or Must We Also Forget?,” Ronel Elul, Federal
Reserve Bank of Philadelphia, and Piero Gottardi,
European University Institute
TRACKING THE PATTERNS OF HOME
EQUITY WITHDRAWAL AMONG RETIREES
The authors study empirically and theoretically
the patterns of home equity withdrawal among retirees,
using a model in which retirees are able to own or rent
a home, save, and borrow against home equity, in the
face of idiosyncratic risks concerning mortality, health,
medical expenditures, and household size and observed
house price changes. The estimated model is found to
successfully replicate the patterns of homeownership
Business Review Q2 2011 31
and the saving/borrowing decisions of retirees. They
use the estimated model for several counterfactual
experiments. There are three main findings. First, the
model predicts that a house price boom suppresses
homeownership and increases borrowing, while a
decline in house prices has the opposite effect. Second,
the costs of home equity borrowing restrict the borrowing of retirees, and thus a reduction of such costs (e.g.,
lower costs of reverse mortgage loans) might significantly raise home equity borrowing. Third, there are two
implications for the retirement saving puzzle. Although
the cost of borrowing against equity in the house affects
the borrowing of retirees, it does not affect total asset
holding, implying that equity borrowing costs do not
seem to offer a quantitatively significant contribution
to resolving the retirement saving puzzle. On the other
hand, the magnitude of the retirement saving puzzle
might be exaggerated because a sizable part of “retirement saving” is due to house price appreciation.
Working Paper 11-15, “Home Equity Withdrawal in
Retirement,” Makoto Nakajima, Federal Reserve Bank of
Philadelphia, and Irina A. Telyukova, University of California—San Diego
32 Q2 2011 Business Review
SOURCES OF THE DECLINE IN EMPLOYMENT
VOLATILITY 1956-2002
This study documents a general decline in the volatility of employment growth during the period 1956 to
2002 and examines its possible sources. The authors use
a panel design that exploits the considerable state-level
variation in volatility during the period. The roles of
monetary policy, oil prices, industrial employment shifts
and a coincident index of business cycle variables are
explored. Overall, these four variables taken together
explain as much as 31 percent of the fluctuations in
employment growth volatility. Individually, each of the
four factors is found to have significantly contributed to
fluctuations in employment growth volatility, although
to differing degrees.
Working Paper 11-16, “The Long and Large Decline
in State Employment Growth Volatility,” Gerald Carlino,
Federal Reserve Bank of Philadelphia; Robert DeFina, Villanova University; and Keith Sill, Federal Reserve Bank of
Philadelphia
www.philadelphiafed.org
@FedFRASER