The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
EDITOR’S NOTE
The papers in this special volume of the Economic Policy Review all focus
on the theme of a 2009 conference on central bank liquidity tools
organized by the Federal Reserve Bank of New York: the evaluation of
central bank programs implemented to address funding shortages in
the markets. Indeed, readers interested in detailed summaries of the
conference papers and their discussions will find the overview by
Matthew Denes and his coauthors very informative.
Two of the papers presented at the conference are included in
this volume: the studies by Stephen G. Cecchetti and Piti Disyatat and
by Erhan Artuç and Selva Demiralp. Both papers examine the past
actions of central banks in the financial crisis. Cecchetti and Disyatat
consider the implications that recent financial developments may have
for the fundamental nature of central banks’ lender-of-last-resort
function and whether the traditional tools at policymakers’ disposal
remain effective in the face of modern liquidity crises. Artuç and
Demiralp investigate whether changes to the Federal Reserve’s
discount window borrowing facility represent a fundamental shift in
the way the Fed traditionally provided liquidity through the primary
credit facility as well as whether the Fed would be well served to retain
these changes to its borrowing facility indefinitely. A third paper,
submitted separately for this volume, also addresses the role of central
banks during the financial turmoil. Asani Sarkar and Jeffrey Shrader
examine the Federal Reserve’s recent actions in terms of the financial
amplification literature.
Three other papers, solicited for this volume, broaden the
discussion by providing perspectives on the future course of financial
policy in the post-crisis era. Matthew Pritsker offers a theoretical view
on the important topic of how regulators can improve the availability
of information; Viral V. Acharya, João A. C. Santos, and Tanju
Yorulmazer analyze ways to incorporate systemic risk into deposit
insurance premiums; and John Geanakoplos discusses implications
of the leverage cycle—whereby leverage is excessive prior to the crisis
and too low during the crisis—for regulatory policy and reform.
We hope you enjoy the rich perspectives offered in this special
volume of the Review.
—The Economic Policy Review Editorial Board
Central Bank Liquidity Tools
A Conference Sponsored by the Federal Reserve Bank of New York
February 19-20, 2009
Agenda
Thursday, February 19
8:45 a.m.
Opening Remarks
Patricia C. Mosser, Federal Reserve Bank of New York
9:00 a.m.
Session 1: Overview of Recent Problems in Liquidity Provision
Chair: Tobias Adrian, Federal Reserve Bank of New York
Central Bank Tools and Liquidity Shortages
Stephen G. Cecchetti, Bank for International Settlements
Piti Disyatat, Bank for International Settlements
Discussant: Bengt Holmstrom, Massachusetts Institute of Technology
10:30 a.m.
Session 2: Funding Liquidity and Market Liquidity
Chair: Til Schuermann, Federal Reserve Bank of New York
Leverage, Moral Hazard, and Liquidity
Viral V. Acharya, New York University and London Business School
S. “Vish” Viswanathan, Duke University
Discussant: Patrick Bolton, Columbia University
Interbank Market Liquidity and Central Bank Intervention
Franklin Allen, University of Pennsylvania
Elena Carletti, European University Institute
Douglas Gale, New York University
Discussant: Adriano A. Rampini, Duke University
Bank Liquidity, Interbank Markets, and Monetary Policy
Xavier Freixas, Universitat Pompeu Fabra
Antoine Martin, Federal Reserve Bank of New York
David Skeie, Federal Reserve Bank of New York
Discussant: Franklin Allen, University of Pennsylvania
2:00 p.m.
Session 3: Policy Responses to Illiquidity
Chair: James J. McAndrews, Federal Reserve Bank of New York
Illiquidity and Interest Rate Policy
Douglas W. Diamond, University of Chicago and National Bureau of Economic Research
Raghuram G. Rajan, University of Chicago and National Bureau of Economic Research
Discussant: Guido Lorenzoni, Massachusetts Institute of Technology
Liquidity Hoarding and Interbank Market Spreads: The Role of Counterparty Risk
Florian Heider, European Central Bank
Marie Hoerova, European Central Bank
Cornelia Holthausen, European Central Bank
Discussant: Gaetano Antinolfi, Washington University
FRBNY Economic Policy Review / August 2010
3
Agenda
Thursday, February 19 (Continued)
3:40 p.m.
Session 4: Collateral and Haircuts
Chair: Simon M. Potter, Federal Reserve Bank of New York
Rollover Risk and Market Freezes
Viral V. Acharya, New York University and London Business School
Douglas Gale, New York University
Tanju Yorulmazer, Federal Reserve Bank of New York
Discussant: Michael Manove, Boston University
Central Bank Haircut Policy
James Chapman, Bank of Canada
Jonathan Chiu, Bank of Canada
Miguel Molico, Bank of Canada
Discussant: Mitchell Berlin, Federal Reserve Bank of Philadelphia
6:00 p.m.
Keynote Address
John Geanakoplos, Yale University
Agenda
Friday, February 20
9:00 a.m.
Session 5: Empirical Evaluation of Central Bank Liquidity Programs—Part I
Chair: Seth B. Carpenter, Board of Governors of the Federal Reserve System
Do Central Bank Liquidity Facilities Affect Interbank Lending Rates?
Jens H. E. Christensen, Federal Reserve Bank of San Francisco
Jose A. Lopez, Federal Reserve Bank of San Francisco
Glenn D. Rudebusch, Federal Reserve Bank of San Francisco
Discussant: Pierre Collin-Dufresne, Columbia University
Repo Market Effects of the Term Securities Lending Facility
Michael Fleming, Federal Reserve Bank of New York
Warren Hrung, Federal Reserve Bank of New York
Frank Keane, Federal Reserve Bank of New York
Discussant: Lasse H. Pedersen, New York University
4
Conference Agenda
Agenda
Friday, February 20 (Continued)
10:40 a.m.
Session 6: Empirical Evaluation of Central Bank Liquidity Programs—Part II
Chair: James Vickery, Federal Reserve Bank of New York
Funding Liquidity Risk: Definition and Measurement
Mathias Drehmann, Bank for International Settlements
Kleopatra Nikolaou, European Central Bank
Discussant: Marie Hoerova, European Central Bank
Provision of Liquidity through the Primary Credit Facility during the Financial Crisis:
A Structural Analysis
Erhan Artuç, Koc University
Selva Demiralp, Koc University
Discussant: Carolyn Wilkins, Bank of Canada
1:15 p.m.
Panel Discussion
Chair: Patricia C. Mosser, Federal Reserve Bank of New York
Panel:
Louis Crandall, Wrightson ICAP
Andrew W. Lo, Massachusetts Institute of Technology
Paul Mercier, European Central Bank
Lasse H. Pedersen, New York University
W. Alexander Roever, J.P. Morgan Chase
FRBNY Economic Policy Review / August 2010
5
Patricia C. Mosser
Conference Opening Remarks
elcome to the Federal Reserve Bank of New York, and
thank you for coming to this conference on central bank
liquidity tools.
As acting manager of the Federal Reserve’s System Open
Market Account (SOMA), I am responsible for reporting to
policymakers on the implementation of monetary policy in
pursuit of the objectives that they have set. This includes the
ways in which the Fed’s balance sheet is being used as well as
the ways in which financial conditions are impacting both
the stance of monetary policy and its transmission to credit
markets. In recent months, of course, this has also included
the impact of what some have called our “alphabet soup” of
liquidity facilities and programs.
I am very pleased to lead off this conference—the first of
many conferences, I am sure—on central bank liquidity tools.
When the organizers put this conference together many
months ago, we knew there would be much to talk about. Little
did we know that the number of liquidity tools and the depth
of the financial crisis would continue to expand and to
challenge us in the intervening months.
The expansion of the Fed’s liquidity tools has been nothing
short of extraordinary. In normal times, we essentially use four
tools to manage the SOMA portfolio: temporary open market
operations (OMOs), permanent OMOs, the discount window,
and securities lending. By March 2008, when this conference
was organized, we had nine tools; now, if we include the Term
Asset-Backed Securities Loan Facility (TALF) and our new
purchase programs, we have sixteen according to my count.
W
We tend to group the Fed’s liquidity tools into three broad
categories. In the first group, we have facilities that provide
term liquidity to financial institutions—particularly to large,
systemically important ones. These exist to reduce the systemic
risk associated with the inability of a financial institution to get
wholesale funding, which could in turn lead to a widespread
deleveraging cycle involving forced asset sales that would
ultimately become self-reinforcing, particularly for the largest
financial institutions. In short, these facilities exist to forestall
runs. These include the Term Auction Facility (TAF), foreign
central bank swap lines, and the Primary Dealer Credit Facility
(PDCF).
In the second group, we have facilities that provide liquidity
directly to borrowers and lenders in key credit markets to prevent
further declines in credit formation. These include the TALF, the
Commercial Paper Funding Facility (CPFF), and the Money
Market Mutual Fund Investor Funding Facility (MMIFF).
In the third group, we have programs involving the direct
purchase of assets, particularly housing-related ones. These
include our purchases of agency debt and mortgage-backed
securities (MBS).
It is no accident that the Fed started with the first group
in the early stages of the crisis. When this conference was
organized, the Fed was addressing the crisis by rearranging its
balance sheet, expanding lending programs, and reducing its
holdings of Treasury securities. Many of the papers in this
conference directly address the use of these types of tools and
their links to funding and market liquidity issues.
Patricia C. Mosser is a senior vice president at the Federal Reserve Bank of
New York; she was acting manager of the Federal Reserve’s System Open
Market Account at the time these remarks were delivered.
The views expressed are those of the author and do not necessarily reflect
the position of the Federal Reserve Bank of New York or the Federal
Reserve System.
FRBNY Economic Policy Review / August 2010
7
Among the many issues that we are hoping this conference
will address are: What has the current crisis taught us about the
use and effectiveness of traditional and new liquidity tools? To
what extent might the expanded toolkits of central banks be
useful for policy implementation in normal circumstances?
Which tools are better kept as extraordinary measures?
Of course, last fall the balance-sheet constraints of large
financial firms and funding pressures became a full-blown
financial crisis with seriously impaired credit formation, a deep
recession, capital assistance to large banks, and a significant
feedback loop between financial and macroeconomic
weakness. In response, the Fed has begun to use the asset side
of its balance sheet to affect credit provision directly in key
markets, such as those for commercial paper and MBS. To the
extent possible, the Fed attempts to do this in a way that
improves market functioning and liquidity, in order to set the
stage for the private sector to return in the future. As a result,
our balance sheet has ballooned with the expansion of both the
size and number of our programs—our alphabet soup.
But a policy of credit easing in the currently very extreme
situation raises a host of questions that I encourage everyone
here to pursue in future research. Among these are: How can
we measure the effectiveness of such policies? In Chairman
Bernanke’s terminology, “How should the central bank think
about the impact and stance of monetary policy when pursuing
a policy of credit easing?” How does one think about the size of
the central bank’s balance sheet? For example, some of the
Fed’s facilities are designed to expand when credit and market
conditions deteriorate sharply, and to contract when
conditions improve. During the last few weeks, for instance,
the swaps program decreased by $150 billion.
As I mentioned at the outset, this will certainly not be the last
conference on this topic. It is fair to say that economists, central
bankers, and historians will be analyzing this financial crisis
and the policy responses to it for decades to come. Nonetheless,
we here at the Federal Reserve Bank of New York—who
sometimes feel we are in the trenches every day—appreciate
the insights that this conference can provide, preliminary
though they may be. Because we are so close to many of these
programs, we also appreciate the distance and perspective that
your research can give. We particularly look forward to your
future work in this area. I am guessing that central banks have
provided you with a rich research agenda.
Again, welcome, and thank you for coming.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
8
Conference Opening Remarks
Matthew Denes, Daniel Greenwald, Nicholas Klagge, Ging Cee Ng,
Jeffrey Shrader, Michael Sockin, and John Sporn1
Conference Overview and
Summary of Proceedings
1. Introduction
T
he financial crisis that emerged in 2007 had many and
varied causes, but one of its most consistent themes has
been the disappearance of liquidity. Indeed, in one of the first
manifestations of the crisis in August 2007, BNP Paribas
announced that it would suspend redemptions from three
hedge funds, noting that a “complete evaporation of liquidity
in certain market segments of the U.S. securitization market”
had made it impossible to value the funds’ assets. Since then,
much economic policymaking has been devoted to
understanding and combating liquidity shortages.1
Although the crisis began less than two years ago, a
significant body of academic work has already attempted to
understand and address its causes and symptoms. Indeed, in
February 2009, the Federal Reserve Bank of New York
organized a Central Bank Liquidity Tools Conference to bring
together some of the world’s leading experts on liquidity to
present their work and discuss its relevance and significance in
the context of the ongoing crisis. While the papers considered a
variety of topics, three critical and related questions unified the
discussions: How do we define and understand liquidity? What
are the causes and consequences of illiquidity? And what is the
proper regulatory response to issues of liquidity?
One goal was to set out a clear definition of liquidity and to
distinguish between different interpretations of the term. In
particular, there is “market liquidity,” which involves the
readiness with which firms can buy or sell assets; “funding
liquidity,” which involves the ability of firms to obtain funding
1
Introduction and panel discussion: Klagge; Session 1: Denes; Session 2: Sporn;
Session 3: Greenwald; Session 4: Sockin; Session 5: Ng; Session 6: Shrader.
Nicholas Klagge is a financial analyst and Ging Cee Ng, Michael Sockin, and
John Sporn are assistant economists at the Federal Reserve Bank of New York;
Matthew Denes is a former research associate and Daniel Greenwald and
Jeffrey Shrader are former assistant economists.
Correspondence: asani.sarkar@ny.frb.org
quickly and easily; and “central bank liquidity,” which involves
the ability of banks to easily borrow and lend reserve balances
at the central bank. Although each of these types of liquidity is
distinct, they are closely linked, and problems with one can
quickly cause problems with the others.
A second goal was to examine the causes and consequences of
liquidity shortages. Shocks to liquidity can be exacerbated,
perpetuated, and spread because of financial market frictions
such as balance-sheet constraints and the maturity mismatch
between assets and liabilities, potentially leading to difficulties in
rolling over sources of funding. In examining the consequences
of illiquidity, many academics have made reference to traditional
models of bank runs, updating them to account for the greater
complexities of the modern financial system. Another common
thread in the recent literature is the issue of systemic risk,
whereby financial market illiquidity can turn firm-specific
problems quickly into system-wide problems.
A third goal was to determine how central banks can best
respond to these problems. Common issues of concern
included the relative merits and effectiveness of ex ante policy
(addressing causes) and ex post policy (addressing
consequences), the need to define and measure policy goals in
the absence of a single clear target such as the overnight rate, and
the proper scope of financial regulation in a system where there
are many major players outside the traditional banking sector.
Ultimately, all of the papers presented sought to answer a
common question: What is the new “normal”? There is a broad
consensus that the post-crisis financial system will not look like
the pre-crisis system, as market participants and regulators
adjust to the issues raised by the present crisis. Because
illiquidity has played a key role in the crisis, an answer to this
The views expressed are those of the authors and do not necessarily reflect
the position of the Federal Reserve Bank of New York or the Federal
Reserve System.
FRBNY Economic Policy Review / August 2010
9
question requires us to develop a better understanding of the
nature of illiquidity, the role of illiquidity in the financial
system, and the most effective policy responses to illiquidity.
2. Session 1: Overview of Recent
Problems in Liquidity Provision
PAPER:
“Central Bank Tools and Liquidity Shortages”
Stephen G. Cecchetti, Bank for International Settlements
Piti Disyatat, Bank for International Settlements
DISCUSSANT:
Bengt Holmstrom, Massachusetts Institute of Technology
Cecchetti and Disyatat examine the role of central banks as
lenders of last resort. They distinguish three types of liquidity:
central bank liquidity, market liquidity, and funding liquidity.
Central bank liquidity consists of deposits from financial
institutions at a central bank, which are often called reserve or
settlement balances. Market liquidity is the ability of market
participants to buy and sell assets in relatively large quantities
without significantly influencing their market price. Funding
liquidity is the ability of an individual or institution to raise
cash by selling assets or borrowing.
Motivated by the definitions of liquidity, the authors
describe three kinds of liquidity shortages. The first is a
shortage of central bank liquidity, which occurs when
institutions find themselves short of the reserve balances that
they wish to hold. This shortage can be caused either by
insufficient aggregate supply of reserves or by problems related
to their distribution, and is not directly related to the solvency
of individual institutions. The second type is an acute shortage
of funding liquidity at a specific institution. This occurs when
an institution is unable to raise funds to meet its short-term
obligations, and is typically associated with solvency concerns.
The third type of liquidity shortage is a systemic shortage of
funding and market liquidity. This is potentially the most
harmful kind of liquidity shortage, and it arises when
coordination failures and an evaporation of confidence among
market participants lead to a breakdown of key financial
markets that affect many institutions simultaneously.
As a lender of last resort, a central bank has two main liquidity
tools: open market operations and institution-specific
transactions. In open market operations, a central bank lends and
borrows or buys and sells assets outright in the open market. In
addition, a central bank may also deal with specific institutions in
order to channel liquidity directly to them.
10
Conference Overview and Summary of Proceedings
The authors go on to examine the use of the two main
liquidity tools to address each type of liquidity shortage. If there
is a shortage of central bank liquidity, the primary aim of
central bank intervention is to maintain the smooth
functioning of the payments system and keep interest rates
near their targets. This is generally accomplished by open
market operations when aggregate supply shortages occur and
through discount window lending directly to specific
institutions when distribution problems arise. When a central
bank is confronted with an acute shortage of funding liquidity
at a specific institution, central bank support is designed to
contain potential contagion and spillover effects to the rest of
the financial system, hence forestalling an institution-specific
problem from becoming a systemic one. The response typically
takes the form of bridge financing in order to allow the
institution time for restructuring. In such situations, the
central bank must tactfully handle communication challenges
to support confidence while staving off panic. Finally, in the
face of a systemic shortage of funding and market liquidity, the
immediate objective of central bank intervention is to restore
market functioning and shore up confidence in the financial
system as a whole. This is likely to entail the broad provision of
liquidity to institutions as well as to specific markets.
In the current crisis, central banks have taken four major
steps to stem systemic shortages of funding and market
liquidity. First, they are providing backstop financing to
financial institutions. Second, central banks are supporting
term funding by lengthening the maturity on refinancing
operations and establishing swap lines between central banks.
Third, they are lending high-quality liquid securities against
lower quality, less liquid securities in an effort to bolster
markets for the latter and ease collateral constraints more
generally. Fourth, central banks are supplying credit to the
nonbank sector directly. These actions have significantly
increased the size of the balance sheets at many central banks,
including the Federal Reserve, the Bank of England, and the
European Central Bank (ECB).
Overall, Cecchetti and Disyatat conclude that the traditional
view of lender of last resort, as originally expounded by Walter
Bagehot, requires modification. Significantly, the appropriate
principles of lender-of-last-resort support by central banks
must be conditioned on the particular type of liquidity shortage
that is taking place. Moreover, given the complexities of the
modern financial system, with large interdependencies
between financial institutions and markets, the lender of last
resort may need to act to support not only institutions, but
certain markets as well.
Holmstrom—Cecchetti and Disyatat’s discussant—drew
lessons from the crisis on the relative merits of liquidity provision
and risk sharing. He motivated his remarks by discussing issues of
aggregate risk sharing, high demand for secure, liquid debt, and
the role of government in supplying and managing liquidity. He
began by noting that even though the originate-and-distribute
model may have led to weaker incentives to supervise lending
standards and tranching of mortgages, where risk is spread to
many investors, one should not jump to the conclusion that the
model is fundamentally flawed.
Holmstrom argued that lack of transparency is a significant
problem now, but that it is a standard, even essential, feature of
liquidity provision. A traditional bank has never been
transparent; there is no mark-to-market accounting and the
balance sheet is quite opaque. In analyzing the nature of
liquidity provision, it is important to recognize the high
velocity of credit markets, a feature that prevents investors
from evaluating the creditworthiness of investments. Such an
evaluation requires that agents have symmetric information
about the value of the instruments they are trading. The natural
way to achieve this is to create information-insensitive
instruments, such as debt, where agents rely on coarse ratings
rather than detailed information about the assets supporting
the debt. Securitization and limited transparency are logical
steps to reduce information intensity.
The current crisis has been spurred by the symbiotic
relationship between excess foreign demand for savings and
demand for subprime loans. However, while the originate-anddistribute model has the ability to distribute systemic risk, it is
now apparent that this risk was not always distributed to those
who wanted to bear it and was in many cases held by liquidity
providers. The distribution of systemic tail risk is the major flaw
in the system, and it arises because systemic risk is not
appropriately priced into liquidity-providing markets. This is a
major challenge going forward. The government also has a role
in providing insurance against systemic risk by injecting liquidity
when there are large negative aggregate shocks. Public insurance
is more efficient than private insurance for rare events, since the
government can insure ex post, while private markets have to
arrange insurance ex ante.
Holmstrom concluded by observing that, in an ideal world,
all idiosyncratic risk would be eliminated through diversification
and systemic risk would be borne by everyone in proportion to
their risk tolerance. No crisis would ever occur in that case. The
reality is far from this ideal, because information and incentive
problems lead to an enormous demand for riskless debt. Though
the originate-and-distribute model could be a step toward the
ideal, and it has been useful in the industry, the problem with
systemic tail risk needs to be resolved. As part of the solution,
there should be a greater focus on regulation of leverage, as well
as maturity mismatches.
One participant asked if it was logical for central banks to
charge lower haircuts than the market does. Holmstrom
responded that his presentation focused on redistribution of
aggregate risk and did not incorporate information on haircuts.
Another asked how to overcome issues of adverse selection in
the securities markets. Holmstrom noted that new innovation
has failed to get beyond this problem. Cecchetti mentioned that
the originate-and-distribute model allows the provider to keep
good assets while selling off bad ones. The same participant
observed that private providers are not ideal for offering
insurance for catastrophic events. Holmstrom indicated that
there is some scope for private insurance, but also for
government insurance. Cecchetti added that this insurance
cannot be supplied by private entities at a reasonable price.
The last question related to why over-the-counter markets
have been disrupted. Cecchetti said he felt that most securities
should be forced onto exchanges. A standardized market
structure would be much more resilient.
3. Session 2: Funding Liquidity
and Market Liquidity
PAPERS:
“Leverage, Moral Hazard, and Liquidity”
Viral V. Acharya, New York University
and London Business School
S. “Vish” Viswanathan, Duke University
DISCUSSANT:
Patrick Bolton, Columbia University
“Interbank Market Liquidity and Central Bank Intervention”
Franklin Allen, University of Pennsylvania
Elena Carletti, European University Institute
Douglas Gale, New York University
DISCUSSANT:
Adriano A. Rampini, Duke University
“Bank Liquidity, Interbank Markets, and Monetary Policy”
Xavier Freixas, Universitat Pompeu Fabra
Antoine Martin, Federal Reserve Bank of New York
David Skeie, Federal Reserve Bank of New York
DISCUSSANT:
Franklin Allen, University of Pennsylvania
FRBNY Economic Policy Review / August 2010
11
3.1 Acharya and Viswanathan
Acharya and Viswanathan address a phenomenon that appears
during times of financial shock: the evaporation of liquidity.
Liquidity was plentiful prior to the crisis, and the problem was
not one of hoarding cash, but rather, which asset class would
absorb the demand from yield-seeking investors. With the
onset of the crisis, however, risk aversion swept through the
financial sector. The authors argue that the short-term debt
with which balance sheets had been financed was a possible
contributor to the market freeze. Firms were dependent on the
ability to raise or roll over short-term debt collateralized by
assets, as well as short-term unsecured commercial paper. If
firms faced liquidation risk, these assets would have to be sold
at “fire-sale” prices that would be much lower than the assets’
fair value. Moreover, the inability of firms to roll over their
existing debt would place a high burden on their ability to
cover liabilities, necessitating fire sales.
Acharya and Viswanathan present two possible
explanations for the increasing amount of leverage firms
carried. The first holds that the downward trend in volatility
prior to the crisis—a phenomenon that has been called the
“Great Moderation”—led to rapid growth and increased
issuance of inexpensive debt. The second explanation centers
on the notion of a “credit bubble” characterized by light
regulation and risk taking among financiers. The paper
provides a model capturing the first theory.
In the model, short-term rollover debt is an optimal form of
financing and the risk-shifting problem tied to leverage limits
the funding of financial institutions that are reliant upon
trading. The model revolves around one parameter: the
maximum borrowing allowable as a result of the ex post risk
shifting.
The key result attempts to explain why adverse shocks
preceded by a prosperous economy tend to be much more
severe. The authors state that when times are good, borrowing
is inexpensive and even firms with low capitalization levels can
leverage themselves in the market. Thus, ex ante there are more
firms that are highly leveraged in the financial sector when
times are good, and as a result there is not much spare debt
capacity ex post in the event of a financial crisis. Only firms that
are not highly leveraged during prosperous economic times
have enough spare debt capacity to buy debt from other firms.
Margin borrowing is usually very high during a prosperous
economy, and as a result, prices are much lower during a
subsequent crisis because once the adverse shocks materialize,
there is a much deeper deleveraging in the economy. The asset
substitution problem plays a key role, because it potentially
rations firms when they are faced with the burden of raising
12
Conference Overview and Summary of Proceedings
cash. In such an environment, the only feasible option is to sell
assets. The authors endogenize both the debt market and asset
market and examine the implications for prices. They also
argue that hard debt contracts and collateral requirements give
lenders higher recoveries and raise prices, outcomes that make
raising debt desirable ex ante.
In his discussion, Bolton related this topic to the theory of
lending booms and liquidity crises. He summarized the
Acharya-Viswanathan paper as follows: the main premise is
that firms may engage in asset sales to meet debt obligations.
The buyers of the assets, however, have limited purchasing
power because of the liquidity shock. The prices are
determined by supply and demand and by the distribution of
leverage in the industry. In a boom, increasing profitability
leads to lower demand for outside liquidity, which is followed
by higher asset prices. Because of greater entry into the market
of lower quality assets, however, there is a larger collapse in asset
prices when a negative shock occurs. Bolton also commented on
the fact that the model does not have any losers ex ante, and that
liquidity crises involve no inefficiencies ex post.
3.2 Allen, Carletti, and Gale
Allen, Carletti, and Gale focus on the interbank market. They
begin by explaining that under normal circumstances, the
interbank market works smoothly. Under some circumstances,
however, it ceases to function properly. As a result, central
banks intervene in the market in an attempt to stabilize prices
and correct market inefficiencies.
The paper develops a simple theoretical framework for
analyzing interbank markets and how central banks should
intervene through open market operations. Banks use the
interbank market to hedge against idiosyncratic and aggregate
liquidity shocks. Hedging opportunities are, however, limited
and markets are incomplete. This implies that market
allocations are inefficient, as they entail excess price volatility
and thus consumption volatility across banks. This is the only
market failure in the model. The authors show that, by
conducting open market operations and fixing the interest rate
in the interbank market, the central bank can implement the
constrained optimal allocation, where all banks can offer the
same consumption to their late depositors irrespective of the
idiosyncratic liquidity shock they face. The central bank is
coupled with a fiscal authority that imposes lump-sum taxes on
(or provides transfers to) depositors to acquire the short (or
long) asset at the initial date and can give a lump-sum transfer
to (or impose a tax on) the later consumers at the final date.
Allen, Carletti, and Gale show that the exact nature of central
bank intervention depends on the type of shocks banks face
and on the initial contract that banks promise to their
depositors.
Discussant Rampini observed that “market freezes” in the
context of the paper manifest themselves through a lack of
trade when all banks have excess liquidity and the central bank
drains excess liquidity by selling the long asset. He considered
this an interesting, albeit somewhat unconventional, notion of
a market freeze. Rampini also argued that the central bank
policy proposed encompasses aspects of fiscal policy, and that
the paper might thus provide a guide to the possibility of
monetary and fiscal policy working in conjunction during a
financial crisis.
3.3 Freixas, Martin, and Skeie
The final paper, by Freixas, Martin, and Skeie, begins by
examining the role of central bank policy in the face of crisis.
One view maintains that the central bank should focus on
inflation and output in the medium and long run and not
respond to the crisis directly. However, in the past, central
banks have aggressively lowered interest rates during crises.
During financial disruptions, banks usually face
considerable uncertainty with regard to their demand for
liquid assets. A state-dependent interest rate, which is low
during times of shock and high during a strong economy, can
help mitigate the risks associated with idiosyncratic shocks.
The paper argues that monetary policy plays a crucial role in
setting low interest rates to facilitate the redistribution of
liquidity during a crisis.
In the authors’ model, the interest rate in the interbank
market plays an important role in two ways. Ex ante, high
interest rates are beneficial because they ensure that banks hold
enough liquid assets, as it is expensive to acquire such assets in
the interbank market. Ex post, however, interest rates need to
be low when an idiosyncratic shock hits to facilitate trading in
the interbank market. Redistribution of liquidity and high
levels of interbank risk sharing are now necessary for the
banking sector. The main challenge for a central bank is to set
the right balance between high expected rates ex ante and low
rates ex post in times of crisis.
Allen’s discussion first reviewed the authors’ model and
then showed its relationship to the traditional model of
Diamond and Dybvig (1983). Allen also pinpointed the
innovative addition to the new model: having two states with
different idiosyncratic bank shocks. An important point was
also raised on the issue of how the central bank should set
interest rates. According to Allen, these models are very
important because they are a building block for understanding
the complexities surrounding both market failures and
stability. In light of the crisis, these models can provide clarity
and a possible course of government intervention.
4. Session 3: Policy Responses
to Illiquidity
PAPERS:
“Illiquidity and Interest Rate Policy”
Douglas W. Diamond, University of Chicago
and National Bureau of Economic Research
Raghuram G. Rajan, University of Chicago
and National Bureau of Economic Research
DISCUSSANT:
Guido Lorenzoni, Massachusetts Institute of Technology
“Liquidity Hoarding and Interbank Market Spreads:
The Role of Counterparty Risk”
Florian Heider, European Central Bank
Marie Hoerova, European Central Bank
Cornelia Holthausen, European Central Bank
DISCUSSANT:
Gaetano Antinolfi, Washington University
4.1 Diamond and Rajan
Diamond and Rajan investigate the relationship between
interest rates and the incentives facing banks regarding illiquid
investments. Their work contributes to the longstanding
debate between those who believe, like Alan Greenspan, that
the Federal Reserve cannot prevent asset price bubbles, only
mitigate their consequences, and those who believe that
asymmetric interest rate policy can encourage behavior that
makes booms and busts more likely.
The authors create a model in which entrepreneurs who
invest in long-term projects must borrow from banks that in
turn borrow from risk-averse households. In the model, there
is no uncertainty about the profitability of projects, which are
predetermined, but there is uncertainty about the households’
income in each period. Liquidity problems can emerge if
households have an unexpectedly high need to withdraw
deposits. This, they assert, can occur either because of an
unexpected decrease in present income or an increase in
expected future income. With a decrease in present income,
FRBNY Economic Policy Review / August 2010
13
households face a higher marginal utility of consumption and
may want to spend their financial assets in order to consume
more today. If, however, households expect significantly higher
income in the future, they may spend their assets today in order
to smooth lifetime consumption.
In either case, unanticipated demand for funds can force
banks to call in loans for long-term projects early. As a result,
the real interest rate must rise in order to equalize household
demand for consumption goods and the supply of
consumption goods from terminated projects whose loans
have been called in. This in turn decreases bank net worth,
since a bank’s loans, which pay off only in the long run, fall in
value as the real interest rate rises, but the bank’s liabilities of
demandable deposits do not have a corresponding fall in value.
If the bank’s net worth becomes negative, the bank can
experience runs, which can be highly inefficient when they
cause the terminations of otherwise profitable projects
financed by bank loans. Thus, an increase in households’
withdrawals, owing either to a current decrease in income or to
a future increase in income, can create fragility in the banking
system that harms the real economy.
One solution to this problem would be to change the
structure of banks so that they were less reliant on demandable
deposits for funding. However, such a change would be very
difficult, as Diamond and Rajan, citing their past work, note.
The authors assert that demandable debt is the cheapest form
of financing available to banks, and that using more long-term
liabilities that are not demandable would reduce the efficiency
of intermediation substantially. Changing the sources of banks’
funds is therefore not viewed as a viable option to reduce
fragility in the banking system.
Another option is to use government intervention to
attempt to stabilize the banking system and prevent bank runs.
As a first possibility, governments can intervene by taxing
households and giving the proceeds directly to banks. But while
such a bailout scenario could certainly be effective in
preventing bank runs and might be necessary in times of crisis
such as the present, Diamond and Rajan argue that the severity
with which property rights are violated under these policies
makes them unsuitable for frequent use.
Instead, they consider an alternative policy measure in
which the government lends or borrows in the market in an
attempt to alter interest rates, and apply this type of policy
to their model. Diamond and Rajan first note that since
government action must be financed by tax revenues, there
are potential issues of Ricardian equivalence. If the
government seeks to lower interest rates by lending funds, it
must raise these funds by increasing taxes. When a
household’s taxes are raised, however, the household is
likely to increase its withdrawals in order to make up for the
14
Conference Overview and Summary of Proceedings
current decline in income, as mentioned earlier, which
would counterproductively push interest rates back up.
The authors’ model shows that as long as the government
finances its lending by taxing only households with deposits,
with the level of deposits exceeding the size of the tax, there is
zero effect on the interest rate. As a result, government
intervention is likely to be ineffective when most or all
households hold large amounts of demandable deposits relative
to the size of the tax. However, if there are households that do
not hold deposits, or if the level of the tax exceeds the amount of
the households’ deposit holdings, then the government action
does have a marginal effect in the model, lowering the real
interest rate and increasing banks’ net worth. Thus, although
households’ actions in response to a government intervention
may reduce its effectiveness, the intervention should still be
effective, provided that it is large enough.
Next, Diamond and Rajan note while there can be benefits
to influencing household and bank behavior if it prevents bank
runs, it is also likely that altering these decisions can have
negative effects. In the model, the authors consider both an
“entrepreneur-friendly” central bank that seeks to lower
interest rates as much as possible and a “household-friendly”
central bank that seeks to raise interest rates as much as
possible. They demonstrate that each type of central bank can
have negative effects when its action is anticipated, even on the
group that it attempted to benefit, owing to the distortions in
behavior that it creates.
Finally, the authors argue that when government policy is
anticipated, it can have an important impact on how banks
choose to allocate their portfolios between liquid and illiquid
investments. In the model, they assume that the government
commits to lowering interest rates in case of liquidity problems
and find that this encourages banks to take on more deposits
and to finance more illiquid projects, making liquidity
shortages more likely. As a result, they claim that commitment
to a “one-sided” policy to intervene only to lower interest rates
when they are too high can lead to distortions in bank decisions
that can have a strongly counterproductive effect and make
liquidity crises much more likely.
For this reason, Diamond and Rajan assert that an optimal
interest rate policy must not only prevent bank runs by
lowering interest rates in times of crisis, but also encourage
banks to make more liquid loans to prevent distortion. To this
end, the central bank should pursue a “two-sided” policy of
interventions, in which the bank not only acts to lower interest
rates to prevent runs when rates are too high, but also pushes
interest rates up when the interest rate would otherwise be low.
This type of intervention would punish illiquid banks, forcing
them to call in loans and decreasing their net worth, but would
not raise rates so much as to cause bank runs. Appropriately
implemented, this incentive against illiquidity could balance
out the incentive in favor of illiquidity caused by the central
bank’s commitment to lower interest rates in times of crisis.
Such a two-sided policy could therefore prevent distortions
and allow banks to make an efficient allocation between liquid
and illiquid investments while still allowing the central bank to
intervene in order to prevent harmful bank runs.
Lorenzoni, in his discussion, offered an adaptation of the
basic model presented by Diamond and Rajan. In the original
model, a bank choosing to liquidate an entrepreneur’s project
must liquidate it entirely. Lorenzoni presented a model of
partial liquidation, in which the bank can choose to terminate
only part of a project early for an immediate payoff, leaving the
rest to mature in the final period.
In this variation, the payoff that the bank gets for a project
that is not completely liquidated is assumed to be a concave
function that represents diminishing returns to the proportion
of the original loan still invested in the project (that is, the
proportion not liquidated). When this payoff is combined with
terms representing the returns from liquidation and the cost of
paying interest on deposits, a profit function for banks can be
formed. First-order conditions can then be taken to find a
bank’s optimal policy with regard to liquidation. Lorenzoni
found three possible regimes, depending on the interest rate: a
no-liquidation regime at a low interest rate, a completeliquidation regime at a high interest rate, and a partialliquidation regime at an interest rate between the two extremes.
The discussant then created a supply function by optimizing
consumers’ utility with respect to the amount of funds loaned
over the two periods and combines it with the demand
function to find the market equilibrium. The result is that in an
“exuberant” state, in which consumers’ second-period
endowments turn out to be very high, the equilibrium interest
rate is also high, because consumers require larger incentives to
transfer consumption from the first period to the second. If the
equilibrium rate is high enough in this scenario, it can lead to a
regime in which no lending takes place and banks go bankrupt
and default on their debt as a result.
Lorenzoni incorporated the government into the modified
model. The government taxes consumers and lends out tax
revenues to banks. Once the loans are repaid, the government
returns the tax revenues, plus interest, to the consumers. If
consumers are free to optimize over any quantity of lending,
including negative quantities (meaning that the consumers
borrow from the banks), then households will simply adjust
their lending to offset the tax. Government intervention
therefore has no effect on the net supply of funds, which is
independent of the size of the tax, and Ricardian equivalence
holds. However, if a constraint is imposed that households may
lend but may not borrow (that is, the amount of lending must
be non-negative), then government intervention may have an
effect on the interest rate. Specifically, if the size of the tax is
larger than the supply of loans under the initial equilibrium so
that consumers cannot simply decrease their lending to offset
the tax, then such a policy will reduce market interest rates.
Lorenzoni then turned to the issue of the optimal choice of
banks’ initial short-term debt, from the standpoint of
maximizing expected payment to customers. More debt
increases the probability of inefficient bankruptcy, but also
increases the payment to consumers in nonbankruptcy states.
The optimal level of debt must therefore find an equilibrium that
balances these two opposing forces in favor of the consumer.
The issue of moral hazard was also considered. Lorenzoni
assumed that the government intervenes ex post to protect
banks in the “exuberant” state. But if this can be expected
ahead of time, the level of debt that banks will take on increases
endogenously. It is also possible, Lorenzoni asserted, for this
distortion to make all parties worse off, reinforcing the
potential problems of government intervention posed by
Diamond and Rajan in their original model.
Overall, the partial-liquidation version of the model
adapted by Lorenzoni is consistent with the main findings of
Diamond and Rajan. This is especially true regarding the
benefits and dangers of interest rate interventions not driven by
cyclical conditions. Therefore, the powerful ex ante effects of
moral hazard and reverse moral hazard present in the initial
version of the model are maintained under the assumption of
partial liquidation.
4.2 Heider, Hoerova, and Holthausen
The session’s second paper sought to explain the recent
tensions and eventual breakdown in the unsecured interbank
lending market in a number of countries around the world.
Much more so than in the past, banks have been keeping
liquidity on their accounts rather than lending excess funds on
the interbank market. Authors Heider, Hoerova, and
Holthausen identify this phenomenon as a clear failure of the
interbank market to efficiently redistribute liquidity.
To explain these developments, they present a three-period
model based on adverse selection caused by the asymmetric
information between banks regarding the risk of illiquid assets.
In the first period, banks must allocate their funds between a
risk-free liquid asset and a risky illiquid asset. The liquid asset
pays off in the next period exactly what was put into it, and is
essentially a form of storage. The illiquid asset will either have
a high return R if it succeeds, or a return of zero if it fails. The
size of the return R is known and is the same for all banks. The
FRBNY Economic Policy Review / August 2010
15
probability of success varies across banks, but is unknown to
banks in the first period. It is assumed that the expected return
from the illiquid asset is greater than 1, making it larger than
the return to the liquid asset.
In the second period, banks face a “liquidity shock” in which
either a small or large amount of deposits is withdrawn by
consumers, which the banks must pay. Banks with a shortage of
liquidity (large withdrawals) can borrow from other banks that
have excess liquidity (small withdrawals), thus forming an
interbank market. However, banks also receive private
information as to whether their illiquid assets are riskier (with
a lower probability of success) or safer (with a higher
probability of success) than expected. If banks have a shortage
of liquidity to pay depositors, they may drop out of the
unsecured interbank market and convert their illiquid assets
into liquidity at a cost. Riskier assets are more illiquid, so banks
with safer assets have better opportunities to obtain (costly)
funding outside the unsecured market.
In the third period, the illiquid assets either succeed or fail,
and interbank loans are repaid when possible. Since the illiquid
asset has zero return when it fails, interbank loans are not
repaid when the borrower’s illiquid asset does not succeed.
This potential for default leads to counterparty risk in the
interbank market.
The study focuses on the role of asymmetric information
about counterparty risk in the functioning of the unsecured
interbank market. Banks with a liquidity shortage have a choice
between borrowing and converting their illiquid assets into
liquidity at a cost. Since safer assets are more liquid than riskier
assets, banks with safe assets will require a lower interbank
interest rate than banks with risky assets to be willing to stay in
the unsecured interbank market. If the interest rate is higher
than what the safe borrowers are willing to pay, they will drop
out of the market. However, the risky borrowers may still be
willing to pay this higher interest rate, leading to a scenario of
adverse selection.
Depending on parameters, reflecting in particular the level
and distribution of counterparty risk among banks, three
different equilibrium “regimes” can arise in the interbank
market. Under the first regime, there is full participation in the
interbank market, and banks do not need to resort to converting
their illiquid assets into liquidity. This is typically the case when
there are low levels of counterparty risk and thus low interbank
interest rates, preventing adverse selection. Under the second
regime, the interbank interest rate is high enough that the safe
borrowers are no longer willing to participate. However, there is
still a market to provide unsecured loans to risky borrowers
willing to pay a higher interest rate. This is the regime in which
adverse selection takes place.
16
Conference Overview and Summary of Proceedings
In the third regime, the interbank market breaks down. This
can occur for one of two reasons. In the first case, the banks with
excess liquidity can refuse to lend, and “hoard” their liquidity
instead. A necessary condition for this to occur is that the illiquid
asset that turns out to be riskier than expected is unprofitable in
expected value. Still, the ex ante expected return on the illiquid
asset is positive and dominates the rate of return on the liquid
asset. In the second case, banks with excess liquidity may be
willing to make loans to the banks with risky assets, but the
market interest rate may be so high that even the risky banks
prefer to drop out of the unsecured interbank market.
Heider, Hoerova, and Holthausen then compare the results
of their model with empirical evidence from the three-month
unsecured interbank market in the euro area from July 2006
and January 2009. They argue that the interbank market did in
fact exhibit the three regimes described above as both the
perceived level and dispersion of risk associated with banks’
illiquid assets rose. The authors first examine the spread
between the three-month unsecured interbank rate in the euro
area (Euribor) and the overnight index swap (OIS) in three
months’ time to show changes in the interbank interest rate.
They also look at the use of the ECB’s deposit facility, where
banks can place their excess funds, but which offers a lower
interest rate than does the interbank market, to demonstrate
liquidity hoarding.
In the first phase, beginning in July 2006, the authors note
both a very low spread and an insignificant utilization of the
deposit facility, consistent with a “full-participation” regime.
In the second phase, beginning in August 2007, the spread rises
significantly, but the deposit facility is still used very rarely,
which they argue is consistent with an “adverse selection”
regime, in which only the “riskier” banks, lacking good-quality
collateral to borrow in the repo market, are willing to pay such
high interest rates in the unsecured interbank market. In the
third phase, beginning in September 2008, the interest rate
increases further, and use of the deposit facility increases
dramatically, showing a breakdown of the interbank lending
market and large amounts of hoarding behavior. The authors
also show that a similar pattern of the three-month interbank
market spread can be observed in the United States in the
aforementioned time period.
Heider, Hoerova, and Holthausen conclude by identifying
policy interventions that could reduce or prevent adverse
selection and thereby increase the efficiency of the interbank
market. These are divided into two types of interventions: ex ante
policies to prevent a dropping out of the good risks from the
unsecured market, and ex post policies to restore the
effectiveness of the interbank market after an unexpected
increase in counterparty risk.
On the ex ante side, the study offers two options: liquidity
requirements and improved transparency. Under the liquidity
requirements option, there would be a limit on the amount of
illiquid assets that banks would be permitted to hold at any
given time. This would generally provide banks with more
liquidity, reducing the demand for liquidity in the interbank
market. As a result, the interbank interest rate would fall, which
would make all banks, particularly banks with safe assets, more
willing to borrow. This outcome in turn would ensure the full
participation of banks in the unsecured market and,
consequently, its smooth functioning. The downside of such a
policy is that with less of the illiquid assets held, banks would
receive lower returns on average from their investments,
because of distortions in banks’ optimal portfolio allocation.
Under the improved transparency option, the government
would work to make banks’ private information about their
portfolios more public. This could allow for banks with excess
liquidity to distinguish between safe and risky lenders, and
potentially offer different lending terms to each. It would
prevent adverse selection, as safe banks with a liquidity
shortage would no longer be pooled with riskier banks and
could instead pay a lower rate that reflects the reduced
counterparty risk taken on by the lending bank. Therefore,
improved transparency could also facilitate interbank lending
and reduce early liquidations.
On the ex post side, the authors present three policy
alternatives for situations when interbank market functioning
has already been impaired. First, the central bank can directly
provide liquidity to banks. This, they argue, can be profitable
for all parties involved, since the central bank can raise liquidity
at a unit cost by “printing money,” in contrast to the private
supply of liquidity that must compete with the returns offered
by the illiquid asset. By supplying liquidity to banks in need, the
central bank could crowd out the private supply of liquidity.
Heider, Hoerova, and Holthausen argue that as a result, the
central bank could offer to take on liquidity from the banks
with excess liquidity. In this case, the central bank would act as
an intermediary in the interbank market.
A second option is for the central bank to guarantee
interbank loans. This would reduce or eliminate counterparty
risk and make banks with excess liquidity more willing to lend
in the interbank market. It would in turn reduce the interbank
interest rate, which would increase borrowing and potentially
reduce adverse selection in the interbank market. However,
such guarantees are costly and must be designed optimally to
minimize the overall costs to the guarantor.
The third option is asset purchases, in which the
government directly purchases illiquid assets from banks. Since
the government can afford to purchase the assets at their
expected value, this would prevent banks from having to sell at
fire-sale prices, which occurs when the amount of illiquid assets
being sold in order to convert them into liquidity exceeds the
amount of liquidity available to purchase them. Such a measure
would not increase interbank lending, and would in fact likely
discourage it, but the measure would reduce the losses faced by
banks that would otherwise have to sell assets at a price
significantly below their expected value.
Antinolfi’s discussion offered a number of avenues for
further inquiry using Heider, Hoerova, and Holthausen’s
model. First, he examined the issue of the deposit arrangement
within the model. The question was posed as to whether the
deposit contract as specified is actually optimal, or if a better
arrangement could be found. Antinolfi also considered the
issue of deposit insurance. Whether deposit insurance is
provided, how much is provided, and who pays for it could all
have an important impact on outcomes in the model.
Next, Antinolfi considered the informational aspect of the
model. The adverse selection in the model is entirely driven by
private information held by banks about their assets that is not
available to the public. Therefore, it is important to make sure
that it is reasonable to assume that banks can in fact ascertain
their own “type” while keeping it unknown to potential lenders.
Finally, the discussant suggested that the authors or
future researchers look into the structure of the banking
sector. The model assumes perfect competition, but it might
yield different results under another arrangement, such as
monopoly or oligopoly.
5. Session 4: Collateral and Haircuts
PAPERS:
“Rollover Risk and Market Freezes”
Viral V. Acharya, New York University
and London Business School
Douglas Gale, New York University
Tanju Yorulmazer, Federal Reserve Bank of New York
DISCUSSANT:
Michael Manove, Boston University
“Central Bank Haircut Policy”
James Chapman, Bank of Canada
Jonathan Chiu, Bank of Canada
Miguel Molico, Bank of Canada
DISCUSSANT:
Mitchell Berlin, Federal Reserve Bank of Philadelphia
FRBNY Economic Policy Review / August 2010
17
5.1 Overview
A conference session on collateral and haircuts featured two
papers examining the theoretical underpinnings of the market
for secured short-term debt.2 Many financial institutions rely
on overnight or short-term secured lending to meet their
liquidity needs and finance longer maturity assets. The
counterparty in these loans is often a central bank or market
participant such as a bank, a money market mutual fund, or an
institutional investor. Collateral used to secure these loans can
vary from Treasury and agency debt securities to corporate
bonds, equities, and bank loans. To protect the lender from
changes in the collateral’s value, an initial discount, or
“haircut,” is applied to the value of the asset that can be
borrowed against, hereafter referred to as the asset’s debt
capacity. The optimal choice of haircuts for central banks is the
topic of the paper by Chapman, Chiu, and Molico while
Acharya, Gale, and Yorulmazer explore changes in an asset’s
debt capacity when the debt must be rolled over.3
In the interbank market, secured lending takes the form of
repurchase agreements, or repos.4 Repos typically have a
maturity ranging from overnight to fourteen days. A central
bank can provide intraday liquidity to financial institutions
through repos and, as Chapman, Chiu, and Molico suggest,
affect the supply of liquidity in the market through its choice of
haircuts. The authors develop a general equilibrium
formulation for the optimal level of haircuts in the presence of
agent liquidity constraints, liquidity shocks, and asset price
volatility. Their model stipulates that haircuts are higher when
a central bank cannot exclusively lend to agents with liquidity
constraints, and that a sudden, temporary increase in haircuts
can be welfare-improving.
Acharya, Gale, and Yorulmazer attempt to explain how
markets for collateralized lending can fail as a result of rollover
risk, the risk that short-term debt cannot be rolled over and the
sponsoring institution will have to sell the underlying asset in a
fire sale. By constructing a regime-switching model for how
investors perceive expectations on news, the authors
demonstrate how an asset’s debt capacity can decline without a
change in its fundamental value and raise the issuing firm’s
counterparty credit risk. This adverse event is equivalent to an
2
Secured lending differs from unsecured lending in that an asset with low
credit risk is pledged by the borrower as collateral to be seized in the event of
default. This form of lending allows an institution to borrow at more attractive
interest rates with a debt ceiling not limited by its own credit risk.
3
Since the maturity of short-term debt in commercial paper markets is often
less than the maturity of the asset being financed, the debt must be reissued, or
“rolled over,” to new investors until the asset matures.
4
In a repurchase agreement, the lender purchases the posted collateral at a
discount and agrees to sell it back at a later date at a higher price that includes
the interest on the loan.
18
Conference Overview and Summary of Proceedings
increase in haircuts and can help explain the market dislocation
observed in the asset-backed commercial paper market during
the subprime mortgage crisis beginning in 2007.
As the discussion following each presentation highlighted,
the issues of liquidity and risk management arising from
maturity mismatch and market shocks in secured lending are
nontrivial. Short-term financing ex ante with loans secured
by assets whose fundamental value is not resolved until ex
post creates uncertainty over ultimate payoffs endogenous to
default and counterparty credit risk. Since the debt capacity
of an asset can change over time, it is important to understand
what drives these changes and how to manage the risks from
both the borrower’s and the lender’s perspective. The
inability to sufficiently manage these risks can lead to
depreciation of both liquid and illiquid assets, unforeseen
liquidity constraints, and catastrophic market failure. The
papers presented draw attention to important considerations
for regulators with regard to participation and intervention in
these markets.
5.2 Acharya, Gale, and Yorulmazer
Acharya, Gale, and Yorulmazer examine how changes in
investor expectations in secured short-term lending markets
can lead to market freezes. The authors focus on the market for
asset-backed commercial paper, where debt must be rolled
over several times before the underlying asset matures and its
value is realized. They construct a regime-switching model for
two possible states of the world, denoted as the “optimistic”
and “pessimistic” states (defined later), and explore how the
debt capacity of an asset changes as debt is rolled over in each
state. The study concludes that the debt capacity of an asset is
determined by the terminal state, where it tends to its
fundamental value if the state of the world is optimistic and
zero if the state of the world is pessimistic. This last result can
explain how short-term debt markets can freeze regardless of
the credit risk of the underlying asset.
The authors interpret their model in the context of a special
investment vehicle that finances an asset-backed security by
issuing short-term debt that must be rolled over a finite
number of times before the asset matures. There exist two
states of the world for investor expectations: an optimistic state
where “no news is good news” and a pessimistic state where
“no news is bad news,” which can switch with a fixed
probability each period. In the optimistic state, by backward
induction, the debt capacity increases with each rollover to
match the asset’s fundamental value. In the pessimistic state,
similarly, the debt capacity tends to zero and leads to a market
freeze, wherein the sponsoring bank takes the asset back onto
its balance sheet and sells it in a fire sale.
Based upon these results, the authors propose an explicit
formula for collateral haircuts by solving for the pledged asset’s
debt capacity. As the number of rollovers becomes unbounded in
the pessimistic state, haircuts tend to reach 100 percent as long as
the recovery rate on the asset is less than full recovery. One policy
implication of these results is that firm failure from market freezes
can potentially be avoided if regulators monitor firm capital
structure for excessive reliance on short-term debt that entails
rollover risk. Another implication is that regulators could help
thaw market freezes by lending against the asset as collateral based
on its value if held to maturity without risk of liquidation.
The ensuing discussion centered on the results of the
model’s pessimistic state. As Manove observed, one
implication is that removing risk of default in one period will
not prevent default in future periods once the asset’s debt
capacity is on the default trajectory. While the paper showed
that mismatching maturities by financing long-term
investments with short-term debt can lead to market failure,
Manove noted that using long-term debt to finance long-term
investments lacked the benefits described in the DiamondDybvig (1983) model. In addition, reducing rollover risk by
financing with more unleveraged equity would be less
profitable than debt financing.
Examining the policy implications of the paper, participants
discussed whether regulators could reduce liquidation costs by
swapping assets for more liquid instruments in addition to
lending against them at their value if held to maturity.
Regulators could also limit leverage by requiring firms to
maintain a minimum level of equity financing. Drawing
parallels with the Diamond-Dybvig model, Manove also
compared the market freezes described in the paper with bank
runs. When one views short-term lenders as depositors and
long-term assets as bank loans, a situation such as a market
freeze in secured lending markets can be seen as analogous to a
bank run. Consequently, if creating deposit insurance through
the Federal Deposit Insurance Corporation (FDIC) helped
prevent bank runs, establishing similar insurance in the
secured lending markets could perhaps prevent market freezes.
5.3 Chapman, Chiu, and Molico
Chapman, Chiu, and Molico examine the optimal central bank
haircut policy for the Canadian Large-Value Payment System.
The authors develop a discrete-time three-market model for an
illiquid and a liquid asset with anonymous agents that face
portfolio allocation uncertainty. They find that central bank
liquidity facilities provide insurance against both liquidity and
downside asset risk, and that setting a haircut involves a tradeoff between satisfying agent liquidity constraints and
depreciation of the liquid asset. This depreciation can lead to
portfolio distortions and increased probability of default on
collateralized loans.
In the first subperiod of each period of the model, agents
choose portfolios of the two assets in an asset market based on
a signal as to whether they will be buyers or sellers in the second
subperiod. In the second subperiod, agents reform portfolios
based on the realization of their type in a decentralized market.
This reformation can lead to liquidity constraints that agents
satisfy with collateralized loans from the central bank. In the
third subperiod, the illiquid asset’s value is resolved and agents
choose whether or not to settle their loans or default in a
centralized market. The optimal choice for central bank
haircuts on collateralized loans minimizes the incidence of
default while providing financing to constrained agents.
The results of this model suggest that haircuts are higher
when a central bank cannot identify which agents actually need
liquidity. In addition, a relationship is established between
collateral haircuts and the nominal interest rate, which is
affected by the injection of the liquid asset into the market
through collateralized loans. As haircuts are lowered, defaults
create inflationary pressure by making this injection
permanent. Lowering haircuts relative to the interest rate also
erodes the liquid asset’s value by making the illiquid asset less
costly to hold.
Chapman, Chiu, and Molico’s paper elicited discussion about
the topic’s relevance in the context of recent changes in central
bank collateral policies brought on by the crisis. This included
the expansion of the Federal Reserve’s range of lending facilities
and the European Central Bank’s concern about accepting too
wide a range of collateral. Berlin, leading the discussion,
emphasized providing more empirical interpretation of the
paper’s assumptions and conclusion. He stipulated methods for
measuring the relevant quantities in determining haircuts and
reconsidering the assumption of endogenous default probability
that can lead to strategic default.
From a policy perspective, participants discussed methods
for refining collateral policies in light of the results of the paper.
Berlin, for instance, suggested that discriminating between
potential borrowers based on measures of liquidity or balancesheet signals could potentially lead to a better outcome by
effectively providing liquidity to constrained agents. He also
introduced the possibility of charging higher borrowing rates
to banks with more illiquid balance sheets and making loan
payments contingent on investment returns to mitigate the
impact of the liquidity injected into the system when defaulted
assets are sold. Participants highlighted the paper’s practical
FRBNY Economic Policy Review / August 2010
19
implications in quantifying the impact of market forces driving
haircut policy calibration.
6. Session 5: Empirical Evaluation
of Central Bank Liquidity
Programs—Part 1
PAPERS:
“Do Central Bank Liquidity Facilities Affect Interbank
Lending Rates?”
Jens H. E. Christensen, Federal Reserve Bank
of San Francisco
Jose A. Lopez, Federal Reserve Bank of San Francisco
Glenn D. Rudebusch, Federal Reserve Bank of San Francisco
DISCUSSANT:
Pierre Collin-Dufresne, Columbia University
“Repo Market Effects of the Term Securities Lending Facility”
Michael Fleming, Federal Reserve Bank of New York
Warren Hrung, Federal Reserve Bank of New York
Frank Keane, Federal Reserve Bank of New York
Discussant:
Lasse H. Pedersen, New York University
6.1 Christensen, Lopez, and Rudebusch
Lopez, presenting on behalf of coauthors Christensen and
Rudebusch, examines the effects of central bank liquidity
operations on interbank lending rates using an arbitrage-free
term structure model that controls for fluctuations in the U.S.
Treasury yield curve and the term structure of risk in financial
corporate bond yields. The paper concludes that central bank
liquidity operations at the close of 2007 helped to lower term
interbank lending rates.
Motivating this paper were the large fluctuations in spreads of
the three-month Libor (London interbank offered rate) over
Treasury yields in mid-December 2007, when the Federal
Reserve introduced two major liquidity operations: reciprocal
swap lines with the European Central Bank and the Swiss
National Bank, and the Term Auction Facility (TAF) program,
whereby the Federal Reserve auctions collateralized loans to
banks facing liquidity constraints. The goal of the Christensen,
Lopez, and Rudebusch paper was to determine if these central
bank policy actions helped increase bank liquidity, reduce
liquidity risk premiums, and thus lower Libor rates.
20
Conference Overview and Summary of Proceedings
Fluctuations in the spread of the three-month Libor over
Treasuries are commonly attributed to movements in credit
and liquidity risk premiums. The authors account for credit
risk premiums by using the entire Treasury curve to control for
risk-free rates and the term structure of financial corporate
debt to control for credit risk premiums. In practice, Treasury
bonds are considered free from credit risk and the most liquid
debt instrument available. The key assumption for the latter is
that Libor rates have credit risk characteristics similar to senior,
unsecured AA-rated debt issued by U.S. financial firms.
Controlling for credit risk allows the authors to isolate
movements attributable to liquidity risk premia in interbank
lending rates.
The authors use a six-factor affine arbitrage-free joint model
of Treasury yields, financial bond yields, and Libor rates. The
Treasury yield curve accounts for three factors: the level, slope,
and curvature. Since movements in Treasury, bank bond, and
Libor rates all share common elements, two of the remaining
factors account for differences between bank debt yields and
Treasuries (levels, slope). The last factor captures the
idiosyncratic nature of term Libor rates, which the authors
assume is independent of the other five factors.
The model specification draws on four major assumptions.
The first is that the Libor-specific factor is independent of the
other factors. The second is that the Treasury level factor is
independent and has no dynamic interaction with the two
credit spread risk factors. The third assumption allows the
Treasury level and curvature factors to individually affect the
Treasury slope factor, but not each other. The fourth posits that
there is no feedback from the credit risk level factor to the
Treasury curvature factor or from the credit risk slope factor to
the Treasury slope factor. The likelihood ratio test on the
specification with the independent Libor factor results in a
failure to reject the null hypothesis that these additional zero
restrictions are reasonable.
The paper presents three major results from the preferred
model specification. First, the persistence of shocks was
generally quite high, although much less for the Libor-specific
factor. Second, the effects of Treasury factors on credit risk
factors seem limited. Third, credit risk factors do have an
influence on Treasury slope and curvature factors.
The presentation focuses on results that had implications
for the interbank market. The estimated Libor-specific factor
had been relatively stable around its historical mean in the precrisis period, but dropped more than two standard deviations
below its mean after the first TAF auction on December 17,
2007. To test the hypothesis that this drop represented a
structural break in the Libor factor, the authors use the Kalman
filter and impose different parameters in the pre and post
periods, at December 21, 2007. The likelihood ratio test rejects
the null hypothesis that no break occurred. The authors find
that the data support the conclusion that central bank liquidity
operations had an effect on the Libor-specific factor after the
first TAF auction had taken place.
Lastly, the authors consider the counterfactual situation—
what if the central bank effects on the Libor-specific factor were
“turned off”?—to determine the magnitude of the effect of
central bank liquidity operations. They generate a
counterfactual Libor path by setting the Libor-specific factor
constant at its mean after December 21, 2007. The average
difference between the observed and counterfactual threemonth Libor spread to Treasuries in the post-crisis period is
more than 70 basis points. This provides additional evidence
suggesting that central bank liquidity operations lowered
interbank lending rates.
In conclusion, the authors find that the results from their
six-factor model demonstrated that the TAF auctions
significantly affected the dynamics of the interbank market via
the structural break in the behavior of the model-implied Libor
factor, and that these operations kept the Libor rate roughly
70 basis points lower than it could have been in their absence.
Discussant Collin-Dufresne questioned the assumptions
and methodology of the paper. He wondered what was driving
the difference between Libor and AA-rated bank yields and
how various possible explanations would influence
interpretation of the results.
Collin-Dufresne also questioned whether an affine model was
well suited for a regime shift since affine models tend to need a
lot of data. Given that much of the activity was found in the
second half of the sample, he wondered if the model would have
picked up a structural break at any point in the second half, and
how intrinsically significant the post-TAF date was compared
with other dates in the post-crisis period (that is, if causality
could be established between the TAF and the regime shift). In
addition, he conjectured a regime shift in the underlying
Treasury rates, implying that the graph of the agency-Treasury
spread may represent anticipation in the market.
6.2 Fleming, Hrung, and Keane
Fleming presented on behalf of coauthors Hrung and Keane.
The presentation focused on the effects of the Term Securities
Lending Facility (TSLF), introduced by the Federal Reserve in
March 2008 to improve liquidity in the financing markets for
Treasury and other collateral. In particular, the paper examines
the supply effects of the program on rates and spreads in the
repurchase agreement (repo) market. The authors find that the
TSLF led to a significant narrowing of spreads between
Treasury (higher quality) collateral and lower quality collateral.
The Federal Reserve introduced the TSLF in the midst of
turbulent financial markets to help promote the liquidity of
secured funding markets. The program auctions loans of
Treasury securities to primary dealers for a period of twentyeight days in exchange for lower quality collateral that, owing
to stressed market conditions, would otherwise be difficult or
unattractive to finance. The TSLF thereby increases the ability
of dealers to obtain financing, especially dealers relying on the
repo market for financing of less liquid collateral.
In addition to improving dealers’ financing capacity, the
TSLF can potentially affect rates in the repo market by altering
collateral supplies. By allowing dealers to swap lower quality
collateral for Treasury securities, the TSLF increases the supply
of Treasury collateral in the market and decreases the supply of
lower quality collateral. The additional Treasury collateral
available to the market is hypothesized to put upward pressure
on Treasury general collateral repo rates while the reduction in
lower quality collateral is hypothesized to put downward
pressure on repo rates for such collateral.
The data examined by the authors cover all thirty-seven
TSLF operations from March 27, 2008, to October 30, 2008.
The authors also use repo rates for Treasury securities, agency
debt securities, and agency mortgage-backed securities (MBS)
from the Federal Reserve Bank of New York’s Trading Desk
and Bloomberg. Additional data employed include Treasury
issuances/redemptions and corporate yield spreads.
Fleming, Hrung, and Keane regress changes in overnight
repo rates and spreads on changes in the amount outstanding
under the TSLF. They focus on settlement days because TSLFinduced changes in the supply of securities should affect
overnight repo rates on those days. The dependent variable,
changes in the amount outstanding under the TSLF, is
calculated as the amount awarded in the operation settling that
day less the amount maturing that day. Dummy variables are
also included for the last and first days of the quarter, on which
repo spreads typically widen and narrow, respectively.
The authors find that the TSLF does in fact narrow
financing spreads between Treasury collateral and lower
quality collateral. Further, the observed narrowing is driven
by an increase in Treasury repo rates as opposed to a decrease
in rates on lower quality collateral. Financing spreads also
widen and narrow on the last and first days of the quarter,
as expected.
Additional results show that the effects of the TSLF are
driven by “Schedule 2” operations, in which dealers can pledge
a wide range of collateral, as opposed to “Schedule 1”
operations, in which eligible collateral is limited to Treasury
securities, agency debt securities, and agency MBS. The results
suggest that that agency debt and agency MBS collateral may be
considered substitutes for Treasury collateral to a large degree,
FRBNY Economic Policy Review / August 2010
21
whereas the lower quality collateral that can be pledged in
Schedule 2 operations is not.
A final set of results shows that the effects of the TSLF on
repo rates and spreads increase with the spread between the fed
funds rate and the Treasury general collateral repo rate. That is,
changes in the amount of collateral made available to the
market have more of an effect when the Treasury repo rate is far
below the fed funds rate rather than when it is close to the rate.
Pedersen’s discussion highlighted the statistical significance
of Schedule 2 collateral and the statistical insignificance of
Schedule 1 collateral, which led him to posit that agency and
agency MBS behave more like Treasuries than the other lower
quality collateral in Schedule 2.
Pedersen maintained that repo spreads are generally mean
reverting, and thus controls are necessary for the level of repos
and repo spreads. He also questioned whether the quantity of
Treasury securities provided by the TSLF is endogenous to the
repo rates and spreads: Do high repo spreads lead to a large TSLF
amount? Is the large reduction in repo spreads due to general
mean reversion or to the TSLF auction? Fleming responded that
he and his coauthors consider their results robust.
Lastly, Pedersen addressed what he thought was the big
question: Does the TSLF help solve the banks’ funding problems
and break liquidity spirals? He questioned whether the results of
increased repo rates under the TSLF alleviated liquidity problems.
The question-and-answer session centered on Pedersen’s
“big question” of whether the TSLF effectively achieved its
program goals. One participant asked whether the Federal
Reserve can effectively work only with primary dealers and
banks to reduce haircuts in the repo market, or whether it
should consider dealing with investors. Other participants
observed that the TSLF is about switching good and bad
collateral, as opposed to reducing haircuts, and urged that the
intent of the program be kept in mind. Pedersen, by contrast,
argued that the program is directly about reducing haircuts,
and that the question is whether or not the Federal Reserve has
been successful in doing that.
7. Session 6: Empirical Evaluation
of Central Bank Liquidity
Programs—Part 2
PAPERS:
“Funding Liquidity Risk: Definition and Measurement”
Mathias Drehmann, Bank for International Settlements
Kleopatra Nikolaou, European Central Bank
DISCUSSANT:
Marie Hoerova, European Central Bank
22
Conference Overview and Summary of Proceedings
“Provision of Liquidity through the Primary Credit Facility
during the Financial Crisis: A Structural Analysis”
Erhan Artuç, Koc University
Selva Demiralp, Koc University
DISCUSSANT:
Carolyn Wilkins, Bank of Canada
7.1 Drehmann and Nikolaou
Throughout the current crisis, central banks have introduced
facilities aimed at addressing liquidity shortages in financial
markets. Despite liquidity’s centrality to the crisis policy
response, however, a debate continues on the term’s precise
definition. Drehmann, presenting on behalf of coauthor
Nikolaou, set out to define one aspect of liquidity: funding
liquidity risk. His presentation focused on providing and
testing a definition of funding liquidity risk that could be
constructed from public information by central banks.
Drehmann and Nikolaou define funding liquidity as the
“ability to satisfy demand for money with immediacy.”
Consequently, funding liquidity risk reflects the potential
inability of a bank to meet money demand over some future
period. With this definition in hand, they laid out the theory
and construction of a publicly available proxy for funding
liquidity risk based on information available from open market
operations in the euro area. The measure is based on the theory
that, in turbulent times (that is, in the presence of market
frictions potentially resulting from asymmetric information,
incomplete markets, and issues of market power), a bank with
a greater need for liquidity will bid more aggressively for
liquidity at the central bank auctions. By looking at the spread
between a bank’s average bid rate and the policy rate weighted
by the volume in a price-discriminating auction, the authors
argue that central bankers can easily construct a measure of
liquidity risk for each bank or for the system as a whole.
To test their measure of funding liquidity risk, Drehmann
and Nikolaou exploit the theoretical relationship between
market liquidity and funding liquidity. Some financial theory
shows that as funding liquidity risk rises and market frictions
become important, downward spirals of funding and market
liquidity can occur. Using an average of liquidity proxies for
market liquidity in various markets as a proxy for overall
market liquidity, Drehmann and Nikolaou demonstrate that
their measure of funding liquidity risk does have the negative
relationship with market liquidity suggested by theory.
Hoerova’s comments on Drehmann and Nikolaou’s
measure focused on data issues and alternative theoretical
considerations. Hoerova pointed out that the measure
proposed by Drehmann and Nikolaou suffers from a number
of potential biases, including selection issues and problems
related to construction. Selection bias could occur because the
choice by banks to participate in the auctions is nonrandom
and likely influenced by liquidity conditions. Furthermore, by
summing across the value-weighted spread for all banks, the
proposed measure could overstate the influence of outliers.
The theoretical concerns focused on factors driving bank
bidding behavior. A bank could potentially increase its bid rate
for a number of reasons unrelated to liquidity risk, such as risk
aversion, differences in the personal value of collateral, and the
need for “window dressing” around important regulatory
dates. Finally, Hoerova suggested that the authors look at
alternative measures of market liquidity when documenting
their downward liquidity spirals.
7.2 Artuç and Demiralp
Alongside the need for new data to evaluate central bank
facilities, another critical issue is the construction of
counterfactuals. What would the world have looked like in the
absence of certain policies or if the credit crisis had manifested
itself in alternative ways? The Federal Reserve made a number
of changes to the discount window during the crisis, including
reductions in the penalty rate and an increase in borrowing
terms. Artuç and Demiralp use model-based counterfactual
estimation to examine the impact of these policy changes.
From the data, it is clear that banks responded to the
discount window changes by increasing their borrowing
substantially, but it is also clear that some cost or stigma was
still associated with discount window borrowing because
many banks were seeking funds in the interbank market at
rates above the discount rate. These trends lead one to
wonder how effective the policy changes were in reducing
market stress during the credit crisis. Using a structural
model of the fed funds market based on each bank’s desire
to hold certain daily and maintenance-period– wide levels of
reserves, Artuç and Demiralp estimate the impact of
aggregate shocks to, and changes in, borrowing terms at the
discount window. They compare these estimates with
simulations in which the cost of borrowing remained
unchanged during the crisis period.
Based on the difference between these two models, Artuç
and Demiralp find that the Federal Reserve’s changes to the
discount window were generally, though not universally,
effective. Namely, the most effective policy changes were the
lengthening of the term of discount window loans and the
addition of new eligible collateral. Less effective were the
reductions in the spread between the target fed funds rate and
the primary credit rate.
In her discussion, Wilkins pointed out three potential
shortcomings of this approach to assessing the Federal Reserve
policy changes. First, although the structural model helps
clarify assumptions and allows for the construction of a
counterfactual, some changes remained potentially conflated.
In particular, the implicit cost of borrowing from the discount
window could come from many sources aside from the stigma
cited by Artuç and Demiralp, and certain assumptions such as
the static nature of the model might not hold in reality. Second,
Wilkins questioned the estimation used to calibrate the model.
From the charts presented by Artuç and Demiralp, it appears
that some discrepancies exist between the in-sample estimation
and the observed data. Also, alternative estimation strategies
were not compared with the one used by the authors. Third,
Wilkins wondered if other changes were occurring aside from
a simple doubling of aggregate shocks that should be included
in a model of the crisis period. Most notably, collateral costs
were likely changing over the period and other Federal Reserve
programs, such as the Term Auction Facility, were introduced
to offer additional nonmarket funds to banks. Overall,
however, Wilkins emphasized the importance of the policy
questions raised by Artuç and Demiralp.
8. Panel Discussion
CHAIR:
Patricia C. Mosser, Federal Reserve Bank of New York
PANELISTS:
Louis Crandall, Wrightson ICAP
Andrew W. Lo, Massachusetts Institute of Technology
Paul Mercier, European Central Bank
Lasse H. Pedersen, New York University
W. Alexander Roever, J.P. Morgan Chase
The final event of the conference brought together participants
from the private sector, academia, and central banking to
discuss the crisis and the policy response. Mosser, moderating
the panel, gave participants the freedom to choose topics of
interest, but she began the session by posing the overarching
question: What are the key policy lessons learned from the
crisis so far?
The panelists represented a broad cross section of
perspectives on the financial world: Louis Crandall, chief
economist at broker Wrightson ICAP; Andrew W. Lo, a
professor at MIT’s Sloan School of Business; Paul Mercier,
FRBNY Economic Policy Review / August 2010
23
deputy director general of market operations at the European
Central Bank; Lasse H. Pedersen, a professor at NYU’s Stern
School of Business; and W. Alexander Roever, a debt strategist
at J.P. Morgan Chase’s short-term fixed-income sales and
trading desk. Each panelist gave a presentation with his
perspective on Mosser’s initial question; the panel then opened
itself to questions from the audience.
As a fixed-income strategist, Roever focused on the
contribution of short-term debt markets to the crisis. He first
demonstrated the massive growth in debt markets over the
years leading up to the crisis, showing that the U.S. bond and
money markets grew 2.5 times faster than GDP from 1998 to
2007. However, Roever said that the figures on money markets
do not include debt issued at floating rates indexed to the
Libor, which are a close substitute for money market funding,
with many of the same characteristics. This development
involved not only an increase in leverage on the part of
financial firms issuing the debt, but also an increased reliance
on a small set of firms, which Roever termed “liquidity
investors,” encompassing money market funds and other
short-term investors with low risk appetites. Within this
particular class of investors, Roever showed, assets are heavily
concentrated in a very small number of the largest firms. Thus,
the risk associated with high levels of leverage was magnified by
borrowers’ reliance on a narrow group of firms for funding.
The crisis thus far has destroyed a large amount of these firms’
assets under management, with Roever estimating the overall
figure at $2 trillion. This decrease in wealth meant a sudden
drop in the amount of money available to fund other financial
firms through money markets, asset-backed securities, and
other short-term debt, exacerbating the other problems of the
crisis. Roever’s primary conclusion from this narrative was that
the scope of financial regulation has been too narrow, and
should be expanded beyond banks to encompass a larger
number of participants in the financial system.
Pedersen, whose research focuses on liquidity risk, spoke on
the issue of systemic risk, and what central banks and other
regulators could do to address it. He began by arguing that the
recent crisis, for all its severity, was not a new kind of crisis—
that the issues of market liquidity and funding liquidity that
came to the fore during the last several months have always
been important for financial stability. The key issue, he said,
was the systemic component of risk, which he defined as “the
joint failure of a significant part of the financial institutions.”
Among the drivers of this risk were liquidity spirals—the way
declines in asset prices can increase the need of financial
institutions for liquidity, causing massive simultaneous sales
and further drops in asset prices. To highlight the difference
between systemic and idiosyncratic risk, Pedersen contrasted
the 2008 failure of Lehman Brothers, with all its systemic
24
Conference Overview and Summary of Proceedings
consequences, with the 1995 failure of the London merchant
bank Barings, which was a large institution, but which had
relatively minor systemic consequences. The response of
regulators, said Pedersen, should be to model and regulate
systemic risk explicitly, treating it as a negative externality like
pollution. He suggested that regulators run simulations of
1 percent systemic tail risk scenarios, gauging institutions’
contributions to losses. Guided by these assessments,
regulators should then impose a systemic capital requirement,
systemic risk fees after the model of the FDIC, and required loss
insurance policies that would be provided by a combination of
the government and the private sector. This set of policies,
Pedersen argued, would introduce incentives to limit systemic
risk and reduce the cost and disruption of bailouts when they
become necessary.
Mercier, whose position at the ECB affords him firsthand
knowledge of the central bank’s transactions with banks,
commented primarily on the structure through which the ECB,
and central banks in general, inject liquidity into the financial
system. Mercier considered a precise concept of liquidity,
defined simply as central bank credit. Under “normal”
financial conditions, he said, the central bank relies on a small
group of large and influential banks to further distribute
central bank credit to the rest of the system. In the euro area,
this group of banks is much larger than the Federal Reserve’s
set of primary dealers. With banks hoarding liquidity and the
subsequent seizing up of interbank markets, however, Mercier
noted that this standard practice started to lose its effectiveness,
causing the ECB to lose some control over short-term interest
rates. This led the ECB to implement a second regime, in which
it made no net change to liquidity over its maintenance
periods, but rather frontloaded its injections of liquidity to
provide banks and, by extension, their counterparties with
more certainty. As the crisis intensified after the collapse of
Lehman Brothers, however, the ECB implemented a third
regime, marked by fixed-rate tenders of unlimited quantity,
which did in fact create a gross increase in liquidity in the system.
In net terms, however, there was no increase in liquidity because
net demand remained unchanged (except for the increase in
banknotes in circulation). While some banks were borrowing
more from the Eurosystem, others were increasing their
deposits. Both sides of the balance sheet of the Eurosystem
increased, leading to a wider exposure toward the banking
system. In essence, the Eurosystem became a major intermediary
between banks that were reluctant to lend to each other.
While this third regime has apparently been effective in
providing financial institutions with needed liquidity, observed
Mercier, it has come at the cost of reduced central bank
influence over money market lending rates. Mercier pointed to
two further lessons from the crisis: first, market psychology
plays a significant role and policymakers need to take it into
account, and second, while it is important to consider what
central banks’ “exit policy” from the crisis should be, it is
equally important to consider what a new stable equilibrium
would look like—as he put it, “an exit to what?”
Lo based his presentation on the premise that financial
crises are unavoidable because of two factors: first, fear and
greed are natural parts of human behavior, and second, the
economy is and should continue to be based on free markets.
As a result of these unavoidable factors, policy efforts should
focus on developing early warning systems for impending
financial crises and developing measures to address them when
they do occur. Much of Lo’s recent work has focused on the
role of hedge funds in the economy, and he noted that they,
along with proprietary trading desks at other financial
institutions, generally exhibit early warning signs of impending
crises, and that regulators should look to glean information
from their activities in the markets. On the question of what
new measures regulators should develop to handle financial
crises when they do occur, Lo emphasized the necessity of
creating a different kind of regulation, rather than just more
regulation. After all, he noted, banking and insurance are two of
the most highly regulated sectors of the U.S. economy, yet they
still played major roles in the recent financial crisis. A major
problem with existing regulation, said Lo, is that the main
language used for regulation is the language of accounting,
which is not well-suited for talking about risk. Accounting, he
argued, is fundamentally focused on backward-looking
realizations, while financial regulation needs to be focused on
risk, which is a fundamentally forward-looking concept.
Crandall, the final presenter, mainly addressed the issue of
the currency composition of liquidity. He showed a graph
demonstrating the enormous increase in U.S. dollar funds sent
from U.S. branches of foreign banks to their home offices over
the course of the crisis, as it became more and more difficult for
the home offices to obtain U.S. dollar funding in the interbank
market. He then showed how this large increase was
significantly mitigated by the removal of size limitations on the
Federal Reserve’s reciprocal currency arrangements with four
major foreign central banks, which provide a nonmarket
channel through which foreign financial institutions can access
U.S. dollar funding. The lesson from this example, according to
Crandall, is that the currency composition of a bank’s liquidity
profile matters. He argued in favor of making the reciprocal
currency arrangements permanent, saying that the fixed-rate
unlimited-quantity auctions conducted by foreign central
banks using the reciprocal currency arrangements had
represented a crucial psychological change in financial
markets, essentially giving every bank in the world access to a
“discount window” denominated in U.S. dollars. Second,
Crandall identified one significant limitation facing
policymakers: central banks only have the power to incentivize
banks, rather than bankers themselves. He pointed out that
within banks, profits are socialized (to the bank as a whole),
whereas losses are privatized (putting the individual’s job at
risk). This makes bankers very risk-averse in the sense of being
unwilling to learn about new things if they are not directly
profitable. Crandall noted that liquidity facilities become more
effective as market participants learn more, but that bankers
are not paid to learn about these facilities. This poses a special
challenge in short-term markets, where less attention may be
paid to in-depth research and learning.
A short question-and-answer session concluded the
conference. One topic of further discussion was the “exit
strategy” that Mercier had brought up in his comments. The
participants talked about how long the Federal Reserve and
other central banks should wait before revoking current
liquidity facilities, many of which are legally allowed to
continue only as long as “unusual and exigent circumstances”
persist. There was broad agreement that the facilities should
remain in place for some time, even after circumstances appear
to have stabilized. The panelists noted that many of the
facilities are “self-liquidating” because they lend freely but at
penalty rates, meaning that market participants will stop
turning to them as conditions normalize. Crandall argued
specifically that the facilities should remain in place through
the period when the Federal Reserve begins to raise rates again.
This would do much to instill confidence and remove
uncertainty associated with monetary tightening.
The participants also went on to discuss the topic of “greed
and fear” that Lo had raised, especially the extent to which such
irrational motivations could play a role in creating financial
crises. Pedersen suggested that the key shortcoming of the
neoclassical model, which posits that irrational agents cannot
move markets away from equilibrium as long as there are a
small number of rational traders participating, is that agents
have funding liquidity constraints. As evidence that funding
constraints have recently been binding, he pointed out that
covered interest rate parity has been failing for the major
currencies because of limited availability of capital and limited
willingness to lend, consistent with the idea that liquidity
spirals are important drivers of the crisis. Lo cited Keynes’
comment that “the market can stay irrational longer than you
can stay solvent.” More specifically, he pointed out that the
neoclassical model requires the posited rational arbitrageurs to
have infinite liquidity, which is a particularly unrealistic
assumption during financial crises.
FRBNY Economic Policy Review / August 2010
25
References
Diamond, D. W., and P. H. Dybvig. 1983. “Bank Runs, Deposit
Insurance, and Liquidity.” Journal of Political Economy 91,
no. 3 (June): 401-19.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
26
Conference Overview and Summary of Proceedings
Stephen G. Cecchetti and Piti Disyatat
Central Bank Tools
and Liquidity Shortages
1. Introduction
T
he global financial crisis that began in mid-2007 has
renewed concerns about financial instability and focused
attention on the fundamental role of central banks in
preventing and managing systemic crises. In response to the
turmoil, central banks have made extensive use of both new and
existing tools for supplying central bank money to financial
institutions and markets. Against this backdrop, there has been
intense interest in the implications that recent financial
developments may have for the fundamental nature of central
banks’ lender-of-last-resort (LOLR) function and whether the
traditional tools that have been at policymakers’ disposal
remain adequate in the face of modern liquidity crises. This
paper addresses these issues, and in doing so provides a view of
recent central bank liquidity operations that is tied more closely
to their underlying purpose from the LOLR perspective.
We begin in Section 2 by defining three types of liquidity
shortages that central banks may need to address in operations
aimed at stabilizing the financial system. In taking this
approach, we emphasize the fact that the conditions under
which central bank liquidity—reserves or central bank
money—is made available should, and do, differ depending on
the underlying nature of the problem officials are trying to
mitigate. This means that there may not be a single set of
principles for central banks’ LOLR function. Recognizing this
goes some way toward reconciling the debate surrounding the
appropriate role of LOLR.1
Stephen G. Cecchetti is economic adviser and head of the Monetary and
Economic Department of the Bank for International Settlements; Piti Disyatat
is a senior economist in the Monetary and Economic Department.
stephen.cecchetti@bis.org
piti.disyatat@bis.org
After providing our definitions, in Section 3 we proceed
with a discussion of the tools that central banks have at their
disposal and how they might be tailored to address each type of
liquidity shortage. Section 4 offers a brief description of how
recent actions by major central banks can be interpreted from
this perspective; Section 5 concludes. We note at the outset that
our focus is on central bank liquidity operations and not on
policymakers’ interest rate responses.
2. Liquidity Shortages and
the Lender of Last Resort
Apart from the conduct of monetary policy, a vital
responsibility of central banks in most countries is to perform
the role of LOLR. At its core, the objective of the LOLR is to
prevent, or at least mitigate, financial instability through the
provision of liquidity support either to individual financial
institutions or to financial markets. The underlying premise is
that shortages of liquidity, by which we mean the inability of an
institution to acquire cash or means of payment at low cost, can
lead to otherwise preventable failures of institutions that then
1
We do not enter the debate over whether the LOLR takes the place of a deposit
insurance system. Recent events, especially the retail bank runs that
accompanied the nationalization of Northern Rock in the United Kingdom,
would appear to have settled the matter in favor of the importance of a rulebased deposit insurance system.
The authors thank Claudio Borio, François-Louis Michaud, Christian Upper,
and conference participants for comments. The views expressed are those
of the authors and do not necessarily reflect the position of the Bank for
International Settlements, the Federal Reserve Bank of New York, or the
Federal Reserve System.
FRBNY Economic Policy Review / August 2010
29
result in spillover and contagion effects that may ultimately
engulf the financial system more broadly with significant
implications on the real economy.2 By signaling its willingness
and ability to act decisively, the central bank demonstrates
its intention to restore confidence in the system by avoiding
“fire sales” of assets and supporting market functioning.
The “classical” doctrine of the LOLR as attributed to
Thornton (1802) and Bagehot (1873) is commonly interpreted
to imply that such lending should be extended freely without
limit, but only to solvent institutions at penalty rates and
against good collateral (for example, see Rochet and Vives
[2004]). This set of principles has been subject to substantial
debate for much of the past thirty years, with many issues yet
to be resolved.3
At their most basic level, the underlying principles of
Bagehot’s original dictum have been subject to a variety of
interpretations. Goodhart (1999), for example, emphasizes
that Bagehot’s criteria for lending were not conditioned on the
individual borrower but on the availability of good collateral.
As such, the distinction between illiquidity and insolvency
would not be an important issue. Similarly, while the
imposition of a penalty rate has traditionally been judged
relative to the prevailing market rate, it can be argued that
Bagehot advocated only that lending take place at a rate higher
than the precrisis level. Given that the LOLR strives to achieve
the good—panic-free—equilibrium, a case can be made that
the penalty ought to be relative to the interest rate during
normal times rather than the higher rate that obtains in the
market during a panic (Goodhart 1999). Indeed, in practice,
LOLR lending has frequently taken place at prevailing market
rates (Giannini 1999).
At a more practical level, the distinction between illiquidity
and insolvency has been largely dismissed on the grounds that
banks generally face liquidity problems when solvency is in
question (Goodhart and Schoenmaker 1995). Indeed, an
individual bank will seek assistance from the monetary
authorities only when it cannot meet its liquidity needs in
financial markets. Since the wholesale interbank money market
is the first stop for most banks, this almost certainly means that
there are significant doubts about the institution’s ultimate
solvency. The proposition that central banks only lend against
good collateral is also undermined by the fact that a bank that
is unable to raise funds in the market must, almost by
definition, lack access to good security for collateralized loans.
As such, emergency lending assistance from the central bank
will likely be against collateral of questionable quality. In
addition, the imposition of a penalty rate has been criticized on
2
This definition of LOLR is quite broad and can, in principle, encompass any
injection of central bank reserves, including routine ones. That said, we focus
primarily on extraordinary interventions driven by unanticipated events.
3
See Davis (2008) and Rochet (2008) for detailed expositions of the various views.
30
Central Bank Tools and Liquidity Shortages
the grounds that such a policy could compound the problem if
it imposes a substantial burden on the troubled institution.
At the same time, another facet of the debate has focused on
the appropriate implementation of LOLR support. Some argue
that in an advanced financial system, LOLR should be
exclusively through open market operations. As long as
systemwide changes in demand for reserves are met through
such operations, the market can direct reserves to those most
in need, thereby avoiding the mispricing that administrative
mechanisms might create (Schwartz 1992; Kaufman 1991;
Goodfriend and King 1988). Such an approach was clearly
successful, for example, in the case of operations associated
with the spikes in liquidity demand during the Y2K episode and
in the aftermath of the stock market crash of October 1987.
However, others argue that LOLR may require direct lending,
not open market operations, as the market may fail to deliver
liquidity to distressed banks whose failure threatens the
financial system (Rochet and Vives 2004; Freixas et al. 2000;
Freixas, Parigi, and Rochet 2000; Goodhart 1999).
2.1 Three Kinds of Liquidity Shortages
Rather than getting mired in the theoretical debate on the
design and role of the LOLR, we take a more pragmatic
approach and outline the broad conditions under which
central banks’ provision of liquidity is undertaken in practice.
From this we derive some general principles that apply
depending on the specific situation. Indeed, once it is
recognized that the nature of the LOLR differs across
circumstances, many of the issues at the center of the
theoretical debate fade.
It is useful at the outset to distinguish between three types of
liquidity: central bank liquidity, market liquidity, and funding
liquidity. Central bank liquidity is the term we use to describe
deposits of financial institutions at the central bank; it is
synonymous with reserves, or settlement balances. These
reserve balances are held by financial institutions to meet
reserve requirements, if any, and to achieve final settlement of
all financial transactions in the payments system. Individual
institutions can borrow and lend these funds in the interbank
market, but, for the system as a whole, the only source of these
funds is the central bank itself.
Market liquidity refers to the ability to buy and sell assets in
reasonably large quantities without significantly affecting price.
This use of the term “liquidity” is closest to the common,
textbook definition: the ease with which an asset can be
converted into means of payment (that is, money or cash).
Finally, there is funding liquidity. This term describes the
ability of an individual or institution to raise cash, or its
equivalent, again in reasonably large quantities, either via asset
sales or by borrowing. As such, market and funding liquidity
are closely linked (see Brunnermeier and Pedersen [2007]).
With this distinction in mind, our discussion of central
banks’ liquidity operations and their appropriate structure
with respect to the fulfillment of the LOLR function is best
premised on the clear separation of three kinds of liquidity
shortages: a shortage of central bank liquidity, an acute
shortage of funding liquidity at specific institutions, and a
systemic shortage of funding and market liquidity. We now
proceed to describe each of these in turn.
Shortage of Central Bank Liquidity
The first kind of liquidity shortage is perhaps the most benign
and occurs when institutions find themselves short of the
reserve balances that they wish to hold, either because of
inadequacies in the aggregate supply of reserves or problems
associated with their distribution within the system. In this
situation, financial institutions risk being unable to fulfill their
immediate payment obligations, creating the potential for
“gridlock” in the payments system. Typically, the tensions
manifest themselves in a spike in the overnight interest rate but
may sometimes also be transmitted to other segments of the
money market as well. For the most part, these problems occur
in the absence of any concern over the solvency of specific
institutions.
When central bank liquidity shortages occur as a result of
problems associated with the distribution of reserves, the
underlying cause is typically technical in nature, having to
do with either technological glitches or mismanagement of
liquidity positions. The computer malfunction at the Bank of
New York on November 20, 1985, which resulted in a large
shortage of cash despite the bank’s patent solvency, and the
September 2001 crisis are examples of such situations. The
immediate problem confronting central banks in each case was
the dislocation of reserves, reflecting a breakdown in payments
systems and the coexistence of institutions unable to lend
excess funds to institutions that desperately needed them.
A shortage of central bank liquidity can also arise from
an inadequate supply of reserves to the system as a whole.4
This may reflect an error in the central bank’s forecast of
autonomous factors affecting liquidity conditions (for
example, as a result of unexpected changes in the Treasury’s
balances with the central bank) or a sudden, unanticipated shift
in demand, or both. At the beginning of August 2007, for
4
Since it assumes that the interbank market is still functioning normally,
this situation is close in nature to the problem envisaged by Goodfriend
and King (1988).
example, a sharp rise in uncertainty over future funding
availability led to an abrupt increase in demand for reserves in
the system as a whole. This put considerable upward pressure
on overnight rates, and many central banks initially found it
harder to achieve their policy targets. The natural policy
response was an immediate increase in the supply of reserves in
an effort to meet what officials hoped would be a brief shortage
of central bank liquidity.
Acute Shortage of Funding Liquidity
at Specific Institutions
The second kind of liquidity shortage occurs when a particular
institution experiences an acute shortage of funding liquidity
associated with solvency concerns as the willingness of
counterparties to trade with the institution dissipates. This
situation can arise as the result of a flawed business strategy—
which becomes evident often only ex post—that has left the
institution exposed to persistent cash drains. Reflecting
substantial perceived insolvency, the shortage of liquidity is
prolonged and the form of assistance needed is essentially
bridge financing that allows time for fundamental
restructuring.
The primary threat posed by an institution-specific acute
liquidity shortage, and hence the main justification for any
official assistance, is that failure may result in contagion and
spillover effects that could put the entire financial system at
risk. The key criterion in the consideration of liquidity support
is then whether the institution in question is systemically
important or not. The distinction between illiquidity and
insolvency is not really relevant. Prominent examples of
situations in which an acute shortage of funding liquidity at
certain institutions necessitated LOLR support include
Continental Illinois in 1984 and the provision of liquidity
support to various bank and nonbank financial institutions
in the current crisis.
Systemic Shortage of Funding and Market Liquidity
The final form of liquidity shortage—a systemic shortage of
both funding and market liquidity—is potentially the most
destructive. It involves tensions emanating from an
evaporation of confidence and from coordination failures
among market participants that lead to a breakdown of key
financial markets. Markets, just as intermediaries, may be
subject to “runs” that are driven by fundamentally similar
forces. As we saw in the immediate aftermath of the September
2008 bankruptcy of Lehman Brothers, the result is a sudden
FRBNY Economic Policy Review / August 2010
31
and prolonged evaporation of both market and funding
liquidity, with serious consequences for the stability of both
the financial system and the real economy.
Such crises are generally associated with a sharp rise in
market participants’ uncertainty about asset values as well as
about the financial strength of potential counterparties.
Because financial markets need participants to function, a
sharp rise in uncertainty that causes many players to disengage
results in illiquid markets (see Caballero and Krishnamurthy
[2008]). As a direct consequence, assets that were thought to
be easily convertible into cash are not, which creates funding
liquidity problems for individuals and institutions. This, in
turn, heightens the credit risk of potential counterparties. The
dynamics of these systemic crises are then driven by a mutually
reinforcing feedback process involving market liquidity,
funding liquidity, and counterparty credit risk.5 The 1987 stock
market crash is an example of such a situation, and systemic
liquidity shortages have been a prominent element of the
current crisis from the very beginning.6
3. Central Bank Tools and
Liquidity Shortages
The three types of liquidity shortages—central bank, acute
institution-specific funding, and systemic funding and
market—do not always occur in isolation. Important
interdependencies exist, and the occurrence of one can lead to
another with dynamics that often reinforce one another. For
example, acute concerns about the viability of a particular
institution can rapidly spread to a loss of confidence in other
institutions, resulting in systemic disruptions in the interbank
market that, in turn, hamper the distribution of reserves
among participants, leading to problems in the payments
system. Indeed, the current crisis that began in mid-2007 has
involved all forms of liquidity shortages.7
In their capacity as LOLR, central banks essentially have
three tools with which they can influence the availability of
liquidity in the financial system. The first is lending or
borrowing in the open market. These operations include the
repos and reverse-repos that are the bread and butter of
liquidity management during normal times. They are not
5
Brunnermeier and Pedersen (2007) provide a formal representation of this
mutually reinforcing process. Freixas, Parigi, and Rochet (2000) and Flannery
(1996) develop models that illustrate how coordination failures can lead to a
systemic seizing up of the interbank market. See also Borio (2004).
6
A detailed exposition of the 1987 crisis can be found in Carlson (2007).
7
A broad analysis of the current crisis is provided by Borio (2008), Bank for
International Settlements (2008a, 2008b), Calomiris (2008), Cecchetti (2008),
and Gorton (2008).
32
Central Bank Tools and Liquidity Shortages
targeted at specific institutions—though they may be
undertaken bilaterally—but are designed to address
systemwide liquidity pressures. The operations are typically
collateralized and conducted at the discretion of the central
bank. The basic function is to regulate the level of aggregate
reserves to ensure smooth functioning of the payments system
and facilitate the attainment of the relevant policy interest rate
target. That said, these operations can be utilized and
structured to address a broader set of problems as well. For
example, through these operations, central banks may lend
not only reserves but also highly liquid securities such as
government bonds.
The second tool is the outright purchase or sale of assets in
the open market. These operations affect the aggregate supply
of central bank money (reserves) on a permanent basis and are
typically conducted in sovereign bonds denominated in either
domestic or foreign currencies. Prior to the current episode,
similar interventions in other asset markets were rare. The
purchases of equities by the Hong Kong Monetary Authority
during the 1997 Asian financial crisis and by the Bank of Japan
in 2002 were notable exceptions. The application of outright
transactions aimed at affecting market prices is quite
controversial and is usually justified in terms of correcting a
fundamental misalignment in asset prices or the provision
of two-way liquidity.
Finally, central banks can conduct transactions directed at
specific institutions instead of markets as a whole. Unlike open
market operations, these transactions can take place at the
discretion of either the central bank or the financial institution
itself, involve the channeling of liquidity directly to or from
particular institutions, and can be either collateralized or
uncollateralized. Examples of such operations include standing
facilities and traditional emergency lending assistance.
The specific institutional setup of each of these three tools
varies a great deal across countries—including differences in
maturity, frequency, counterparty arrangements, and eligible
collateral. These variations can have significant implications
for how financial institutions manage their own liquidity
positions as well as for the liquidity characteristics of various
assets themselves.8 Moreover, the specific setup of each of these
tools crucially determines their function during a liquidity
crisis. Depending on their structure, each can in principle
contribute to the alleviation of all three types of liquidity
shortages discussed earlier. The key features that characterize
their application to various types of crises are set out below and
are summarized in the table. Unsurprisingly, the choice of tool
to be employed will depend on the type of liquidity shortage
that has arisen. Critically, this means that unlike the framework
8
Markets Committee (2008) contains detailed descriptions of the specific
practices for a large cross-section of countries.
Principles of Lender-of-Last-Resort Support
Type of Liquidity Shortage
Nature of Liquidity Support
Distinction between illiquidity and solvency
Directed lending or open market
Lending or outright
Ambiguity of access
Penalty relative to market rate
Quality of collateral/degree of central bank
risk exposure
Term of support
Public announcement of support
Separation from monetary policy
Coordination with fiscal authority
Shortage of Central Bank
Liquidity
Chronic Shortage of Funding
Liquidity at Specific Institutions
Systemic Shortage of Funding
and Market Liquidity
Yes
Either
Lending
No
No, if aggregate shortage
Yes, if institution-specific
No
Directed
Lending
Yes
No
Both
Both
No
No
No
High/negligible
Very short (overnight)
No
Yes
No
Low/high
Long
Depends
Yes
Yes
set out by Bagehot in the nineteenth century, there is no unique
set of principles that governs how the LOLR should respond.
Before describing how central banks use their tools to
respond to each of the aforementioned liquidity shortages, it is
useful to note some key implications for their balance sheets.
The fulfillment of the LOLR function typically involves
changing the composition of assets held by the central bank,
the overall size of its balance sheet, or both. In doing so, central
banks will normally offset any impact on reserve balances
outstanding in order to maintain the policy interest rate near
its target. The main exceptions to this are: 1) if there is an
aggregate shortage of central bank liquidity; 2) if the policy rate
is zero; or 3) if reserves are remunerated at the policy rate.
Whether the overall size of the balance sheet expands or not
then depends on the choice of offsetting operations. If the latter
is achieved by allowing one asset to substitute for another, then
balance-sheet size is unchanged. However, if the offset is
achieved through the issuance of various forms of central bank
liability, such as an increase in the size of the government’s
deposit balance or the sale of central bank bills, balance-sheet
size increases. Typically, the latter becomes necessary as the
scale of liquidity support rises beyond a certain point.
3.1 Shortage of Central Bank Liquidity
When central banks are faced with a shortage of reserves in the
banking system as a whole, the primary aim of their
intervention is to maintain the smooth functioning of the
payments system and keep interest rates near their targets. If
the problem is largely one of insufficient aggregate supply, all
Low-high/low-high
Short to medium
Yes
No
Yes
three forms of central bank intervention can be employed to
address the situation. Generally, however, the preferred option
is to accommodate the extra demand for reserves by lending in
the open market and relying on the market to distribute
reserves to those most in need. The provision of additional
reserves would typically not be at a penalty rate since the
maintenance of the appropriate aggregate supply of reserves is
an important remit of central banks. Moreover, the underlying
cause cannot generally be attributed to mismanagement on the
part of banks. The sharp pickup in demand for liquidity buffers
that began in August 2007, for example, reflected a general rise
in uncertainty regarding future funding needs that was largely
unforeseen.
If the shortage of reserves is caused by problems related to
their distribution within the banking system—a situation
associated with frictional payment shocks that leave some
institutions suddenly and unexpectedly short of funds—the
LOLR function can be implemented through directed liquidity
support. Standing facilities, where banks can either deposit
excess balances or borrow additional balances directly from the
central bank at prespecified rates at the end of the day, are
designed to handle these situations. Since the nature of the
problem envisaged is largely transitory, this type of liquidity
support is designed to be extended for a very short term,
usually overnight. Moreover, to maintain the incentive for
financial institutions to transact in markets, central banks tend
to make access to standing facilities at penalty rates of interest.
Finally, standing facilities can exert a stabilizing influence on
markets without any funds actually being lent, since their mere
presence can act to assure banks of orderly access to overnight
funds. This effect is ensured by making access unambiguous.
FRBNY Economic Policy Review / August 2010
33
Regardless of whether the central bank liquidity shortage is
systemwide and institution-specific, the operations conducted
to address it are designed explicitly to minimize the impact on
market prices of all securities other than the overnight interest
rate. As such, their implementation has no bearing on, nor is it
in conflict with, the official stance of policy. Furthermore, since
the terms are very short and all loans are fully collateralized, the
central bank faces virtually no credit risk. The principles
behind standing facility lending are in fact very much in line
with conventional interpretations of Bagehot’s instructions to
lend freely to solvent institutions, against good collateral, at a
penalty rate. As emphasized by Paul Tucker, much of the
central bank lending that was discretionary in Bagehot’s day
has, in effect, become “hard coded” into the operating
framework (Tucker 2004).
While these operations work well most of the time, the
current crisis has highlighted some potential constraints that
may arise in the use of both open market operations and
traditional standing facilities. For one, financial institutions
may not have sufficient access to the types of assets that the
central bank regards as being of acceptable quality to serve as
collateral. In addition, the institutions most in need of central
bank liquidity may not have direct access to the central bank
itself. As recent experience has shown, development of more
global capital markets has made it more likely that disturbances
will originate in markets and involve counterparties that are
several steps removed from the central bank’s sphere of direct
operation. Finally, when financial institutions lose confidence
in nearly all potential counterparties, bringing their soundness
into question, access to standing facilities can become
stigmatized, impairing the effectiveness of these facilities as a
liquidity backstop. This was particularly evident in the United
States during 2007 and 2008, when market rates at times rose
well above the interest rates on the facilities (see Committee on
the Global Financial System [2008]). As we discuss in more
detail in Section 4, central banks have addressed these
problems by widening the pool of eligible assets, broadening
the range of institutions with which they are willing to transact
directly, and assuring market participants that borrowing from
standing facilities should not be regarded as a sign of weakness.
3.2 Acute Shortage of Funding Liquidity
at Specific Institutions
When the official sector confronts an institution facing an
acute shortage of funding liquidity, the justification for
intervention must be that failure threatens the stability of the
entire financial system. In such a circumstance, the solvency
34
Central Bank Tools and Liquidity Shortages
of the institution will be of secondary importance. Instead,
central bankers are faced with a decision whether to exercise
discretionary authority to provide emergency lending
assistance to a particular institution. Clearly, this situation is
distinct from the one just described, in which an institution
finds itself short of funds at the end of the day. Rather, the
problem is one of large-scale and potentially prolonged
shortages of funding liquidity against which the use of standing
facilities is inadequate or inappropriate. Furthermore, given
the institution-specific nature of the intervention, emergency
lending assistance can be clearly separated from the monetary
policy stance.
Any liquidity support extended in this situation will likely
expose the central bank to credit risk, since an institution in
need of a loan of last resort will typically have exhausted its
stock of both marketable assets and acceptable collateral. So the
assets pledged to the central bank are likely to be some part of
the borrowing bank’s loan book, or illiquid securities, or some
physical asset whose value is uncertain. To the extent that a
loan extended under this circumstance is, in the end, simply
bridge financing while a takeover or major restructuring of
the recipient institution is organized, it will generally be
accompanied by a plan for private sector (Bear Stearns) or
government (Northern Rock) support or recapitalization. This
acts, at least in principle, to limit the central bank’s exposure to
substantial losses.
A key factor determining the scope and scale of emergency
lending to an institution facing an acute shortage of funding
liquidity is the central bank’s ability to absorb losses. In this
context, the current crisis highlights serious potential resource
limitations. As financial institutions have become increasingly
globalized, the scale of any potential support required has
grown tremendously, requiring the joint participation of fiscal
authorities. Moreover, in cases such as Iceland in 2008, it can
even stretch beyond the limits of the entire official sector.
Because of the moral hazard implications, officials are
tremendously hesitant to grant such loans. When they do, they
not only charge high rates of interest to mitigate taxpayer
exposure but have the ability to write down shareholder equity
as well as replace management. Insofar as the institution is
unable to obtain funding on its own in the market, however,
the provision of liquidity support cannot necessarily be
deemed punitive relative to the market rate.9 As a further
counterbalance to moral hazard, the provision of support to
acutely illiquid institutions is on a discretionary basis so that
the market does not take it for granted. Such “constructive
ambiguity” does not necessarily mean, however, that the
9
The imposition of a penalty rate is determined largely by the degree of moral
hazard that is associated with the provision of liquidity support. We discuss this
further in Section 3.4.
general set of principles that would justify emergency lending
assistance should not be made explicit. Taylor (2009), for
example, argues that uncertainty about what the government
would do to aid financial institutions, and under what
circumstances, was a key factor in the deterioration that
marked the current crisis.
Once an emergency loan is granted, communication can be
critical in determining the chances of success. On the one hand,
the announcement of assistance may work to assure the public
that the financial system is sound, thereby boosting confidence
among market participants. On the other hand, news of
liquidity support may confirm public fears about potential
failures, and the institution receiving support may suffer a
further loss of reputation. In the United Kingdom in 2007,
news of LOLR support to Northern Rock precipitated a retail
deposit run, which was stopped only by announcement of a
government guarantee. In the wake of this incident, banks
understandably became unwilling to access central bank
lending facilities even for more benign liquidity needs, for fear
of reputational consequences. The result was a further
tightening up of the money market, which worsened an already
bad situation.
While stigma is surely not a relevant issue for an ostensibly
failing institution seeking emergency lending assistance, the
central bank’s decision to grant a request may worsen the
stigma associated with all forms of direct lending, complicating
liquidity management. Confidentiality may help to prevent
knowledge of LOLR support from giving rise to panic, but
maintaining it is difficult in practice since banks usually know
the approximate condition of their competitors, and the scale
of such operations would typically necessitate public oversight.
3.3 Systemic Shortage of Funding
and Market Liquidity
The limits of the central bank’s LOLR function are most
severely tested in a systemic liquidity crisis, not least because
such situations are likely to be accompanied by the other
two types of liquidity shortages as well. In this circumstance,
the underlying aim of official intervention is to shore up
confidence in the financial system as a whole, restoring market
functioning through the reestablishment of both funding and
market liquidity. This will help forestall asset fire sales, facilitate
the orderly reduction in borrowing, support the process of
price discovery in markets, and restore credit flows. Succeeding
will almost surely require utilization of all of the forms of
central bank liquidity intervention described earlier and may
involve substantial modifications in standard practices and
procedures. In addition, as is fairly clear, the central bank could
well become exposed to considerable market and credit risk.
In a systemic liquidity crisis, the key challenge facing
central banks is to find ways to contain flight-to-quality and
re-engage the private sector in the intermediation process.
Such re-engagement will occur only as agents’ uncertainty
over outcomes is reduced. To this end, the central bank will
have to perform an intermediating role, and its actions may
be designed to supplement the role of banks or even bypass
banks altogether. Indeed, whereas the primary function of the
LOLR in traditional discussions is to liquefy the balance sheet
of banks, the current crisis has highlighted that when faced
with a systemic crisis in a market-based financial system, the
scope of LOLR support is likely to be much broader and
involve interventions more akin to liquefying the limit order
book of a particular market.
Typically, this will require a broadening of the central bank’s
provision of liquidity both in terms of accessibility and
structure. Tensions in the term funding market, for example,
can be alleviated by the central bank both directly (through
greater provision of term funding that offsets some of the
shortfall in market supply) and indirectly (through the
assurance of access to liquidity directly from the central bank).
To the extent that the latter helps to ease intermediaries’
concerns about rollover risk, they may become more willing to
extend term loans. At the same time, the set of institutions with
which the central bank transacts may need to be expanded to
ensure that the interventions reach those most in need.
A basic thrust of liquidity operations during a systemic crisis
is to accommodate the increase in demand for assets of
unquestionable quality while at the same time financing those
institutions that find it hard to borrow in the market. This
involves shifting the asset composition of central banks’
balance sheets away from highly liquid assets (primarily
government securities) toward less liquid ones (typically
private sector debt). In some instances, it may be necessary to
sidestep the banking system and provide funding directly to
borrowers and investors in key credit markets. This may be
accomplished through outright purchases of, or lending
against, specific classes of debt linked to particular market
segments (for example, mortgages or corporate bonds). By
reassuring investors that a committed buyer is in the market,
such interventions may reduce the liquidity premium on
various asset classes and boost the flow of credit. More
generally, market prices may be influenced through the
portfolio balance effect, whereby the change in the relative
supplies of imperfectly substitutable private and public
securities will lower the premium that the private sector
demands for holding risky private securities at the margin.
In addition, by making an asset eligible for central bank
FRBNY Economic Policy Review / August 2010
35
operations, the liquidity premium that might otherwise be
needed to induce investors to hold that asset will be reduced.
Because the purpose of these policies is to affect market
pricing of specific assets independently of the overnight rate, it
will be difficult to distinguish them from the stance of monetary
policy per se. They also represent a departure from the
conventional view that monetary policy should refrain from
directly influencing relative prices by not targeting specific asset
prices. Indeed, whether yield spreads are too wide or whether
specific bonds are rationally priced given the amount of risk
inherent in the prevailing economic outlook is largely a
subjective assessment. Justification for such policy actions, then,
rests on the same logic that has been used to motivate foreign
exchange interventions—the enhancement of two-way liquidity
or the attempt to move a misaligned asset price.
Ultimately, though, a systemic crisis is less amenable to
central bank intervention. Central bank tools are much more
limited in this context, since the fundamental problem is more
greatly removed from monetary policymakers’ sphere of
influence. The bulk of market and funding liquidity is
generated through transactions among private entities and, as
such, is created endogenously in the financial system. In an
environment where there is pervasive uncertainty about banks’
balance sheets, both because asset valuations of various types
become problematic and because of incomplete knowledge
about what assets each bank holds, a central bank’s liquidity
operations can ease these problems only indirectly, alleviating
the symptoms rather than the cause. Central banks can provide
liquidity by transacting with market participants, but they are
not able to directly ensure that private agents will transact with
each other.
In the end, whether central bank actions are effective in
attenuating the impact of a systemic crisis and restoring the
functioning of markets depends on the extent to which they
have a catalytic effect on mutually voluntary private sector
transactions. A key aim would be to generate a virtuous cycle
that relies primarily on the private sector to re-establish
liquidity in interconnected markets. In this respect,
announcements of intended actions can be sufficient if they are
credible. During the 1987 crisis, for example, the Federal
Reserve not only encouraged banks and securities firms to
make credit available to brokers and dealers but also issued very
public statements affirming its commitment to providing
liquidity. Carlson (2007) argues that the latter was critical to
stabilizing the situation.
By extension, ambiguity of access to central bank liquidity
facilities is likely to be counterproductive during a systemic
crisis. On the contrary, uniform access for all financial
institutions, irrespective of their condition and systemic
importance, is more likely to alleviate heightened counterparty
36
Central Bank Tools and Liquidity Shortages
fears. Standing facilities and loan guarantees are examples of
intervention that can have this kind of catalytic effect without
the liquidity actually being drawn upon. For example, several
of the new facilities introduced by the Federal Reserve in the
current crisis are available at the discretion of market
participants (the PDCF, AMLF, CPFF, MMIFF, and TALF),
while others appear to have been structured to encourage
market intermediation of credit.10
Importantly, the implementation of such measures involves
an intricate balancing act. To the extent that an expanded
intermediation role discourages financial institutions from
dealing with one another, the central bank’s response may
create countervailing forces between catalyzing market activity
on the one hand and substituting for it on the other. The onus
then falls on the design of an appropriate pricing structure and
well-defined exit strategies, both of which can be difficult to
achieve in practice.
Finally, in a situation of generalized market failure, it makes
less sense for liquidity support to be provided at a penalty rate
relative to prevailing market rates since no particular
institution is benefiting relative to others. In fact, liquidity
support will often, and probably should, be provided at a
subsidized rate when it involves an illiquid asset in which a
market price cannot be found. That said, liquidity facilities may
be designed in ways so that accessing them is not punitive when
markets are dysfunctional and is punitive when normal activity
returns.11 Doing so would also naturally lead to an automatic
run-off of liquidity support as markets stabilize.
3.4 Lender of Last Resort and Moral Hazard
The creation of moral hazard is a long-standing concern
associated with LOLR operations. Goodhart (2007), for
example, argues that generous provision of liquidity by central
banks, in normal times and in times of crisis, has made banks
careless in managing their liquidity risks. With this in mind, it
is useful to assess the nature of moral hazard in light of the
different types of liquidity shortages we set out here. As will
become apparent, we view the moral hazard created by the
LOLR as either relatively unimportant in practice or an issue
10
The TALF (Term Asset-Backed Securities Loan Facility), for example,
provides term credit against newly issued asset-backed securities rather than
outright purchases, which creates an incentive for participants to establish
sound collateral for the securities since they are likely to be kept on their books.
The PDCF is the Primary Dealer Credit Facility; the AMLF is the Asset-Backed
Commercial Paper Money Market Fund Liquidity Facility; the CPFF is the
Commercial Paper Funding Facility; the MMIFF is the Money Market Investor
Funding Facility.
11
Many of the Federal Reserve’s new facilities in the current crisis are designed
this way. The CPFF, for example, charges a fixed spread over the three-month
market rate that should become unattractive in normal times.
that is best addressed by other facets of policy not directly
associated with the provision of liquidity support itself.
With respect to shortages of central bank liquidity, the
potential for moral hazard arises if the provision of liquidity
support reduces the incentive for financial institutions to
devote resources to enhancing the efficiency and effectiveness
of their daily liquidity management operations. Moreover,
excessive reliance on the central bank for daily liquidity
management would substantially undermine private interbank
market activity. Central banks have generally responded to
these issues successfully through the establishment of a pricing
structure that preserves the incentive for market participants to
trade with one another before going to the central bank’s
standing facility.
Looking at the case of an acute shortage of funding liquidity
at specific institutions, we note that the underlying moral
hazard concern is that the extension of liquidity assistance
could establish precedents that lead to lax risk management
and make financial institutions generally more vulnerable to
shocks. Attempts to address these concerns have centered on
both the prevention of potential problems through regulatory
frameworks such as prompt corrective action and the
imposition of highly punitive financial and nonfinancial
penalties on management and shareholders in the process of
crisis resolution. The latter makes it unlikely that expectations
of liquidity support will directly contribute to the taking on of
excessively risky activities. Nevertheless, to the extent that
creditors are protected from losses, the exercise of market
discipline is weakened. This in and of itself may facilitate
(rather than cause) the pursuit of excessively aggressive
business strategies.
Finally, in situations of systemic crisis, the underlying
coordination failures that trigger the crisis cannot be easily
attributed to anticipation by private agents of government
support measures in the event of a financial meltdown, so it is
difficult to see how it could have been the outcome of moral
hazard. Indeed, if one views the evaporation of liquidity in key
financial markets as a form of market failure—associated with
the inability of markets to cope with aggregate, as opposed to
idiosyncratic, liquidity shocks—a case can be made that the
provision of liquidity support in systemic crises serves to
enhance social welfare (see, for example, Kearns and Lowe
[2008]).
At the same time, expectations of generalized liquidity
provision by the central bank in systemic crises may lead
institutions to neglect the task of building buffers that can be
run down during such events. In this way, the inherent
financial fragility that potentially contributes to making
systemic crises more likely may be partly attributable to
complacencies in risk management associated with
anticipation of central bank intervention. This does not,
however, constitute grounds for the central bank to refrain
from providing support should a systemic crisis occur, nor
does it suggest that provision at that time should be on highly
punitive terms. Economically and politically, authorities have
little choice but to act in the midst of a crisis, and any ex ante
stance precluding provision of such support cannot be made
credible. Thus, even if the existence of the central bank’s
liquidity facilities creates moral hazard, efforts to mitigate it are
more productively channeled elsewhere. Insofar as crises are
associated with complacency in risk management, mistaken
assumptions about asset price trajectories that become evident
only ex post, skewed compensation arrangements, limited
liability, and overall financial conditions that encourage risktaking, the burden of their prevention falls more naturally on
the appropriate management of macroeconomic policies and
regulatory structures than on the specifics of the framework for
emergency liquidity provision.
4. Liquidity Operations
during the Current Crisis:
The LOLR Perspective
In the face of the widespread financial market dislocations that
began in August 2007, central banks have expanded liquidity
operations, actively deploying their balance sheets to address all
three types of liquidity shortages. While the inherent cause of
the current crisis may be rooted in coordination failures and
informational asymmetries—and so is not new—the scale and
scope of the problem have necessitated measures in some
countries that are clearly unprecedented. In particular, because
institutions have come to depend on market-based sources of
liabilities, replacing lost funding liquidity now requires
interventions on a scale that is large relative to the size of the
central bank’s balance sheet in normal times. This section
outlines the general thrust of central banks’ actions from the
perspective of their LOLR function.12
Each of the measures central banks have undertaken since
the fall of 2007 can be seen as addressing directly or indirectly
at least one of the three types of liquidity shortages. With
respect to addressing shortages of central bank liquidity, the
focus has been on accommodating the greater instability in the
demand for reserves and alleviating distributional problems.
These have been addressed by varying the size and frequency of
operations—conducting them outside their regular schedule
and in larger than usual amounts—broadening the number
12
For further details on central bank actions, see Bank for International
Settlements (2008b) and Committee on the Global Financial System (2008).
FRBNY Economic Policy Review / August 2010
37
and type of counterparties, and enlarging the scope of eligible
collateral. A key objective of these interventions has been to
contain deviations of market rates from the official policy
stance (Chart 1).
For acute shortages of funding liquidity at specific
institutions, central banks have extended emergency lending
assistance to various financial institutions. This involved, for
example, the extension of credit to Northern Rock by the
Bank of England; the Federal Reserve’s support for Bear
Stearns, AIG, and Citigroup; and the Swiss National Bank’s
financing of the transfer of distressed assets out of UBS. These
actions were undertaken jointly with the fiscal authority and
generally structured to minimize the financial risk to the
central bank.
Finally, there have been four broad components to efforts
aimed at alleviating systemic shortages of funding and market
liquidity. First, central banks have sought to ensure the
availability of backstop liquidity to key financial institutions as
reflected, for example, in the creation of the Federal Reserve’s
PDCF, which established overnight funding for primary
dealers. Second, there has been an effort to provide greater
assurance of the availability of term funding through the
lengthening of the maturity on refinancing operations as well
as the establishment of inter–central-bank swap lines to ensure
the availability of (primarily) dollar funding in offshore
markets. Third, policymakers have worked to provide highquality securities—usually sovereign ones—in exchange for
lower quality, less liquid securities in order to encourage
trading in the latter. The Federal Reserve and the Bank of
England, for example, established facilities to lend government
securities in exchange for less liquid market securities. Fourth,
there have been initiatives aimed at ensuring the availability of
credit to non-banks in cases where particular financial markets
had become inoperative. The Federal Reserve’s extension of
credit through the CPFF and the TALF, direct purchases of
mortgage-backed securities issued by key government agencies,
and the Bank of Japan’s outright purchases of commercial
paper are examples of such an approach.13
Over the past sixteen months, central bank actions have
covered this broad spectrum through two main phases. During
13
It is useful to emphasize that these somewhat unconventional liquidity
operations can be applied regardless of the level of the policy rate itself. Central
bank balance sheets can expand aggressively even when interest rates are
positive, contrary to the widely held view that such expansion can take place
only at the cost of pushing rates to zero. The latter view is often based on
Japan’s “quantitative easing” experience; however, the ability to expand the
balance sheet without compromising targets for interest rates is constrained
only by central banks’ capacity to offset the impact on bank reserves. Indeed,
Asian central banks that have seen their balance sheets expand in recent years
with the sustained accumulation of foreign reserves have, on the whole, been
able to maintain their interest rate targets. Disyatat (2008) provides further
discussion of these issues.
38
Central Bank Tools and Liquidity Shortages
Chart 1
Policy Rates and Reference Market Rates
Percent
6 United States
5
Policy ratea
4
Reference market rateb
3
2
1
0
-1
6
Euro area
5
Policy ratea
4
3
2
Reference market rateb
1
0
-1
7
United Kingdom
6
5
Policy ratea
4
3
Reference market rateb
2
1
0
-1
2007
2008
2009
Sources: Bloomberg; national data.
a
For the United States, federal funds target rate; for the euro area,
minimum bid rate; for the United Kingdom, official bank rate.
b
For the United States, effective federal funds rate; for the euro area,
Eonia; for the United Kingdom, overnight Libor rate.
the first phase (through mid-September 2008), central bank
efforts were undertaken by varying the asset composition of
their balance sheets while keeping the overall size largely
unchanged. As the crisis intensified following the collapse of
Lehman Brothers, central bank operations entered a second
phase that involved a rapid expansion of the size of their
balance sheets. In particular, as central banks increased the size
and scope of their efforts to support market functioning and
undertook larger emergency lending assistance, offsetting
operations on the asset side of their balance sheets became
constrained and it was necessary to expand the capacity of
reserve-draining instruments on the liability side.
During the fall of 2008, the assets of the Federal Reserve and
the Bank of England more than doubled in a matter of weeks,
while those of the European Central Bank increased by more
than 30 percent (Chart 2). In the case of the Federal Reserve,
the growth in assets was driven by larger term operations, new
lending facilities, and dollar swaps with other central banks.
For the European Central Bank and the Bank of England, the
expansion was driven mainly by repos and auctions of dollar
liquidity. On the liability side, the increase in balance-sheet
capacity of the Federal Reserve came from bank reserves and a
one-off injection in the Treasury account (Chart 3). For the
European Central Bank, the primary offsetting instrument has
been the deposit facility, whereas the Bank of England has
increasingly relied on the issuance of central bank bills.
Chart 2
Central Bank Assets
Chart 3
Billions of National Currency
Central Bank Liabilities
2,500 Federal Reserve
Billions of National Currency
Total assets
Commercial paper funding facility
2,000
2,500
Federal Reserve
Foreign exchange swaps
1,500
2,000
1,000
1,500
U.S. Treasury
supplementary account
Reserve balances
Securitiesa
500
Total liabilities
1,000
Lendingb
0
500
2,500
Notes in circulation
European Central Bank
Total assets
0
2,000
Securitiesd
1,500
1,000
2,250
European Central Bank
2,000
1,750
Total liabilities
Other euro
liabilitiesa
1,500
Foreign currency claimsc
1,250
500
Deposit facility
1,000
Lendinge
Current accounts
750
0
500
300
Notes in circulation
250
0
Bank of England
250
Other assetsc
300
Bank of England
200
250
Securities
Other liabilitiesb
150
Total assets
200
Short-term open
market operationsc
100
150
50
Total liabilities
f
Lending
100
0
2007
2008
2009
Sources: Central banks; Datastream.
a
Reserve balances
50
Notes in circulation
0
2007
Securities held outright (including Term Securities Lending Facility).
2008
b
Repurchase agreements, term auction credit, and other loans.
c
Including U.S. dollar liquidity auctions.
d
Of euro area residents and general government debt in euros.
e
Including repos and other lending in euros.
b
Including to central banks.
f
Short- and long-term reverse sterling repos.
c
Including issuance of Bank of England sterling bills.
2009
Sources: Central banks; Datastream.
a
To other euro-area and non-euro-area residents, including central
banks.
FRBNY Economic Policy Review / August 2010
39
5. Conclusion
One hundred and thirty-five years ago, Walter Bagehot wrote
that, to stay a banking panic, 1) the bank supplying reserves
“must advance freely and vigorously to the public,” 2) “these
loans should only be made at a very high rate of interest,” and
3) “at this rate these advances should be made on all good
banking securities, and as largely as the public ask for them”
(1873, pp. 74-5). From these basic principles, central banks
derived the theory of the lender of last resort. But Bagehot lived
in a different world—not only were there no automobiles,
airplanes, or computers, but there were very few central
banks—fewer than 20, whereas today there are more than 170.
Since central banks are essentially a twentieth-century
phenomenon, it is natural to ask whether Bagehot’s
nineteenth-century doctrine still applies.
In this paper, we have argued that Bagehot’s view of the
lender of last resort requires modification. As the financial
system has gained in complexity, so have all facets of the role of
central banks. Following the trail blazed by Bagehot, we refine
the theory of the LOLR by identifying three types of liquidity
shortages that can occur in the modern financial system:
1) a shortage of central bank liquidity, 2) an acute shortage
of funding liquidity at a specific institution, and 3) a systemic
shortage of funding and market liquidity.
Our analysis leads us to conclude that the appropriate
principles for central banks’ LOLR support must be
40
Central Bank Tools and Liquidity Shortages
conditioned on the particular type of liquidity shortage that
is taking place. When confronted with a simple shortage of
central bank liquidity, for example, Bagehot’s dictum applies.
By contrast, a systemic event almost surely requires lending at
an effectively subsidized rate compared with the market rate
while taking collateral of suspect quality.
In the same way, any discussion of communication policy
in the potential future application of LOLR policy, such as
the desirability of constructive ambiguity, must be linked to
a specific type of liquidity shortage. So, for example, while
ambiguity of access to central bank liquidity may be an
important countervailing force against moral hazard in
situations of acute institution-specific liquidity shortages, it is
likely to be counterproductive when it comes to dealing with
general shortages of central bank liquidity or while in the midst
of a systemic crisis.
In terms of the debate outlined earlier on the appropriate
form of LOLR lending, the current crisis has made it
abundantly clear that the argument that only open market
operations are needed to meet the liquidity needs of
fundamentally sound banks is flawed since money markets
themselves can fail to function properly. This is even more so
in light of recent developments in the financial system that
have increased the interdependencies between financial
institutions and markets, and made it more imperative that
central banks be prepared for situations in which both
experience problems simultaneously.
References
Bagehot, W. 1873. Lombard Street: A Description of the Money
Market. London.
Disyatat, P. 2008. “Monetary Implementation: Misconceptions and
Their Consequences.” BIS Working Paper no. 269, December.
Bank for International Settlements. 2008a. BIS 78th Annual Report
2007/08, chapter VI.
Flannery, M. J. 1996. “Financial Crises, Payment System Problems,
and Discount Window Lending.” Journal of Money, Credit,
and Banking 28, no. 4, part 2 (November): 804-24.
———. 2008b. BIS Quarterly Review, December.
Borio, C. 2004. “Market Distress and Vanishing Liquidity: Anatomy
and Policy Options.” BIS Working Paper no. 158, July.
———. 2008. “The Financial Turmoil of 2007–?: A Preliminary
Assessment and Some Policy Considerations.” BIS Working
Paper no. 251, March.
Brunnermeier, M. K., and L. H. Pedersen. 2007. “Market Liquidity and
Funding Liquidity.” NBER Working Paper no. 12939, February.
Caballero, R. J., and A. Krishnamurthy. 2008. “Musical Chairs:
A Comment on the Credit Crisis.” Banque de France Financial
Stability Review: Special Issue on Liquidity, no. 11
(February): 9-11.
Calomiris, C. 2008. “The Subprime Turmoil: What’s Old, What’s New,
What’s Next.” Paper presented at the Federal Reserve Bank of
Kansas City Symposium on Maintaining Stability in a Changing
Financial System, Jackson Hole, Wyoming, August 21-23.
Carlson, M. 2007. “A Brief History of the 1987 Stock Market Crash
with a Discussion of the Federal Reserve Response.” Board of
Governors of the Federal Reserve System Finance and
Economics Discussion Series, no. 2007-13, May.
Freixas, X., C. Giannini, G. Hoggarth, and F. Soussa. 2000. “Lender
of Last Resort: What Have We Learned since Bagehot?” Journal
of Financial Services Research 18, no. 1 (October): 63-84.
Freixas, X., B. M. Parigi, and J.-C. Rochet. 2000. “Systemic Risk,
Interbank Relations, and Liquidity Provision by the Central Bank.”
Journal of Money, Credit, and Banking 32, no. 3, part 2
(August): 611-38.
Giannini, C. 1999. “‘Enemy of None but a Common Friend of All’? An
International Perspective on the Lender-of-Last-Resort Function.”
Princeton Essays in International Finance, no. 214.
Goodfriend, M., and R. G. King. 1988. “Financial Deregulation,
Monetary Policy, and Central Banking.” Federal Reserve Bank
of Richmond Economic Review, May/June: 3-22.
Goodhart, C. A. E. 1999. “Myths about the Lender of Last Resort.”
International Finance 2, no. 3 (November): 339-60.
———. 2007. “Liquidity Risk Management.” Financial Markets
Group Research Centre Special Paper no. 175, October.
———. 2008. “Liquidity and Money Market Operations.” Financial
Markets Group Research Centre Special Paper no. 179, April.
Cecchetti, S. 2008. “Monetary Policy and the Financial Crisis
of 2007-2008.” Centre for Economic Policy Research Policy
Insight, no. 21, April.
Goodhart, C., and D. Schoenmaker. 1995. “Should the Functions of
Monetary Policy and Banking Supervision Be Separated?” Oxford
Economic Papers 47, no. 4 (October): 539-60.
Committee on the Global Financial System. 2008. “Central Bank
Operations in Response to the Financial Turmoil.” CGFS
Publications, no. 31, July.
Gorton, G. B. 2008. “The Subprime Panic.” NBER Working Paper
no. 14398, October.
Davis, E. P. 2008. “Liquidity, Financial Crises, and the Lender of Last
Resort—How Much of a Departure Is the Sub-Prime Crisis?” Paper
presented at the Reserve Bank of Australia 2008 Conference on
Lessons from the Financial Turmoil of 2007 and 2008, July 14-15.
Kaufman, G. 1991. “Lender of Last Resort: A Contemporary
Perspective.” Journal of Financial Services Research 5, no. 2
(October): 95-110.
FRBNY Economic Policy Review / August 2010
41
References (Continued)
Kearns, J., and P. Lowe. 2008. “Promoting Liquidity: Why and How?”
Reserve Bank of Australia Research Discussion Paper no. 2008-06,
October.
Schwartz, A. J. 1992. “The Misuse of the Fed’s Discount Window.”
Federal Reserve Bank of St. Louis Review 74, no. 5 (September/
October): 58-69.
Markets Committee. 2008. “Monetary Policy Frameworks and Central
Bank Market Operations.” Markets Committee Compendium, June.
Taylor, J. 2009. “The Financial Crisis and the Policy Responses: An
Empirical Analysis of What Went Wrong.” NBER Working Paper
no. 14631, January.
Rochet, J.-C. 2008. Why Are There So Many Banking Crises? The
Politics and Policy of Bank Regulation. Princeton, N.J.:
Princeton University Press.
Thornton, H. 1802. An Enquiry into the Nature and Effects of
the Paper Credit of Great Britain. London: Hatchard.
Rochet, J.-C., and X. Vives. 2004. “Coordination Failures and the
Lender of Last Resort: Was Bagehot Right after All?” Journal of
the European Economic Association 2, no. 6 (December):
1116-47.
Tucker, P. 2004. “Managing the Central Bank’s Balance Sheet: Where
Monetary Policy Meets Financial Stability.” Remarks delivered to
mark the fifteenth anniversary of the founding of Lombard Street
Research, July 28.
The views expressed are those of the authors and do not necessarily reflect the position of the Bank for International Settlements,
the Federal Reserve Bank of New York, or the Federal Reserve System. The Federal Reserve Bank of New York provides no
warranty, express or implied, as to the accuracy, timeliness, completeness, merchantability, or fitness for any particular
purpose of any information contained in documents produced and provided by the Federal Reserve Bank of New York in any
form or manner whatsoever.
42
Central Bank Tools and Liquidity Shortages
Erhan Artuç and Selva Demiralp
Provision of Liquidity
through the Primary Credit
Facility during the Financial
Crisis: A Structural Analysis
1. Introduction
I
n response to the liquidity crisis that began in August 2007,
central banks designed a variety of tools for supplying
liquidity to financial institutions. The Federal Reserve
introduced several programs, such as the Term Auction
Facility, the Term Securities Lending Facility, and the Primary
Dealer Credit Facility, while enhancing its open market
operations and discount window. This paper focuses on the
financial market effects of changes to the discount window
borrowing facility. Specifically, we investigate whether the
changes represent a fundamental shift in the way the Federal
Reserve traditionally provided liquidity through the primary
credit facility as well as whether the Fed would be well served to
retain these changes to its borrowing facility indefinitely.
In January 2003, the Federal Reserve revised its discount
window lending program. The revision was designed to
improve the operation of the facility, which had experienced
declines in usage. Before 2003, borrowing from the Fed took
place at a rate below the market rate, known as the discount
rate. Fed officials applied a non-price funds-rationing
mechanism by asking potential borrowers detailed questions
about their financial well-being before lending funds. This
administrative process deterred depository institutions from
Erhan Artuç is an assistant professor of economics and Selva Demiralp
an associate professor of economics at Koç University, Turkey.
Correspondence: eartuc@ku.edu.tr
using the discount window because borrowing from the Fed
was perceived as a signal of financial weakness by market
participants.1
The revised discount window borrowing facility was
designed to eliminate the reluctance to borrow from the Fed
by including a new “no-questions-asked” policy for eligible
borrowers. However, despite Fed assurance that the new facility
would eliminate all administrative costs of borrowing, some
argued that the stigma could not be eliminated completely
(see, for example, Furfine [2001, 2003]). However, Artuç and
Demiralp (2010) recently showed that the stigma of borrowing
declined substantially in the post-2003 period, following the
easing of the Fed’s administrative policy and restrictions.
In this paper, we assess the effects of changes to the primary
credit facility since August 2007 by performing out-of-sample
simulations based on a model developed by Artuç and
Demiralp (2010). Our results are highly consistent with the
predictions of our 2008 study—that is, the revised discount
window is effective and plays an essential role in moderating
volatility in the federal funds market.
1
See, for example, Goodfriend (1983), Pearce (1993), Dutkowsky (1993),
Peristiani (1998), Clouse and Dow (1999), Furfine (2003), Dow (2001),
and Darrat et al. (2004).
The authors thank Seth Carpenter, Refet Gürkaynak, Carolyn Wilkins, and
conference participants for comments. Financial support for this paper was
provided by the Scientific and Technological Research Council of Turkey
(TUBITAK). Demiralp’s research was also funded by the Turkish Academy
of Sciences (TUBA). The views expressed are those of the authors and do not
necessarily reflect the position of the Federal Reserve Bank of New York or
the Federal Reserve System.
FRBNY Economic Policy Review / August 2010
43
2. Recent Changes to the Primary
Credit Facility
The primary credit facility, as revised by the Federal Reserve
in 2003, offered credit to financially sound banks at a rate
100 basis points above the Federal Open Market Committee’s
target federal funds rate (the primary credit rate). Primary
credit was made available to depository institutions at an
above-market rate but with very few administrative restrictions
and no limits on the use of proceeds (see Madigan and Nelson
[2002]). Because the interest rate charged on primary credit
was above the market price of funds, it replaced the rationing
mechanism for obtaining funds from the central bank and
eliminated the need for administrative review by the Federal
Reserve.
Amid the onset of the liquidity crisis in August 2007, the
Federal Reserve lowered the spread between the primary credit
rate and the target funds rate from 100 basis points to 50 basis
points and extended the maximum term of loans to thirty days.
In March 2008, the Fed once again narrowed the spread, this
time to 25 basis points, and extended the loan term to ninety
days. The moves were motivated by the desire to make discount
window credit more accessible to depository institutions.
The Federal Reserve’s actions led to an increase in the
volume of discount window borrowing during the crisis
(Chart 1). The upper panel of the chart shows total primary
credit outstanding since the establishment of the revised facility
in 2003. The middle and lower panels, which split the sample at
August 2007, illustrate the enormous rise in borrowing that
occurred.
While the massive increase in the volume of borrowing
supports the argument that the stigma of borrowing had been
eliminated, one should be cautious when interpreting this
result. Chart 2, which plots the highest traded funds rate
against the primary credit rate, shows that despite the
expansion in borrowing, some trades in the funds market took
place at rates above the primary credit rate. What is reassuring
about these findings, however, is their consistency with the
predictions of our earlier work. As Artuç and Demiralp (2010)
describe, reluctance to borrow from the Fed has several
components. The non-price mechanism is the component
attributable to the Federal Reserve’s implementation of
discount lending. Artuç and Demiralp show that this
component declined significantly after the establishment of
the revised facility in 2003. Meanwhile, a second type of stigma
arises from the asymmetric information problems associated
with discount window borrowing. Specifically, while most
banks borrow from the discount window, the facility is also
used by troubled or failing institutions. Because market
participants cannot fully differentiate sound from troubled
44
Provision of Liquidity through the Primary Credit Facility
Chart 1
Primary Credit Outstanding
Weekly Average
Millions of dollars
20,000
Since 2003
16,000
12,000
8,000
4,000
0
2003
800
04
05
06
07
08
2003-August 2007
700
600
500
400
300
200
100
0
2003
20,000
04
05
06
07
August 2007-September 2008
16,000
12,000
8,000
4,000
0
A
S
O N
2007
D
J
F
M
A
M J
2008
J
A S
Source: Federal Reserve Statistical Release H.4.1.
borrowers, they may view borrowing as a potential sign of
weakness of any bank that visits the window. If this type of
stigma increases at the early stages of a financial crisis, when
institutions are trying to signal their good health, it could
explain the spikes in the funds rate over the primary credit rate
shown in Chart 1.2 In addition, it is plausible that the capital
crunch during the financial crisis left some institutions without
sufficient collateral to apply for primary credit loans and thus
forced them to bid for higher rates in the federal funds market,
which is unsecuritized.
average level of reserve balances ( R ), which sets the balance
of bank i equal to:
Chart 2
Daily High Funds Rate and Primary Credit Rate
Weekly Average
(1)
Basis points
16
14
Funds rate
12
10
8
6
4
2
0
Credit rate
A
M
J
J
A S
2007
O
N D
J
F M
A M J
2008
J A
Sources: Daily high funds rate: Federal Reserve Bank of New York
(http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
primary credit rate: Federal Reserve Statistical Release H.15.
3. The Model
The model we describe closely resembles the one developed
in Artuç and Demiralp (2010), which can be viewed as an
extension of the model proposed by Clouse and Dow (1999).
Hence, our discussion relies heavily on Section 3 of Artuç and
Demiralp (2010). We consider a framework in which bank i’s
goal is to keep its daily reserves holdings at a level L 1. Daily
reserve balances vary over the course of the maintenance
period (see Carpenter and Demiralp [2006]). However, from
the borrower’s perspective, a bank’s decision to borrow from
the Fed is static based on liquidity conditions each day.
Therefore, we do not differentiate across days of the
maintenance period except for the settlement Wednesday,
which may necessitate higher borrowing because banks have
less flexibility in absorbing any reserve shortages on the last day
of the maintenance period. On this day, the desired level of
reserve holdings is determined by L 2 .
Banks’ balance holdings follow a stochastic process. During
the day, there are aggregate and individual shocks to the
2
Indeed, a closer look at the days with spikes in the funds rate reveals that
market commentaries are consistent with elevated asymmetric information
problems. An extreme example is October 25, 2007, when the highest traded
funds rate exceeded the primary credit rate by almost 10 percentage points.
On that date, Wrightson ICAP reported that “the stigma of discount window
borrowing was heightened by the news that the New York Fed had extended
$400 million of secondary-credit loans on Wednesday. If word got out that
a given institution had tapped the window on Thursday, the market might
speculate that the bank in question was the same one that had been forced
to make use of the higher-cost secondary credit program for shaky institutions
the day before. The reputational damage of a leak of that nature would be
disastrous” (Wrightson ICAP, Fed Funds Monitor, http://www.wrightson.com).
i
i
R t = R + Ut + V t ,
where Ut ∼ N [ 0,X U ] is an aggregate shock3 and
i
Vt ∼ U [ – X v ,X v ] is an individual shock where X U is the
standard deviation of the aggregate shock while X V represents
the support of the mean zero uniform distribution. Hence, the
i
individual bank becomes a lender in the funds market if R t > L
i
and demands funds if R t < L for L = L 1,L 2 .
Banks that are short of reserves have two options: they can
either borrow from the funds market or from the Federal
Reserve. If the bank chooses to borrow φ t dollars from the
funds market, the cost per dollar is the market interest rate rt .
Alternatively, if the bank borrows φ t dollars from the Federal
Reserve, the cost per dollar is the discount rate (or the primary
credit rate after 2003), rf , plus a fixed cost c. Thus, total cost per
dollar is rf + ---c- . Because of the fixed cost, partial borrowing
φt
from the Federal Reserve is not optimal, and a bank either
prefers to borrow entirely from the Federal Reserve or from the
funds market.4
In addition to borrowing from the Federal Reserve because
of market conditions, banks borrow because of technical
difficulties, such as network problems that force them to use
the Fed regardless of market conditions. To capture this type of
borrowing, we assume that a random fraction of banks, pt , will
face a technical problem in the system where pt has a uniform
distribution: pt ∼ U [ 0, F ] .
We assume that there is a continuum of banks, indexed
from 0 to 1. Thus, there are an infinite number of banks with
zero individual measure whose measure integrates to 1. We
index according to reserve balance levels, such that a bank with
the lowest level of reserve balances is indexed to 0 and one with
the highest level of reserve balances is indexed to 1.
Total demand for funds has two components: It can be met
in the funds market or it can be met at the discount window.
The equilibrium federal funds rate, rt , is determined by the
market equilibrium when the total supply of funds is equal to
the total demand for funds. In modeling borrowing behavior,
our focus is on individual trades in the funds market and on
days of market tightness because borrowing from the Fed on
3
Because the original model is estimated by removing the outliers, we subtract
0.5 percent from the tails of the normal distribution.
4
Without loss of generality, one may think of the fixed costs of borrowing as
varying by bank, reflecting each bank’s relative reluctance to borrow from the
discount window based on factors such as the size of the borrowing, the history
of borrowing, or the availability of credit lines in the funds market. While
we model it in a homogenous manner for simplicity, modeling it in a
heterogeneous manner is also trivial and does not change any implications
of our model.
FRBNY Economic Policy Review / August 2010
45
these days is more likely. Therefore, we set the daily high funds
rate equal to:
(2)
rt
high
= rt + ω t where ω t ∼ ( 0s, ) .
Equation 2 shows that the maximum funds rate registered
for a given day will differ from the equilibrium funds rate
depending on the reserves need and the bargaining power faced
by the counterparties of that particular trade, represented
by ω t .
Turning to the days without market tightness, we note that
trades are almost always cleared in the funds market unless
there is a technical problem. For that reason and without loss
of generality, if supply is larger than demand, we simply set
the funds rate ( rt ) equal to the marginal benefit of holding
balances, as in Clouse and Dow (1999). Hence, a bank can offer
reserves in the funds market if the market rate is greater than
the marginal benefit of holding balances.
If the fixed costs of borrowing decline in the period after
2003, then, all else equal, it implies a decline in the volatility of
the funds rate in the post-2003 period and an increase in the
sensitivity of discount window borrowing to the funds rate.
(A more detailed discussion of the implications of the model
can be found in Artuç and Demiralp [2010].) This implied
change in volatility and the revival of the borrowing function
allow us to identify the size of the implicit cost before and
after 2003.
If we could attribute the entire decline in fed funds volatility
to the revised discount facility, we could proceed with
estimation without any second thoughts. However, the decline
in fed funds volatility is also influenced by other developments,
such as enhanced liquidity management by the Federal
Reserve’s Trading Desk (see Demiralp and Farley [2005]),
improvements in internal information systems (including
those that track a bank’s Federal Reserve account balance),
or banking industry consolidation. To minimize the effects of
such factors on fed funds volatility, we keep our sample period
relatively recent, starting it in 1998. Furthermore, to control for
any remaining effects of such factors, we allow the distributions
i
of Ut and V t to widen or narrow in a linear fashion over time.
That is, we let:
(3)
X U = A + D × t and X V = B + E × t ,
where t is the time trend, X U and X V are defined after equation 1.
To identify the potential decrease in the stigma associated
with discount window borrowing, we consider the following
specification for the implicit borrowing cost c :
c = c 1 , prior to 2003
c = c 1 + c 2 , after 2003.
Note that the above specification treats the implicit costs
of borrowing as exogenously determined. An alternative and
more plausible strategy would be to model the costs of
46
Provision of Liquidity through the Primary Credit Facility
borrowing as a function of the amount borrowed from the
Federal Reserve. However, modeling the cost of borrowing
endogenously cannot be identified in this study, so the issue
remains a topic for future research.5
To estimate our model, we rely on “indirect inference,”
which uses the estimates of an auxiliary model (rather than
moments) to compare actual and simulated data. Because we
can think of data moments as the parameters of a simplified
auxiliary model, Method of Simulated Moments (or GMM)
can be considered special cases of indirect inference. An
auxiliary model does not need to be “correct” for indirect
inference to yield consistent results. As long as the selected
auxiliary model summarizes the data well, the estimates of the
actual model will be consistent and asymptotically normal.
This is because the auxiliary model is used only to extract
information on the underlying data-generating process and,
provided that the parameter estimates from the actual data are
close to those from the simulated data, whether both estimates
are biased or not is of secondary importance. In other words,
the auxiliary parameter estimates themselves do not carry
much meaning other than being indicators of how closely the
simulations match the data (see Artuç and Demiralp [2010];
for a more technical reading on indirect inference, see
Gourieroux, Monfort, and Renault [1993] and Smith [1993]).
We contemplate a simple borrowing function as the
auxiliary model. The auxiliary borrowing function summarizes
how borrowing from the Fed changed over time and after the
policy change in 2003 through a simple ordinary least squares
(OLS) regression, shown in equation 4. In addition to OLS
estimates, we use the mean and the variances of borrowing and
the spread between the daily high funds rate and the target as
part of the auxiliary model (equations 5-8). We also add the
lowest 50 percent of the spread between the daily high funds
rate and the target to capture funds rate volatility in the absence
of market tightness (equations 9 and 10). The estimation
strategy is to find the parameters that will make the simulations
of the model and the actual data look as similar as possible with
respect to the auxiliary model’s OLS estimates and moments.
Specifically, our auxiliary model is:
(4) BR t = β 1 + β 2 r̃t + β 3 t + β 4 tr̃t + β 5 D
2003
(5)
BR t = β 7 + ε 2t
(6)
r̃t = β 8 + ε 3t
(7)
( BR t – β 7 ) = β 9 + ε 4t
(8)
(r̃t – β 8 ) = β 10 + ε 5t
5
r̃t + β 6 D
Settl.
Wed.
SettlWed
2
2
We thank Carolyn Wilkins for bringing this point to our attention.
+ εt
(9)
r̃2t = β 11 + ε 6t
(10)
( r̃2t – β 11 ) = β 12 + ε 7t
2
and
β = [β 1 ,β 2 … β 12],
where BR t is the amount of borrowing from the Fed
normalized by required operating balances, r̃t is the spread
between the funds rate and the funds rate target, t is the time
2003
trend, D
is a dummy for days after the policy change,
SettlWed.
⋅Wed
Settl.
is a dummy for the settlement Wednesday, ε t is
D
an iid random shock, T is sample size, and r̃2t is the lowest
50 percent of r̃t .
Let β̂ be an OLS estimate of β from the actual data
and β̃ be an estimate of from the simulated data. We select
the model’s parameters [ ABc 1 c 2 DEFRsL 2 ] such that
(β̂ – β̃ ) W (β̂ – β̃ )' is minimized, where W is the weighting
matrix that is equal to the inverse of the covariance matrix of β .
In estimating the model, we exclude those days on which
the daily high funds rate exceeds the target rate by more than
25 percent to obtain a more realistic distribution of shocks.
Our estimation results, presented in the appendix, suggest that
the implicit fixed cost of borrowing declines about 90 percent
(from 0.054 to 0.007) after the policy change in 2003. This
result offers strong evidence that the Fed’s new policy was
indeed successful in reducing the stigma associated with
discount window borrowing.
Our estimation period captures a period of a relatively stable
structural environment. The sample starts on June 30, 1998,
with the switch from contemporaneous reserves accounting to
lagged reserves accounting. It ends on March 19, 2007, a few
months prior to the onset of the liquidity crisis in August 2007.
Indeed, if we use the estimates from our model for out-ofsample simulations, the severity of the crisis and the model’s
inability to forecast this environment become clear. Charts 3
and 4 compare actual data with the model’s out-of-sample
simulations for the deviation of the daily high funds rate from
the target and for primary credit outstanding, respectively.
Chart 3
Daily High Funds Rate Less Target
1.2
1.0
Actual data
0.8
0.6
0.4
0.2
Simulation (no crisis)
0
March
2003
In this section, we use our model to analyze the role of the
Federal Reserve’s primary credit lending facility in stabilizing
the money markets in the face of the liquidity crisis.
Specifically, we ask the following questions:
What are the effects of the establishment of the revised
lending facility on total borrowing and interest rates?
In particular, how would the crisis picture look if the
implicit costs of borrowing had not been reduced with
the new regime in 2003?
2.
What are the implications of the increased term of
discount lending in the funds markets?
3.
What are the effects of narrowing the spread between
the primary credit rate and the target rate in stabilizing
the money markets?
4.
What are the implications for discount window
borrowing of paying interest on reserves?
Recall that the model described earlier is designed to
capture the “normal times” of healthy functioning markets.
June
2008
Sources: Federal Reserve Bank of New York (daily high funds rate:
http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
(target: http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
simulated series: authors’ calculations.
4. Simulation Analysis
1.
March
2007
Chart 4
Primary Credit Outstanding
Normalized with Required Operating Balances
35
Actual data
30
25
20
15
Simulation
(no crisis)
10
5
0
March
2003
March
2007
June
2008
Sources: Primary credit outstanding: Federal Reserve Statistical
Release H.4.1; required operating balances: Federal Reserve Board;
simulated series: authors’ calculations.
Note: Required operating balances is the sum of required reserve
balances and contractual clearing balances.
FRBNY Economic Policy Review / August 2010
47
Chart 5
Chart 6
Daily High Funds Rate Less Target
(with Benchmark Simulation)
Primary Credit Outstanding
(with Benchmark Simulation)
Normalized with Required Operating Balances
1.8
1.6
35
Benchmark simulation
1.4
Actual data
30
1.2
25
1.0
20
0.8
Actual data
0.6
15
0.4
10
0.2
0
March
2003
Benchmark simulation
5
March
2007
June
2008
Sources: Federal Reserve Bank of New York (daily high funds rate:
http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
(target: http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
simulated series: authors’ calculations.
0
March
2003
March
2007
June
2008
Sources: Primary credit outstanding: Federal Reserve Statistical
Release H.4.1; required operating balances: Federal Reserve Board;
simulated series: authors’ calculations.
Note: Required operating balances is the sum of required reserve
balances and contractual clearing balances.
While it is a daily model, we present the results in terms of
monthly averages for visual clarity. The vertical line in each
chart corresponds to the end of our estimation period in
March 2007. There is a wide discrepancy between the data and
the model’s simulations, indicating that the period after
August 2007 represents quite extraordinary circumstances
that cannot be captured by the estimates prior to August 2007,
as we would expect.
The sizable discrepancy between the data and the
simulations for the crisis period suggests that we should
incorporate the crisis circumstances into our model before we
can conduct counterfactual experiments on the efficiency of
the Federal Reserve’s policies. At this point, we make several
assumptions to replicate conditions during the crisis. To
capture the overall need for short-term liquidity, we increase
the volatility of the aggregate shock Ut . Furthermore, the
increase in the term of the borrowing is expected to reduce the
implicit costs of borrowing by making it more convenient to
lengthen the duration of a loan.6
To match the moments of the data, we double the standard
error of the aggregate shock Ut and reduce the costs of
borrowing by one half, which allows us to obtain more
reasonable estimates for the interest rate spread and the volume
of borrowing during the crisis period (Charts 5 and 6). We call
these simulations the “benchmark simulations.” In evaluating
6
The Federal Reserve may have also reduced the implicit costs of discount
borrowing indirectly by introducing several other lending facilities and making
overall borrowing more accessible.
48
Provision of Liquidity through the Primary Credit Facility
the model’s performance, one should be careful not to use the
“eyeball metric” to compare the simulated series with the actual
data, because it gives the wrong impression that the model’s
goal is to match the actual data on a day-by-day basis. Instead,
our goal is to match the underlying data-generating process,
and our estimation results, presented in the appendix, show
that the model does reasonably well in achieving this goal.
Indeed, even if we match the underlying data-generating
process perfectly, the simulated series will differ from the actual
data because of the presence of random shocks.
We now analyze the questions raised at the beginning of this
section. The first involves the effects of the 2003 policy change
on mitigating the crisis after 2007. In other words, had the Fed
not changed its lending policy in 2003, how would the funds
market look? Based on our findings in Artuç and Demiralp
(2010), we would expect funds market volatility to worsen
significantly in the absence of the new regime because the
current practice allows institutions in need of funds to utilize
this service without much hesitation. Chart 7 confirms our
expectations. The chart plots the actual spread between the
daily high funds rate and the target (the dashed line) as well as
the simulations generated by our benchmark model (the blue
line). In addition, it shows the spread under the counterfactual
experiment, where the cost of borrowing remains at its
pre-2003 level (the gray line). As the chart reveals, the
counterfactual series skyrockets during the crisis period,
suggesting that the Federal Reserve’s switch to the new lending
Chart 7
Chart 8
Daily High Funds Rate Less Target
(with No Policy Change in 2003)
Daily High Funds Rate Less Target
(with No Decrease in Cost of Borrowing)
20
3.0
No policy change
in 2003
No decrease in cost
of borrowing
2.5
15
2.0
10
1.5
Actual
data
Benchmark
simulation
1.0
Actual data
5
Benchmark simulation
0
March
2003
March
2007
0.5
June
2008
0
March
2003
Sources: Federal Reserve Bank of New York (daily high funds rate:
http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
(target: http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
simulated series: authors’ calculations.
regime was very effective in containing the severity of the crisis
in the money markets.
Turning from prices to quantities, we note that the volume
of borrowing cannot differ between the two regimes because
in our model banks have to borrow the necessary amount of
reserve balances to avoid an overdraft or a reserve deficiency.
For this reason, in reporting our simulation results, we present
only the spread between the daily high funds rate and the target
rate and not the borrowing behavior when the latter is
unaffected under different scenarios.
Next, we analyze the effectiveness of changes in the primary
credit facility that the Federal Reserve introduced at the
beginning of the crisis. Recall that our benchmark model
implies a 50 percent decline in borrowing costs during the crisis
period. In assessing the implications of extended terms of
borrowing, we keep the fixed cost of borrowing at its precrisis
level and simulate the interest rate spread under this scenario.
Chart 8 displays our results. The elevated volatility under the
counterfactual scenario indicates that extending the borrowing
term was an effective action in reducing the implicit costs of
borrowing and hence controlling funds market volatility.
In addition to extending the borrowing term, the Federal
Reserve also narrowed the spread between the primary credit
rate and the target rate from 100 basis points to 25 basis points
during the crisis. Our earlier findings in Artuç and Demiralp
(2010) would suggest that the primary credit rate works as an
upper bound in the absence of market stigma and that a decline
in this rate should decrease deviations of the funds rate from
the target. Our next simulation keeps the spread between the
primary credit rate and the target unchanged at 100 basis
March
2007
June
2008
Sources: Primary credit outstanding: Federal Reserve Statistical
Release H.4.1; required operating balances: Federal Reserve Board;
simulated series: authors’ calculations.
Note: Required operating balances is the sum of required reserve
balances and contractual clearing balances.
points. As shown in Chart 9, the counterfactual spread is at
least as high as the benchmark simulation, if not higher.
This elevated volatility suggests that the narrowing of the
spread was an effective action, even though the difference
between the counterfactual and benchmark simulations is not
as outstanding as in the previous exercises, probably because of
the increased need for collateral under the crisis conditions.
That is, because federal funds borrowing is unsecuritized,
whereas discount window borrowing requires collateral,
Chart 9
Daily High Funds Rate Less Target
(with No Change in Spread)
1.6
1.4
No change in spread
1.2
1.0
0.8
Actual data
0.6
0.4
0.2
0
March
2003
Benchmark
simulation
March
2007
June
2008
Sources: Federal Reserve Bank of New York (daily high funds rate:
http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
(target: http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
simulated series: authors’ calculations.
FRBNY Economic Policy Review / August 2010
49
certain banks may still need to borrow in the funds market
and pay a higher premium if they lack sufficient collateral for
discount borrowing.
Recently, the Federal Reserve has been granted the authority
to pay interest on reserve balances. In addition to placing a
theoretical lower bound on the funds rate, interest payments
on reserve balances may increase the demand for balances
simply because the cost of holding these balances has been
reduced. Our last exercise considers the impact of a higher level
of balances on controlling funds rate volatility. While it is
difficult to estimate the precise magnitude of the change in
reserve balances, we increase the average normalized reserve
balances by 10 percent in our counterfactual experiment.
Chart 10 shows that control over interest rates improves while
Chart 11 shows that the need for borrowing declines if the
average balance holdings increase, as predicted under this new
regime. Together, these results suggest that any policy change
that leads to an increase in reserve holdings, such as interest
payments on reserves, is useful in stabilizing the money
markets.
Chart 11
Primary Credit Outstanding
(with 10 Percent Higher Average Reserves)
Normalized with Required Operating Balances
35
Actual data
30
25
20
10 percent higher
average reserves
15
10
Benchmark simulation
5
0
March
2003
March
2007
June
2008
Sources: Federal Reserve Bank of New York (daily high funds rate:
http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
(target: http://newyorkfed.org/markets/omo/dmm/fedfundsdata.cfm);
simulated series: authors’ calculations.
5. Conclusion
Chart 10
Daily High Funds Rate Less Target
(with 10 Percent Higher Average Reserves)
1.6
Benchmark simulation
1.4
1.2
10 percent higher
average reserves
1.0
0.8
Actual data
0.6
0.4
0.2
0
March
2003
March
2007
June
2008
Sources: Primary credit outstanding: Federal Reserve Statistical
Release H.4.1; required operating balances: Federal Reserve Board;
simulated series: authors’ calculations.
Note: Required operating balances is the sum of required reserve
balances and contractual clearing balances.
50
Provision of Liquidity through the Primary Credit Facility
This paper analyzes the effectiveness of various changes
adopted by the Federal Reserve since the onset of the liquidity
crisis in August 2007. We show that the steps taken to reduce
the implicit costs of borrowing were more effective in
stabilizing the money markets while the narrowing of the
spread between the primary credit rate and the target was not
as effective.
Would the Federal Reserve be well served to retain these
changes to its borrowing facility indefinitely? Our results
suggest that the spread between the primary credit rate and the
target rate could be increased back to 100 basis points without
much impact on the financial markets. Meanwhile, the recent
policy change of paying interest on reserves should make it
easier for the Federal Reserve’s Trading Desk to maintain the
target permanently, not only by placing a lower bound on the
funds rate, but also by increasing the level of reserve balances—
which should reduce the demand for borrowing and ease the
resulting tightness in the funds markets.
Appendix
Panels A and B of the table on the next page present ordinary
least squares estimates of the auxiliary model parameters using
actual as well as simulated data along with the mean and the
variance of borrowing and r̃t . Comparing columns 2 and 4 of
panel A, we note that the auxiliary model’s estimates from the
simulated data and the actual data are fairly similar, as the
algorithm minimizes the distance between those two estimates.
However, they are not identical, as the auxiliary model has
more parameters than the true underlying model. As shown in
row 5, borrowing responsiveness to the interest rate spread ( r̃t )
increases significantly after the Federal Reserve policy change
in 2003, consistent with a decline in market stigma associated
with discount window borrowing and the revival of the
borrowing function. Panel B provides a similar comparison
between the moments generated by the actual data (column 2)
and those computed from the simulated data (column 3).
Similar to panel A, the two sets of statistics display a strong
resemblance.
Panel C presents the parameter estimates of the true
underlying model and their standard errors. The most
interesting parameters for our purposes are displayed in rows 1
and 2. Notice that the implicit fixed cost of borrowing declines
about 90 percent (from c 1 = 0.054 to c 1 + c 2 = 0.007 ) after the
policy change in 2003. This result provides strong evidence that
the Fed’s new policy was indeed successful in reducing the
stigma associated with discount window borrowing. In
addition to estimating the fixed cost of borrowing from the
discount window, we are also interested in determining
whether this implicit cost exhibits any gradual changes over
time. In particular, one may expect a gradual decline in the
implicit cost of borrowing in the post-2003 period because of
the market’s slow adjustment to the new regime. To address
this issue, we experimented with an alternative model that
allows for a time trend in the implicit cost of borrowing prior
to and after 2003 (not shown). However, the trend terms
associated with the implicit cost of borrowing were not
significant in either sample. This finding suggests that there
may not be a gradual adjustment to the new regime in the
second sample. Our results may also be driven by the fact that
we may not have a sufficient number of observations to identify
such a time trend.
Row 3 of panel C shows that the aggregate reserve shock Ut
ranges between -0.43 and 0.43 in the beginning of the sample,
i
while row 4 shows that the bank-specific reserve shock V t
varies between -0.34 and 0.34 initially. Rows 5 and 6 show that
there is a significant time trend in these shocks. In fact, when
we substitute the estimates for D and E in equation 3, we
observe that the aggregate reserve shock exhibits a negative
trend while the bank-specific shock exhibits a positive trend.
The estimate of E implies that the standard error of Ut
decreases about 0.05 percent per year while the estimate of D
i
implies that the range of V t increases about 15 percent each
year. The mild negative time trend in the aggregate shock, Ut ,
could reflect improvements in the Federal Reserve Trading
Desk’s reserve management ability over time, as we observe in
this paper.
Row 7 of panel C shows that the estimated ratio of banks
that incur a technical problem, and thus are forced to borrow
from the Fed rather than the markets, varies from 0 to 0.04.
This result indicates that no more than 4 percent of banks are
affected by this type of condition at any time. Row 10 indicates
that banks seek to attain a higher level of balances on the last
day of the maintenance period, consistent with our
expectations.
FRBNY Economic Policy Review / August 2010
51
Appendix (Continued)
Auxiliary Model and Indirect Inference Estimations
Panel A: Auxiliary Model Ordinary Least Squares Regression
Actual Data
1. Constant
2. r̃ t
3. t
4. t × r̃ t
5. D
2003
6. D
× r̃ t
Settl⋅ Wed⋅
Coefficient
Standard Error
0.48**
0.26**
-0.04**
0.06**
0.07
0.05
0.01
0.01
0.53**
0.35**
-0.05**
0.05**
0.04
0.03
0.004
0.005
0.86**
0.22
1.01**
0.10
0.46**
0.11
0.48**
0.07
Panel B: Auxiliary Model Moments
Actual Data
1. Mean(BR)
2. Mean( r̃ t )
3. Mean( r̃ 2t )
4. Var( ζ )
5. Var( r̃ t )
6. Var( r̃ 2t )
0.46
0.42
0.25
3.01
1.14
0.14
Panel C: Indirect Inference Estimation
Coefficient
1. c 1
2. c 2
3. A
4. B
5. D
6. E
7. F
8. R
9. s
10. L 2
Where:
BR
t
r̃
r̃ 2t
t
2003
D
Settl Wed
D ⋅ ⋅
c1
c2
A
B
D
E
F
R
s
L2
Simulated Data
0.0541**
-0.0485**
0.4257**
0.3432**
-0.0010**
0.2001**
0.0421**
0.8594**
0.0027
0.4828
Simulated Data
0.45
0.37
0.25
2.15
1.93
0.08
Standard Error
0.002
0.004
0.001
0.0008
0.0005
0.0007
0.0004
0.0034
0.00001
0.0016
normalized borrowing from the Federal Reserve
daily high funds rate minus target rate
lowest 50 percent of daily high funds rate less target rate
time trend
dummy variable for period after January 6, 2003
dummy variable for last day of maintenance period
implicit cost prior to 2003
implicit cost after 2003
interval parameter for aggregate shock
interval parameter for bank-specific shock
time trend parameter for aggregate shock
time trend parameter for bank-specific shock
interval parameter for probability of technical problem
average reserve balances
variance of noise parameter for daily high funds rate
implicit reserve target on last day of maintenance period
Source: Authors’ calculations.
* Statistically significant at the 95 percent confidence level.
** Statistically significant at the 99 percent confidence level.
52
Provision of Liquidity through the Primary Credit Facility
Coefficient
Standard Error
References
Artuç, E., and S. Demiralp. 2010. “Discount Window Borrowing after
2003: The Explicit Reduction in Implicit Costs.” Journal of
Banking and Finance 34, no. 4 (April): 825-33.
———. 2003. “Standing Facilities and Interbank Borrowing:
Evidence from the Federal Reserve’s New Discount Window.”
International Finance 6, no. 3 (winter): 329-47.
Carpenter, S., and S. Demiralp. 2006. “The Liquidity Effect in the
Federal Funds Market: Evidence from Daily Open Market
Operations.” Journal of Money, Credit, and Banking 38,
no. 4 (June): 901-20.
Goodfriend, M. 1983. “Discount Window Borrowing, Monetary
Policy, and the Post–October 6, 1979 Federal Reserve Operating
Procedure.” Journal of Monetary Economics 12, no. 3
(September): 343-56.
Clouse, J. A., and J. P. Dow, Jr. 1999. “Fixed Costs and the Behavior of
the Federal Funds Rate.” Journal of Banking and Finance 23,
no. 7 (July): 1015-29.
Gourieroux, C., A. Monfort, and E. Renault. 1993. “Indirect Inference.”
Journal of Applied Econometrics 8, December (Supplement):
S85-S118.
Darrat, A. F., K. Elkhal, G. Banarjee, and M. Zhong. 2004. “Why Do
U.S. Banks Borrow from the Fed? A Fresh Look at the ‘Reluctance’
Phenomenon.” Applied Financial Economics 14, no. 7
(August): 477-84.
Madigan, B., and W. Nelson. 2002. “Proposed Revision to the Federal
Reserve’s Discount Window Lending Programs.” Federal Reserve
Bulletin 88, no. 7 (July): 313-9.
Demiralp, S., and D. Farley. 2005. “Declining Required Reserves,
Funds Rate Volatility, and Open Market Operations.” Journal
of Banking and Finance 29, no. 5 (May): 1131-52.
Dow, Jr., J. P. 2001. “The Recent Behavior of Adjustment Credit at the
Discount Window.” Journal of Macroeconomics 23, no. 2
(spring): 199-211.
Dutkowsky, D. H. 1993. “Dynamic Implicit Cost and Discount
Window Borrowing: An Empirical Investigation.” Journal
of Monetary Economics 32, no. 1 (August): 105-20.
Pearce, D. K. 1993. “Discount Window Borrowing and Federal Reserve
Operating Regimes.” Economic Inquiry 31, no. 4 (October):
564-79.
Peristiani, S. 1998. “The Growing Reluctance to Borrow at the
Discount Window: An Empirical Investigation.” Review
of Economics and Statistics 80, no. 4 (November): 611-20.
Smith, Jr., A. A. 1993. “Estimating Nonlinear Time Series Models
Using Simulated Vector Autoregressions.” Journal of Applied
Econometrics 8, December (Supplement): S63-S84.
Furfine, C. H. 2001. “The Reluctance to Borrow from the Fed.”
Economics Letters 72, no. 2 (August): 209-13.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / August 2010
53
Asani Sarkar and Jeffrey Shrader
Financial Amplification
Mechanisms and the
Federal Reserve’s Supply
of Liquidity during
the Financial Crisis
1. Introduction
O
ne of the primary questions associated with the recent
financial crisis is how losses on subprime mortgage assets
of approximately $300 billion1 led to rapid and deep declines in
the value of a wide range of other financial assets and,
increasingly, real economic output. The disproportionate
amount of total losses compared with the relatively small size
of the initial trigger points to the presence of amplification
mechanisms that allowed losses centered in one market to
cause a systemwide downturn. A further question is why
subprime mortgage-backed securities (MBS) in particular,
rather than any other asset, led to the downturn. Identifying
key factors leading to the crisis, Blanchard (2009) cites the
interaction between general market conditions, such as high
leverage, underpricing of risk, and high interconnectedness,
and certain features of subprime MBS, such as opacity, as well
as investors’ belief in ever-rising housing prices.2
1
See the International Monetary Fund’s “Global Financial Stability Report,”
April 2008.
2
Acharya and Richardson (2009), Adrian and Shin (forthcoming),
Brunnermeier (2009), and Gorton (2008), among others, also describe the
genesis of the crisis and provide explanations for how it was propagated.
Asani Sarkar is a research officer at the Federal Reserve Bank of New York;
Jeffrey Shrader is a former assistant economist at the Bank.
asani.sarkar@ny.frb.org
In this paper, we examine how the conditions identified by
Blanchard and other researchers led to widespread losses in
financial markets. Our study focuses on two financial
amplification mechanisms of relevance to the crisis: balancesheet amplifiers and adverse-selection amplifiers.3 We also
interpret the actions of the Federal Reserve in the context of the
literature on financial amplification mechanisms as well as
provide new empirical evidence on the effectiveness of the
Fed’s liquidity supply during the crisis.
The balance-sheet mechanism is often cited as an
explanation for liquidity crises. For example, it has been used
to explain the stock market crash of 1987 (Brunnermeier and
Pedersen 2009), the Long-Term Capital Management (LTCM)
crisis of 1998 (Gromb and Vayanos 2002), and the current
crisis (Bernanke 2009). Indeed, the Bank of England
incorporates this mechanism into its quantitative Risk
3
For our discussion, a financial amplification mechanism represents the
process whereby an initial shock occurring within the financial sector triggers
substantially larger shocks elsewhere in the sector and in the real economy. A
number of other mechanisms have been proposed in the literature. Examples
are the maturity mismatch between assets and liabilities (Diamond and Dybvig
1983), Knightian uncertainty (Krishnamurthy forthcoming; Pritsker 2010),
and interdependency from credit chains, whereby firms simultaneously
borrow and lend (Kiyotaki and Moore 1997b).
The authors thank Viral Acharya, Mark Flannery, Gary Gorton, Arvind
Krishnamurthy, and Lasse Pedersen for comments. The views expressed are
those of the authors and do not necessarily reflect the position of the Federal
Reserve Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / August 2010
55
Assessment Model for Systemic Institutions, or RAMSI
(Aikman et al. 2009). In all of these cases, the initial trigger was
relatively small in magnitude and local (for example, the
Russian default in 1998 and news associated with mergers and
acquisitions in 1987), but the crisis spread rapidly and globally
to other markets. The amplification underlying these events is
understood to operate as follows: an initial shock tightens
funding constraints, causing the net worth of institutions to
decrease and funding conditions to tighten further. We discuss
the different ways proposed in the literature for funding shocks
to reduce net worth, such as higher margins, lower collateral
value, lower asset market prices, and higher volatility. Since the
literature is extensive, we focus on a small number of key
contributions that introduce alternative feedback loops
between funding shocks and changes in net worth (or, more
generally, balance-sheet conditions).
Central banks appear well placed to mitigate funding
constraints as lenders of last resort (LOLRs). Since banks
typically fund long-term assets with short-term money, a loss
of confidence would force them to engage in asset “fire sales.”
By providing a liquidity backstop, central banks work to
avoid potential fire sales. Bernanke (2009) describes the stages
of the Federal Reserve’s responses to the current crisis. The
first-stage programs—the Term Auction Facility (TAF),
central bank liquidity swaps, the Term Securities Lending
Facility (TSLF), and the Primary Dealer Credit Facility
(PDCF), all introduced between December 2007 and March
2008 (see exhibit)—involved the provision of short-term
liquidity to sound financial institutions, in line with the Fed’s
traditional role of LOLR.4
We describe the Federal Reserve’s first-stage liquidity
programs and discuss available evidence on their effectiveness.
The evidence is consistent with the view that the Fed mitigated
funding stresses by charging lower effective rates on
collateralized funds compared with rates in the private market.
The Fed was able to take such action because, as a patient
investor, it required a lower liquidity risk premium than
private lenders did.
Next, we focus on the adverse-selection mechanism, which
differs from the balance-sheet mechanism in terms of the role
played by credit risk. The balance-sheet mechanism focuses on
“collateralizable” net worth (Bernanke and Gertler 1989) and
secured financing. Here, while credit risk may trigger the initial
funding shock, it plays no role in the amplification mechanism.
Clearly, though, in addition to this balance-sheet effect,
feedback from asymmetric information and credit risk is also a
potentially important amplifier in crisis periods. Indeed, as the
crisis evolved, concerns about the credit risk of financial
institutions and bank capital came increasingly to the fore.
Amplifications from adverse selection appear to be
particularly relevant in the later stages of a crisis. We provide a
brief survey of the literature that focuses mainly on those effects
and their explicit policy implications, particularly for the
current crisis. The literature finds that when borrowers have
private information about their own asset values, private
funding markets may break down, as safe borrowers exit the
markets and lenders, faced with an adverse selection of risky
borrowers, reduce their lending. The market failure provides a
role for liquidity supply by central banks. However, the
literature is also skeptical of the efficacy of such intervention in
the face of asymmetric information.
The Federal Reserve’s crisis interventions evolved along
with the changing nature of the crisis. The second-stage
programs—the Asset-Backed Commercial Paper Money
Market Mutual Fund Liquidity Facility (AMLF), the
Commercial Paper Funding Facility (CPFF), the Money
Market Investor Funding Facility (MMIFF), and the Term
Asset-Backed Securities Loan Facility (TALF), all rolled out
starting in September 2008 (see exhibit)—went beyond
providing liquidity and addressed the funding needs of
borrowers in selected credit markets. With these facilities, the
Fed accepted a certain amount of credit risk, which it managed
through the imposition of haircuts on the collateral given to it.
The increased credit risk that the Fed accepted is attributable to
the longer maturity of the loans (up to five years for TALF
loans, for example), the nonrecourse nature of the loans in the
case of the AMLF and TALF, and the broader set of
counterparties (any U.S. company with eligible collateral can
borrow at the TALF, for instance). Given the relatively late date
of the introduction of these programs, examination of the
programs and their effectiveness remains at an early stage.
Our study concludes by providing fresh evidence on the
effect of changes in the Federal Reserve’s supply of liquidity on
changes in the three-month spread between the London
interbank offered rate and the overnight indexed swap rate,
better known as the Libor-OIS spread.5 In contrast to previous
work that focuses on announcement date effects, our paper
examines changes in the amount outstanding of funds supplied
by the Fed through the TAF and the swap facilities. We control
for credit risk, the uncertainty regarding credit risk, and
liquidity risk, guided by the literature. We distinguish between
periods of increasing and decreasing supplies of funds by the
Fed, and find that increases tend to reduce interest rates during
5
4
We do not consider the Fed’s term financing to JPMorgan Chase for the
acquisition of Bear Stearns on March 14, 2008, to be a liquidity program,
but rather a one-time transaction.
56
Financial Amplification Mechanisms
Libor is a benchmark unsecured interbank interest rate published by the
British Bankers’ Association; OIS represents the expected average of the
overnight fed funds rate over the term of the loan. The spread is widely used
to measure interbank market stress.
Market Events and Federal Reserve Actions
December 2007-December 2008
Fed policy announcements
8/17
Primary credit
penalty rate
reduced and
loan term
extended
9/18
Target rate
to 4.75%
10/31
Target rate
to 4.5%
1/22
Target rate
to 3.5%
4/30
Target rate
to 2%
3/18
Target rate
to 2.25%
Jan.
Feb.
Mar.
10/29
Target rate
to 1%
10/21
Money Market
Investor Funding
Facility (MMIFF)
12/16
Target rate
to 0-25 basis
points
10/8
Target rate
to 1.5%
3/11
Term Securities
Lending Facility
(TSLF)
12/11
Target rate
to 4.25%
Market events
Asset-Backed
Commercial Paper
Money Market
Mutual Fund
Liquidity Facility
(AMLF)
Primary credit
penalty rate
reduced again
12/12
Term Auction
Facility (TAF)
and swap lines
2007
9/19
Fed to purchase
short-term
agency debt
3/16
Primary Dealer
Credit Facility
(PDCF)
1/30
Target rate
to 3%
11/25
Fed to
purchase
long-term
GSE MBS
11/15
10/7
Commercial Paper Term Asset-Backed
Securities Loan
Funding Facility
Facility (TALF)
(CPFF)
April
May
June
July
Aug.
Sept.
Oct.
Nov.
Dec.
2008
3/14
Bear Stearns
gets emergency
loan from Fed
3/16
JPMorgan moves
to purchase
Bear Stearns
7/11
Run on
IndyMac
9/15
10/14
BoA purchases Nine banks
Merrill Lynch
agree to
Treasury
Lehman files
capital
for bankruptcy
injection
AIG debt
11/23
downgraded
Citigroup receives
government
9/16
assistance
RWC money
market fund
“breaks the buck”
9/25
WaMu closed
by OTS
9/29
Systemic risk
exception granted
to Wachovia
periods of high funding liquidity risk. Surprisingly, decreases in
the supply of funds also appear to be associated with lower
spreads. Moreover, the impact of the funds supply on the
spread has diminished over time, a result that is helpful in
evaluating the impact of the Fed’s potential future exit from
its liquidity programs.
2. The Balance-Sheet Amplification
Mechanism
The literature on balance-sheet mechanisms focuses on the
principal agent problem between borrowers and lenders that
arises from delegated investment. Households invest in hedge
funds and mutual funds that invest in securities; these funds
may in turn invest with more specialized investors with
expertise in sophisticated trading strategies.6 The principal
FRBNY Economic Policy Review / August 2010
57
agent problem is defined as a deviation from first best
outcomes associated with the necessity of external financing
(Bernanke and Gertler 1989), and a consequence is that the
intermediary’s investments come to depend on external
financing terms and the intermediary’s balance-sheet
conditions.
The balance-sheet amplification channel involves a positive
feedback between funding constraints and changes in the asset
values or cash flow of intermediaries. An early example is
provided in Bernanke and Gertler (1989), who show how
funding shocks reduce borrowers’ cash flows and impair their
ability to finance investments from retained earnings, thereby
increasing the cost of new investments. They propose a model
in which borrowers have better information about project
quality than potential lenders do.7 The resulting agency cost
creates a wedge between the borrower’s costs of internal and
external funds. Moreover, the external funds premium is
greater when borrower net worth is lower, as in periods of
financial distress. This inverse relationship arises because
agency costs are higher when borrower cash flows are lower
and consequently the external funds premium must be greater
to compensate the lender. Reduced investments result in lower
output and cash flows, creating a “financial accelerator” effect
of cash flows on investments attributable to countercyclical
agency costs.
In literature subsequent to Bernanke and Gertler, emphasis
is placed on the effect of funding shocks on asset prices (instead
of cash flows), which affects firm net worth through changes in
the value of assets and liabilities (Kiyotaki and Moore 1997a;
Shleifer and Vishny 1997; Gromb and Vayanos 2002;
Brunnermeier and Pedersen 2009). Since asset prices are
forward looking, persistent shocks that impact them can have
potentially large wealth effects.
The generic balance-sheet constraint for time t can be
expressed (following Krishnamurthy [forthcoming]) as:
(1)
mt θt ≤ wt ,
where m is broadly interpreted as a “margin” requirement per
unit of asset holding, θ is the number of units of assets, and w
is the value of equity capital. This interpretation of m is
consistent with Gromb and Vayanos (2002) and Brunnermeier
and Pedersen (2009).8 In other words, the firm’s equity capital
6
For example, the “Fund of Funds” strategy is used by hedge funds that invest
in other hedge funds.
7
The superior information arises because the lender is assumed to pay a fixed
auditing cost in order to observe the borrower’s realized return, whereas the
borrower observes a return for free.
8
Margin constraints are perhaps the most common example of a balance-sheet
constraint, but other constraints are possible. For example, in He and
Krishnamurthy (2008), incentive conflicts limit the amount of coinvestment
by outsiders in a mutual fund.
58
Financial Amplification Mechanisms
must be sufficient to cover its total margins. Higher margins
reduce asset prices, which in turn lower w and cause the
constraint to tighten further; this is the feedback loop between
funding conditions and asset market prices.
An alternative interpretation of m is obtained from Kiyotaki
and Moore (1997a), in which lenders limit the debtor’s
investments based on pledged collateral. Suppose that
borrowers pledge θ units of assets to borrow γθ P , where P is
the asset price and γ < 1 . Then, the borrower’s budget
constraint is:
(2)
θ t P t ≤ γθ t P t + w t .
Or, rewriting,
(3)
( 1 – γ )P t θ t ≤ w t .
Here, γ can be viewed as the “haircut” on the collateral. If we
write m = ( 1 – γ )P , then equations 3 and 1 are equivalent
expressions of the budget constraint.
In Kiyotaki and Moore (1997a), credit constraints arise
because borrowers can only borrow against assets that can be
pledged as security for the loan. The pledgable assets have a
dual capacity: as factors of production and as collateral. An
initial productivity shock reduces the net worth of constrained
firms, resulting in lower investments and lower prices of
pledgable collateral assets. As asset prices fall, constrained firms
suffer a capital loss on their collateral asset, and the magnitude
of this loss is large because of leverage. The subsequent
reduction in borrowing capacity leads to further rounds of
decreased investments, asset prices, and borrower net worth.
While Bernanke and Gertler (1989) and Kiyotaki and
Moore (1997a) are concerned with “collateralizable” net
worth, they acknowledge but do not consider the market
liquidity of the collateral. This issue is addressed by Shleifer and
Vishny (1997), Gromb and Vayanos (2002), and Brunnermeier
and Pedersen (2009). These papers are also concerned with the
two-way feedback between borrowing limits and asset prices
present in Kiyotaki and Moore. However, they also introduce
the idea of a positive feedback between funding illiquidity and
market illiquidity. Funding illiquidity is the marginal investor’s
scarcity value (or shadow cost) of capital; market illiquidity is
the difference between the transaction price of a security and its
fundamental value. The amplification mechanism discussed in
these papers may be used to understand purely financial crises,
independent of any effects on the real economy, such as the
stock market crash of 1987 and the LTCM crisis of 1998.
Shleifer and Vishny (1997) examine the effect of
intertemporal wealth constraints on the incentives of
arbitrageurs to eliminate mispricings between two securities
with identical cash flows. They consider the agency relationship
between arbitrageurs with specialized market knowledge, such
as hedge funds, and the investors who fund them, such as
wealthy individuals, banks, and endowments. If investors chase
returns, they are likely to withdraw capital from arbitrageurs
when prices are falling. In turn, arbitrageurs lacking capital are
unable to reduce mispricing. This phenomenon is referred to as
the “limits of arbitrage.”
Gromb and Vayanos (2002) provide a welfare analysis of
competitive arbitrage. In the process, they formalize many of
the intuitions of Shleifer and Vishny (1997). The possibility of
arbitrage arises because of segmented asset markets: some
investors are able to invest in one risky asset but not in another
(identical) risky asset. Arbitrageurs can invest in both assets
and act as intermediaries: by exploiting price discrepancies,
they facilitate trade among investors, effectively providing
liquidity to them. Thus, arbitrage activity benefits all investors.
It is assumed that arbitrageurs must have separate margin
accounts for the two assets (that is, there is no crossmargining).9 This implies that arbitrageur positions are wealth
constrained. Gromb and Vayanos show that if changes in
arbitrageur wealth are insufficient to cover variations in both
margin accounts, arbitrageurs may be unable to take a position
large enough to eliminate price discrepancies. Further,
arbitrageurs may choose not to invest up to their wealth
constraint if the capital gain from the arbitrage position is
expected to be risky.10 They can also increase price volatility by
liquidating their positions in the event that price discrepancies
widen further.
The feedback loop in Kiyotaki and Moore (1997a) and
Gromb and Vayanos (2002) may be called an illiquidity spiral:
reductions in collateral values result in lower asset prices and
further reductions in collateral values. In terms of equation 3,
the feedback is between θ P and w, for given m. By comparison,
Brunnermeier and Pedersen (2009) derive a margin spiral, in
which lower asset prices reduce arbitrageur net worth through
higher margins. In terms of equation 1, the feedback is between
m and w, for given θ . While this distinction is useful for
expositional reasons, changes in m and θ are clearly
interdependent.
Brunnermeier and Pedersen examine the relationship
between margin conditions and market illiquidity. In their
model, customers with offsetting demand shocks arrive
sequentially to the market. Speculators smooth the temporal
order imbalance and thereby provide liquidity. The speculators
9
The authors argue that this assumption captures the notion that a custodian
of the margin account in one market might refuse to accept a position in a
different market as collateral. This assumption may not hold in all asset
markets, however. For example, an arbitrageur with a simultaneous position in
Treasury spot and futures markets generally cannot cross-margin.
10
This follows from the possibility that the price discrepancy may grow wider
and result in capital losses for arbitrageurs.
borrow using collateral from financiers who set margins
(defined as the difference between the security’s price and its
collateral value) to control their value-at-risk. Financiers can
reset margins every period, so speculators face funding
liquidity risk from the possibility of higher margins or losses on
existing positions. A margin spiral occurs as follows: Suppose
markets are initially highly illiquid and margins are increasing
in market illiquidity.11 A funding shock to speculators lowers
market liquidity and results in higher margins, which cause
speculators to delever, further tightening their funding
constraints. Therefore, market liquidity falls even further.
There is no default risk in balance-sheet models, as loans are
fully collateralized.12 Thus, amplification works through fund
flows and liquidity risk. The fact that inefficiencies can arise in
the absence of credit risk suggests the positive role of central
banks in alleviating funding and capital constraints during
periods of crisis.
3. The Balance-Sheet Amplification
Mechanism: Implications
for Central Banks
The welfare analysis of Gromb and Vayanos (2002) shows that
arbitrageurs may not take on an optimal level of risk, in part
because they fail to internalize the effect on prices of changing
their positions.13 For example, arbitrageurs may underinvest in
an arbitrage opportunity because they do not consider the
possibility that larger positions in the current period would
reduce price discrepancies in future periods. Thus, the key
source of allocative inefficiency is the negative externality from
changes in an arbitrageur’s positions on other arbitrageurs.
11
This occurs if financiers are unsure if price changes are attributable to news
shocks or liquidity shocks, and if volatility is time varying. Under these
conditions, liquidity shocks lead to higher volatility, which increases financiers’
expectations of future volatility; this in turn leads to higher margins. In
contrast, if financiers know for sure that price changes are linked to
fundamental news shocks, they realize that prices will revert in the future,
making arbitrage positions in the current period profitable. This reduces the
incentives of financiers to increase margins when liquidity decreases.
12
This is explicit in Kiyotaki and Moore (1997a). Bernanke and Gertler (1989)
explain that their model is about “collateralizable” net worth. The models of
Gromb and Vayanos (2002) and Brunnermeier and Pedersen (2009) rule out
default because margin accounts must be fully collateralized.
13
An important reason why arbitrageur position changes are “Paretoimproving”—that is, they make some people better off without making anyone
worse off—is that price changes result in wealth redistributions, and market
segmentation implies that agents’ marginal rates of substitution differ (as
shown by Geanakoplos and Polemarchakis [1986] in a general, incomplete
market setting). Arbitrageurs prefer to receive more wealth earlier while other
investors prefer to receive wealth later; this creates the potential for Paretoimproving wealth redistributions across time and states.
FRBNY Economic Policy Review / August 2010
59
An implication of Gromb and Vayanos is that regulatory
intervention may affect arbitrageurs’ financial constraints by
reducing their capital and margin requirements or by
providing financing to those institutions that provide capital to
arbitrageurs.14 Since the ex ante choice of leverage may be
suboptimal, there is scope for prudential capital and liquidity
requirements and, more generally, regulation of financial
sector balance sheets. In addition, ex post policy actions to
address the allocative inefficiency should be welfare improving,
although they need not be unanimously approved (because of
distributional effects).
In Bernanke and Gertler (1990), the optimal policy is a
“debtor bailout,” whereby the government redistributes
endowment (via lump-sum taxes) from lenders to borrowers
until the agency cost disappears. The policy works by directly
addressing the problem of low net worth of borrowers
(financial firms such as brokers, banks, and clearinghouses).
Further, such transfers need not be direct, rather, they could be
channeled through financial intermediaries under the
assumptions that the latter can identify legitimate borrowers
and that the government ensures that funds are channeled to
successful projects. The moral-hazard problem is addressed by
recommending bailouts only in response to large aggregate or
systemic shocks over which borrowers have no control.
Brunnermeier and Pedersen (2009) discuss the ability of
central banks to enhance market liquidity by controlling
funding liquidity. If a central bank is effective at distinguishing
news shocks and liquidity shocks and it conveys this distinction
to financiers, the financiers may ease their margin
requirements. Alternatively, the central bank can directly ease
speculator funding conditions during a crisis, either by
providing emergency funding at lower margins or simply by
stating its intention to do so. If the statement is credible,
financiers may loosen margin requirements, because their
worst-case scenarios have a lower probability of occurring.15
14
When regulators have limited control over financial constraints, they may
prefer to tighten them in some cases to reduce overinvestment (for example, by
limiting entry into the arbitrage industry). Overinvestment occurs if
arbitrageurs are initially fully invested in the arbitrage opportunity. If demand
by other investors increases, the price discrepancy widens and arbitrageurs
suffer capital losses on their current positions. If arbitrageurs reduce their
positions, they limit losses and can provide liquidity in future periods by
trading more aggressively, a practice that mitigates the price wedge.
15
Allen, Carletti, and Gale (2009) provide another rationale for central bank
intervention. When markets are incomplete, the authors show that the price of
the long-lived asset may exhibit excessive volatility. By using open market
operations appropriately to set interest rates, the central bank can prevent the
price volatility and implement the constrained efficient solution. Thus, the
central bank effectively completes the market, and open market operations are
sufficient to address systemic liquidity crises.
60
Financial Amplification Mechanisms
4. The Federal Reserve as Lender
of Last Resort during the Early
Stages of the Crisis
We turn to an assessment of the Federal Reserve’s ex post
interventions during the financial crisis, viewed in the context
of the balance-sheet literature. From equations 1 and 3, we
observe that a regulator has three types of instruments at its
disposal:
• reducing m, the required margins on new funds;
• increasing γ , the value of pledgable assets;
• increasing w, the equity capital.
We focus on the Fed’s efforts to reduce m and increase γ
during the early stages of the crisis. Traditional LOLR policies
advocate lending to solvent institutions against good collateral at
a penalty rate (Rochet and Vives 2004). However, Cecchetti and
Disyatat (2010) argue that, when there is generalized market
failure, it may not make sense to provide liquidity at a penalty
rate over the market rate because no institution benefits relative
to others. The authors conclude that “liquidity support will
often, and probably should, be provided at a subsidized [relative
to the market] rate when it involves an illiquid asset in which a
market price cannot be found.”
Normally, the Fed provides reserves to a small number of
primary dealers that distribute the funds to banks via interbank
markets; in turn, banks lend to ultimate borrowers. When the
markets are disrupted, the Fed relies on the discount window
facility to provide short-term backup funding to eligible
depository institutions. In the current crisis, interbank markets
were dysfunctional, especially for term lending. The Fed
encouraged banks to borrow from the discount window, but
the banks were reluctant, perhaps in part because of the
“stigma” associated with such borrowing.16
Responding to these concerns, the Fed introduced a number
of programs (the aforementioned stage-one group) between
December 2007 and March 2008 designed to provide shortterm liquidity to sound financial institutions.17 In the context
of the balance-sheet literature, the programs can be viewed as
easing balance-sheet constraints and thereby breaking the
illiquidity spiral. An example is the TSLF, which allows dealers
to exchange illiquid securities (say, MBS) for liquid Treasury
securities that the dealers can subsequently use as collateral to
16
For example, Furfine (2003) presents evidence consistent with potential
borrowers staying away from the discount window, perhaps out of concern
that such borrowing would be viewed as a sign of higher credit risk. Armantier
et al. (2010) provide evidence that a discount window stigma existed
throughout the financial crisis.
17
See Armantier, Krieger, and McAndrews (2008), Adrian, Burke, and
McAndrews (2009), and Fleming, Hrung, and Keane (2009) for descriptions of the
TAF, PDCF, and TSLF programs, respectively. For descriptions of other Federal
Reserve programs, see http://www.federalreserve.gov/monetarypolicy/bst.htm.
borrow funds. The dealer pays a smaller haircut (say,
H_Treasury) when borrowing against liquid Treasuries than
what it pays (say, H_Illiquid) when borrowing against illiquid
securities. Of course, the TSLF also charges a haircut (say,
H_TSLF). However, as long as H_TSLF < (H_Illiquid H_Treasury), the facility lowers the dealer’s net funding costs.
Thus, the TSLF may be viewed as increasing γ in equation 3.
Other stage-one programs may be viewed as breaking the
margin spiral (reducing m in equation 1). For example, the
TAF auctioned credit to eligible depository institutions for a
term of twenty-eight days initially and up to eighty-four days
by August 2008. A similar program, the PDCF, issued credit to
primary dealers. The international counterpart to TAF is
bilateral currency swap arrangements with foreign central
banks, which allow the banks to provide dollars to institutions
in their own jurisdictions. These programs may bring down m
in two ways: They may provide financing when private
financing is simply unavailable, or when private financing is
available only at more expensive terms.
How effective were these programs in reaching their
objectives? To answer this question, we examine one liquidity
risk proxy: the spread between overnight repo rates on MBS
and Treasury securities.18 Because both MBS and Treasury
repo loans are collateralized and are issued for a short
(overnight) maturity, the spread between them mainly reflects
the relative illiquidity of the two assets. In particular, during the
crisis, investors sought safety in the Treasury market while
agency MBS became relatively illiquid, leading to an increase in
the spread between agency MBS and Treasury repos.19 The
repo markets are important for bank financing (Hordahl and
King 2008). In addition, if the secured financing market is
stressed, it is highly likely that the unsecured financing market
is also under duress. For these reasons, the MBS-Treasury repo
spread provides a good proxy for funding illiquidity in the
economy, not just in the secured financing markets.
The source for the MBS-Treasury spread data is the Federal
Reserve Bank of New York’s primary dealer survey. The Trading
Desk at the New York Fed collects information each morning
from dealers on the average overnight general collateral repo rate
at which each dealer has financed its positions in Treasury
securities, agency debt securities, and agency MBS, as well as the
quantity of securities financed. An overall weighted average is then
calculated for each collateral type.
18
Overnight repo rates on MBS are general collateral repo rates that reference
nonspecific government securities with the lowest level of counterparty risk
(Hordahl and King 2008). In contrast, specific collateral rates reference
particular types of collateral, such as an on-the-run bond.
19
Brunnermeier (2009) uses the repo spread (although not of the overnight
maturity) to illustrate liquidity risk during the financial crisis. Gorton and
Metrick (2009) discuss the role of repo markets during the crisis.
Chart 1
Liquidity Risk during the Financial Crisis
Basis points
350
300
TAF and TSLF
swap lines
CPFF
MMIFF
TALF
PDCF
250
200
150
100
50
0
2007
2008
Overnight MBS–Treasury repo spread
2009
Sources: Federal Reserve Bank of New York; Haver Analytics.
Notes: MBS is mortgage-backed securities. Full names of the liquidity
facilities appear in the exhibit on page 57.
Providing evidence on the effectiveness of the TSLF and
PDCF programs, the spread between overnight agency MBS
repo rates and Treasury collateral repo rates decreased after the
TSLF program was implemented (Chart 1). Fleming, Hrung,
and Keane (2009) show that this reduction is statistically
significant. They further show that the narrowing of the repo
spread is primarily attributable to increases in the Treasury
repo rate and less so to decreases in the MBS repo rate.
However, as the authors note, increases in the Treasury repo
rate are important for the liquidity of the market.20 Since the
overnight repo spread may be attributable to the reduced
collateral value (from lower market liquidity) of MBS relative
to Treasuries, or to the increased collateral value of Treasuries
(from higher market liquidity) relative to MBS, the reduction
in the spread suggests an increase in γ .
The top panel of Chart 2 shows the difference between the
Libor, which is taken to be the benchmark borrowing rate in
the private markets, and the discount window borrowing rate
(the primary credit rate).21 The discount window rate was
20
Treasury securities are widely used as collateral for secured funding, so
improved liquidity for Treasuries is likely to have a beneficial effect on secured
funding rates in general. In addition, Fleming, Hrung, and Keane (2009) observe
that an “unusually low Treasury general collateral repo rate puts downward
pressure on repo rates for individual Treasury securities, increasing the
likelihood of settlement problems” (also see Fleming and Garbade [2004, 2005]).
21
The Libor is used for unsecured funding while the prime rate and the stopout rate are used for secured funding. However, much of the collateral posted
to the Fed was illiquid and could not be used to obtain secured funding
elsewhere. Therefore, the Libor closely approximates the opportunity cost of
funds for TAF participants.
FRBNY Economic Policy Review / August 2010
61
The Federal Reserve’s success in easing funding constraints
during the crisis likely had a beneficial effect on the real
economy, via the channels suggested in Bernanke and Gertler
(1989) and Kiyotaki and Moore (1997a). Del Negro et al.
(2009), who extend the model of Kiyotaki and Moore (2008),
study the impact of a large shock of the order of magnitude
observed in the 2008 financial crisis. Their model simulations
suggest that the Fed’s policy interventions in 2008-09
prevented a repeat of the Great Depression.
Chart 2
Cost of Borrowing from the Federal Reserve
Relative to the Market
Basis points
300 Discount window
TAF and
swap lines
TALF
PDCF
200
TSLF
100
5. The Adverse-Selection
Amplification Mechanism:
Implications for Central Banks
0
MMIFF
CPFF
-100
Three-month Libor–primary credit spread
300
Term Auction Facility
TAF and TSLF PDCF
swap lines
CPFF
TALF
200
100
0
MMIFF
-100
2007
2008
Libor-TAF stop-out spread
2009
Sources: Federal Reserve Bank of New York; Haver Analytics; British
Bankers’ Association; Bloomberg.
Notes: Libor is the London interbank offered rate. Solid circles represent
the one-month Libor–twenty-eight-day TAF stop-out spread; open circles
represent the three-month Libor–eighty-four-day TAF stop-out spread.
For twenty-eight-day TAF auctions, the Libor-TAF spread is calculated as
the spread between the one-month Libor and the twenty-eight-day TAF;
for eighty-four-day TAF auctions, the spread is calculated as the spread
between the three-month Libor and the eighty-four-day TAF. Full names
of the liquidity facilities appear in the exhibit on page 57.
The Federal Reserve’s first-stage liquidity programs exposed it
to minimal credit risk. The Fed’s loans to banks and primary
dealers through the various facilities are overcollateralized and
made with recourse to the borrowing firm.23 In the case of the
currency swap lines, the foreign central banks are responsible
for payments; moreover, the Fed receives and holds an
equivalent amount of foreign exchange for the dollars it
provides to the central banks.
As the crisis evolved, concerns about the credit risk of
financial institutions and bank capital came increasingly to the
fore. The Fed’s stage-one programs were dependent on solvent
institutions to intermediate credit flow from the central bank
to the economy.24 As these intermediaries themselves became
impaired, they were less willing to lend. In addition, certain
credit markets, such as commercial paper, were particularly
afflicted. Consequently, the Fed decided to lend directly to
some affected borrowers and markets. Thus, with its secondstage programs, the Fed was forced to take on and manage a
certain amount of credit risk.
To understand the intent behind these programs, we
examine amplification mechanisms based on asymmetric
information between borrowers and lenders. In contrast to our
22
initially above the Libor, a development that partly explains
banks’ reluctance to use the window early in the crisis. The
bottom panel of the chart illustrates the difference between the
Libor and stop-out rates in the twenty-eight- and eighty-fourday TAF auctions. It shows that the Libor generally exceeded
the stop-out rates, indicating that the Fed was successful in
providing credit at below-market rates. In addition, evidence
indicates that the TAF and the swap line programs reduced
interest rate spreads.22
62
Financial Amplification Mechanisms
McAndrews, Sarkar, and Wang (2008) study the effect of the TAF on the
Libor-OIS spread. McAndrews (2009) and Coffey, Hrung, and Sarkar (2009)
analyze the effect of swap lines: the former on the Libor–fed funds spread, the
latter on deviations from covered interest rate parity. Cetorelli and Goldberg
(2009) examine the effect of liquidity programs on the internal capital markets
of global banks.
23
For a description of the required collateral, see http://www.federalreserve.gov/
monetarypolicy/bst_ratesetting.htm.
24
The Federal Reserve’s objective was to improve the distribution of liquidity
across financial intermediaries, as stated in its announcement of the TAF
program on December 12, 2007 (http://www.federalreserve.gov/newsevents/
press/monetary/20071212a.htm). This objective could not have been achieved
by way of a generalized injection of liquidity, such as through the purchase of
Treasury debt.
review of balance-sheet amplifiers, we focus here on the role of
credit risk and the distribution of credit risk across borrowers.
The papers surveyed in this discussion find a role for central
bank intervention when adverse-selection problems lead to
market breakdowns. However, they also raise concerns that
central bank liquidity provision might crowd out private
market liquidity.
Heider, Hoerova, and Holthausen (2009) build a model of
the effect of counterparty risk on unsecured interbank markets
with asymmetric information.25 Banks need liquidity, as
customers may withdraw deposits on demand (as in Diamond
and Dybvig [1983]). The interbank markets distribute funding
from banks with excess reserve balances to those with reserve
shortages. Counterparty risk exists because banks have risky
long-term assets and may be unable to repay their interbank
loans. Asymmetric information about counterparty risk exists
because banks have private information about the riskiness of
their long-term assets.
The authors show that different regimes occur in the
interbank markets depending on the level and distribution of
counterparty risk. Because lenders cannot distinguish between
safe and risky banks, the interest rate contains a risk premium.
In the “good” regime, the risk premium is small compared with
the opportunity cost of funds, so the interbank markets
perform smoothly with low interest rates. If, however, the risk
premium is too high, safe borrowers exit the interbank
markets. Consequently, in this “worst” regime, lenders face an
adverse selection of risky borrowers and the interest rate is
high. In this regime, both the level and the dispersion of credit
risk are high;26 as a result, the interbank markets stop
functioning. Either lenders find it unprofitable to lend (even at
high interest rates) and thus hoard funds,27 or risky borrowers
find the interest rate too high and exit.
What are the implications of this model for central bank
liquidity supply? 28 Suppose credit risk increases unexpectedly
and lenders face an adverse selection of borrowers (but the
market is still functioning). If the central bank has the same
information as the market, it can offer liquidity to all banks at
the highest interest rate that safe banks are willing to accept. As
25
Flannery (1996) also studies asymmetric information problems and
identifies a “winner’s curse” facing new lenders in banking markets. He shows
that private loan markets can fail because lenders become less certain about
how to distinguish between illiquid and insolvent banks.
26
If p s ( pr ) is the probability that the long-term investment has a higher-
(lower-) than-expected chance of success, dispersion is defined as p s – p r .
27
Liquidity hoarding can also arise if banks fear that they will be unable to
finance projects and trading strategies because of uncertainty in the aggregate
demand for liquidity (Allen, Carletti, and Gale 2009). In such a case, central
bank intervention may not be needed because banks hold sufficient liquidity to
meet their own needs without accessing the interbank markets (Allen and
Carletti 2008).
in Flannery (1996), this rate is discounted relative to the market
rate, and the central bank’s supply of liquidity mitigates the
private liquidity shortage. The cost is that the central bank does
not distinguish between sound and risky institutions, a concern
also raised by Goodfriend and King (1988). Moreover, the
private supply of liquidity is crowded out.
Bolton, Santos, and Scheinkman (2009) also raise the
possibility that central bank liquidity crowds out private
liquidity.29 Their model features two types of investors: shortrun investors, who invest in valuable risky projects that
typically mature early, and long-run investors, who invest in
higher return long-term assets. The ex ante efficient solution is
for short-run investors to sell risky assets to long-run investors
(to obtain “outside” liquidity) and for trading not to occur too
quickly. However, short-run investors have private
information about the assets. If investors are concerned about
adverse-selection problems that may undermine secondary
markets in the future, they may trade too soon and at fire-sale
prices.
A central bank may step in and provide liquidity (in the
form of price support) to mitigate the fire sale. The
effectiveness of liquidity supply depends on whether the central
bank can accurately time the supply. If it delays liquidity
provision, it crowds out outside liquidity and undermines the
incentives of short-run investors to obtain outside liquidity by
selling assets for cash. However, if the central bank acts quickly,
its liquidity can complement private market liquidity. In this
case, the central bank plays the role of market maker of last
resort by inducing short-run investors to obtain liquidity
through asset sales.
28
There is a vast literature on central bank or government intervention to
address market failures in the face of asymmetric information, moral hazard,
and monopoly power. Holmstrom and Tirole (1998) and Diamond and Rajan
(2005) analyze the optimal (central bank) provision of liquidity when
interbank markets face aggregate liquidity shocks and contagious failures
generated by the illiquidity of bank assets. Gorton and Huang (2006)
rationalize the LOLR function of central banks with the need to monitor banks
and provide them with liquidity during crises in order to prevent inefficient
panics. Acharya, Gromb, and Yorulmazer (2008) examine how the strategic
power of an interbank lender might force a liquidity-constrained borrower to
sell at fire-sale prices. The strategic power is the market failure that justifies
central bank intervention.
29
Bolton, Santos, and Scheinkman (2009) build on the literature that integrates
financial intermediaries and securities markets in a single framework. In
Diamond (1997), banks coexist with securities markets because households face
costs associated with switching between banks and securities markets. Fecht
(2004) introduces segmentation on the asset side between financial
intermediaries’ investments in firms and claims issued directly by firms to
investors through securities markets. Allen and Gale (2004) introduce securities
markets into a general-equilibrium theory of institutions. Intermediaries provide
liquidity insurance, as in Diamond and Dybvig (1983), and risk-sharing services
by packaging existing claims for investors that lack access to markets. The
financial system is efficient as long as markets are complete.
FRBNY Economic Policy Review / August 2010
63
6. Adverse Selection and the Fed’s
Actions during the Later Stages
of the Crisis
The Fed’s second-stage programs were designed to provide
funding in a targeted manner to borrowers and investors in key
credit markets (Bernanke 2009). These programs, rolled out
starting in September 2008, came in two varieties (see exhibit).
Continuing its LOLR role, the Federal Reserve provided a
liquidity backstop to money market mutual funds and to
commercial paper borrowers. The Fed developed a facility to
finance bank purchases of high-grade asset-backed commercial
paper from money market mutual funds, which helped the
funds to meet redemption demands without having to sell
assets at distress prices. Through another facility, the Fed
bought high-quality (A1-P1) commercial paper at a term of
three months, which reduced the risk of commercial paper
borrowers being unable to roll over maturing issues.
The second type of Federal Reserve programs went beyond
providing liquidity to address the funding needs of borrowers
in selected asset-backed markets. The TALF, representing a
joint effort with the U.S. Treasury, provides three- or five-year
term loans to investors against (mostly) new issuances of AAArated securities. With the Treasury providing funding, the
facility allows the Fed to accept a certain amount of credit risk.
The Fed manages the credit risk through the imposition of
haircuts on the collateral put to it. The objective of the program
is to revive private lending by enabling lenders to securitize new
loans. In addition, by stimulating market activity, the facility
potentially increases the valuation of existing loans by reducing
the illiquidity premium.
The design of the TALF program appears to address the
concern that the Fed might crowd out the private supply of
liquidity in the affected markets. The program leverages private
originations of asset-backed securities, consistent with Bolton,
Santos, Scheinkman (2009). Further, it offers funding at
different rates for various asset classes (as the haircuts differ by
asset). This feature appears to alleviate the moral-hazard
problems inherent in offering a flat rate to all investors
independent of credit risk, a concern raised by Goodfriend and
King (1988) and Heider, Hoerova, and Holthausen (2009).
Given the relative newness of these programs, rigorous
empirical evidence on their effectiveness is scarce. An
exception is Ashcraft, Garleanu, and Pedersen (2009), who
report the results of a survey of financial institutions on how
the institutions’ bid prices for securities depend on Federal
Reserve financing. The Fed, by offering loans at lower margins
than the market, effectively lowers the required return for
holding securities put to the TALF. Consistent with this idea,
the surveyed bid price increases as the Fed reduces its offered
64
Financial Amplification Mechanisms
margins. This evidence is consistent with the expected effect of
lower margins on asset prices.
7. Evolution of Credit and Liquidity
Risk during the Crisis
As the crisis progressed, the relative importance of the balancesheet and adverse-selection mechanisms likely changed. This
evolution is implicit in the timing of the Fed’s responses. In
particular, the Fed’s stage-one programs emphasized the
provision of liquidity to solvent institutions, suggesting that at
this early point in the crisis the Fed viewed a lack of access to
funding as a greater risk to the economy than counterparty
credit risk. In contrast, the second-stage programs reflected the
Fed’s views on the increasing importance of credit risk. In this
section, we estimate proxies for liquidity risk, credit risk, and
the distribution of credit risk across banks to examine the
changing importance of the financial amplification
mechanisms over time.
The adverse-selection effects operate via credit risk
and its distribution across banks (Heider, Hoerova, and
Holthausen 2009). The credit risk measures considered here
are the CDX IG index of credit default swap (CDS) spreads and
the dispersion in Libor panel quotes. The CDX IG index,
provided by Markit, is composed of spreads on five-year CDS
contracts for 125 North American companies; it provides
information on the average default risk of major global firms.
Because the index tends to rise with increases in the level of
economy-wide credit risk, we expect a positive relationship
between the index and adverse selection.
The Libor panel dispersion, provided by the British Bankers’
Association via Bloomberg, is defined as the difference between
the maximum and minimum three-month quote of the sixteen
Libor panel banks each day; it proxies for uncertainty about
counterparty credit risk. The quote dispersion shows the extent
to which some Libor panel banks report greater borrowing
costs, an indicator of higher counterparty risk compared with
the typical Libor panel bank. Our uncertainty measure is
consistent with those proposed in Heider, Hoerova,
Holthausen (2009) and Pritsker (2010), that is, the spread in
default probabilities assigned by lenders to a borrower’s
investments. Again, the expected relationship between the
quote dispersion and adverse selection is positive.
Balance-sheet effects operate according to illiquidity and
margin conditions. To measure liquidity risk, we use the spread
overnight MBS and Treasury general collateral repo rates. As
discussed in Section 4, the spread between these two rates
should primarily reflect the relative illiquidity of MBS relative
Chart 3
Risk Evolution during the Crisis
Basis points
400
TAF and
swap lines
TSLF
CPFF
PDCF
TALF
MMIFF
300
Overnight
MBS–Treasury
repo spread
CDX spread
200
Three-month
Libor–OIS spread
100
Libor quote
dispersion
0
2007
2008
2009
Sources: Federal Reserve Bank of New York; Haver Analytics, Markit; British Bankers’ Association.
Notes: MBS is mortgage-backed securities; Libor is the London interbank offered rate. The overnight MBS–Treasury repo spread is the liquidity risk proxy;
the CDX spread and the Libor quote dispersion are the credit risk proxies. Full names of the liquidity facilities appear in the exhibit on page 57.
to Treasuries. The credit risk component of these two rates is
minimal because of the secured nature of the transaction, the
short duration of the loan, and haircuts that are generally set in
advance. In contrast, the daily repo rate on a given day reflects
supply and demand pressures in the market. During the
financial crisis, there was a rush to buy Treasuries, which
increased the demand for these securities. The greater demand
likely lowered the risk of a repo buyer being unable to sell the
Treasuries in the event of counterparty default. Impairment in
the MBS market, however, meant that the same was not true for
buyers accepting MBS securities as collateral. Therefore, the
differences in these two rates reflect the ability of buyers to
quickly and easily sell the collateral from their repo
transactions—in other words, the two securities are relatively
liquid. We compare these series to the three-month Libor-OIS
spread, which contains credit and noncredit risk premia.
Arbitrage should normally ensure that the spread is close to
zero, but the spread has widened dramatically during the crisis
(Chart 2).30 The variable considered here takes Libor quotes
reported on day t+1 and the OIS rate reported on date t, both
at a term of three months. We use t+1 Libor rates because the
rate is fixed each morning at 11:00 a.m. London time while the
OIS rate is determined at the end of the business day, U.S.
Eastern time.
Chart 3 illustrates the evolution of liquidity risk (the MBSTreasury repo spread) and credit risk (the CDX IG index and
Libor quote dispersion) during the crisis, along with the LiborOIS spread. All values are in basis points. The evolution of risk
proxies is consistent with the view that, at the beginning of the
crisis, liquidity risk was relatively more important than credit
risk, but credit risk became more prominent as the crisis
progressed, gaining particular importance after April 2008 and
especially during September 2008. The initial months of the
crisis were characterized by large spikes in liquidity risk, but
only a modest rise in credit risk. After April 2008, however,
liquidity risk fell while the CDX spread remained elevated.
After mid-September 2008, both types of risk increased, but the
two credit risk proxies increased relatively more and remained
elevated longer.
The Libor-OIS spread appears to co-move with both the
credit and liquidity risk variables during the crisis period. We
examine changes in the spread more formally in the next
section.
30
The arbitrage works as follows: loan $X for, say, three months, fund the loan
by borrowing $X each day in the fed funds market and, finally, hedge the
interest rate risk by purchasing an OIS contract (Gorton and Metrick 2009).
FRBNY Economic Policy Review / August 2010
65
8. Effectiveness of the Fed’s
Liquidity Supply: Methodology
Here, we investigate the relationship between the Libor-OIS
spread and the supply of funds through the Federal Reserve’s
TAF and swap facilities. We focus on the latter facilities because
they are the longest running new programs introduced during
the crisis, and because both were meant to provide dollar
funding to the interbank markets (in contrast to other stageone liquidity programs, such as the TSLF).
We interpret the TAF and swap programs as primarily
intending to decrease liquidity risk. Because the Libor-OIS
spread contains credit and noncredit risk components, we
control for credit risk to obtain meaningful correlations
between the spread and the supply of funds by the Fed. To
isolate the supply effects, we consider changes in the amount
of funds outstanding, which are the net effect of changes in
the Fed’s supply of funds and repayment of funds by
participating banks. During the first ten months of TAF
operation, the Fed raised the maximum amount offered at
auction four times, introduced longer term auctions, and
increased the frequency of auctions. The swap facility
underwent similar changes, such as increases in size and
adjustments in frequency. These changes worked mainly to
increase the size of the programs; more recently, the Fed has
been reducing their size.
Our maintained assumption is that changes in the TAF and
in the swap amount outstanding are exogenous. Before
October 2008, the Fed and other central banks determined the
maximum offering amount for the TAF and the swap lines well
in advance of the auctions, and banks fully subscribed to each
auction. Thus, changes in the amount outstanding for these
facilities were not influenced by market conditions concurrent
with the supply announcement dates. Although the offer
amounts were known in advance, uncertainty remained about
whether the auctions would be fully subscribed; therefore,
changes in the amount outstanding were not fully anticipated
by banks. We calculate changes in the amount outstanding to
occur on the day of disclosure rather than on the date of funds
disbursement (generally two days later) to maximize the
“news” content of our measure.
Since October 2008, the TAF offer amount was increased
to $150 billion per auction and the auctions became
undersubscribed. At almost the same time, the swap lines were
uncapped and foreign banks were allowed to bid for any
quantity of funds. These changes meant that market conditions
around auction dates likely played a larger role in determining
the actual amount of funds disbursed. For this reason,
endogeneity problems are likely to be greater since October
2008. To mitigate this concern, we include the Treasury-MBS
66
Financial Amplification Mechanisms
general collateral repo spread to help control for changes in
bank demand for TAF and swap loans.
McAndrews, Sarkar, and Wang (2008) decompose the
Libor-OIS spread into its credit risk and non–credit-risk
components for the January 2007-April 2008 period. They
find that the non–credit-risk component was the major part
of the spread in 2007. The credit risk component was high
and volatile in 2008. However, because the CDS market
became highly illiquid at this time, part of the credit risk
component is likely to reflect liquidity risk as well.
Consistent with the importance of liquidity risk, the authors
find that the Fed’s announcements of new supplies of TAF
funds significantly reduced the Libor-OIS spread during
their sample period.
Our analysis differs from that study’s approach in four
primary respects. First, we use changes in the actual supply
of funds through the TAF and swap facilities rather than
announcement dates. The amount outstanding variable, being
continuous, is able to capture variations in supply changes,
unlike the auction date variables used by McAndrews, Sarkar,
and Wang, which are binary. Second, our examination of a
longer time series enables us to analyze recent decreases in
the size of these facilities, potentially allowing us to draw
implications for the Fed’s exit strategies. Third, we look at
the TAF and swap facilities simultaneously, a natural approach
because of the facilities’ high degree of similarity. Both are
intended to provide dollar funding to a broad range of
counterparties, both were introduced at the same time and
relatively early in the crisis, and both correspond closely in
terms of the timing, terms, and magnitude of auctions. Finally,
we employ an expanded set of covariates to control for credit
and liquidity risk.
We examine interactions between binary variables over four
periods and the TAF and swap amounts outstanding to allow
for changes in the importance of liquidity risk over time.31 The
periods are chosen to correspond to the turning points of the
crisis and to encompass TAF and swap auctions that occurred
around these points. Period 1 starts on August 1, 2007, roughly
the beginning of the crisis, and ends on March 9, 2008. Period 2
begins on March 10, 2008, the date of the last TAF auction
before the acquisition of Bear Stearns by JPMorgan Chase, and
ends on September 9, 2008. Period 3 captures the Lehman
bankruptcy and its aftermath, beginning on September 10,
2008, and ending on December 31, 2008. The final period runs
from January 1, 2009, through July 31, 2009, a period when
markets were normalizing.
31
The effect of risk variables on the Libor-OIS spread could also change over
time. Unreported results from regressions allowing the risk variable
coefficients to vary over different crisis periods indicate no qualitative changes
to our estimates for the amounts outstanding of the TAF and swap variables.
Table 1
Variables Used in Regressions
Variable
Description
Three-month Libor-OIS spread on date t
TAF outstanding
Non-negative component of TAF outstanding
Non-positive component of TAF outstanding
Swap outstanding
Non-negative component of swap outstanding
Non-positive component of swap outstanding
Period 1
Period 2
Period 3
Period 4
CDX spread
Three-month Libor quote dispersion on date t
VIX
Overnight MBS–Treasury spread
Three-month Libor on date t+1 minus three-month OIS rate on date t
Outstanding value of TAF funds on award announcement date
Equal to the maximum of TAF outstanding and 0
Minimum of 0 and TAF outstanding
Outstanding value of all swap lines on award announcement date
Maximum of swap outstanding and 0
Minimum of 0 and swap outstanding
Binary variable equal to 1 for dates between August 1, 2007,
and March 9, 2008; 0 otherwise
Binary variable equal to 1 for dates between March 10, 2008,
and September 9, 2008; 0 otherwise
Binary variable equal to 1 for dates between September 10, 2008,
and December 31, 2008; 0 otherwise
Binary variable equal to 1 for dates between January 2, 2009,
and July 31, 2009; 0 otherwise
CDX IG index
Difference between maximum and minimum quote of banks in three-month
Libor panel on date t+1
Options-implied volatility in equities market
Overnight MBS rate minus Treasury general collateral repo rate
Unit
Basis points
Billions of U.S. dollars
Billions of U.S. dollars
Billions of U.S. dollars
Billions of U.S. dollars
Billions of U.S. dollars
Billions of U.S. dollars
—
—
—
—
Basis points
Basis points
Basis points
Basis points
Note: Libor is the London interbank offered rate; TAF is the Term Auction Facility; MBS is mortgage-backed securities.
We estimate the following equation, where Δ is the daily
change in the variable:
(4) Δ ( Libor – OIS t ) = β 1 + β 2 Δ TAF t∗ Period1
+ β 3 Δ TAF t∗ Period2 + β 4 Δ TAF t∗ Period3
+ β 5 Δ TAF t∗ Period4
+ β 6 Δ SWAPt ∗ Period1
+ β 7 Δ SWAPt ∗ Period2
+ β 8 Δ SWAPt ∗ Period3
+ β 9 Δ SWAPt ∗ Period4
+ β 10 Δ CDXt + β 11 Δ LIBOR – DISP t
+ β 12 Δ VIXt + β 13 Δ MBS – TRSY – REPO t
+ εt .
The equation relates changes in the Libor-OIS spread to
changes in the amounts outstanding at the Fed’s TAF (denoted
Δ TAF ) and swap (denoted Δ SWAP ) facilities. We control for
credit risk using the CDX index (Δ CDX ) and the Libor quote
dispersion variable ( Δ LIBOR – DISP ). We control for general
market risk using options-implied volatility in the equity
market ( Δ VIX ). Because VIX has been found to be a
significant determinant of asset prices in several markets, we
use it to account for financial market risk broadly.32 Finally, we
control for banks’ balance-sheet funding risk using the overnight
MBS-Treasury repo spread (Δ MBS – TRSY – REPO ). We use
changes in variables to account for deterministic time-series
effects, such as trends. All variables are summarized in Table 1.
TAF auction results are from the Federal Reserve Board
website; swap line results are from participating central bank
websites.33 VIX data are from Bloomberg.
In a related regression, we decompose the TAF and swap
line amounts outstanding into positive and negative changes.
To be specific, we replace Δ TAF in equation 4 with the
following terms:
.( , Δ TAF ).
.( , Δ TAF ) and Δ TAFN = minN0
Δ TAFP = maxX0
32
VIX has been shown to be a significant determinant of prices of foreign
exchange (Brunnermeier, Nagel, and Pedersen 2008) and sovereign CDS
(Longstaff et al. 2007).
FRBNY Economic Policy Review / August 2010
67
Further, we replace Δ SWAP in equation 4 with the
following terms:
. ( , Δ SWAP ), and
Δ SWAPP = maxX0
Δ SWAPN = minX0
.( , Δ SWAP ) .
Table 2
Changes in Amounts Outstanding at Federal
Reserve Facilities, and the Libor-OIS Spread
August 2007-July 2009
Dependent Variable: Change in Three-Month Libor-OIS Spread
The balance-sheet constraint is predicted to bind on the
down side (when intermediaries are capital constrained) but
not on the up side (when capital is widely available). This
predicted asymmetry implies that increases in the supply of
funds by the Fed should decrease spreads, whereas reductions
in the supply should have little impact on them.
Explanatory Variable
Change in TAF outstanding
Period 1: August 1, 2007–March 9, 2008
Period 2: March 10, 2008–September 9, 2008
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
9. The Effectiveness of the Fed’s
Liquidity Supply: Results
Table 2 presents our results from estimating equation 4. The
results indicate that the supply of funds from both the TAF and
the swap line programs was associated with a reduction in the
Libor-OIS spread during the early phase of the crisis (up to
March 9, 2008). In particular, an increase of $1 billion in the
supply of TAF and swap line funds outstanding is associated
with an average decline in the spread of 0.1 to 0.5 basis point
during this period. This result is consistent with the operation
of the balance-sheet amplification mechanism in the early stage
of the crisis.
We find that in subsequent periods, the supply of TAF and
swap funds is not a significant predictor of the interest rate
spread. The sign of the TAF supply coefficient remains negative
in Periods 2 and 3, but it is not significant.34 In the next section,
we show that this apparent lack of significance may be
attributable to an averaging of the separate effects of increases
and decreases in the supply of funds. The sign of the swap line
coefficient is negative in Periods 1 and 3. Overall, considering
33
http://www.federalreserve.gov/monetarypolicy/taf.htm
http://www.ecb.int/mopo/implement/omo/html/index.en.html
http://www.snb.ch/en/ifor/finmkt/id/
finmkt_usdollars?LIST=lid1&EXPAND=lid1&START=1
http://www.bankofengland.co.uk/markets/other/dollarrepo/index.htm
http://www.boj.or.jp/en/type/release/adhoc/mok0812b.pdf
http://www.rba.gov.au/MarketOperations/Domestic/ExcelFiles/usd_repos.xls
http://www.riksbank.com/templates/ItemList.aspx?id=30117
http://www.norges-bank.no/templates/pagelisting____73626.aspx
http://www.nationalbanken.dk/DNUK/MarketInfo.nsf/side
USD_auction!OpenDocument
http://www.bok.or.kr/broadcast.action?menuNaviId=1562
http://www.banxico.org.mx/sitioingles/portalesEspecializados/tiposCambio
US_dollar_auctions_results.html
34
The difference between the TAF coefficient in the early crisis period (Period 1)
and Period 2 is not statistically significant, but the Period 1 coefficient is
significantly different from the estimates for Periods 3 and 4. The early crisis
swap coefficient is significantly different from all later swap coefficients.
68
Financial Amplification Mechanisms
Change in swap outstanding
Period 1: August 1, 2007–March 9, 2008
Period 2: March 10, 2008–September 9, 2008
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
Credit risk
Change in CDX spread
Change in three-month Libor quote dispersion
Liquidity risk
Change in overnight MBS–Treasury spread
Market risk
Change in VIX
Constant
2
Adjusted R
Observations
Coefficient
-0.130***
(0.037)
-0.167
(0.110)
-0.031
(0.036)
0.009
(0.018)
-0.481***
(0.150)
0.048
(0.065)
-0.047
(0.064)
0.019
(0.016)
0.140***
(0.042)
0.160***
(0.050)
0.025*
(0.014)
0.511***
(0.139)
0.091
(0.286)
0.17
607
Source: Authors’ calculations, based on data from the British Bankers’
Association, Haver Analytics, the Board of Governors of the Federal
Reserve System, foreign central banks, the Federal Reserve Bank of
New York, and Markit.
Notes: Newey-West standard errors (five lags) are in parentheses. The full
sample is daily observations from January 3, 2007, to July 31, 2009. TAF
is the Term Auction Facility. See Table 1 for a description of variables.
*** p<0.01.
** p<0.05.
* p<0.1.
the TAF and swap line results together, we conclude that the
supply of liquidity by the Fed was most effective in the early
stages of the crisis and the effectiveness moderated over time.
The credit risk variables are of the expected sign, with the
Libor quote dispersion and the CDX spread being positively
and significantly associated with the Libor-OIS spread. A
1 basis point change in either credit risk variable is associated
with about a 0.15 basis point change in the Libor-OIS spread.35
The overnight repo spread is also positively associated with the
Libor-OIS spread during the crisis, but the estimate is only
significant at the 10 percent level. As we discussed, the marginal
significance of the repo spread might be explained by the Fed’s
action to reduce the spread through the PDCF and TSLF
programs. Finally, changes in VIX are also significantly and
positively associated with the Libor-OIS spread.36
Results from the regressions provide an indication of when
the Fed might expect its liquidity facilities to help improve
funding conditions. Comparing the coefficient estimates with
the results in Chart 3, we observe that the facilities were most
effective during periods of high liquidity risk and relatively low
credit risk. The facilities did not appear to be effective during
periods of extremely elevated credit risk, such as the months
just after the Lehman failure in 2008, and during periods of low
liquidity risk, such as the first half of 2009. This is consistent with
the stated intentions of the TAF and swap facilities: to provide
short-term funding to banks. As such, these facilities were not
expected to have a direct effect on the credit risk of banks.
10. Asymmetric Market Responses
to the Fed’s Liquidity Supply
We next report estimates using TAF and swap outstanding
variables decomposed into positive and negative changes.
Chart 4 presents the time-series plots of the two main variables
of interest: changes in TAF and swap amounts outstanding.
Note that the TAF has experienced negative changes in
amounts outstanding since Period 3, while the swap lines have
experienced both increases and decreases during each period
since the crisis began. The share of negative changes in the TAF
and swap lines combined, compared with the total number of
changes, is small in Periods 1 and 2, and rises to 40 percent in
Period 3 and 80 percent in Period 4.
35
Similar specifications with indexes of Libor bank CDS spreads instead of the
CDX index yielded highly similar results for the TAF and swap variables of
interest, but results for the Libor-based indexes were insignificant.
36
We also considered the term premium, defined as the spread between the
five- and two-year on-the-run Treasury yields, but this variable was not a
significant predictor of the Libor-OIS spread.
Chart 4
Changes in Term Auction Facility (TAF)
and Swap Line Amounts Outstanding
Billions of dollars
150
Pre-crisis
Period 1
Period 2
Period 3
Period 4
100
Change in
TAF amounts
outstanding
50
0
-50
Change in
swap line amounts
outstanding
-100
2007
2008
2009
Sources: Federal Reserve Bank of New York; foreign central banks.
Note: Vertical lines correspond to the period divisions used in the
estimations.
The results from estimation of the second regression are
presented in Table 3. Symmetric responses of the Libor-OIS spread
are indicated by negative changes to both increases and decreases
in the amount outstanding—that is, reductions (increases) in the
spread in response to a decrease (increase) in the amount
outstanding. By comparison, asymmetric responses are indicated
by different signs of the coefficient depending on whether the
change in amount outstanding is positive or negative.
In the pre–Bear Stearns period (Period 1), expansion of the
TAF and swap lines in the early part of the crisis tended to be
associated with a reduction in the Libor-OIS spread, consistent
with prior results. Further, reductions in the swap line amount
outstanding resulted in an increase in the spread. Therefore, the
effect of the Fed’s funds supply is symmetric during this period.
In contrast, during the post–Lehman periods (Periods 3 and 4),
the effect of liquidity supply by the Fed is asymmetric.
In particular, decreases in the TAF and swap amounts
outstanding are associated with declines in the Libor-OIS
spread, whereas increases in the TAF and swap lines are also
associated with decreases in the spread during this period.
These results are statistically significant for changes in the TAF
amount outstanding. This asymmetry suggests that the lack of
significance in the overall TAF coefficients during Periods 3
and 4 in Table 2 may be attributable to an averaging of the
positive and negative changes (which are of roughly equal
magnitude). Hence, to understand responses of interest rates to
changes in the supply of funds by the Fed during the post–
Lehman period, it is important to account for this asymmetry.
FRBNY Economic Policy Review / August 2010
69
Table 3
Positive and Negative Changes in Amounts
Outstanding at Federal Reserve Facilities
August 2007-July 2009
Dependent Variable: Change in Three-Month Libor-OIS Spread
Explanatory Variable
Coefficient
Positive changes in TAF outstanding
Period 1: August 1, 2007–March 9, 2008
Period 2: March 10, 2008–September 9, 2008
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
Negative changes in TAF outstanding
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
Positive changes in swap outstanding
Period 1: August 1, 2007–March 9, 2008
Period 2: March 10, 2008–September 9, 2008
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
Negative changes in swap outstanding
Period 1: August 1, 2007–March 9, 2008
Period 2: March 10, 2008–September 9, 2008
Period 3: September 10, 2008–December 31, 2008
Period 4: January 2, 2009–July 31, 2009
Constant
-0.093**
(0.045)
-0.033
(0.078)
-0.134***
(0.020)
-0.108**
(0.045)
0.150***
(0.016)
0.034**
(0.015)
-0.957***
(0.050)
0.036
(0.066)
-0.084
(0.083)
0.204
(0.161)
-0.304***
(0.036)
-0.087*
(0.050)
0.063
(0.045)
0.021
(0.015)
0.252
(0.264)
Risk variables included?
Yes
2
0.19
475
Adjusted R
Observations
Source: Authors’ calculations, based on data from the British Bankers’
Association, Haver Analytics, the Board of Governors of the Federal
Reserve System, foreign central banks, the Federal Reserve Bank of
New York, and Markit.
Notes: Newey-West standard errors (five lags) are in parentheses. Negative changes in TAF outstanding did not occur until Period 2. The full
sample is daily observations from January 3, 2007, to July 31, 2009. TAF
is the Term Auction Facility. See Table 1 for a description of variables.
*** p<0.01.
** p<0.05.
* p<0.1.
70
Financial Amplification Mechanisms
The existence of balance-sheet constraints that bind only on
the downside implies a negative relationship between the
Libor-OIS spread and positive changes in the TAF and swap
lines and no relationship for negative changes. We find,
however, that declines in the TAF amount outstanding actually
improved the Libor-OIS spread in Periods 3 and 4. This
association might reflect reduced pressure on funding markets
at this time, leading to declining demand at the Fed facilities
and a reduced spread. Indeed, the two declines in the TAF
amount outstanding during Period 3 occur in December 2008,
when risk factors were already beginning to normalize. In
Chart 3, one can see that by December 2008 liquidity risk had
declined, as had the Libor quote dispersion, although the CDX
index had remained elevated.
The results in Table 3 also shed light on the Fed’s exit
strategy from these programs. First, the decline in outstanding
value that has occurred since the beginning of 2009 likely
reflects a return by participants to market sources for funding
as interbank market rates have fallen. Chart 2 supports this
view by showing that the spread between Libor and the Fed
facilities has been steadily decreasing since early 2009. The view
is further supported by the coefficient estimates on the negative
changes in the TAF and swap amounts outstanding in 2009
(Table 3), indicating that the reductions in the programs were
not adversely affecting market interest rates. This result
represents a potentially positive sign for the market, as it
indicates that reductions in the supply of funds by the Fed have
not been a negative shock.
11. Conclusion
The financial crisis has led to large reductions in asset prices and
in new issuances of primary securities while affecting a wide
variety of markets and institutions. Yet the magnitude of these
effects appears to be disproportionate to the relatively small
losses that occurred in the subprime mortgage markets. To
explain this seeming disparity, our paper surveys the literature
on financial amplification mechanisms, focusing on the balancesheet and adverse-selection channels. It then discusses and
interprets the Federal Reserve’s actions during the crisis in terms
of the literature. We show that the Fed’s early-stage liquidity
programs were mainly designed to dampen the balance-sheet
amplification arising from the positive feedback between
financial constraints and asset prices. The Fed’s later-stage crisis
programs take into account the adverse-selection amplification
that operates through increases in credit risk and the externality
imposed by risky borrowers on safe ones.
We also examine how changes in the Fed’s supply of
liquidity (the amount of funds outstanding at the TAF and
swap facilities) are associated with changes in interest rate
spreads, after controlling for credit risk and short-term funding
conditions. We find that an increase in the supply of funds is
associated with a reduction in the Libor-OIS spread early in the
crisis. During more recent periods, the Fed has been gradually
withdrawing funds from these programs. We find that the
reduced supply of funds has had no significant impact on
interest rate spreads in these periods. These results suggest that
the potential withdrawal of liquidity by the Fed may not have
an adverse effect on market prices.
FRBNY Economic Policy Review / August 2010
71
References
Acharya, V., D. Gromb, and T. Yorulmazer. 2008. “Imperfect Competition
in the Inter-Bank Market for Liquidity as a Rationale for Central
Banking.” Unpublished paper, London Business School.
Acharya, V., and M. Richardson, eds. 2009. Restoring Financial
Stability: How to Repair a Failed System. John Wiley & Sons.
Adrian, T., C. R. Burke, and J. J. McAndrews. 2009. “The Federal
Reserve’s Primary Dealer Credit Facility.” Federal Reserve Bank
of New York Current Issues in Economics and Finance 15,
no. 4 (August).
Adrian, T., and H. Shin. Forthcoming. “The Changing Nature of
Financial Intermediation and the Financial Crisis of 2007-2009.”
Annual Review of Economics.
Bernanke, B. 2009. “The Crisis and the Policy Response.” Stamp Lecture,
London School of Economics, London, England, January 13.
Bernanke, B., and M. Gertler. 1989. “Agency Costs, Net Worth, and
Business Fluctuations.” American Economic Review 79, no. 1
(March): 14-31.
———. 1990. “Financial Fragility and Economic Performance.”
Quarterly Journal of Economics 105, no. 1 (February): 87-114.
Blanchard, O. 2009. “The Crisis: Basic Mechanisms and Appropriate
Policies.” International Monetary Fund Working Paper no. 09/80,
April.
Bolton, P., T. Santos, and J. A. Scheinkman. 2009. “Inside and Outside
Liquidity.” NBER Working Paper no. 14867, April.
Aikman, D., P. Alessandri, B. Eklund, P. Gai, S. Kapadia, E. Martin,
N. Mora, G. Sterne, and M. Willison. 2009. “Funding Liquidity Risk
in a Quantitative Model of Systemic Stability.” Bank of England
Working Paper no. 372, June.
Brunnermeier, M. K. 2009. “Deciphering the Liquidity and Credit
Crunch 2007-2008.” Journal of Economic Perspectives 23, no. 1
(winter): 77-100.
Allen, F., and E. Carletti. 2008. “The Role of Liquidity in Financial
Crises.” In Maintaining Stability in a Changing Financial
System. Proceedings of the 2008 Jackson Hole Economic
Symposium, Federal Reserve Bank of Kansas City, August 21-23.
Brunnermeier, M. K., S. Nagel, and L. H. Pedersen. 2008. “Carry
Trades and Currency Crashes.” In D. Acemoglu, K. Rogoff, and
M. Woodford, eds., NBER Macroeconomics Annual 2008,
313-47. Chicago: University of Chicago Press.
Allen, F., E. Carletti, and D. Gale. 2009. “Interbank Market Liquidity
and Central Bank Intervention.” Journal of Monetary
Economics 56, no. 5 (July): 639-52.
Brunnermeier, M. K., and L. H. Pedersen. 2009. “Market Liquidity and
Funding Liquidity.” Review of Financial Studies 22, no. 6
(June): 2201-38.
Allen, F., and D. Gale. 2004. “Financial Intermediaries and Markets.”
Econometrica 72, no. 4 (July): 1023-61.
Cecchetti, S. G., and P. Disyatat. 2010. “Central Bank Tools and
Liquidity Shortages.” Federal Reserve Bank of New York
Economic Policy Review 16, no. 1 (August): 29-42.
Armantier, O., E. Ghysels, A. Sarkar, and J. Shrader. 2010. “Stigma
in Financial Markets: Evidence from Liquidity Auctions and
Discount Window Borrowing during the Crisis.” Unpublished
paper, Federal Reserve Bank of New York.
Armantier, O., S. Krieger, and J. McAndrews. 2008. “The Federal Reserve’s
Term Auction Facility.” Federal Reserve Bank of New York Current
Issues in Economics and Finance 14, no. 5 (July).
Ashcraft, A., N. Garleanu, and L. H. Pedersen. 2009. “Haircuts or
Interest Rate Cuts: New Evidence on Monetary Policy.”
Unpublished paper, Federal Reserve Bank of New York, University
of California at Berkeley, and New York University.
72
Financial Amplification Mechanisms
Cetorelli, N., and L. Goldberg. 2009. “Following the Money in Global
Banks: Internal Transfers during the Crisis.” Unpublished paper,
Federal Reserve Bank of New York.
Coffey, N., W. B. Hrung, and A. Sarkar. 2009. “Capital Constraints,
Counterparty Risk, and Deviations from Covered Interest Rate
Parity.” Federal Reserve Bank of New York Staff Reports, no. 393,
October.
Del Negro, M., G. Eggertsson, A. Ferrero, and N. Kiyotaki. 2009. “The Great
Escape? A Quantitative Evaluation of the Fed’s Non-Standard
Policies.” Unpublished paper, Federal Reserve Bank of New York.
References (Continued)
Diamond, D. W. 1997. “Liquidity, Banks, and Markets.” Journal of
Political Economy 105, no. 5 (October): 928-56.
Diamond, D. W., and P. H. Dybvig. 1983. “Bank Runs, Deposit
Insurance, and Liquidity.” Journal of Political Economy 91,
no. 3 (June): 401-19.
Gorton, G. B. 2008.“The Panic of 2007.” In Maintaining Stability
in a Changing Financial System, 131-262. Proceedings of
the Federal Reserve Bank of Kansas City’s 2008 Jackson Hole
Symposium. Available at http://www.kc.frb.org/publicat/sympos/
2008/Gorton.03.12.09.pdf.
Diamond, D. W., and R. G. Rajan. 2005. “Liquidity Shortages and
Banking Crises.” Journal of Finance 60, no. 2 (April): 615-47.
Gorton, G. B., and L. Huang. 2006. “Banking Panics and the
Endogeneity of Central Banking.” Journal of Monetary
Economics 53, no. 7 (October): 1613-29.
Fecht, F. 2004. “In the Stability of Different Financial Systems.”
Journal of the European Economic Association 2, no. 6
(December): 969-1014.
Gorton, G. B., and A. Metrick. 2009. “Securitized Banking and the Run
on the Repo.” Yale University, International Center for Finance
Working Paper no. 09-14, November.
Flannery, M. J. 1996. “Financial Crises, Payment System Problems,
and Discount Window Lending.” Journal of Money, Credit,
and Banking 28, no. 4, part 2 (November): 804-24.
Gromb, D., and D. Vayanos. 2002. “Equilibrium and Welfare in Markets
Fleming, M. J., and K. D. Garbade. 2004. “Repurchase Agreements
He, Z., and A. Krishnamurthy. 2008. “A Model of Capital and Crises.”
Unpublished paper, Northwestern University.
with Negative Interest Rates.” Federal Reserve Bank of New York
Current Issues in Economics and Finance 10, no. 5 (April).
———. 2005. “Explaining Settlement Fails.” Federal Reserve Bank of
New York Current Issues in Economics and Finance 11, no. 9
(September).
Fleming, M. J., W. B. Hrung, and F. H. Keane. 2009. “The Term
Securities Lending Facility: Origin, Design, and Effects.” Federal
Reserve Bank of New York Current Issues in Economics and
Finance 15, no. 2 (February).
Furfine, C. H. 2003. “Standing Facilities and Interbank Borrowing:
Evidence from the Federal Reserve’s New Discount Window.”
International Finance 6, no. 3 (winter): 329-47.
Geanakoplos, J., and H. M. Polemarchakis. 1986. “Existence,
Regularity, and Constrained Suboptimality of Competitive
Allocations when the Asset Market Is Incomplete.” In W. P. Heller,
R. M. Starr, and D. A. Starrett, eds., Uncertainty, Information,
and Communication, vol. 3, 65-95. Cambridge University Press.
Goodfriend, M., and R. C. King. 1988. “Financial Deregulation,
Monetary Policy, and Central Banking.” In W. S. Haraf and
R. M. Kushmeider, eds., Restructuring Banking and
Financial Services in America. Washington, D.C.: American
Enterprise Institute for Public Policy Research.
with Financial Constrained Arbitrageurs.” Journal of Financial
Economics 66, nos. 2-3 (November-December): 361-407.
Heider, F., M. Hoerova, and C. Holthausen. 2009. “Liquidity Hoarding
and Interbank Market Spreads: The Role of Counterparty Risk.”
European Central Bank Working Paper no. 1126, December.
Holmstrom, B., and J. Tirole. 1998. “Private and Public Supply of
Liquidity.” Journal of Political Economy 106, no. 1
(February): 1-40.
Hordahl, P., and M. R. King. 2008. “Developments in Repo Markets
during the Financial Turmoil.” BIS Quarterly Review,
December: 37-53.
Kiyotaki, N., and J. Moore. 1997a. “Credit Cycles.” Journal of
Political Economy 105, no. 2 (April): 211-48.
———. 1997b. “Credit Chains.” Unpublished paper, London School
of Economics.
———. 2008. “Liquidity, Monetary Policy, and Business Cycles.”
Unpublished paper, Princeton University.
Krishnamurthy, A. Forthcoming. “Amplification Mechanisms
in Liquidity Crises.” American Economic Journal:
Macroeconomics.
FRBNY Economic Policy Review / August 2010
73
References (Continued)
Longstaff, F. A., J. Pan, L. H. Pedersen, and K. J. Singleton. 2007.
“How Sovereign Is Sovereign Credit Risk?” NBER Working
Paper no. 13658, December.
McAndrews, J. 2009. “Segmentation in the U.S. Dollar Money Markets
during the Financial Crisis.” Unpublished paper, Federal Reserve
Bank of New York.
McAndrews, J., A. Sarkar, and Z. Wang. 2008. “The Effect of the
Term Auction Facility on the London Inter-Bank Offered Rate.”
Federal Reserve Bank of New York Staff Reports, no. 335, July.
Pritsker, M. 2010. “Informational Easing: Improving Credit
Conditions through the Release of Information.” Federal Reserve
Bank of New York Economic Policy Review 16, no. 1 (August):
77-87.
Rochet, J.-C., and X. Vives. 2004. “Coordination Failures and the
Lender of Last Resort: Was Bagehot Right after All?” Journal of
the European Economic Association 2, no. 6 (December):
1116-47.
Shleifer, A., and R. W. Vishny. 1997. “The Limits of Arbitrage.”
Journal of Finance 52, no. 1 (March): 35-55.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
74
Financial Amplification Mechanisms
Matthew Pritsker
Informational Easing:
Improving Credit
Conditions through
the Release of Information
1. Introduction
T
o ensure repayment of borrowed funds, lenders require that
borrowers undergo costly credit evaluations. In the financial
sector, credit often flows along chains of borrowers and lenders
who are already familiar with each other’s creditworthiness—a
process that minimizes the cost of credit evaluations. However, if
the creditworthiness of key participants along a chain is called into
question, the chain can break and cut off the flow of credit to final
borrowers. If enough chains in the economy break, a financial
crisis can ensue, investment by final borrowers can dry up, and
output can decline.
The flow of credit can stop because a lender believes a borrower
has a high default probability or because a lender is uncertain
about whether a borrower has a high default probability. The latter
may often be the more likely scenario. For example, in a classic
bank run, it is unlikely that depositors know the probability that
their bank will become insolvent, but it is likely that they worry
about the possibility that their bank has high default probability
and withdraw their deposits as a precaution.1
More generally, a decision maker faces risk if the outcomes in
his decision problem are random; he faces uncertainty if the
outcomes are random and he does not know the probabilities of
the outcomes.2 For example, when lenders are uncertain, they
Matthew Pritsker is a senior economist in the Division of Research and
Statistics of the Board of Governors of the Federal Reserve System.
matthew.pritsker@frb.gov
cannot assign a single figure to a borrower’s default probability, so
they instead assign a range. During economic expansions, this
range may be small, such as 1/4 to 1/2 percent; however, during
economic downturns, the range may be 2 to 5 percent. If a lender
is uncertainty-averse in the sense of Gilboa and Schmeidler (1989),
it will charge spreads based on the high end of its range. This
decision will be unimportant during expansions, when the range
is narrow, but during downturns the required spread may be so
high that a borrower cannot afford a loan—and the flow of credit
from that borrower to any borrowers farther along the lending
chain will be cut off.3
This paper addresses how central banks can resuscitate
lending chains by providing information that reduces
1
Easley and O’Hara (2009) argue that deposit insurance was instituted to
eliminate bank runs motivated by uncertainty among small depositors because
it allays the worries of small depositors that their bank will become insolvent.
In a similar vein, Caballero and Krishnamurthy (2008) model an excessive
flight to quality and flight to liquid assets that can occur when there is
uncertainty over the timing of liquidity shocks—and argue for government
intervention aimed at reversing the flight.
2
For examples of different methods of modeling decision making under
uncertainty in nondynamic settings, see the discussion and approach in Rigotti
and Shannon (2005) as well as the approaches in Klibanoff, Marinacci, and
Mukerji (2005) and Easley and O’Hara (2009). For an overview of uncertainty
in dynamic settings, see Hansen and Sargent (2007) and their references.
3
In this paper, the terms “lending chain” and “credit chain” are used
interchangeably.
The views expressed are those of the author and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Board of Governors
of the Federal Reserve System.
FRBNY Economic Policy Review / August 2010
77
uncertainty about participants along the chains. This action has
been taken before: the Bank Holiday of 1933, declared by
President Franklin Delano Roosevelt, resolved uncertainty
about the health of individual banks by using bank inspections
to publicly identify which banks were sound. This event
restored the flow of funds to the banking sector and facilitated
bank lending. During the 2007-09 financial crisis, the Federal
Reserve used “stress tests” to measure and report on the health
of large banks in the U.S. banking system and to identify those
banks that required shoring up through capital injections.
In addition to providing information to the financial sector,
central banks have other tools at their disposal to revive
lending. When credit chains involve financial intermediaries
such as banks, central banks can lower their target rates to
reduce intermediaries’ costs of borrowing, accept a wider range
of collateral, guarantee interbank loans, or shore up banks’
health through capital injections. Alternatively, they can bypass
intermediaries altogether and lend directly to final borrowers
in credit chains.
Each of these tools has merit in some situations—but none
is perfect. Monetary easing may lower target rates to 0, but if
credit spreads remain too high, lending along credit chains may
still cease. Broadening the range of acceptable collateral, loan
guarantees, and government-sponsored capital injections
increases lending, but it can also increase the central bank’s
exposure to credit and market risk. Direct lending outside the
financial sector may reduce lending efficiency, because such
intermediation is not a central bank’s usual function.
Under conditions of less uncertainty, many of these efforts
would be less costly and more effective. This statement is
intuitive, as it is easier to convince potential lenders that a
solvency problem has been fixed if they have better
information about the scope of the problem. It follows that
during a crisis, steps to reduce uncertainty through
information provision should be taken as soon as possible.
In theory, information designed to reduce uncertainty could
be provided privately by borrowers. However, because
borrowers may have an incentive to exaggerate their financial
strength during economic downturns, private information
provision may lack credibility. Moreover, uncertainty
reduction by borrowers upstream in a credit chain may
generate external benefits to borrowers downstream that are
not internalized by private information providers. As a result,
the private sector may provide less than the socially optimal
level of uncertainty reduction. For both these reasons,
situations may arise in which government information
provision to reduce uncertainty may be warranted.
The remainder of this paper is divided into two sections. In
Section 2, we provide a model of credit chain lending that
illustrates how uncertainty can cause credit chains to break and
how government policies that reduce uncertainty can restore
78
Informational Easing
the flow of credit. Section 3 considers potential future uses of
uncertainty reduction policies.
2. The Model
Our stylized model of a credit chain has four participants: A, B,
C, and D, and three dates: 0, 1, and 2. Participant A is a shortterm depositor who has excess funds at date 1 that he wants to
lend until date 2. Participants B and C are banks that make longterm loans at date 0 and short-term loans at date 1. Both loan
types mature at date 2. Participant D is a short-term borrower
who unexpectedly needs a loan at date 1 that matures at date 2.
We assume that some participants are familiar with each
other’s credit risk based on a previous bilateral lending
relationship, while others are not. In particular, A has previously
loaned funds to B, B to C, and C to D. These relationships suggest
a natural basis for a credit chain to form at date 1. D could
borrow from A, B, or C. Since A and B are unfamiliar with D, a
costly credit evaluation would be needed before either would
extend a loan to D. Instead, C is the logical lender to D; but if C
does not have the funds, then C will need to turn to A or B for
funding. Because of previous relationships, B is the logical lender
to C, and if B needs funds then A is the logical source. Thus, a
short-term loan from saver A is intermediated to borrower D
along a credit chain in which bank B makes a loan to bank C
through the interbank market (Exhibit 1).
Because many loans are intermediated through the
interbank market, the functioning of the market is important
for credit extension. C can lend to D only if the maximum rate
that D can afford to pay C for a loan, denoted R D , is less than
C’s cost of funds. When C borrows from B, its cost of funds is
equal to the risk-free rate R f plus a spread S C that reflects its
credit risk. Therefore, D will be able to borrow from C only if:
Exhibit 1
Short-Term Lending Chain
Interbank lending
Bank B
Bank
deposit
Saver A
Bank C
Loan
Borrower D
Note: Borrower D needs a short-term loan, and saver A has excess short-term
funds. Because A and B, B and C, and C and D have had previous borrowing
relationships, a lending chain from A to B to C to D is the least expensive way
to fund D’s loan since there is no need for costly credit evaluations.
(3)
Exhibit 2
Broken Lending Chain
Bank B
Breakdown of interbank lending
Trapped
deposits
Bank C
Funding
Shortfall
Borrower D
Saver A
Note: If there is a breakdown of lending from B to C in the interbank market,
the funds are trapped with bank B, and borrower D is short of funds.
(1)
Rf + SC < RD .
Under normal economic conditions, the spreads that banks
charge each other for loans are small and would not typically be
an impediment to D’s borrowing. However, during the
financial crisis of 2007-09, interbank spreads increased
markedly, and lending through the interbank market declined.
A consequence of high interbank spreads is that funds can
become trapped at the wrong place, such as with bank B instead
of borrower D (Exhibit 2). Whether interbank spreads increase
at date 1 depends on B’s assessment of C’s default risk as of
date 1. This in turn depends on C’s long-run asset portfolio and
capital structure, both chosen at date 0.
At date 0, banks B and C both choose their long-run asset
portfolios and capital structures. Since the main concern is B’s
willingness to lend to C, we focus only on C’s portfolio choices
hereafter. For simplicity, C’s long-run asset portfolio consists
only of loans to wheat farmers (w) and oat farmers (o). The
loans generate gross returns R w and R o at date 2 per dollar
invested at date 0. The return on the loans is assumed to be
multivariate normal.4 Bank C’s portfolio weights are ω w and
ω o and its portfolio generates return R p :5
(2)
Rp = ωw Rw + ωo Ro .
To finance its long-run portfolio, C is endowed with equity
capital E and insured certificates of deposit with face value F
that mature at date 2 and pay gross interest R 0C,2 at maturity.
At date 1, information I 1 about the return on the long-term
loans arrives. Conditional on this information, the returns on
the loan portfolio are distributed normally with mean μ 1 and
variance σ 12 :
4
Pritsker (2009) illustrates conditions under which the average return on loans
to a diversified group of borrowers can be approximately normally distributed
even if the returns to individual borrowers are not.
5
The portfolio weights are each assumed to be greater than or equal to 0 and
to sum to 1.
R p I 1 ∼ N ( μ 1, σ 12 ) ,
where the parameters μ 1 and σ 12 depend on the portfolio
weights as well as the means, standard deviations, and
correlation of the assets’ returns, given the information
available at date 1 (see Appendix A).
Additionally, recall that at date 1 bank C has the opportunity
to extend a short-term loan to D that matures at date 2, which
it needs to fund in the interbank market by borrowing from B.6
The spread that bank C pays on its interbank loans depends
on bank B’s perception of the probability that C will default on
its debt at date 2. We assume that bank C’s long-term loan
portfolio is so much larger than its short-run lending
opportunities that the performance of its short-run loans and
their funding does not affect whether C will default. Under this
assumption, C will default only if the value of its long-term
loan portfolio at date 2 is less than what is owed on its deposits:
(4)
( F + E )R p < FR 0C,2 .
From this expression, we show that bank C’s probability of
default—and therefore the loan spread that B charges C—
depends on C’s portfolio weights, financial leverage L
( L = F ⁄ E ) , and the parameters of the return distribution of
C’s loan portfolio.
We assume that the risk inherent in both types of loans is
known by bank B, as is C’s leverage, since leverage information
is usually readily available. However, B does not know C’s
portfolio weights. There are two cases to consider: The first is
that B has beliefs about C’s portfolio that are sufficiently well
formed as to be described by a unique prior probability
distribution, which means that for each portfolio that C could
hold, B assigns a single probability number to the likelihood
that C could hold that portfolio. In this first case, B’s
assessment of C’s probability of default is just a single number
given by the sum of C’s default probability for each portfolio it
could hold multiplied by B ’s belief about the probability that
C will hold that portfolio.7 Because B’s assessment of C’s
default probability is a single figure, B is not uncertain about
C’s default probability.
The second case is that B does not know enough about C’s
portfolio weights, and B’s beliefs cannot be described by a
unique prior probability distribution. Instead, B may be
uncertain about the portfolio weights and thus may hold
6
Bank C may fund some of its short-term loans in the interbank market
because it did not fully anticipate the short-term loan demand or because the
interbank market is usually an inexpensive funding source.
7
For example, suppose B believes C holds only one of two portfolios, 1 or 2, and
the probability that C holds 1 or 2 is 0.3 and 0.7, respectively. Also suppose the
probability that portfolio 1 defaults is .01 and the probability that portfolio 2
defaults is .02. Then, B believes the probability that C defaults is given by
PD = 0.3 × 0.01 + 0.7 × 0.02 .
FRBNY Economic Policy Review / August 2010
79
multiple priors over the weights. Thus, B assigns a range of
probabilities to some or all of the portfolio holdings that C may
have. For example, if bank B is asked about the probability that
C holds a portfolio with a weight of 0.4 in loans to oat farmers
and 0.6 in loans to wheat farmers, B might respond that it is
unsure, but it believes the probability ranges from 10 to
20 percent.8
There are many reasons why B might be uncertain about C’s
portfolio composition. For example, C may have a very complex
portfolio, and thus researching C’s holdings in extensive detail
may be very expensive. This may be true for C’s portfolio because
it consists of loans to farmers, and it may be very difficult for B
to verify which loans are to oat or wheat farmers because this
information may not be readily available, and it may be costly to
obtain.9 Information costs are important because many of the
most active banks in the U.S. interbank market have more than
$1 trillion of assets on their balance sheets, and ascertaining the
loan composition, or even learning enough to form a unique
prior probability distribution about the balance-sheet
composition, can be very expensive.
A simple and parsimonious way to model multiple priors is
to assume that bank B knows C makes only long-term loans to
oat and wheat farmers, and that B knows C has risk
concentration limits that prevent it from making more than
60 percent of its loans to one type of farmer—and that is all B
knows about C’s portfolio. Given its information, bank B
knows that C could have a set of possible portfolios, and that
the weight on wheat is a number t between 0.4 and 0.6 and that
the weight on oats is 1 – t . Given bank B’s information, it does
not know the probability that C will hold any particular
portfolio, but it does know the probability that C will default
on each portfolio that it could hold. From this information,
bank B can compute a range of possible default probabilities
for bank C. The range can be written as
PD ∈ [ PD, PD ] ,
meaning that based on bank B’s information about bank C,
bank B believes C’s default probability lies within a range
between a lower bound PD and an upper bound PD .
The fact that B assigns a range of possible default
probabilities to C is precisely the type of situation described in
the introduction to this paper. The above logic, formally
derived in Appendix A, shows that the result of B’s uncertainty
about C’s portfolio weights is that B assigns a range of possible
values to C’s probability of default. The spread that B charges C
8
Knowledge of bank B’s portfolio weight in one of the risky assets is sufficient
to describe its portfolio because its weight in the other risky asset is 1 minus the
weight of the first asset.
9
Gorton (2008, 2009) argues that uncertainty about the types of assets
collateralizing asset-backed securities was an important factor behind the
2007-09 credit crisis.
80
Informational Easing
will depend on the range of uncertainty that B has about C and
on B’s preferences. In particular, if bank B sets spreads in an
uncertainty-averse fashion, as in Gilboa and Schmeidler
(1989), then B will set C’s spread as if it believes C’s default
probability is equal to PD , the upper end of its range. Other
decision rules for setting spreads in the face of uncertainty are
plausible. It seems reasonable to believe that for many rules, all
else equal, B would charge a higher spread when the upper end
of the range of possible default probabilities increases.
For illustrative purposes, we assume that in the face of
uncertainty, there are many banks like B that set spreads in an
uncertainty-averse fashion. As a consequence, banks like C will
pay a premium for uncertainty. More specifically, let PD∗
denote C’s true default probability, and for simplicity assume
that bank B is risk-neutral and uncertainty-averse. In this
circumstance, if at date 1 bank B can invest at the risk-free rate
between dates 1 and 2, or bank B can lend to bank C at
interbank rate R 1C, 2 , then for B to be indifferent between the
two, R f = R 1C, 2 ( 1 – PD ) , which implies S C = R 1C,2 – R f is
given by:
PD
S C = R f ----------------- .
1 – PD
Suppose C’s true PD at time 1 based on all information is
PD ∗ . Then if PD ∗ was known by B, C’s spread based on risk
alone but not uncertainty would be
PD ∗
S C∗ = R f -------------------- .
1 – PD ∗
Because of uncertainty and uncertainty aversion, bank C’s
spread will consist of the risk premium S C∗ plus an additional
uncertainty premium given by:
PD ∗ - .
PD - – ------------------S C – S C∗ = R f ---------------1
–
PD ∗
1 – PD
If B sets its spread based on its worst-case-scenario beliefs
about C’s default probability, then the uncertainty premium
will always be positive. The size of the uncertainty premium
paid by bank C depends on C’s capital structure as well as the
conditional expected return and volatility of its loan portfolio.
To analyze the uncertainty premium, we compute the
premium when C’s loan portfolio is split evenly between oats
and wheat. Our analysis shows that the uncertainty premium
can be very low when leverage is low, but it can also be low
when leverage is high, provided that economic conditions are
favorable enough. In particular, all else equal, for reasonable
parameter values, uncertainty premia are lower when the
volatility of the returns on both types of loans is low, when the
expected returns on both types of loans is high, or in both
circumstances (Charts 1 and 2).10 This explains how banks can
10
The simulations are for illustrative purposes. Details are available from the
author upon request.
Chart 1
Chart 2
Uncertainty Premium as a Function of Leverage
and Loan Volatility
Uncertainty Premium as a Function of Leverage
and Average Loan Return
points)
Uncertainty premium (basis 40 45
35
5 10 15 20 25 30
0.6
Vol
atili
t
0.9 y relat
0 0 Leve
.65
ra
0.7 ge loa
0 0
.75 n retur
n
0.8
0 0
.8
ive
t
1.3 o bas
elin
1,7 e
5 0
.90
ts)
Uncertainty premium (basis poin
80 100 120 140
60
40
20
1
1
6
6
21
16 ge
a
11 ever
L
21
16 ge
a
11 ever
L
28
28
0.6
31
31
0.4
0.8
Volatility relative to baseline
1.0
1.2
1.4
1.6
1.8
2.0
0.055 0.060
8
2
0.065
Average loan return
0.070 0.075 0.080
0.085
0.090 0.095
4
6
12
10
12
Leverage
14
Leverage
8
16
20
16
18
24
20
22
28
24
26
28
30
32
Source: Author’s calculations.
32
Source: Author’s calculations.
Note: For the stylized risky loan portfolio held by bank C, the chart presents
surface and contour plots of the uncertainty premium that bank C pays for its
short-term unsecured interbank borrowing as a function of C’s leverage and
as a function of the average return on its loans when the average return on
each loan in its portfolio is increased or decreased by the same amount.
Note: For the stylized risky loan portfolio held by bank C, the chart presents
surface and contour plots of the uncertainty premium that bank C pays for its
short-term unsecured interbank borrowing as a function of C’s leverage and
as a function of the volatility (standard deviation) of C’s assets relative to
their baseline volatility.
FRBNY Economic Policy Review / August 2010
81
Chart 3
Uncertainty Premium as a Function
of Sector Performance
Uncertainty premium (basis points)
500
400
300
200
100
0
0.07
0.06
Exp
ect 0.05
0.04
ed
wh
0.03
eat
retu
rn 0.02 0.02
0.1
0.08
0.06
eturn
0.04
heat r
ard w
d
n
a
t
S
Source: Author’s calculations.
Note: For the stylized loan portfolio held by bank C, the chart presents C’s
uncertainty premium as a function of the performance of loans to wheat
farmers, one of two types of long-term loans extended by C. The chart shows
that the uncertainty premium grows when loans to wheat farmers become
more risky, and when the expected return on loans to wheat farmers decreases.
often be uncertain about each other’s portfolio composition,
and yet because of their choice of capital structure they can
usually lend and borrow from each other while charging low
spreads. The analysis also shows that banks may be able to take
on very significant leverage during very prosperous times, and
still pay only a small uncertainty premium. In fact, this roughly
describes the situation prior to the global financial crisis of
2007-09, because before that time volatility was considered
very low by historical standards, the spread paid by banks was
low, and yet bank leverage was fairly high (Chart 1, bottom
panel).
During the crisis, the bursting of the housing bubble
heralded the arrival of bad news about the housing sector.
Interbank spreads increased appreciably because of uncertainty
over which banks were exposed to housing—and especially
uncertainty over which banks were exposed to subprime loans.
To understand the same effects for bank C, suppose the bad
news is a wheat blight that increases the likelihood that wheat
farmers default on their loans, and thereby increases the
volatility and decreases the expected returns on loans to wheat
farmers. For given leverage, these changes can have a dramatic
effect on the uncertainty premium paid by bank C. As
illustrated in Chart 3, the bank’s uncertainty premium ranges
from near 0 when volatility is low and expected returns are high
82
Informational Easing
to several hundred basis points when expected returns are low
and volatility is high. The result of the elevated premium is high
interbank spreads that cause borrowers such as D to lose access
to their funding.
A government-sponsored stress test would reveal
information on bank C’s solvency, through a publicly released
assessment of C’s financial health, the release of summary
information on C’s risk exposures, or a combination of the two.
There is a strong case for doing both. For example, recall that
government action may be needed to reduce uncertainty when
the private costs of providing information to reduce uncertainty
are too high. There are two sources of costs: The first is the cost
of compiling and disclosing the information on risk exposures at
the finer level of detail that is required during economic
downturns. This is a nontrivial cost for very large banks. The
second is the cost of processing the information on risk
exposures to make inferences about the bank’s solvency risk. If
the second cost is high enough, then some potential lenders to C
will not be able to process the information on exposures, and
thus would be unwilling to lend to C. For this reason, the
government may have to intervene to provide processed
information on the bank’s health, which it did as part of the
recent Supervisory Capital Assessment Program stress testing in
the United States. In that case, the information provided was the
amount of capital injection required by banks to ensure capital
adequacy during a particular stress scenario that was common
across banks. The case for releasing better information on
exposures is that the information provides more detail on bank
portfolios that further reduces the uncertainty premia charged
by lenders that can process the exposure information.
Under ideal circumstances, C’s true condition would be
revealed by the stress tests, all uncertainty about its risk
exposures would be eliminated, and its uncertainty spread
would decrease to 0. More realistically, stress tests will reduce,
but not eliminate, uncertainty spreads because although they
may eliminate uncertainty over risk exposures, other sources of
uncertainty may remain (such as uncertainty over the correct
form of pricing models for some assets).
If the information revealed by the stress test about C is
favorable enough, then C will be able to borrow from B to lend
to D and the chain of credit will be restored. If instead it is
learned that C’s balance sheet is weak, or its loans are not
performing, then additional steps, such as bank equity
injections or temporary government-sponsored guarantees on
interbank lending, may be warranted.
Equity injections and government-sponsored loan
guarantees can both be implemented without a stress test.
The value-added benefit of the stress test is its ability to
make these other steps more cost-effective if lenders are
uncertainty-averse.
Chart 4
Size of Equity Injection Required to Restore Lending
Required equity injection (percent)
60
50
40
30
20
10
0
−10
0.4
0.45
0.5
0.55
Omega (wheat)
0.6
0.65
Source: Author’s calculations.
Note: For the set of different long-term loan portfolios that bank C could
possibly hold, indexed by omega—the fraction of long-term loans extended
to wheat farmers—the chart presents the percentage increase in C’s equity
(the required equity injection) that would be needed to restore C’s ability to
acquire a short-term loan from bank B to finance a loan to borrower D. If
bank B is uncertainty-averse, and does not know C’s portfolio, it will require a
conservative equity injection of 50 percent before lending. If B becomes
familiar with C’s portfolio, the required equity injection will be smaller, and
could be negative.
Consider first an equity injection into bank C. If bank C is
to restart lending to borrowers such as D, a sufficient amount
of equity must be injected to bring bank C’s spread down to the
level
because C’s holdings can be no worse than the worst case, the
amount of equity it will need to inject is smaller. For example,
if C’s true portfolio is split evenly in each type of loan, the size
of the required equity injection would be only about
20 percent, and in some cases no equity injection would be
required.11
For similar reasons, stress tests reduce the costs and increase
the effectiveness of government programs that guarantee
interbank loans. To illustrate, we note that interbank loan
guarantees are very expensive because they transfer credit risk
from the banking system to the government. Therefore, in the
United States the guarantees offered by the Federal Deposit
Insurance Corporation were limited as to the amount of new
interbank lending that was guaranteed, and banks that
participated in the program were charged a fee based on the
amount borrowed. While the fees and limitations on the
amount of new loans that are covered reduce the government’s
exposure as a guarantor of interbank loans, they also limit
banks’ ability to borrow under these programs.
If a stress test is conducted before the loan guarantee
program is put in place, then the test may help the market
distinguish low- from high-risk banks. The banks that are
identified as low risk may then be able to borrow more at better
rates than the loan guarantee program could provide; thus,
they could potentially increase lending while saving money.
Finally, stress tests and other programs to restart lending
may work better in combination than alone. For example, in
equation 1, lowering R f to 0 may be insufficient to restart
lending, and eliminating the uncertainty spread without
lowering R f may also be insufficient—but both actions
together may be sufficient.
SC = R D – Rf .
If the equity injection occurs before the stress test, then B
remains uncertain about C’s portfolio, and consequently a
large amount of equity will be required to bring C’s loan spread
down. This scenario is depicted in Chart 4, with details
provided in Appendix B. In the chart, C needs to inject enough
equity to bring its perceived probability of default down to
2 percent. If B is uncertainty-averse, it will charge spreads based
on the most pessimistic beliefs about C’s portfolio, which
correspond to a portfolio invested 60 percent in wheat,
attributable to the wheat blight. In this case, C will need to
increase the equity in the bank by about 50 percent to drive
down B’s lending rate sufficiently so that C can lend to D.
If the stress test was instead conducted before the equity
injection, then B would discover C’s portfolio holdings,
eliminating the uncertainty. If B is uncertainty-averse, then
3. Conclusion
When credit is provided along chains of borrowers and lenders,
uncertainty over borrowers’ economic conditions can
sometimes cause the flow of credit to break down. However,
when a breakdown occurs, a central bank can take action to
restart the flow of credit. One such action is to reduce
uncertainty through government provision of information on
financial intermediaries, such as banks, that are key links in
lending chains. Information provision works by reducing those
components of borrowers’ credit spreads attributable to
uncertainty over their economic conditions. Because
information provision can reduce the interest rates paid by
borrowers, it can be viewed as a substitute for easing interest
11
For details, see Appendix B.
FRBNY Economic Policy Review / August 2010
83
rates by other means, such as lowering central bank target rates,
and may prove especially useful when central bank target rates
are at their lower bounds.
Although government-sponsored information provision
may improve the flow of credit ex post, its use has been—and
probably should be—relatively infrequent for two reasons.
First, gathering information is costly, and the benefits of
providing it, in terms of lower spreads, will probably not exceed
the costs in many circumstances. Second, government
84
Informational Easing
provision of information is a two-edged sword: It may be
needed to reduce uncertainty ex post because private incentives
to do the same are inadequate ex ante. However, government
information provision ex post may further worsen private
incentives to choose capital structures and transparent
portfolio holdings that reduce uncertainty spreads. Thus, in the
future, perhaps central banks should be concerned with
uncertainty reduction ex post and with efforts to improve
private incentives to reduce uncertainty ex ante.
Appendix A: Details of Model Derivation
We show how bank B calculates a range of possible default
probabilities for bank C when bank B is uncertain about C’s
portfolio holdings.
As we discuss in the text, the returns on two types of loans,
to wheat farmers (w) and to oat farmers (o), conditional on the
information known at date 0, are multivariate normal. At
date 1, news arrives. Conditional on I 1 , the information that is
known at date 1, the return on bank C’s assets is multivariate
normal with means μ w and μ o , standard deviations σ w and
σ o , and correlation parameter ρ w,o . Therefore, the conditional
distribution of the return on the long-term loan portfolio is
given in equation 2, with parameters μ 1 and variance σ 12 as
follows:
(A1)
μ1 = ωw μw + ωo μo ,
(A2)
σ 12 = ω w2 σ w2 + ω o2 σ o2 + 2 ω w ω o σ o σ w ρ w,o .
Bank C will default at date 2 under the condition in
equation 4. The bank’s probability of default conditional on
the information known at date 1 is given by:
(A3)
L
------------- R 0C,2 – μ 1
1
+
L
PD ( ω w, ω o, L, 1 ) = Φ ------------------------------------ ,
σ 12
where ω w and ω o are bank C’s portfolio weights, μ 1 and σ 12
are the mean and variance of the portfolio’s return distribution
given the portfolio weights (equations A1 and A2), and Φ(.) is
the cumulative distribution function of the standard normal
distribution.
To model uncertainty about C’s portfolio weights, we
assume that B knows that C could have a set of possible
portfolios, and that the weight on wheat is some number t
between 0.4 and 0.6 and the weight on oats is 1 – t . More
formally, C’s possible portfolios can be written as
ω w = t , ω o = 1 – t , t ∈ [ 0.4, 0.6 ] .
Given the available information, bank B does not know the
probability that C will hold any particular portfolio; however,
from equation A3 bank B does know the probability that C will
default on each portfolio that it could hold. The set of default
probabilities is given by the probability of default in the
equation below for different choices of t:
(A4) PD ( t ) = PD ( ω w = t, ω o = 1 – t, L, 1 ) , t ∈ [ 0.4, 0.6 ] .
Therefore, given the set of possible portfolios, we have a
range of possible default probabilities that bank C could have.
FRBNY Economic Policy Review / August 2010
85
Appendix B: Solving for the Size of Bank C’s Required Equity Injection
We solve for the size of the equity injection needed to sustain
interbank lending from bank B to bank C when there is
uncertainty about bank C’s portfolio holdings and when
bank B knows C’s portfolio composition because it has been
revealed as part of a stress test.
When there is uncertainty about bank C’s portfolio
holdings, an uncertainty-averse lender will assess the default
risk as equal to PD , which is the highest default probability
that bank C could have, given its possible portfolio holdings:
PD =
m axXxPD ( ω w = t, ω o = 1 – t, L, 1 ) ,
t ∈ [ 0.4, 0.6 ]
where PD(.) is defined in equation A3. Provided that PD < 0.5 ,
which is very plausible, Pritsker (2009) shows that PD(.) is a
convex function of the portfolio weights. Therefore, the
problem of solving for PD maximizes a convex objective
function over a convex set. It follows that the solution is on the
boundary, at either t = 0.4 or t = 0.6.
Using the expression for PD(.), PD can be expressed as
L
------------- R 0C,2 – μ 1
1
+
L
PD = Φ ------------------------------------ ,
σ 12
where L = F ⁄ E ; ω w and ω o are the portfolio weights for the
portfolio that generates the maximum probability of default;
and μ 1 and σ 1 are the mean and standard deviation,
respectively, of the return on the portfolio that maximizes C’s
default probability.
Solving the above equation for E, it then follows that the
original amount of equity capital in bank C, denoted E0, is:
F [ R 0C, 2 – ( PD σ 1 + μ 1 ) ]
E 0 = --------------------------------------------------------.
PD σ 1 + μ 1
86
Informational Easing
Let PDT, “the PD target,” denote the required maximum
level of PD for which it is possible to support an interbank loan
between banks B and C when B is uncertainty-averse. From the
above equation it follows that, holding F constant, the amount
of equity in C’s capital structure needed to reduce its maximum
level of PD to PDT is
T
F [ R 0C, 2 – ( PD σ 1 + μ 1 ) ]
E T, NI = ----------------------------------------------------------T
PD σ 1 + μ 1
when information to reduce uncertainty is not provided (NI is
no information).
When information is provided to reduce uncertainty,
revealing C’s portfolio weights, then the amount of equity
needed in C’s capital structure is
T
F [ R 0C, 2 – ( PD σ 1 + μ 1 ) ]
-,
E T, I = ----------------------------------------------------------T
( PD σ 1 + μ 1 )
where I is information.
When uncertainty is unresolved, the percentage equity
injection that is required is 100 × ( E T, NI ⁄ E 0 – 1) ; when
information is provided that resolves uncertainty, it is
100 × ( E T, I ⁄ E 0 – 1) . The percentage equity injections are
reported in Chart 4 in the text for different initial portfolios
ω(t) .
References
Caballero, R. J., and A. Krishnamurthy. 2008. “Collective Risk
Management in a Flight to Quality Episode.” Journal of
Finance 63, no. 5 (October): 2195-230.
Easley, D., and M. O’Hara. 2009. “Ambiguity and Nonparticipation:
The Role of Regulation.” Review of Financial Studies 22, no. 5
(May): 1817-43.
Gilboa, I., and D. Schmeidler. 1989. “Maxmin Expected Utility with
Non-Unique Prior.” Journal of Mathematical Economics 18,
no. 2: 141-53.
Gorton, G. B. 2008. “The Panic of 2007.” In Maintaining Stability
in a Changing Financial System, 131-262. Proceedings of
the Federal Reserve Bank of Kansas City’s 2008 Jackson Hole
Symposium. Available at http://www.kc.frb.org/publicat/sympos/
2008/Gorton.03.12.09.pdf.
———. 2009. “Slapped in the Face by the Invisible Hand: Banking and
the Panic of 2007.” Paper prepared for the Federal Reserve Bank of
Atlanta’s 2009 Financial Markets Conference on Financial
Innovation and Crisis. Available at http://www.frbatlanta.org/
news/CONFEREN/09fmc/gorton.pdf.
Hansen, L. P., and T. J. Sargent. 2007. Robustness. Princeton, N.J.:
Princeton University Press.
Klibanoff, P., M. Marinacci, and S. Mukerji. 2005. “A Smooth Model of
Decision Making under Ambiguity.” Econometrica 73, no. 6
(November): 1849-92.
Pritsker, M. 2009. “Knightian Uncertainty and Interbank Lending.”
Unpublished paper, Board of Governors of the Federal Reserve
System.
Rigotti, L., and C. Shannon. 2005. “Uncertainty and Risk in Financial
Markets.” Econometrica 73, no. 1 (January): 203-43.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or
the Board of Governors of the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or
implied, as to the accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information
contained in documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / August 2010
87
Viral V. Acharya, João A. C. Santos, and Tanju Yorulmazer
Systemic Risk and Deposit
Insurance Premiums
1. Introduction
hile systemic risk—the risk of wholesale failure of banks
and other financial institutions—is generally considered
to be the primary reason for supervision and regulation of the
banking industry, almost all regulatory rules treat such risk in
isolation. In particular, they do not account for the very features
that create systemic risk in the first place, such as correlation
among banks’ investments (Acharya 2009; Acharya and
Yorulmazer 2007, 2008); the large size of some banks (O’Hara
and Shaw 1990),1 which leads to “fire-sale”–related pecuniary
externalities; and bank interconnectedness (Allen and Gale
2000; Kahn and Santos 2005). In this paper, we aim to fill this
important gap in the design of regulatory tools by providing a
normative analysis of how deposit insurance premiums could
best be structured to account for systemic risk.
Demand deposits are explicitly or implicitly insured in most
countries up to some threshold amount per individual (or
deposit account). While regulators in some countries have
realized the need to establish a deposit insurance fund only
during the 2007-09 financial crisis, others have established
funds much earlier. Demirgüç-Kunt, Karacaovali, and Laeven
(2005) show that most countries provide deposit insurance.
Furthermore, during the crisis of 2007-09, some countries,
including developed countries such as Australia and New
Zealand, introduced guarantees for the first time, whereas a
significant majority of others increased their insurance
coverage. In most cases, the capital in these deposit insurance
funds is the reserve built up over time through the collection of
insurance premiums from banks that receive the benefits of
deposit insurance. Yet how should such premiums be charged?
We argue that the extent of systemic risk in the financial
sector is a key determinant of efficient deposit insurance
premiums. The basic argument is as follows. When a bank with
insured deposits fails, the deposit insurance fund takes over the
bank and sells it as a going concern or piecemeal. During
periods of widespread bank failure, it is difficult to sell failed
banks at attractive prices because other banks are also
experiencing financial constraints (Shleifer and Vishny 1992;
Allen and Gale 1994). Hence, in a systemic crisis, the deposit
insurance fund suffers from low recovery from the liquidation
of failed banks’ assets. This, in turn, leads to higher drawdowns
per dollar of insured deposits. This argument gives our first
result: the actuarially fair deposit insurance premium—the
premium that exactly covers the expected cost to the deposit
insurance provider—should not only increase in relation to
individual bank failure risk but also in relation to joint bank
failure risk.2
1
Only recently, the Federal Deposit Insurance Corporation (FDIC) announced
a special assessment, to be collected on September 30, 2009, that will be
computed based on total assets (minus “tier 1” capital). See http://
www.fdic.gov/deposit/insurance/assessments/proposed.html.
2
Viral V. Acharya is a professor of finance at the London Business School and
New York University; João A. C. Santos is an assistant vice president and Tanju
Yorulmazer a senior economist at the Federal Reserve Bank of New York.
vacharya@stern.nyu.edu
joao.santos@ny.frb.org
tanju.yorulmazer@ny.frb.org
The authors thank Douglas Gale, Kenneth Garbade, Todd Keister, George
Pennacchi, Asani Sarkar, Ingo Walter, and Michelle Zemel for helpful
suggestions. The views expressed are those of the authors and do not
necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System.
W
Pennacchi (2006) shows that if insurance premiums are set to a bank’s
expected losses and fail to include a systematic risk premium, banks that make
investments with higher systematic risk enjoy a greater financing subsidy
relative to banks that make investments with lower systematic risk.
FRBNY Economic Policy Review / August 2010
89
In addition, the failures of large banks lead to greater firesale discounts. This occurrence has the potential to generate
a significant pecuniary externality that can have adverse
contagion-style effects on other banks and the real economy
(compared with the effects stemming from the failure of
smaller banks).3 Hence, the resolution of large banks is more
costly for the deposit insurance regulator, directly in terms
of losses from liquidating large banks and indirectly from
contagion effects. This leads to our second result: the premium
for large banks should be higher per dollar of insured deposit
compared with that for small banks.
Furthermore, bank closure policies reflect a timeinconsistency problem (see, for example, Mailath and Mester
[1994] and Acharya and Yorulmazer [2007, 2008]). In
particular, regulators ex ante would like to commit to being
tough on banks even when there are wholesale failures to
discourage banks from ending up in that situation. However,
this strategy is not credible ex post, and regulators show greater
forbearance during systemic crises. While such forbearance
among most regulators around the world has been a feature
of the current crisis, it has a strong precedent. For example,
Hoggarth, Reidhill, and Sinclair (2004) study resolution
policies adopted during thirty-three international banking
crises from 1977 to 2002. They document that when faced with
individual bank failures, authorities have typically sought a
private sector resolution in which the losses have been passed
on to existing shareholders, managers, and sometimes
uninsured creditors—but not taxpayers. Still, government
involvement has been an important feature of the resolution
process during systemic crises: at early stages, liquidity support
from central banks and blanket government guarantees have
been granted, usually at a cost to the budget; bank liquidations
have occurred very infrequently, and creditors have rarely
suffered any losses.
Such forbearance during systemic crisis creates incentives
for banks to herd and become interconnected; thus, when they
fail, they do so with others—and this increases their chance of
a bailout. Given this collective moral hazard, we obtain our
third and final result: the incentive-efficient premium that
discourages banks from excessive correlation in their investments
features a higher charge for joint bank failure risk than the
actuarially fair premium. In other words, from a normative
standpoint, the deposit insurance premium charged to banks
is increasing in systemic risk.
The remainder of our paper is organized as follows.
Section 2 offers a brief history of the FDIC and deposit
insurance premiums. In Section 3, we describe a model we have
developed to provide normative analysis of deposit insurance
premiums. Section 4 derives the actuarially fair deposit
insurance premium as a function of systemic risk separately for
large and small banks. In Section 5, we consider the role of
forbearance and derive the incentive-efficient deposit
insurance premium—taking into account all potential costs
associated with the resolution of failed banks, such as the cost
of inefficient liquidations and bailouts—and compare it with
the actuarially fair premium. Section 6 concludes.
2. The FDIC and Deposit
Insurance Premiums
While the three principles used to determine efficient deposit
insurance premiums apply generally, it is useful to consider
them in the context of how premiums have been priced in the
United States. Accordingly, we briefly discuss the Federal
Deposit Insurance Corporation—the U.S. deposit insurance
regulator—and the premium schemes that have been used in
the United States.4
In response to the devastating effects of the Great
Depression, the U.S. government established the FDIC in 1933
to insure deposits of commercial banks and prevent banking
panics. The FDIC’s reserves began with a $289 million capital
injection from the U.S. Treasury and the Federal Reserve in
1934. Throughout most of the FDIC’s history, deposit
insurance premiums have been independent of bank risk,
mainly because of the difficulty assessing that risk. Between
1935 and 1990, the FDIC charged flat deposit insurance
premiums at the rate of approximately 8.3 cents per $100 of
insured deposits. However, in 1950, the FDIC began to rebate
some of the collected premiums. The rebates have been
adjusted to target the amount of reserves in the FDIC’s deposit
insurance fund (DIF).
While the banking industry usually wanted deposit
insurance assessments to be set at a relatively low level, the
FDIC preferred that premiums be high enough for the reserves
to cover future claims from bank failures. In 1980, the DIF was
given a range of 1.1 percent to 1.4 percent of total insured
deposits. However, as a result of a large number of bank failures
during the 1980s, the DIF was depleted. Subsequently, the
Financial Institutions Reform, Recovery, and Enforcement Act
of 1989 mandated that the premiums be set to achieve a
1.25 percent designated reserve ratio (DRR) of reserves to total
insured deposits. (Chart 1 shows total deposits insured by the
FDIC; Chart 2 displays the balances of the DIF and the reserve
ratio for the 1990-2008 period.)
3
Such effects have epitomized the current crisis—especially the failures of
Lehman Brothers and (effectively) AIG, although they are not deposit-insured
entities.
90
Systemic Risk and Deposit Insurance Premiums
4
Our discussion is based largely on Pennacchi (2009) and Cooley (2009);
also see Saunders and Cornett (2007).
Chart 1
Chart 2
Total Deposits Insured by FDIC
Balances of Deposit Insurance Fund
and the Reserve Ratio
Billions of dollars
4,500
Billions of dollars
Percentage of insured deposits
80
70
4,000
3,500
1.6
1.4
Reserve ratio
(right scale)
60
1.2
50
1.0
40
0.8
30
3,000
0.6
Fund balances
(left scale)
20
2,500
1990
92
94
96
98
00
02
04
06
08
Source: Federal Deposit Insurance Corporation (FDIC).
The bank failures of the 1980s and early 1990s led to reforms
in the supervision and regulation of banks; these included the
Federal Deposit Insurance Corporation Improvement Act
(FDICIA) of 1991, which introduced several nondiscretionary
rules. In particular, the FDICIA required the FDIC to set
risk-based premiums, whereby premiums differed according to
three levels of bank capitalization (well capitalized, adequately
capitalized, undercapitalized) and three supervisory rating
groups (ratings of 1 or 2, a rating of 3, ratings of 4 or 5).
However, the new rules have not been as effective as possible
in differentiating between banks; indeed, from 1996 to 2006,
more than 90 percent of all banks were categorized in the
lowest risk category (well capitalized, with a rating of 1 or 2).
Furthermore, the FDICIA and the Deposit Insurance Act of
1996 specified that if DIF reserves exceed the 1.25 percent DRR,
the FDIC is prohibited from charging insurance premiums to
banks in the lowest risk category. During the 1996-2006 period,
DIF reserves were above 1.25 percent of insured deposits and,
because the majority of banks were classified in the lowest risk
category, these banks did not pay for deposit insurance.
The Federal Deposit Insurance Reform Act of 2005 brought
some changes to the setting of insurance premiums. In
particular, the Act gave the DRR a range of 1.15 percent to
1.50 percent, instead of a hard target of 1.25 percent. When DIF
reserves exceed 1.50 percent (1.35 percent), 100 percent
(50 percent) of the surplus is rebated to banks. If DIF reserves
fall below 1.15 percent, the FDIC must restore the fund and
raise premiums to a level sufficient to return reserves to the
DRR range within five years.
During the financial crisis of 2007-09, DIF reserves were
hard-hit. The reserves fell to 1.01 percent of insured deposits
on June, 30, 2008, and they decreased by $15.7 billion
(45 percent) to $18.9 billion in the fourth quarter of 2008—
0.4
10
0.2
0
0.0
-10
-20
1990
-0.2
-0.4
92
94
96
98
00
02
04
06
08
Source: Federal Deposit Insurance Corporation.
plunging the reserve ratio to 0.4 percent of insured deposits, its
lowest level since June 30, 1993.5 In the first week of March
2009, the FDIC announced plans to charge 20 cents for every
$100 of insured domestic deposits to restore the DIF.6 On
March 5, 2009, Sheila Bair, Chairperson of the FDIC, said that
her agency would lower the charge to approximately 10 basis
points if the FDIC’s borrowing authority were increased.7
Subsequently, U.S. senators Christopher Dodd and Michael
Crapo introduced a bill that would permanently raise the
FDIC’s borrowing authority to $100 billion, from $30 billion,
as well as temporarily allow the agency to borrow as much as
$500 billion in consultation with the President and other
regulators.
This short discussion confirms our earlier assertion that
deposit insurance premiums have either been risk-insensitive
or relied only on individual bank failure risk and never on
systemic risk. Furthermore, even when premiums have been
risk-sensitive, the focus has been on maintaining reserves at an
“appropriate” level. For example, when the deposit insurance
fund’s reserves become sufficiently high relative to the size of
insured deposits, the FDIC in effect returns premiums to
banks. This type of approach to premiums is divorced from
incentive properties. The rationale for charging banks a
5
More recently, two additional failures have depleted the insurance fund
further. On May 1, 2009, federal regulators shut down Silverton Bank, the
fifth-largest bank to fail during the financial crisis of 2007-09. The FDIC
estimates that the failure would cost the DIF $1.3 billion. On May 21, 2009,
federal regulators seized BankUnited FSB, at an estimated cost of $4.9 billion
to the DIF.
6
“Bair: Without Fee, Fund May Go Dry.” American Banker, March 5, 2009.
7
“FDIC to Slash Special Fee.” American Banker, March 6, 2009.
FRBNY Economic Policy Review / August 2010
91
premium on a continual basis according to individual and
systemic risk, regardless of the deposit insurance fund’s size, is
that it causes banks to internalize the costs of their failures on
the fund and rest of the economy. Since a systemic crisis would
most likely make the fund fall short and require the use of
taxpayer funds, the incentive-efficient use of excess fund
reserves is a return to taxpayers rather than to insured banks.
3. The Model
Our model is purposely simple. It is meant to illustrate the
straightforward nature of our three results on the efficient
design of premiums. In practice, quantifying systemic risk can
be a challenge, but recent advances on this front (see, for
instance, Adrian and Brunnermeier [2008] and Acharya et al.
[2009]) present the opportunity to employ them in revisions
of future deposit insurance schemes.
Our paper is related to the literature on the pricing of
deposit insurance (Merton 1977, 1978; Marcus and Shaked
1984; McCulloch 1985; Ronn and Verma 1986; Pennacchi
1987a; Flannery 1991), the difficulty (Chan, Greenbaum, and
Thakor 1992) and nondesirability (Freixas and Rochet 1998)
of pricing deposit insurance fairly, deposit insurance and
the degree of government regulatory control over banks
(Pennacchi 1987b), and, more closely, deposit insurance
pricing in the presence of regulatory forbearance in the closing
of banks (Allen and Saunders 1993; Dreyfus, Saunders, and
Allen 1994). However, our paper differs importantly from the
literature cited, as our main purpose is to analyze the pricing of
deposit insurance in a way that accounts for systemic risk as
well as important features that contribute to systemic risk, such
as correlation among banks’ investments; the large size of some
banks, which leads to fire-sale–related pecuniary externalities;
and bank interconnectedness (also see Pennacchi [2006]).
We use the set-up in Acharya and Yorulmazer (2007). We
consider an economy with three dates – t = 0 , 1, 2 , two banks –
Bank A and Bank B, bank owners, depositors, outside investors,
and a regulator. Each bank can borrow from a continuum of
depositors of measure 1. Bank owners as well as depositors are
risk-neutral, and obtain a time-additive utility w t , where w t is
the expected wealth at time t. Depositors receive a unit of
endowment at t = 0 and t = 1. Depositors also have access to a
reservation investment opportunity that gives them a utility of
1 per unit of investment. In each period, that is, at date t = 0
and t = 1, depositors choose to invest in this reservation
opportunity or in their bank.
Deposits take the form of a simple debt contract with a
maturity of one period. In particular, the promised deposit rate
92
Systemic Risk and Deposit Insurance Premiums
is not contingent on investment decisions of the bank or on
realized returns. In order to keep the model simple and yet
capture the fact that there are limits to equity financing, we do
not consider any bank financing other than deposits.
Banks require one unit of wealth to invest in a risky
technology. The risky technology can be thought of as a
portfolio of loans to firms in the corporate sector. The
performance of the corporate sector determines its random
output at date t + 1. We assume that all firms in the sector can
either repay fully the borrowed bank loans or they default on
these loans. In the case of a default, we assume for simplicity
that there is no repayment.
Suppose R is the promised return on a bank loan. We denote
the random repayment on this loan as R̃ , R̃ ∈ { 0 , R}. The
probability that the return from these loans is high (R) in
period t is α t :
Riwithiprobabilityi α t
(1)
R̃ =
0iwithiprobabilityi1 – α t .
We assume that the returns in the two periods are independent
but allow the probability of high return to be different in the
two periods. This helps isolate the effect of each probability on
our results.
In addition to banks and depositors, there are outside
investors who always have funds to purchase banking assets
were these assets to be liquidated. However, outsiders do not
have the skills to generate the full value from banking assets. To
capture this, we assume that outsiders cannot generate R in the
high state but only ( R – Δ ) . Thus, when the banking assets are
liquidated to outsiders, there may be a social welfare loss due to
misallocation of these assets. We revisit this point in Section 5,
when we investigate whether actuarially fair deposit insurance
can prevent systemic risk.
The notion that outsiders may not be able to use banking
assets as efficiently as the existing bank owners is akin to the
notion of asset-specificity, first introduced in the corporate
finance literature by Williamson (1988) and Shleifer and Vishny
(1992). In summary, this literature suggests that firms whose
assets tend to be specific, that is, whose assets cannot be readily
redeployed by firms outside of the industry, are likely to
experience lower liquidation values because they may suffer
from fire-sale discounts in cash auctions for asset sales, especially
when firms within an industry simultaneously become
financially or economically distressed.8 Regarding the evidence
of such specificity for banks and financial institutions, James
8
There is strong empirical support for this idea in the corporate finance
literature, as shown, for example, by Pulvino (1998) for the airline industry and
by Acharya, Bharath, and Srinivasan (2007) for the entire universe of defaulted
firms in the United States from 1981 to 1999 (see also Berger, Ofek, and Swary
[1996] and Stromberg [2000]).
(1991) studies the losses from bank failures in the United States
from 1985 through mid-1988 and documents that “there is
significant going concern value that is preserved if the failed bank
is sold to another bank (a ‘live bank’ transaction) but is lost if the
failed bank is liquidated by the FDIC.”
In addition, our model includes the presence of a regulator.
The deposits are fully insured by the regulator and the
regulator charges deposit insurance premiums. Since deposits
are fully insured, they are riskless. Hence, the rate of return on
deposits is equal to the rate of return from the storage
technology, that is, the deposit rate is equal to 1 in both periods.
For simplicity, we assume that banks pay the insurance
premiums using their retained earnings from earlier
investments before t = 0.
If the return from the first-period investment is high, then
the bank operates for one more period and makes the secondperiod investment.9 For a bank to continue operating, it needs
one unit to pay old deposits and an additional one unit to
undertake the second-period investment, a total of two units.
Since available deposits for a bank amount to only one unit (the
endowment of its depositors), if the return from the firstperiod investment is low, then the bank is in default, it is closed,
and its assets are sold (we discuss bailouts and recapitalization
in Section 5).10 We assume that if there is a surviving bank,
then it has resources from its first-period profits to purchase
the failed bank.
The possible states at date 1 are given as follows, where S
indicates survival and F indicates failure:
SS: Both banks had the high return, and they operate in the
second period.
SF: Bank A had the high return, while Bank B had the low
return. Bank B is acquired by Bank A.
FS: This is the symmetric version of state SF.
FF: Both banks failed.
3.1 Correlation of Bank Returns
A crucial aspect of our model is that banks can choose the
correlation of the returns from their investments by selecting
the industries they invest in. At date 0, banks borrow deposits
and then choose the composition of loans that compose their
respective portfolios. This choice determines the level of
correlation between the returns from their respective
9
For simplicity, we assume that the bank does not reinvest its profits from the
first investment, for example, distribute them as dividends.
10
In this model, the asset to be sold is the franchise value of the bank, that is,
the expected future profit from the second-period investment the bank can
take.
Joint Probability of Bank Returns
Same Industry
Different Industries
Bank B
Bank B
High (R)
Low (0)
High (R)
Low (0)
α t ( 1 –α t )
High (R)
αt
0
2
αt
Low (0)
0
1 – αt
αt ( 1 – αt )
Bank A
( 1 – αt )
2
Source: Authors’ calculations.
investments. We refer to this correlation as “interbank
correlation.”
Suppose there are two possible industries in which banks
can invest, denoted as 1 and 2. Bank A (B) can lend to firms A1
and A2 (B1 and B2) in industries 1 and 2, respectively. If in
equilibrium banks choose to lend to firms in the same industry,
specifically they either lend to A1 and B1, or they lend to A2 and
B2, then their returns are assumed to be perfectly correlated.
However, if they choose different industries, then their returns
are less than perfectly correlated, say, independent. When
banks invest in the same industry, the correlation of banks’
returns is ρ = 1, whereas when they invest in different
industries, we have ρ = 0 . This gives us the joint probability
distribution of bank returns as presented in the table. Note that
the individual probability of each bank succeeding or failing is
constant ( α 0 and 1 – α 0 , respectively), irrespective of the
correlation in their returns.11
4. Actuarially Fair Insurance
without Bailouts
In this section, we assume that the regulator sells the assets of
the failed banks. (We analyze regulatory intervention in the
form of recapitalization and bailouts in Section 5.)
Next, we show that the actuarially fair deposit insurance
premium, the premium that is equal to the expected value of
the payments from the insurance fund to the bank’s depositors,
depends on the correlation structure in banks’ investments.
Since deposits are fully insured, the deposit rate in both
periods is equal to 1.
In state FF, both banks fail, and sale to another bank is not
an option. Thus, the failed banks’ assets are sold to outsiders,
which can also be thought of as the liquidation of the banks’
11
Our results hold as long as the probabilities of states SS and FF are higher
when banks invest in the same industry, rather than in different industries.
FRBNY Economic Policy Review / August 2010
93
assets. Note that the outsiders cannot generate R from the
banking assets but only R – Δ . They are therefore willing to pay
a price of at most p for the failed banks’ assets where
(2)
p
FF
= p = α1 ( R – Δ – 1 ).
We can think of p as the liquidation value of the bank.
In states SF or FS, the surviving bank can acquire the failed
bank’s assets. Note that a surviving bank can generate the full
value of R from these assets. Thus, these assets are worth p for
the surviving bank, where
(3)
p = α1 ( R – 1 ).
Note that p > p . We assume that neither the regulator nor the
surviving bank has the full bargaining power for the sale of the
SF
failed banks’ assets. Thus, the price, denoted as p , lies
SF
between p and p , that is, p ∈ ( p, p) .
When banks invest in the same industry with probability
( 1 – α 0 ), both banks fail and the proceeds from the sale of the
FF
failed banks’ assets are equal to p = p . Let qs be the insurance
premium when banks invest in the same industry, where
(4)
qs = ( 1 – α0 ) ( 1 – p ) .
When banks invest in different industries, with probability
α 0 ( 1 – α 0 ) , only one bank fails and the proceeds from the sale
SF
2
are p ; with probability ( 1 – α 0 ) , both banks fail and the
proceeds from the sale of the failed banks’ assets are equal to
FF
p = p . Let q d be the insurance premium when banks invest
in different industries, where
(5)
SF
2
qd = α0 ( 1 – α0 ) ( 1 – p ) + ( 1 – α0 ) ( 1 – p )
SF
= qs – α0 ( 1 – α0 ) ( p – p ) .
Since the proceeds from the sale of failed banks’ assets are lower
when both banks fail, the loss to the insurance fund is higher
when both banks fail. Thus, the actuarially fair insurance
premiums should be higher when banks invest in the same
industry, that is, q s > q d .
Result 1—(Correlation and actuarially fair insurance premiums):
The actuarially fair insurance premium depends on the
correlation between banks’ returns and should be higher when
banks invest in the same industry, and is given as
SF
qs = qd + α0 ( 1 – α0 ) ( p – p ) > qd .
Next, we show that the insurance premium should depend
on bank size as well. Suppose that instead of two banks of equal
size, we let Bank A be the large bank, with the size of depositors
much larger than 1, while we keep the size of Bank B at 1.
We assume that if the regulator decides to liquidate the
small bank, the large bank (or some other bank in the industry)
94
Systemic Risk and Deposit Insurance Premiums
has enough funds to purchase the small bank and can run it
efficiently. Thus, assuming that all bargaining power does not
lie with the regulator or the acquiring bank, when the small
bank is liquidated the liquidation value is assumed to be
psmall ∈ ( p, p ) .
However, the size of Bank A is large enough so that the small
bank cannot acquire and run the large bank efficiently. Thus,
when the large bank is liquidated, it can be purchased only by
outside investors and the price per unit of the large bank’s
assets is pbig = p . Hence, the actuarially fair insurance
premiums depend on the size of the bank. In particular, we
obtain:
Result 2—(Size and actuarially fair insurance premiums): The
actuarially fair premium per dollar of insured deposits for the
large bank is higher compared with that of the small bank and
(6)
q small = ( 1 – α 0 ) ( 1 – p small ) < ( 1 – α 0 ) ( 1 – p ) = q big .
So far, we have restricted the actions of the regulator to the
provision of deposit insurance and the resolution of bank
failures only through sales. Since the failure of large banks or
many banks at the same time can result in more adverse effects
on the rest of the economy, it is more likely that, in such cases,
regulators show forbearance or intervene in the form of
bailouts or capital injections, resulting in fiscal costs. This,
in turn, strengthens our argument that size and correlation
should be an important component of insurance premiums.
In the next section, we analyze insurance premiums in the
presence of bailouts and recapitalizations, taking into account
costs associated with the resolution of failed banks, such as the
costs of liquidations and bailouts.
5. Resolution of Bank Failures
and Insurance Premiums
In this section, we first analyze the problem of resolving bank
failures when the regulator can bail out and recapitalize failed
banks as well as sell the failed banks to a surviving bank (if any)
or to outsiders. We show that in the case of a joint failure of
banks, the regulator may prefer to bail out or recapitalize failed
banks ex post (“too-many-to-fail” guarantees). However, such
guarantees create incentives for banks to herd and make
correlated investments, which makes the joint failure state—
that is, the state of systemic crisis—more likely in the first place.
Next, we derive the full-cost insurance premiums that take
into account all social costs of bank failures, including costs of
inefficient liquidations and bailouts, and show that these
premiums should be higher than the actuarially fair insurance
premiums derived in Section 4. Furthermore, we analyze how
the regulator can use insurance premiums as a tool to minimize
the occurrence of systemic crisis by preventing banks from
choosing highly correlated investments. We use the term
incentive-efficient full-cost insurance premiums to describe the
premiums that take into account all social costs of bank failures
while giving banks incentives to choose the low correlation.
5.1. Resolution of Bank Failures
Since there is no social welfare loss when assets remain in the
banking system, the regulator has no incentive to intervene (in
the form of bailouts) in states SS, SF, and FS. However, in state
FF, the assets of failed banks can be purchased only by outside
investors, resulting in misallocation costs. Hence, the regulator
compares the welfare loss resulting from asset sales to outsiders
with the cost of bailing out the failed banks. If it turns out that
the welfare loss from inefficient liquidation is greater, then the
regulator may decide to intervene in the form of bailouts and
recapitalizations. The regulator’s ex post decision is thus more
involved in state FF, and we examine it fully. In order to analyze
the regulator’s decision to bail out or close failed banks, we
make the following assumptions:
1) The regulator incurs a cost of f ( x ) when it injects x units
of funds into the banking sector. We assume that this cost
function is increasing, f ′ > 0 , and for simplicity we consider a
linear cost function: f ( x ) = ax ,a > 0 . While we do not model
this cost explicitly, we have in mind fiscal and opportunity
costs to the regulator from providing funds with immediacy to
the banking sector. Thus, if the regulator bails out only one
bank (both banks), it incurs a bailout cost of a ( 2a ).
The fiscal costs of providing funds to the banking sector
with immediacy can be linked to a variety of sources, most
notably: a) the distortionary effects of tax increases required to
fund deposit insurance and bailouts and b) the likely effect of
government deficits on the country’s exchange rate, manifested
in the fact that banking crises and currency crises have often
occurred as “twins” in many (especially emerging market)
countries. Ultimately, the fiscal cost we have in mind is one of
immediacy: Government expenditures and inflows during the
regular course of events are smooth, relative to the potentially
rapid growth of “off-balance-sheet contingent liabilities,” such
as the costs of bank bailouts.12
2) If the regulator decides not to bail out a failed bank, the
existing depositors are paid back through deposit insurance
and the failed bank’s assets are sold to outsiders. The crucial
difference between bailouts and asset sales from an ex post
standpoint is that proceeds from asset sales lower the fiscal cost
from the immediate provision of deposit insurance, whereas
bailouts produce no such proceeds. In other words, bailouts
entail an opportunity cost to the regulator in fiscal terms.
Under these assumptions, the regulator’s resolution policy
can be characterized as follows. The regulator’s objective in
state FF is to maximize the total expected output of the banking
sector net of any bailout or liquidation costs. We denote this as
ff
E (Π 2 ) . Thus, if both banks are closed, the regulator’s objective
function takes the value
(7)
ff
E ( Π 2 ) = 2 [ α 1 ( R – Δ ) – 1 ],
which is the liquidation value of banking assets to outsiders.
This equals [2 ( α 1 R – 1) – 2 α 1 Δ ], the difference between the
banking sector output in each of the states SS, SF, and FS,
minus the liquidation costs from closing both banks.
If both banks are bailed out, then the regulator’s objective
function takes the value
(8)
ff
E ( Π 2 ) = 2 (α 1 R – 1 ) – f ( 2 ) ,
as the bailout costs are now based on the total amount of funds,
2, injected into the banking sector with immediacy.13
Comparing these objective-function evaluations, we obtain
the following resolution policy for the regulator in state FF.
It has the intuitive property that if liquidation costs ( α 1 Δ ) are
sufficiently high and/or the costs of bailouts ( f (.)) are not too
steep, then there are “too many (banks) to fail” and the
regulator prefers to rescue failed banks.
Resolution : When both banks fail (state FF), the regulator takes
the following actions:
• If α 1 Δ ≤ f ( 1) , then both banks’ assets are sold to outsiders.
• If α 1 Δ > f ( 1) , then the regulator bails out both banks.
12
See, for example, the discussion of fiscal costs associated with banking
collapses and bailouts in Calomiris (1998). Hoggarth, Reis, and Saporta (2002)
find that the cumulative output losses have amounted to a sizable 15 percent to
20 percent of annual GDP in the banking crises of the past twenty-five years.
Caprio and Klingebiel (1996) argue that the bailout of the thrift industry cost
$180 billion (3.2 percent of GDP) in the United States in the late 1980s. They
also document that the estimated cost of bailouts was 16.8 percent for Spain,
6.4 percent for Sweden, and 8 percent for Finland. Honohan and Klingebiel
(2000) find that countries spent 12.8 percent of their GDP to fix their banking
systems, whereas Claessens, Djankov, and Klingebiel (1999) set the cost at
15 percent to 50 percent of GDP. Also see Panageas (2009) for an analysis of the
optimal financing of government interventions.
13
With the linear fiscal cost function f (.), the regulator either bails out both
banks or liquidates both. With a strictly convex fiscal cost function f (.), there
may be cases in which it is optimal to bail out one bank and liquidate the other,
since the marginal cost of bailouts increases as more banks are bailed out.
See Acharya and Yorulmazer (2007) for a discussion.
FRBNY Economic Policy Review / August 2010
95
Thus, the expected second-period profits of the bank depend
on the regulator’s decision:
(9)
0 if α 1 Δ ≤ f ( 1 )
ff
.
E (π 2 ) =
p if α 1Δ > f ( 1 )
Note that in either case, in state FF there is a social welfare
loss resulting from bailout or liquidation, whereas no such cost
arises in states SF or FS. Thus, the socially optimal outcome is
achieved when the probability of state FF is at a minimum, that
is, when banks invest in different industries.
5.2. Systemic Risk and Insurance Premiums
First, we derive the full-cost insurance premiums—the
premiums that take into account all social costs of bank failures
including costs of liquidations and bailouts. Note that the
actuarially fair insurance premiums in Section 4 take into
account only the expected payments to depositors; thus, they
fail to account for the social costs of bank failures, such as costs
of liquidations and bailouts.
We can show that the full-cost insurance premiums q̃ s and
q̃d when banks invest in the same industry and in different
industries, respectively, are given as:
(10)
(11)
E ( π1 ( ρ ) ) + E ( π2 ( ρ ) ) ,
(13)
where discounting has been ignored since it does not affect any
of the results. Recall that if banks invest in different industries,
then interbank correlation ρ equals 0, or else it equals 1.
Note that when banks invest in the same industry,
Pr(SF) = 0, so that
(14)
ss
q̃d = α 0 ( 1 – α 0 )( 1 – p SF )
2
We can obtain the relationship between these insurance
premiums as follows:
SF
(12) q̃s = q̃d + α 0( 1 – α 0 )[ ( p – p ) + min {α 1 Δ , f ( 1)} ] > q̃d .
As in the case of actuarially fair insurance premiums, the
loss to the regulator through the insurance fund is higher when
both banks fail. Furthermore, the joint failure state is always
associated with social costs, such as costs from inefficient
liquidations or bailouts, whereas these costs can be avoided in
the individual failure states. Thus, the full-cost insurance
premiums are higher than the actuarially fair insurance
premiums, that is, q̃ s > qs and q̃d > qd . Furthermore, the wedge
between the insurance premiums q̃s and q̃d is higher compared
with the corresponding wedge for the actuarially fair insurance
premiums, that is, q̃ s – q̃d > qs – qd .
Next, we investigate banks’ choice of correlation in their
investments and find the incentive-efficient insurance premiums
q̂ s and q̂ d that induce banks to choose the low correlation.
Also, we combine our results with those of the previous
Systemic Risk and Deposit Insurance Premiums
ff
E ( π 2 ( 1 ) ) = α 0 E ( π 2 ) + ( 1 – α 0 )E (π 2 ) – q̂ s .
When banks invest in different industries, we obtain that
(15)
2
ss
sf
E ( π 2 ( 0 ) ) = α 0 E (π 2 ) + α 0 ( 1 – α 0 ) E (π 2 )
2
ff
+ ( 1 – α 0 ) E (π 2 ) – q̂ d .
sf
ss
SF
We know that E (π 2 ) = E (π 2 ) + ( p – p ) . Thus, we can write
(16)
ss
SF
E ( π 2 ( 0 ) ) = α 0 E (π 2 ) + α 0 ( 1 – α 0 ) ( p – p )
2
ff
+ ( 1 – α 0 ) E (π 2 ) – q̂ d ,
q̃ s = ( 1 – α 0 )[ ( 1 – p ) + min { α 1 Δ , f ( 1)} ] > qs , and
+ ( 1 – α 0 ) [ ( 1 – p ) + min {α 1 Δ , f ( 1)} ] > qd .
96
discussion to find the incentive-efficient full-cost insurance
premiums that take into account all costs associated with the
resolution of failed banks while incentivizing banks to choose
the low correlation.
In the first period, both banks are identical. Hence, we
consider a representative bank. Formally, the objective of each
bank is to choose the level of interbank correlation ρ at date 0
that maximizes
which gives us
ff
SF
(17) E (π 2 ( 1 ) ) – E (π 2 ( 0 ) ) = α 0 ( 1 – α 0 ) [ E (π 2 ) – ( p – p ) ]
+ q̂d – q̂ s .
Hence, the only terms that affect the choice of interbank
ff
correlation are the subsidy that failed banks receive ( E (π 2 ) )
from a bailout in state FF, the discount the surviving bank
receives in state SF from acquiring the failed bank’s assets, and
the deposit insurance premiums q̂ s and q̂d . Therefore, for
banks to choose the low correlation, the premium charged
when banks invest in the same industry has to be at least:
(18)
ff
SF
q̂ s = α 0 ( 1 – α 0 ) [ E (π 2 ) – ( p – p ) ] + q̂d .
Note that when the regulator chooses to liquidate the failed
bank, rather than bail it out, there is no bailout subsidy and the
full-cost insurance premiums q̃ s and q̃d are at the same time
incentive-efficient, that is, they induce banks to choose the low
correlation. However, when the regulator bails out failed
banks, the subsidy from the bailout creates a wedge between the
incentive-efficient premium q̂ s and the full-cost insurance
premium q̃ s . Combining this with our previous result on the
insurance premium, we obtain the incentive-efficient full-cost
premiums as q̃d when banks invest in different industries and
qs = max { q̃ s , q̂ s} when banks invest in the same industry.
When the regulator charges the premiums ( qs , q̃d ) , banks
choose the low correlation (incentive-efficient) and pay for the
entire expected costs associated with their failure, including the
costs of inefficient liquidations and bailouts. We obtain the
following result:
Result 3 —(Incentive-efficient full-cost premiums): The insurance
premiums that induce banks to choose the low correlation and
that cover all expected costs associated with bank failures are q̃d
and qs = max { q̃ s , q̂ s} when banks invest in different industries
and the same industry, respectively. Furthermore, we obtain
q̃ s > q s and q̃d > qd , and the wedge between the insurance
premiums qs and q̃d is higher compared with the corresponding
wedge for the actuarially fair insurance premiums, that is,
qs – q̃d > qs – qd .
Note that the insurance premiums with regulatory
intervention in the form of bailouts are different from the ones
without such regulatory intervention. Given that the regulator
may not be credible in closing banks during systemic crises,
which creates incentives for banks to invest in the same
industry ex ante, deposit insurance premiums may act as a tool
to alleviate the time-inconsistency problem inherent in the
regulator’s policy.
We observe government bailouts during banking crises,
more so when a crisis is systemic. Thus, banks may have private
benefits from choosing correlated investments such as possible
bailouts. In those cases, the actuarially fair premium (which
may no longer be fair from a social welfare point of view) may
not be enough to prevent banks from choosing highly
correlated investments. If we believe that the social costs of
bank failures (either misallocation costs due to liquidation and
destruction of value, or costs of bailouts) increase in a convex
fashion as the number of failures increases, then the regulator
would like to prevent states in which many banks fail, that is,
the regulator would like to prevent banks from being
overexposed to common risk factors. In those cases, the
actuarially fair premium may not prevent banks from investing
in the same industry, that is, it may not prevent systemic bank
failures. Thus, for the regulator to prevent systemic risk, all
costs of failures should be priced in, and the premium imposed
when banks invest in the same industry should be higher.
The practical design of regulatory tools to address
important contributors to systemic risk, such as correlation
and size, can be difficult and potentially costly from a political
point of view. An alternative way to address these issues is
through the use of closure rules. One possibility, as Acharya
and Yorulmazer (2008) argue, is to use taxpayer funds, not to
guarantee bank debt, but to make transfers to healthier
institutions and enable the institutions to acquire failed banks
at higher costs than they would using only private funds. Such
mechanisms, however, have their limits, as larger banks emerge
from crisis resolution and closure rules are generally negatively
affected by time-inconsistency problems.
6. Conclusion
This paper has shown that the efficient setting of deposit
insurance premiums would be most effective if it took into
account systemic risk, which justifies the existence of such
insurance in the first place. Some of the major factors that lead
to systemic risk are correlation among banks’ returns, bank
size, and bank interconnectedness. These factors need to be
explicitly and continually considered when setting deposit
insurance premiums.
Our focus has been on the pricing of deposit insurance.
Although the same principles apply to the design of other
regulatory tools, such as capital and liquidity requirements
(Acharya 2009), an interesting question is the effectiveness of
different regulatory rules in addressing different sources of
systemic risk.14 Systemic risk is a negative externality arising
from one financial institution’s failure on other institutions
and the economy; it entails significant welfare costs when it
materializes in the form of widespread failures. Regulation is
required to maintain efficient levels of systemic risk—much
like pollution is regulated through the imposition of certain
taxes. However, such regulation will be effective only if it is tied
to the extent of systemic risk.
14
Sharpe (1978) shows that in the absence of moral hazard and information
frictions, there is an isomorphism between risk-based insurance premiums and
risk-related capital standards. Flannery (1991), however, shows that when
there is asymmetry of information, this isomorphism no longer holds.
FRBNY Economic Policy Review / August 2010
97
References
Acharya, V. V. 2009. “A Theory of Systemic Risk and Design of
Prudential Bank Regulation.” Journal of Financial Stability 5,
no. 3 (September): 224-55.
Chan, Y., S. I. Greenbaum, and A. V. Thakor. 1992. “Is Fairly Priced
Deposit Insurance Possible?” Journal of Finance 47, no. 1
(March): 227-45.
Acharya, V. V., S. T. Bharath, and A. Srinivasan. 2007. “Does IndustryWide Distress Affect Defaulted Firms? Evidence from Creditor
Recoveries.” Journal of Financial Economics 85, no. 3
(September): 787-821.
Claessens, S., S. Djankov, and D. Klingebiel. 1999. “Financial
Restructuring in East Asia: Halfway There?” World Bank
Financial Sector Discussion Paper no. 3, September.
Cooley, T. F. 2009. “A Captive FDIC.” Forbes.com, April 15.
Acharya, V. V., L. Pedersen, T. Philippon, and M. Richardson. 2009.
“Regulating Systemic Risk.” In V. V. Acharya and M. Richardson,
eds., Restoring Financial Stability: How to Repair a Failed
System, Chapter 13. John Wiley & Sons.
Demirgüç-Kunt, A., B. Karacaovali, and L. Laeven. 2005. “Deposit
Insurance around the World: A Comprehensive Database.”
World Bank Policy Research Working Paper no. 3628, June.
Acharya, V. V., and T. Yorulmazer. 2007. “Too-Many-To-Fail—
An Analysis of Time-Inconsistency in Bank Closure Policies.”
Journal of Financial Intermediation 16, no. 1 (January):
1-31.
Dreyfus, J.-F., A. Saunders, and L. Allen. 1994. “Deposit Insurance and
Regulatory Forbearance: Are Caps on Insured Deposits Optimal?”
Journal of Money, Credit, and Banking 26, no. 3, part 1
(August): 412-38.
———. 2008. “Cash-in-the-Market Pricing and Optimal Resolution
of Bank Failures.” Review of Financial Studies 21, no. 6
(November): 2705-42.
Flannery, M. J. 1991. “Pricing Deposit Insurance When the Insurer
Measures Bank Risk with Error.” Journal of Banking and
Finance 15, no. 4-5 (September): 975-98.
Adrian, T., and M. K. Brunnermeier. 2008. “CoVar.” Federal Reserve
Bank of New York Staff Reports, no. 348, September.
Freixas, X., and J.-C. Rochet. 1998. “Fair Pricing of Deposit Insurance.
Is It Possible? Yes. Is It Desirable? No.” Research in Economics 52,
no. 3 (September): 217-32.
Allen, F., and D. Gale. 1994. “Limited Market Participation and
Volatility of Asset Prices.” American Economic Review 84, no. 4
(September): 933-55.
———. 2000. “Financial Contagion.” Journal of Political
Economy 108, no. 1 (February): 1-33.
Allen, L., and A. Saunders. 1993. “Forbearance and Valuation of
Deposit Insurance as a Callable Put.” Journal of Banking
and Finance 17, no. 4 (June): 629-43.
Berger, P., E. Ofek, and I. Swary. 1996. “Investor Valuation of the
Abandonment Option.” Journal of Financial Economics 42,
no. 2 (October): 257-87.
Calomiris, C. W. 1998. “Blueprints for a Global Financial Architecture,
Reform Proposals.” International Monetary Fund.
Caprio, G. Jr., and D. Klingebiel. 1996. “Bank Insolvencies: CrossCountry Experience.” World Bank Policy Research Working Paper
no. 1620, July.
98
Systemic Risk and Deposit Insurance Premiums
Hoggarth, G., J. Reidhill, and P. Sinclair. 2004. “On the Resolution of
Banking Crises: Theory and Evidence.” Bank of England Working
Paper no. 229.
Hoggarth, G., R. Reis, and V. Saporta. 2002. “Costs of Banking System
Instability: Some Empirical Evidence.” Journal of Banking and
Finance 26, no. 5 (May): 825-55.
Honohan, P., and D. Klingebiel. 2000. “Controlling Fiscal Costs
of Bank Crises.” World Bank Policy Research Working Paper
no. 2441, September.
James, C. 1991. “The Losses Realized in Bank Failures.” Journal
of Finance 46, no. 4 (September): 1223-42.
Kahn, C. M., and J. A. C. Santos. 2005. “Liquidity, Payment, and
Endogenous Financial Fragility.” Unpublished paper, Federal
Reserve Bank of New York.
References (Continued)
Mailath, G. J., and L. J. Mester. 1994. “A Positive Analysis of Bank
Closure.” Journal of Financial Intermediation 3, no. 3
(June): 272-99.
———. 2006. “Deposit Insurance, Bank Regulation, and Financial
System Risks.” Journal of Monetary Economics 53, no. 1
(January): 1-30.
Marcus, A. J., and I. Shaked. 1984. “The Valuation of FDIC Deposit
Insurance: Empirical Estimates Using the Options Pricing
Framework.” Journal of Money, Credit, and Banking 16,
no. 4, part 1 (November): 446-60.
———. 2009. “Deposit Insurance.” Paper prepared for the AEI
Conference on Private Markets and Public Insurance Programs.
March.
McCulloch, J. H. 1985. “Interest-Risk-Sensitive Deposit Insurance
Premia: Stable ACH Estimates.” Journal of Banking and
Finance 9, no. 1 (March): 137-56.
Merton, R. C. 1977. “An Analytical Derivation of the Cost of Deposit
Insurance and Loan Guarantees: An Application of Modern
Option Pricing Theory.” Journal of Banking and Finance 1,
no. 1 (June): 3-11.
———. 1978. “On the Cost of Deposit Insurance When There Are
Surveillance Costs.” Journal of Business 51, no. 3 (June):
439-52.
O’Hara, M., and W. Shaw. 1990. “Deposit Insurance and Wealth
Effects: The Value of Being ‘Too Big to Fail.’” Journal of
Finance 45, no. 5 (December): 1587-1600.
Panageas, S. 2009. “Too Big to Fail, but a Lot to Bail: Optimal
Financing of Large Bailouts.” Unpublished paper, University
of Chicago.
Pennacchi, G. G. 1987a. “Alternative Forms of Deposit Insurance:
Pricing and Bank Incentive Issues.” Journal of Banking and
Finance 11, no. 2 (June): 291-312.
Pulvino, T. C. 1998. “Do Asset Fire Sales Exist? An Empirical
Investigation of Commercial Aircraft Sale Transactions.”
Journal of Finance 53, no. 3 (June): 939-78.
Ronn, E. I., and A. K. Verma. 1986. “Pricing Risk-Adjusted Deposit
Insurance: An Option-Based Model.” Journal of Finance 41,
no. 4 (September): 871-95.
Saunders, A., and M. M. Cornett. 2007. Financial Institutions
Management: A Risk Management Approach.
Irwin/McGraw-Hill.
Sharpe, W. F. 1978. “Bank Capital Adequacy, Deposit Insurance, and
Security Values.” Journal of Financial and Quantitative
Analysis 13, no. 4 (November): 701-18.
Shleifer, A., and R. W. Vishny. 1992. “Liquidation Values and Debt
Capacity: A Market Equilibrium Approach.” Journal of
Finance 47, no. 4 (September): 1343-66.
Stromberg, P. 2000. “Conflicts of Interest and Market Illiquidity in
Bankruptcy Auctions: Theory and Tests.” Journal of Finance 55,
no. 6 (December): 2641-92.
Williamson, O. E. 1988. “Corporate Finance and Corporate
Governance.” Journal of Finance 43, no. 3 (July): 567-92.
———. 1987b. “A Reexamination of the Over- (or Under-) Pricing
of Deposit Insurance.” Journal of Money, Credit, and
Banking 19, no. 3 (August): 340-60.
The views expressed are those of the authors and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / August 2010
99
John Geanakoplos
Solving the Present
Crisis and Managing
the Leverage Cycle
1. Introduction
T
he present crisis is the bottom of a leverage cycle.
Understanding that tells us what to do, in what order,
and with what sense of urgency. Public authorities have acted
aggressively, but because their actions were not rooted in (or
explained with reference to) a solid understanding of the causes
of our present distress, we have started in the wrong place and
paid insufficient attention and devoted insufficient resources
to matters—most notably, the still-growing tidal wave of
foreclosures and the sudden deleveraging of the financial
system—that should have been first on the agenda.
In short and simple terms, by leverage cycle I mean this.
There are times when leverage is so high that people and
institutions can buy many assets with very little money down
and times when leverage is so low that buyers must have all or
nearly all of the money in hand to purchase those very same
assets. When leverage is loose, asset prices go up because buyers
can get easy credit and spend more. Similarly, when leverage is
highly constrained, that is, when credit is very difficult to
obtain, prices plummet. This is what happened in real estate
and what happened in the financial markets. Governments
have long monitored and adjusted interest rates in an attempt
to ensure that credit did not freeze up and thereby threaten
the economic stability of a nation. However, leverage
John Geanakoplos is the James Tobin Professor of Economics at Yale
University, an external professor at the Santa Fe Institute, and a partner in
Ellington Capital Management, which trades primarily in mortgage securities.
john.geanakoplos@yale.edu
(equivalently, collateral rates) must also be monitored and
adjusted if we are to avoid the destruction that the tail end of an
outsized leverage cycle can bring.
Economists and the public have often spoken of tight credit
markets, meaning something more than high interest rates, but
without precisely specifying or quantifying exactly what they
meant. A decade ago, I showed that the collateral rate, or
leverage, is an equilibrium variable distinct from the interest
rate.1 The collateral rate is the value of collateral that must be
pledged to guarantee one dollar of loan. Today, many
businesses and ordinary people are willing to agree to pay bank
interest rates, but they cannot get loans because they do not
have the collateral to put down to convince the banks their loan
will be safe.
Huge moves in collateral rates, which I have called “the
leverage cycle,” are a recurring phenomenon in American
financial history.2 The steps we must take at the end of the
current cycle emerge from understanding what makes a
leverage cycle swing up, sometimes to dizzying extremes,
and then come crashing down, often with devastating
consequences.
1
Geanakoplos (1997, 2003).
The history of leverage is still being written, because until recently it was
not a variable that was explicitly monitored. But work by Adrian and Shin
(forthcoming) and others is helping to restore the historical record.
2
On October 3, 2008, the author presented to Ben Bernanke and the Federal
Reserve Board of Governors the substance of this paper’s proposal. The author
is grateful to Susan Koniak for very helpful and detailed comments and advice,
as well as for allowing him to use material from their two New York Times
editorials. He is also appreciative of the very fine comments received from
Asani Sarkar and two referees, and is deeply indebted to Joseph Tracy for
editorial advice far beyond the call of duty, encompassing tone, style, and
content. Needless to say, neither he nor anyone else is responsible for, or even
necessarily in agreement with, the views expressed here. The views expressed
are those of the author and do not necessarily reflect the position of the
Federal Reserve Bank of New York or the Federal Reserve System.
FRBNY Economic Policy Review / August 2010
101
All leverage cycles end with: 1) bad news that creates
uncertainty and disagreement, 2) sharply increasing collateral
rates, and 3) losses and bankruptcies among the leveraged
optimists. These three factors reinforce and feed back on each
other. In particular, what begins as uncertainty about
exogenous events creates uncertainty about endogenous
events, like how far prices will fall or who will go bankrupt,
which leads to further tightening of collateral, and thus further
price declines and so on. In the aftermath of the crisis, we
always see depressed asset prices, reduced economic activity,
and a collection of agents that are not yet bankrupt but
hovering near insolvency. How long the aftermath persists
depends on how deep the crisis was and how effective
government intervention is.
Once the crisis has started, the thematic solution is to
reverse the three symptoms of the crisis: contain the bad news,
intervene to bring down margins, and carefully inject
“optimistic” equity back into the system. As with most difficult
problems, a multi-pronged approach is generally the most
successful. To be successful, any government plan must respect
all three remedial prongs, and their order. The unusual
government interventions in this cycle have in many respects
been quite successful in averting a disaster—precisely, I would
argue, because they embodied some of the novel leverage cycle
principles I describe here. The effectiveness of the interventions
could be increased even further by respecting the priorities of
the problem.
In what follows, I explain what happens in the leverage cycle
and why it is so bad for the economy that it must be monitored
and controlled by the government. I show how this last cycle
fits the pattern and I further explain why this leverage cycle is
worse than all the others since the Depression. I point out that
the now-famous counterparty risk problem, which has
received so much attention of late, is also a matter of collateral.
Next, I present details on how to intervene to pull out of a
leverage cycle crisis like the one we are passing through now;
this discussion is divided into three sections, corresponding to
the three symptoms of every leverage cycle crisis. I advocate a
permanent lending facility that will stand ready, should
another crisis arise, to give loans with less collateral than the
market demands. In another section, I suggest that principal
reduction (partial debt forgiveness) by private lenders is a key
tool in dealing with the many agents, like homeowners today,
that fall underwater at the bottom of a deep leverage crisis. In
the third section, I assemble the many pitfalls the government
must be watchful of if it feels obliged to rescue drowning firms
or it is tempted to buy assets at “fire-sale” prices in the darkest
days of the crisis. I conclude with a list of recommendations
for managing the leverage cycle in its ebullient period that
might prevent the next cycle from reaching such a devastating
crisis stage.
102
Solving the Present Crisis and Managing the Leverage Cycle
2. Margins, the Leverage Cycle,
and Asset Prices
Traditionally, governments, economists, as well as the general
public and the press, have regarded the interest rate as the most
important policy variable in the economy. Whenever the
economy slows, the press clamors for lower interest rates from
the Federal Reserve, and the Fed often obliges. But sometimes,
especially in times of crisis, collateral rates (equivalently,
margins or leverage) are far more important than interest rates.
The Fed could be managing collateral rates all through the
leverage cycle, but especially in the ebullient and the crisis
stages.
The use of collateral and leverage is widespread. A homeowner (or a big investment bank or hedge fund) can often
spend $20 of his own cash to buy an asset like a house for $100
by taking out a loan for the remaining $80 using the house as
collateral. In that case, we say that the margin or haircut or
down payment is 20 percent, the loan to value is $80/$100 =
80 percent, and the collateral rate is $100/$80 or 125 percent.
The leverage is the reciprocal of the margin, namely, the
ratio of the asset value to the cash needed to purchase it, or
$100/$20 = 5. All of these ratios are different ways of saying
the same thing.
In standard economic theory, the equilibrium of supply and
demand determines the interest rate on loans. But in real life,
when somebody takes out a secured loan, he must negotiate
two things: the interest rate and the collateral rate. A proper
theory of economic equilibrium must explain both. Standard
economic theory has not really come to grips with this problem
for the simple reason that it seems intractable: how can one
supply-equals-demand equation for a loan determine two
variables—the interest rate and the collateral rate? There is not
enough space to explain the resolution of this puzzle here, but
suffice it to say that ten years ago I showed that supply and
demand do indeed determine both. Moreover, the two
variables are influenced in the equilibration of supply and
demand mainly by two different factors: the interest rate
reflects the underlying impatience of borrowers, and the
collateral rate reflects the perceived volatility of asset prices and
the resulting uncertainty of lenders.3 Another factor
influencing leverage in the long run is the degree of financial
innovation. Since scarce collateral is often an important
limiting factor, the economy will gradually devise ways of
stretching the collateral, by tranching (so the same collateral
backs several loans) and pyramiding loans (so the same
3
In Geanakoplos (1997), I show how supply and demand can indeed
simultaneously determine the interest rate and the collateral rate. In
Geanakoplos (2003), I show how intertemporal changes in volatility lead to
changes in the equilibrium leverage over time as part of what I call a leverage
cycle. In Geanakoplos (1997) and Geanakoplos and Zame (2009), I emphasize
the scarcity of collateral and the role of tranching and pyramiding.
collateral can be used over and over to back loans backed by
loans).
Practitioners, if not economists, have long recognized the
importance of collateral and leverage. For a Wall Street trader,
leverage is important for two reasons. The first is that if he is
leveraged λ times, then a 1 percent change in the value of the
collateral means a λ percent change in the value of his capital.
(If the house in our example goes from $100 to $101, then after
selling the house at $101 and repaying the $80 loan, the investor
is left with $21 of cash on his $20 investment, a 5 percent return.)
Leverage thus makes returns riskier, either for better or for
worse. Second, a borrower knows that if there is no-recourse
collateral, so that he can walk away from his loan after giving up
the collateral without further penalty, then his downside is
limited. The most the borrower can lose on the house loan is his
$20 of cash, even if the house falls in value all the way to $0 and
the lender loses $80. No-recourse collateral thus effectively gives
the borrower a put option (to “sell” the house for the loan
amount). Recently, several commentators have linked leverage
to the crisis, arguing that if banks were not so leveraged in their
borrowing they would not have lost so much money when prices
went down, and that if homeowners were not so leveraged, they
would not be so far underwater now and so tempted to exercise
their put option by walking away from their house. Of course,
these two points are central to my own leverage cycle theory; I
discuss them in more detail later. But there is another, deeper
point to my theory that has so far not received as much attention,
which I think is the real story of leverage.
The main implication of my leverage cycle theory is that
when leverage goes up, asset prices go up, and when leverage
goes down, asset prices go down.4 For many assets, there is a
class of natural buyers or optimists who are willing to pay much
more for the asset than the rest of the public. They may be more
risk-tolerant. Or they may simply be more optimistic. Or they
may like the collateral (for example, housing) more.5 If they
can get their hands on more money through borrowing, they
will spend it on the assets and drive those asset prices up. If they
lose wealth, or lose the ability to borrow, they will be able to buy
less of the asset, and the asset will fall into more pessimistic
hands and be valued less.
It is useful to think of the potential investors arrayed on a
vertical continuum, in descending order according to their
willingness to buy, with the most enthusiastic buyers at the top
(see exhibit). Whatever the price, those at the top of the
continuum above a threshold will value the asset more and
become buyers, while those below will value it less and sell. The
4
Leverage is like more money in making prices go up, but, unlike money, it
affects only prices of goods that can serve as collateral; printing more money
tends to increase all prices, including those of food and other perishables.
5
Two additional sources of heterogeneity are that some investors are more
expert at hedging assets, and that some investors can more easily obtain the
information (like loan-level data) and expertise needed to evaluate the assets.
Natural Buyers Theory of Price
Natural buyers
Marginal buyer
Public
marginal buyer is the agent at the threshold on the cusp of
selling or buying and it is his opinion that determines the price.
The higher the leverage, the smaller the number of buyers at the
top required to purchase all the available assets. As a result, the
marginal buyer will be higher in the continuum and therefore
the price will be higher.
It is well known that a reduction in interest rates will
increase the prices of assets such as houses. It is less
appreciated, but more obviously true, that a reduction in
margins will raise asset prices. Conversely, if margins go up,
asset prices will fall. A potential homeowner who in 2006 could
buy a house by putting 3 percent cash down might find it
unaffordable to buy now that he has to put 30 percent cash
down, even if the Fed managed to reduce mortgage interest
rates by 1 percent or 2 percent. This has diminished the
demand for housing, and therefore housing prices. What
applies to housing applies much more to the esoteric assets
traded on Wall Street (such as mortgage-backed investments),
where the margins (that is, leverage) can vary much more
radically. In 2006, the $2.5 trillion of so-called toxic mortgage
securities could be bought by putting $150 billion down and
borrowing the other $2.35 trillion.6 In early 2009, those same
securities might collectively have been worth half as much, yet
a buyer might have had to put nearly the whole amount down
in cash. In Section 3.1, I illustrate the connection between
leverage and asset prices over the current cycle.
Economists and the Federal Reserve ask themselves every
day whether the economy is picking the right interest rates. But
one can also ask the question whether the economy is picking
the right equilibrium margins. At both ends of the leverage
cycle, it does not. In ebullient times, the equilibrium collateral
rate is too loose; that is, equilibrium leverage is too high. In bad
times, equilibrium leverage is too low. As a result, in ebullient
times asset prices are too high, and in crisis times they plummet
too low. This is the leverage cycle.
6
This number is calculated by applying the bank regulatory capital
requirement (based on bond credit rating) to each security in 2006 at its
2006 credit rating.
FRBNY Economic Policy Review / August 2010
103
The policy implication of the leverage cycle is that the Fed
could manage systemwide leverage, seeking to maintain it
within reasonable limits in normal times, stepping in to curtail
it in times of ebullience, and propping it up as market actors
become anxious, and especially in a crisis. To carry out this
task, of course, the Fed must first monitor leverage. The Fed
must collect data from a broad spectrum of investors, including
hitherto secretive hedge funds, on how much leverage is being
used to buy various classes of assets. Moreover, the amount of
leverage being employed must be transparent. The accounting
and legal rules that govern devices, such as structured
investment vehicles, that were used to mask leverage levels
must be reformed to ensure that leverage levels can be more
readily and reliably discerned by the market and regulators
alike. As we shall see, the best way to monitor leverage is to do
it at the security level by keeping track of haircuts on all the
different kinds of assets used as collateral, including in the repo
market and in the housing market. Also very useful, but less
important, is monitoring the investor leverage (or the debtequity ratio) of big firms.
The leverage cycle is no accident, but a self-reinforcing
dynamic. Declining margins, or, equivalently, increasing
leverage, are a consequence of the happy coincidence of
universal good news and the absence of danger on the horizon.
With markets stable and the horizon looking clear, lenders are
happy to reduce margins and provide more cash. Good, safe
news events by themselves tend to make asset prices rise. But
they also encourage declining margins, which in turn cause the
massive borrowing that inflates asset prices still more.
Similarly, when the news is bad, asset prices tend to fall on the
news alone. But the prices often fall further if the margins are
tightened. Sudden and dramatic increases in margins are relatively
rare. They seem to happen once or twice a decade. Bad news
arrives much more often than that, so it is not bad or even very bad
news alone that drastically raises margins. Bad news lowers
expectations, and, like all news, usually clarifies the situation.
Every now and then, bad news, instead of clarifying matters,
increases uncertainty and disagreement about the future. It is
this particular kind of “scary bad” news that increases margins.
For example, when an airline announces the plane will be ten
minutes late, the passengers start to worry the delay might be
an hour. When a bank announces a $5 billion loss, investors
worry that more losses might be on the way. In 2006, people
disagreed about whether losses from defaults on prime
mortgages would be 1/4 percent or 1/2 percent, and whether
losses on subprime mortgages would be 1 percent or 5 percent.
By contrast, after the scary news of 2007, people disagreed
about whether some subprime losses would be 30 percent or
80 percent. Even from their low, many lenders were afraid
many assets could lose even more value, maybe all their value.
The present became worse, and the future more uncertain.
104
Solving the Present Crisis and Managing the Leverage Cycle
The upshot of increased uncertainty and disagreement is
that margins go up drastically. Lenders are typically more
pessimistic than buyers. Otherwise, they too would be buying,
instead of lending. Even if the optimists are not worried much
about more losses, the lenders are, and they will demand high
margins. When the lenders are worried about 80 percent losses
from current levels, they will lend only if margins are at least
90 percent, or not lend at all.
As we have just witnessed, the rapid increase in margins
always comes at the worst possible time. Buyers who were
allowed to massively leverage their purchases with borrowed
money are forced to sell when bad news drives asset prices
lower. But when margins rise dramatically, more modestly
leveraged buyers are also forced to sell. Tightening margins
turn willing buyers into forced sellers, driving prices further
down. We enter the crisis stage I discuss below.
The dynamic of the leverage cycle cannot be stopped by
a tongue lashing of greedy Wall Street investors or overly
ambitious homeowners in the ebullient stage of the cycle, nor
by exhortations not to panic in the crisis stage. The cycle
emerges even if (in fact, precisely because) every agent is acting
rationally from his individual point of view. It is analogous to
a prisoner’s dilemma, where individual rationality leads to
collective disaster. The government must intervene.
The intervention becomes all the more necessary if agents
are irrationally exuberant and then irrationally panicked, or are
prone to short-sighted greed, or to the “keeping up with the
Jones” syndrome. If greedy investors want higher expected
returns, no matter what the risk, competition will force even
conservative fund managers to leverage more. For example, an
investor comes to a hedge fund and says, “the fund down the
block is getting higher returns.” The fund manager counters
that the competitor is just using more leverage. The investor
responds, “well whatever he’s doing, he’s getting higher
returns.” Pretty soon, both funds are leveraging more. Housing
prices can rise in the same way. When some families borrow a
lot of money to buy their houses, housing prices rise and even
conservative homeowners are forced to borrow and leverage so
they too can live in comparable houses, if keeping up with their
peers is important to them. At the bottom end, nervous
investors might withdraw their money, forcing hedge fund
managers to sell just when they think the opportunities are
greatest. However, of all the irrationalities that exacerbated this
leverage cycle, I would not point to these or to homeowners
who took out loans they could not really afford, but rather to
lenders who underestimated the put option and failed to ask
for enough collateral.
The observation that collateral rates are even more
important outcomes of supply and demand than interest rates,
and even more in need of regulation, was made over 400 years
ago. In The Merchant of Venice, Shakespeare depicted
accurately how lending works: one has to negotiate not just an
interest rate but the collateral level too. And it is clear which of
the two Shakespeare thought was the more important. Who
can remember the interest rate Shylock charged Antonio? But
everybody remembers the “pound of flesh” that Shylock and
Antonio agreed on as collateral. The upshot of the play,
moreover, is that the regulatory authority (the court)
intervenes and decrees a new collateral level—very different
from what Shylock and Antonio had freely contracted—
“a pound of flesh, but not a drop of blood.” The Fed, too, could
sometimes decree different collateral levels (before the fact, not
after, as in Shakespeare).
The modern study of collateral seems to have begun with
Kiyotaki and Moore (1997), Bernanke, Gertler, and Gilchrist
(1996, 1999), Holmstrom and Tirole (1997), Geanakoplos
(1997, 2003), and Geanakoplos and Zame (2009).7 Bernanke,
Gertler, and Gilchrist and Holmstrom and Tirole emphasize
the asymmetric information between borrowers and lenders as
the source of limits on borrowing. For example, Holmstrom
and Tirole argue that the managers of a firm would not be able
to borrow all the inputs necessary to build a project, because
lenders would like to see them bear risk, by putting their own
money down, to guarantee that they exert maximal effort.
Kiyotaki and Moore (1997) and Geanakoplos (1997) study the
case where the collateral is an asset such as a mortgage security,
where the buyer/borrower using the asset as collateral has no
role in managing the asset, and asymmetric information is
therefore not important. The key difference between Kiyotaki
and Moore and Geanakoplos (1997) is that in Kiyotaki and
Moore, there is no uncertainty, and so the issue of leverage as a
ratio of loan to value does not play a central role; to the extent
it does vary, leverage in Kiyotaki and Moore goes in the wrong
direction, getting higher after bad news, and dampening the
cycle. In Geanakoplos (1997, 2003), I introduce uncertainty
and solve for equilibrium leverage and equilibrium default
rates; I show how leverage could be determined by supply and
demand, and how under some conditions, volatility (or more
precisely, the tail of the asset return distribution) pins down
leverage. In Geanakoplos (2003), I introduce the leverage cycle
in which changes in the volatility of news lead to changes in
leverage, which in turn lead to changes in asset prices. This line
of research has been pursued by Gromb and Vayanos (2002),
Fostel and Geanakoplos (2008), Brunnermeier and Pedersen
(2009), and Adrian and Shin (forthcoming), among others.
7
Minsky (1986) was a modern pioneer in calling attention to the dangers of
leverage. But to the best of my knowledge, he did not provide a model or formal
theory. Tobin and Golub (1998) devote a few pages to leverage and the
beginnings of a model.
2.1 Investor Heterogeneity, Equilibrium
Leverage, Default, and Maturity
Without heterogeneity among investors, there would be no
borrowers and lenders, and asset prices would not depend on
the amount of leverage in the economy. It is interesting to
observe that the kind of heterogeneity influences the amount of
equilibrium leverage, and hence equilibrium asset prices, and
equilibrium default.
When investors differ only in their optimism about future
events in a one-dimensional manner, then the equilibrium
leverage will consist of the maximum promise that does not
permit default.8 For example, suppose an asset will be worth
either 1 or .2 next period. Suppose further that risk-neutral
investors differ only in the probability h that they assign to the
outcome being 1. The most optimistic investor h = 1 is sure that
the asset will be worth 1, and the most pessimistic investor h = 0
is sure the asset will be worth .2. At any asset price p, the
investors with h big enough that h*1 + (1-h)*(.2) > p will want
to buy the asset, while the rest will want to sell the asset. The
buyers with high h will want to borrow money in order to get
their hands on what they regard as cheap assets, while the
sellers with low h will not need the money and so will be willing
to lend. How much will the borrowers be able to promise using
the asset as collateral, assuming the promise is not contingent
on the state? The answer is .2, precisely the maximum promise
that does not lead to default in either state.9
Thus, when the heterogeneity stems entirely from onedimensional differences in opinion, equilibrium leverage
entails no default. A consequence of this is that the loans will be
very short term. The longer the maturity of the loan, the more
that can go wrong in the meantime, and therefore the smaller
the loan amount can be if it avoids any chance of default.
Investors who want to borrow large amounts of money will be
driven to borrow very short term. The repo market displays
these characteristics of short, one-day loans, on which there is
almost never any default, even in the worst of crises.
Much the same analysis holds when investors differ only in
their risk aversion. For the most risk-averse investors, an asset
that pays 1 or .2 will be regarded as too dangerous, while
8
See Geanakoplos (2003).
At first glance, it would seem that the most optimistic buyers might be willing
to promise, say, .3 in both states, in order to get more money today to invest in
a sure winner of an asset. But since this promise will deliver .3 in the good state
but only .2 in the bad state (assuming no-recourse collateral), the lenders will
not want to pay much for this debt: this risky debt is very much like the asset
they do not want to hold, and so they will pay very little more for it than the
(.2,.2) promise, where (g,b) denotes a payoff of g if the good state occurs and b
if the bad state occurs. Since the borrowers would have to give up .3 > .2 in the
state they think is likely to occur, they will choose to use their scarce collateral
to back the (.2,.2) promise instead of the (.3,.3) promise.
9
FRBNY Economic Policy Review / August 2010
105
investors with greater risk tolerance will find it attractive at the
right price. These risk-tolerant investors will leverage their
purchases, by borrowing money to buy the asset, using it as
collateral for their loan. Once again, the equilibrium leverage
will rise to the point that the promises made will be (.2,.2) but
no more (see footnote 9 for an explanation of notation). To be
more concrete, suppose contrary to the previous case, that all
the agents regard the outcomes 1 and .2 as equally likely. But
suppose that untraded endowments rise and fall together with
the asset payoffs. Then risk-averse agents on the margin will
regard an extra penny when the asset pays 1 as less valuable
than an extra penny when the asset pays .2; on the margin, they
would prefer a penny when the asset pays .2. Hence, they will
behave as if they regarded the payoff of 1 as less likely, exactly
the same way the pessimists behaved, despite having the same
beliefs as the risk-tolerant agents. Equilibrium leverage with
heterogeneous risk aversion becomes the same as with
heterogeneous beliefs.
The situation changes when some investors simply like
owning the asset for its own sake in the period they buy it, such
as when a homeowner likes living in the house. A similar
situation arises if a producer can get more output from the
asset than can be recovered if the lender takes it over.
Somewhat surprisingly, in these cases the equilibrium leverage
might be to promise (1,1) even when the asset will only deliver
(1,.2) with probabilities everyone agrees on. If there are
multiple states, and a cost of seizing the collateral, then the
equilibrium promise will be somewhere between the
maximum and minimum delivery. Contrary to the previous
two cases, equilibrium leverage will involve a distinctly positive
probability of default. Furthermore, in order to avoid the
default costs of seizing the collateral, the equilibrium loans will
be longer term, as in the mortgage market, where we see
defaults and long-maturity loans.
scary kind that does not clarify but obscures the situation and
produces widespread uncertainty and disagreement about
what will happen next. Suddenly, lenders increase the margins
and thus deliver the fatal blow. At that point, even modestly
leveraged buyers are forced to sell. Prices plummet. The assets
eventually make their way into hands that will take them only
at rock-bottom prices.
During a crisis, margins can increase 50 percent overnight,
and 100 percent or more over a few days or months. New
homeowners might be unable to buy, and old homeowners
might similarly be unable to refinance even if the interest rates
are lowered. But, holding long-term mortgages, at least they do
not have to put up more cash. For Wall Street firms, the
situation is more dire. They often borrow for one day at a time
in the repo market. If the margins double the next day, then
they immediately have to double the amount of cash they hold
for the same assets. If they do not have all that cash on hand,
they will have to sell the assets. This is called deleveraging.
All this would happen even if traders were completely
rational, processing information dispassionately. When we add
the possibility of panic and the turmoil created by more and
more bankruptcies, it is not surprising to see lending
completely dry up.
2.3 The Aftermath of the Crisis
After the crisis ends, many businesses and individuals will be
broke and unemployed. Parts of the economy will be disrupted,
and some markets may be on the verge of shutting down. The
government will then face the choice of who to assist, and at
what cost. This assistance will typically be very inefficient,
causing further losses to economic productivity. Doubts about
which firms will survive will create more uncertainty,
contributing to a difficult lending environment.
2.2 The Crisis Stage
The crisis stage of the leverage cycle always seems to unfold in
the same way. First there is bad news. That news causes asset
prices to fall based on worse fundamentals. Those price
declines create losses for the most optimistic buyers, precisely
because they are typically the most leveraged. They are forced
to sell off assets to meet their margin restrictions, even when
the margins stay the same. Those forced sales cause asset prices
to fall further, which makes leveraged buyers lose more. Some
of them go bankrupt. And then typically things shift: the loss
spiral seems to stabilize—a moment of calm in the hurricane’s
eye. But that calm typically gives way when the bad news is the
106
Solving the Present Crisis and Managing the Leverage Cycle
2.4 What Is So Bad about the Leverage Cycle?
The crisis stage is obviously bad for the economy. But the
leverage that brings it on stimulates the economy in good
times. Why should we think the bad outweighs the good? After
all, we are taught in conventional complete-markets economics
that the market decides best on these types of trade-offs. In
Geanakoplos (2010), I discuss eight reasons why the leverage
cycle may nevertheless be bad for the economy. The first three
are caused by the large debts and numerous bankruptcies that
occur in big leverage cycles.
First, optimistic investors can impose an externality on the
economy if they internalize only their private loss from a
bankruptcy in calculating how much leverage to take on. For
example, managers of a firm calculate their own loss in profits
in the down states, but sometimes neglect to take into their
calculations the disruption to the lives of their workers when
they are laid off in bankruptcy. If, in addition, the bankruptcy
of one optimist makes it more likely in the short run that other
optimists (who are also ignoring externalities) will go
bankrupt, perhaps starting a chain of defaults, then the
externality can become so big that simply curtailing leverage
can make everybody better off.
Second, debt overhang destroys productivity, even before
bankruptcy, and even in cases when bankruptcy is ultimately
avoided. Banks and homeowners and others who are
underwater often forgo socially efficient and profitable
activities. A homeowner who is underwater loses much of the
incentive to repair a house, even if the cost of the repairs is less
than the gain in value to the house, since increases in the value
of the house will not help him if he thinks he will likely be
foreclosed eventually anyway.10
Third, seizing collateral often destroys a significant part of
its value in the process. The average foreclosure of a subprime
loan leads to recovery of only 25 percent of the loan, after all
expenses and the destruction of the house are taken into
account, as I discuss later. Auction sales of foreclosed houses
usually bring 30 percent less than comparable houses sold by
their owners.
The next four reasons stem from the swings in asset prices
that characterize leverage cycles. A key externality that
borrowers and lenders in both the mortgage and repo markets
do not recognize is that if leverage were curtailed at the high
end of the leverage cycle, prices would fall much less in the
crisis. Foreclosure losses would then be less, as would
inefficiencies caused by agents being so far underwater. One
might argue that foreclosure losses and underwater
inefficiencies should be taken into account by a rational
borrower and lender and be internalized: it may be so
important to get the borrower the money, and the crisis might
ex ante be so unlikely, that it is “second best” to go ahead with
the big leverage and bear the cost of the unlikely foreclosure.
But that overlooks the pecuniary externality: by going into
foreclosure, a borrower lowers housing prices and makes it
more likely that his neighbor will do the same.
Fifth, asset prices can have a profound effect on economic
activity. As James Tobin argues with his concept of Q, when the
prices of old assets are high, new productive activity, which often
involves issuing financial assets that are close substitutes for the
old assets, is stimulated. When asset prices are low, new activity
might grind to a halt.11 When asset prices are well above the
10
See Myers (1977) and Gyourko and Saiz (2004).
complete-markets price, because of the expectation by the
leveraged few that good times are coming, a huge wave of
overbuilding usually results. In the bad state, this overbuilding
needs to be dismantled at great cost and, more importantly, new
building nearly stops. To make the point a bit more dramatically,
very high leverage means that the asset prices are set by a small
group of investors. If agent beliefs are heterogeneous, why
should the prices be determined entirely by the highest outliers?
In the current crisis, as I observed earlier, the $2.5 trillion of toxic
mortgage securities were purchased with about $150 billion in
cash and $2.35 trillion in loans. As of 2006, just two men, Warren
Buffet and Bill Gates, between them had almost enough money
to purchase every single toxic mortgage security in the whole
country. Leverage allows the few to wield great influence on
prices and therefore on what is produced.12
Sixth, a large group of small businesspeople who cannot buy
insurance against downturns in the leverage cycle can easily sell
loans to run their businesses or pay for their consumption in
good times at the height of the leverage cycle, but have a hard
time at the bottom. Government policy may well have the goal of
protecting these people by smoothing out the leverage cycle.13
Seventh, the large fluctuations in asset prices over the
leverage cycle lead to massive redistributions of wealth and
changes in inequality. When leverage λ = 30, there can be wild
swings in returns and losses. In the ebullient stage, the
optimists become rich as their bets pay off, while in the down
states, they might go broke. Inequality becomes extreme in
both kinds of states.14
The eighth problem with the leverage cycle is caused by the
inevitable government responses to the crisis stage. In an effort
to mitigate the crisis, the government often intervenes in
inefficient ways. In the current crisis, the government is
supporting the financial sector by holding the federal funds rate
near zero. The government’s foreclosure prevention efforts
have created financial subsidies for households that opt not to
move, which can create inefficiencies in labor market
adjustment.15 Government bailouts, even if they were all for
the public good, cause resentment from those who are not
bailed out. The agents in the economy do not take into account
that by leveraging more and putting the economy at greater
11
See Tobin and Golub (1998).
Standard economics does not really pay any attention to the case where
agents have different beliefs, and median beliefs are closer to the truth than
extreme outliers.
13
Here I rely on Tobin’s Q and the absence of insurance markets. The small
businessmen cannot insure themselves against the crisis stage of the leverage
cycle. In conventional complete-markets economics, they would be able to
buy insurance for any such event. Geanakoplos and Polemarchakis (1986)
offer a proof that when insurance markets are missing, there is almost always
a government intervention in the existing markets that will make everyone
better off.
14
This is a purely paternalistic reason for curtailing leverage.
15
See Ferreira et al. (forthcoming).
12
FRBNY Economic Policy Review / August 2010
107
risk, they create more inefficient government interventions.
And of course, the expectation of being assisted by the
government, should things go wrong, causes many agents to
be more reckless in the first place.16
Chart 1
Housing Leverage Cycle
Margins Offered (Down Payments Required) and Home Prices
Case-Shiller
national HPI
Down payment
for mortgage (percent)
200
0
3. The Leverage Cycle of 2000-09
Fits the Pattern
3.1 Leverage and Prices
By now, it is obvious to everybody that asset prices soared from
1999 (or at least after the disaster period that began September 11,
2001) to 2006, and then collapsed from 2007 to 2009. My thesis
is that this rise in prices was accompanied by drastic changes in
leverage, and was therefore just part of the 1999-2006 upswing
in the leverage cycle after the crisis stage in 1997-98 at the end
of the last leverage cycle. I do not dispute that irrational
exuberance and then panic played a role in the evolution of
prices over this period, but I suggest that they may not be as
important as leverage; certainly, it is harder to regulate animal
spirits than it is leverage.
Let us begin with the housing bubble, famously documented
by Robert Shiller. In Chart 1, I display the Case-Shiller national
housing index for 2000-09. It begins at 100 in 2000:1, reaches
190 in 2006:2, and falls to 130 by 2009:1, as measured on the
right vertical axis. But I superimpose on that graph a graph of
leverage available to homeowners each month. This is
measured on the left vertical axis and labeled “Down payment
for mortgage,” which is 100 percent minus the loan-to-value
(LTV) ratio. To compute this, I begin by looking house by
house each month from 2000-09 at the ratio of all the
outstanding mortgage loans (usually a first and sometimes a
second lien) to the appraised value of the house at the moment
a first mortgage was issued for every subprime and alt-A house
available in the First American CoreLogic LoanPerformance
Data Base. I then average over the 50 percent houses with the
highest LTV levels.17 In this way, I obtain a robust estimate of
leverage offered to homeowners. By leaving out the bottom
50 percent, I ignore homeowners who clearly chose to leverage
less than they could have, and by including all homes in the top
50 percent, I ensure that the leverage measure was really
available and not just a special deal for a few outliers. If
anything, my numbers underestimate the offered leverage.18
16
This mechanism has been formalized in Farhi and Tirole (2009).
These data were compiled and analyzed by the research team at the hedge
fund Ellington Capital Management.
17
108
Solving the Present Crisis and Managing the Leverage Cycle
Average down payment
for 50 percent lowest
down payment, subprime/
alt-A borrowers
5
180
160
(Left scale)
10
140
Case-Shiller national
home price index (HPI)
15
120
(Right scale)
100
20
2000
02
04
06
08
09
Sources: First American CoreLogic LoanPerformance Data Base;
Ellington Capital Management.
Notes: The down payment axis has been reversed, because lower down
payment requirements are correlated with higher home prices. For every
alt-A or subprime first-lien loan origination from 2000:1 to 2008:1, the
down payment percentage was calculated as appraised value (or sale price,
if available) minus total mortgage debt, divided by appraised value. For
each quarter, the down payment percentages were ranked from highest
to lowest, and the average of the bottom half is shown. This number is
an indicator of the down payment required; clearly, many homeowners
put down more than they had to, which is why the top half is dropped
from the average. A 13 percent down payment in 2000:1 corresponds
to leverage of about 7.7, and a 2.7 percent down payment in 2006:2
corresponds to leverage of about 37. Subprime/alt-A issuance ended
in 2008:1.
It is striking how correlated prices and leverage are, rising
and then falling together. Especially noteworthy is that leverage
peaks in 2006:2, with 2.7 percent down, exactly when housing
prices peak, and heads down much faster than housing prices.
In Chart 2, I present the history of the J.P. Morgan AAA
prime floater mortgage index from about 2000 to 2009. The
index is measured on the right vertical axis. The prime
mortgages underlying the bonds in the index were taken out by
investors with pristine credit ratings, and the bonds are also
protected by some equity in their deals. For most of its history,
this index stays near 100, but starting in early 2008, it falls
rapidly, plummeting to 60 in early 2009. The cumulative losses
on these prime loans even today are still in the single digits; it is
hard to imagine them ever reaching 40 percent (which would
mean something like 80 percent foreclosures with only
50 percent recoveries). It is of course impossible to know what
people were thinking about potential future losses when the
index fell to 60 in late 2008 and early 2009. My hypothesis is
that leverage played a big role in the price collapse.
18
At the peak of nonprime lending in mid-2005, these loans represented
45 percent of the flow of new mortgage borrowing (correspondence with
editors).
Chart 2
Chart 3
Securities Leverage Cycle
VIX Index
Margins Offered and AAA-Rated Securities Prices
Margin down payment required
to purchase securities (percent)
90
80
Price
0
10
20
30
40
50
100
70
90
50
80
40
60
Average margin on portfolio of collateralized
mortgage obligations rated AAA at issuance
(Left scale)
30
Estimated
average margin
70
(Left scale)
60
Prime fixed prices
70
80
1998 99
60
(Right scale)
50
00
01
02
03
04
05
06
07
10
0
1990
92
94
96
98
00
02
04
06
08
10
08 09
Sources: Ellington Capital Management; J.P. Morgan.
Notes: The chart represents the average margin required by dealers
on a hypothetical portfolio of bonds subject to certain adjustments
described below. The margin axis has been reversed, because lower
margins are correlated with higher prices. The portfolio evolved over
time, and changes in average margin reflect changes in composition
as well as changes in margins of particular securities. In the period
following August 2008, a substantial part of the increase in margins
is attributable to bonds that could no longer be used as collateral
after being downgraded, or for other reasons, and hence count as
100 percent margin.
On the left vertical axis, I give the loan-to-value, or,
equivalently, the down payment or margin, offered by Wall
Street banks to the hedge fund Ellington Capital Management
on a changing portfolio of AAA mortgage bonds.19 As I noted
earlier, it is astonishing that the Fed itself does not have such
historical data. Fortunately, the hedge fund Ellington, which I
have worked with for the past fifteen years, does keep its own
data. The data set is partly limited in value by the fact that the
data were only kept for bonds Ellington actually followed, and
these changed over time. Some of the variation in average
margin is due to the changing portfolio of bonds, and not to
changes in leverage. But the numbers, while not perfect,
provide substantial evidence for my hypothesis and tell a
fascinating story. In the 1997-98 emerging markets/mortgage
crisis, margins shot up, but quickly returned to their previous
levels. Just as housing leverage picked up over the period after
1999, so did security level leverage. Then in 2007, leverage
dramatically fell, falling further in 2008, and leading the drop
in security prices. Very recently, leverage has started to increase
again, and so have prices.
19
20
These are the offered margins and do not reflect the leverage chosen by
Ellington, which since 1998 has been drastically smaller than what was offered.
Chart 3 displays the history of implied volatility for the S&P
500, called the VIX index. Volatility in equities is by no means
a perfect proxy for volatility in the mortgage market, but it is
striking that the VIX reached its peak in 2008 at the crisis stage
of the current leverage cycle, and reached a local peak in 1998
at the bottom of the last leverage cycle in fixed-income
securities. The VIX also shot up in 2002, but there is no
indication of a corresponding drop in leverage in the Ellington
mortgage data.
3.2 What Triggered the Crisis?
The subprime mortgage security price index collapsed in
January 2007. The stock market kept rising until October 2007,
when it too started to fall, losing eventually around 57 percent
of its value by March 2009 before rebounding to within
27 percent or so of its October peak in January 2010. What, you
might wonder, was the cataclysmic event that set prices and
leverage on their downward spiral?
The point of my theory is that the fall in prices from scary
bad news is naturally going to be out of proportion to the
significance of the news, because the scary bad news
precipitates and feeds a plunge in leverage. A change in
volatility, or even in the volatility of volatility, is enough to
prompt lenders to raise their margin requirements. The data
show that that is precisely what happened: margins were raised.
But that still begs the question, what was the news that
indicated volatility was on the way up?
One obvious answer is that housing prices peaked in mid2006, and their decline was showing signs of accelerating in the
beginning of 2007. But I do not wish to leave the story there.
Housing prices are not exogenous; they are central to the
leverage cycle. So why did they turn in 2006?
FRBNY Economic Policy Review / August 2010
109
3.3 Why Did Housing Prices Start to Fall?
Many commentators have traced the beginning of the
subprime mortgage crisis to falling housing prices. But they
have not asked why housing prices started to fall. Instead, they
have assumed that housing prices themselves, fueled on the
way up by irrational exuberance and on the way down by a
belated recognition of reality, were the driving force behind the
economic collapse.
I see the causality going in the other direction, starting with
the turnaround in the leverage cycle. The leverage cycle was of
course greatly exacerbated by the terrible consequences of
falling housing prices, which then fed back to cause further
housing declines.
As I hope I have made clear, in my view housing prices
soared because of the expansion of leverage. Greater leverage
enabled traditional buyers to put less money down on a bigger
house, and therefore pushed up housing prices. It also enabled
people to buy houses who previously did not have enough cash
to enter the market, pushing housing prices up even further.
There is, however, a limit on how much leverage can
increase, and on how many new people can enter the market.
Though negative amortizing loans pushed the envelope, no
money down is a natural threshold beyond which it is hard to
move. And as more and more households entered the market
with less and less money down, lenders began to become
apprehensive that these people were less reliable and more
inclined to exercise their put option to walk away from the
house if housing prices fell. The rapidly expanding supply of
new housing demand, fueled by access to easy mortgages,
began to slow for completely rational reasons, not because of a
sudden pricking of irrational exuberance. This naturally led to
a peak in housing prices by 2006:2. But this does not explain
why housing prices should steeply decline. Indeed, over the
next two quarters, prices and leverage waffled, both moving
slightly in a negative direction: During the last half of 2006,
housing down payment requirements rose slightly, from
2.7 percent to 3.2 percent, and prices fell slightly, by
1.8 percent.
At that point, bad news appeared in the securities market in
the form of rising delinquencies. Charts 4 and 5 show losses
and delinquencies of Countrywide deals by vintage.20 (These
deals are fairly representative of the whole subprime market.)
One can see in Chart 4 that by January 2007, losses for the
2005 vintage were just 0.2 percent and losses for the 2006
vintage were nonexistent. But the 2005 and 2006 delinquencies
displayed in Chart 5 were already approaching 5 percent, more
than double those of previous vintages. More disturbing, they
showed no signs of leveling off. This is precisely the kind of
scary news that creates wide uncertainty about what might
20
Data were provided by Ellington Capital Management.
110
Solving the Present Crisis and Managing the Leverage Cycle
Chart 4
Cumulative Loss of Original Balance
Cumulative loss (percent)
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Beginning of subprime crisis
2004
2003
2005
2006
2007
2003
04
05
06
07
Source: Ellington Capital Management.
Chart 5
Delinquencies on Original Balance
OTS delinquent 90+ (percent)
16
Beginning of subprime crisis
14
2007
12
10
8
2006
6
2005
4
2003
2
0
2004
2003
04
05
06
07
Source: Ellington Capital Management.
happen next. With that new information, how much
extrapolation should a buyer from 2006 have made in his
expectations of losses and delinquencies going forward?
The ABX index for 2006 vintage subprime bonds began to
fall in November 2006 with the smallest trickle of bad news
about homeowner delinquencies, then spiked downward in
January 2007 after the year-end delinquency report (Chart 6).
This price drop of 2006 BBB bonds to below 80 implied that the
market was suddenly anticipating huge losses on subprime
deals on the order of 10 percent. Recall that for a pool of
mortgages to lose 9 percent or 10 percent of its value, the
market must anticipate that something like 30 percent of the
homeowners will be thrown out of their houses, with 30 percent losses on the mortgage on each home sold (30 percent x
30 percent = 10 percent). This expectation turned out to be not
pessimistic enough, but at that time it was a heroic
extrapolation from the observed delinquencies of less than
5 percent.21
Chart 6
ABX Index
ABX-HE:
Percent
AAA 06-1
AAA 06-2
AA 06-1
AA 06-2
A 06-1
A 06-2
BBB 06-1
BBB 06-2
BBB- 06-1
BBB- 06-2
AAA 07-1
AA 07-1
A 07-1
BBB 07-1
BBB- 07-1
AAA 07-2
AA 07-2
A 07-2
BBB 07-2
BBB- 07-2
120
100
80
60
40
20
0
2006
07
08
09
10
My contention is that this sudden drop in prices, and the
further price declines later, were not simply the result of a drop
in expected payoffs (that is, in fundamentals) by the same old
buyers, but also the result of a change in the marginal buyer. A
critical new downward force entered the market for mortgage
securities. Standardized credit default swaps (CDS) on
mortgage bonds were created for the first time in late 2005, at
the very height of the market. The volume of CDS expanded
rapidly throughout 2006 and especially in 2007 (Chart 7).22
A CDS is an insurance contract for a bond. By buying the
insurance, the pessimists for the first time could leverage their
negative views about bond prices and the houses that backed
them. Instead of sitting out of the subprime securities market,
pessimists could actively push bond prices down. Their
purchase of insurance is tantamount to the creation of more
(“synthetic”) bonds; naturally, the increase in supply pushed
the marginal buyer down and thus the price down.
In January 2007, after the dramatic fall in BBB subprime
mortgage prices, housing prices were still only 1.8 percent off
their peak. Though the peak of the housing market preceded
the peak of the securities market, the collapse in securities
prices preceded the significant fall in housing prices. Thus, in
my view the trigger for the downturn in bonds was the bad
news about delinquencies and the concurrent creation of the
standardized CDS market in subprime mortgage indexes,
which then spilled over into the housing market.
The downward pressure on bond prices from credit default
swaps and worrisome delinquency numbers meant that new
securitizations became more difficult to underwrite.
Securitizers of new loans looked for better loans to package in
order to continue to back bonds worth more than the loan
amounts they had to give homeowners. They asked for loans
with more collateral. As Chart 7 shows, from 2006:4 to 2007:4,
21
50
The collapse of the ABX index in January 2007 is a powerful illustration of
the potency of market prices to convey information. This first market crash
should have been enough to alert our government to the looming foreclosure
disaster, but three years later we still have not taken decisive action to mitigate
foreclosures.
22
Chart 7 is derived from data provided in “ISDA Market Survey: Historical
Data,” available at www.isda.org/statistics/historical/html. Unfortunately, it
includes all CDS, not just CDS on mortgages. The data on mortgage CDS seem
difficult to find, since these CDS were traded bilaterally and not on an
exchange. It seems very likely to me that the mortgage CDS increased even
more dramatically from 2004-05 to 2006-07.
Chart 7
Volume of Credit Default Swaps
Trillions of dollars
70
60
40
30
20
10
0
2000
01
02
03
04
05
06
07
08
09
Source: “ISDA Market Survey: Historical Data.”
FRBNY Economic Policy Review / August 2010
111
the required down payment on houses rose dramatically from
3.2 percent to 15.9 percent (equivalently, LTV fell from 96.8
percent to 84.1 percent). This meant that potential new
homeowners began to be closed out of the market, which of
course reduced home prices. In that same period, housing
prices began to fall rapidly, declining by 8.5 percent.
But more insidiously, the desire by lenders to have more
collateral for each dollar loaned kept homeowners from
refinancing because they simply did not have the cash: given
the drop in the permissible LTV ratio, and the fall in housing
prices, they suddenly needed to put down 25 percent of their
original loan in cash to refinance. Refinancing virtually stopped
overnight. Until 2007, subprime bondholders could count on
70 percent or so of subprime borrowers refinancing by the end
of their third year.23 These homeowners began in pools that
paid a very high rate of interest because of their low credit
rating. But after two years of reliable mortgage payments, they
would become eligible for new loans at better rates, which they
traditionally took in vast numbers. Of course, a prepayment
means a full payment to the bondholder. Once refinancing
plummeted and this sure source of cash disappeared, the bonds
became much more at risk and their prices fell more. Margins
on bonds began to tighten.
Mortgagees who had anticipated being able to refinance
were trapped in their original loans at high rates; many
subsequently became delinquent and entered foreclosure.
Foreclosures obviously lead to forced sales and downward
pressure on housing prices. And falling home prices are a
powerful force for further price reductions, because when
house values fall below the loan amount, homeowners lose the
incentive to repay their loans, leading to more defaults,
foreclosures, and forced selling. All this leads back to falling
security prices and tighter margins on securities.
The feedback from falling security prices to higher margins
on housing loans to lower house prices and then back to tougher
margins on securities and to lower security prices and then back
again to housing is what I call “the double leverage cycle.”
4. Why this Leverage Cycle Is the
Worst since the Great Depression
Every leverage cycle has the same broad features. The crisis
stage of every leverage cycle is bad. But the current crisis is far
worse than the crises we saw in the two previous leverage cycles.
There are a number of reasons why this cycle is worse than all
previous cycles since the Depression, but the unifying theme
23
Seventy-four percent of all subprime loans issued in or before 2004 had
refinanced by the end of their third year, according to the First American
CoreLogic LoanPerformance Data Base.
112
Solving the Present Crisis and Managing the Leverage Cycle
behind all of them is a failure to put up enough collateral to
back promises.
4.1 Securities Leverage Got Higher then
Fell Farther than Ever Before
In this cycle, leverage on traditional collateralizable assets
increased to more than the highs from the previous cycle. That
can be seen in the history of one mortgage hedge fund’s
margins (haircuts) over the last eleven years (Chart 2). Note in
the chart that before the crisis of 1997-98 that ended the last
leverage cycle, leverage was about 10 to 1 (margins were about
10 percent). During the 1998 crisis, margins jumped to
40 percent, staying there about two months, before returning
to their previous levels of 10 percent. In the “great moderation”
in the nine years afterward, when volatility got very low,
leverage increased from about 10 to 1 to about 20 to 1 (the
margins fell from 10 percent to 5 percent).
Beginning in 2007 (after reaching its peak in 2006), leverage
collapsed, with margins going from 5 percent to 70 percent on
average. Two years after the collapse, leverage was still low,
whereas in 1998 the crisis was all over in two months.
The most dramatic change in margins has come from assets
that were rated AAA, and that have been, or are about to be,
downgraded. Previously, one could borrow 90 or even 98.4 cents
on a dollar’s worth of AAA assets, and now one cannot borrow
anything at all with these assets as collateral. According to
Moody’s, AAAs are supposed to have a 1 in 100 risk of default
over a ten-year period.24 We are now seeing over 50 percent of all
alt-A and subprime AAA bonds partially defaulting, and we will
see virtually 100 percent of all AAA collateralized debt
obligations (CDOs) partially default. Even when some assets
have little or no chance of losing more than a few percent of their
value, the market no longer trusts the AAA rating, and lenders
will not lend anything on them.
In the run-up to the present crisis, financial innovation
enabled many new kinds of assets to become usable as
collateral. Thus, even if margins had not declined on old
collateral, the leverage of the economy as a whole would have
increased because there was new borrowing backed by
previously unusable collateral, which brings us to pooling and
securitization.
The process of pooling and securitization has been a crucial
source of new collateral and increased leverage. Imagine a
single subprime mortgage loan. Even in the days when it was
believed that the expected loss from such a mortgage was
between 1 percent and 4 percent, people still recognized that
24
See Backman and O’Connor (1995).
there was a nontrivial chance of a much bigger loss on a single
loan. Lenders, inherent pessimists, would not have considered
lending using a single subprime mortgage as collateral. But
now consider a pool of subprime mortgages from around the
country. If one believed that the loans were independent, so
that a housing price decline in Detroit did not imply a housing
price decline in California, then on a big enough pool of loans,
the chance for more than 30 percent default might be
considered less than 1 in 10,000. Even a very pessimistic lender
who believed in a 4 percent expected loss per loan would be
willing to lend 70 percent of the value of the entire pool,
provided that he got paid before anyone else. Thus, a buyer of
the pool of mortgages could imagine borrowing 70 percent of
their collective value, when it would have been impossible to
borrow anything on the individual loans.
Securitization took this borrowing on pools one step further
by converting the loans into long-term loans. The underwriter
of the pool typically issued different bonds, whose payments
depended on the homeowners’ payments on their loans.
Consider, for example, a bond structure with just two
“tranches” of bonds. The senior tranche might pay interest
slightly above the riskless government rate on the best
70 percent of the loans. As long as losses on the pool are below
30 percent, the senior tranche holder continues to get paid his
interest and eventually his principal. The junior bondholder
receives what is left from the pool after the senior bondholder
is paid. The whole securitized structure can be interpreted as if
the buyer of the junior piece actually bought the whole pool,
using a long-term loan from the buyer of the senior piece,
collateralized by the whole pool. Once one understands the
juniors as effectively borrowing from the seniors, it becomes
clear how the rapid spread of securitization over the last thirty
years, but especially over the last ten years, dramatically
increased the leverage in the system.
Another factor that dramatically increased overall leverage
in the system is the credit default swap, which I discuss shortly.
4.2 Housing and the Double Leverage Cycle
Leverage on houses got to be much higher in this leverage cycle.
In the recent leverage cycles, ending in 1994 and 1998,
homeowner leverage did not get remotely as high as it did in the
recent cycle. In 2006, many homeowners were borrowing with
basically no money down, or as little as 3 percent, as we saw in
Chart 1.25 New mortgages like option arms were invented,
which abetted this mad rush to loan homeowners all or nearly
all of the purchase price. Whereas previous cycles’ leverage
25
See Haughwout, Peach, and Tracy (2008) for details on leverage used for
nonprime borrowers from 2001 to 2007.
involved many financial institutions, it never involved such a
large fraction of the general population. When housing prices
and securities prices fell, millions of homeowners as well as
many of the most venerable financial institutions in America
found themselves underwater, owing more money than the
value of their assets.
Thus, the current cycle is really a double leverage cycle: not
only are the mortgage securities subject to the leverage cycle,
but their “fundamental” cash flows (namely, homeowner
mortgage payments) are also subject to the leverage cycle.
These two cycles feed off each other. When margins are raised
on homeowners, it becomes more difficult to get a new
mortgage and home prices fall, jeopardizing mortgage
securities backed by houses. But more importantly, it becomes
more difficult for homeowners to refinance their old loans,
putting these loans and the securities they back in much more
jeopardy of defaulting. Similarly, when margins on securities
are raised and their prices fall, then in order to sell the securities
for higher prices, underwriters demand better underlying
mortgages, that is, more money down from home buyers.
4.3 Credit Default Swaps
The current cycle has been more violent because of the
standardization/creation of the derivative credit default swap
market for mortgages in 2005, just at the top of the leverage
cycle. One reason for the abruptness of the fall is that CDS
allowed pessimists to leverage at just the worst time. Once CDS
emerged, they were bound to put downward pressure on
prices, because they allowed pessimists to express their views
for the first time and indeed leverage those views. Had the CDS
market for mortgages been around from the beginning, asset
prices might never have gotten so high. But their appearance at
the very top of the cycle guaranteed that there would be a fall.
Not only did CDS allow pessimists to leverage for the first
time, it also allowed them to leverage more than optimists.
When a bond trades near 100, but there is a perceptible chance
of a big drop in price, then in a rational world the writer of
insurance is almost always going to be asked to put up much
more collateral than the buyer of insurance, because his
potential liability is so high. A small group of pessimists can
therefore have an outsized negative impact on prices by
leveraging their CDS positions, since traders on the other side
will need far more capital to offset those positions.
A second reason why CDS made the fall much worse is that
in practice they allowed optimists to leverage even more than
they had before. To the extent that CDS did not lower prices
before any bad news, it was because leveraged optimists
increased their leverage by taking the other side of the CDS, on
FRBNY Economic Policy Review / August 2010
113
top of their leveraged purchases of the underlying assets. But
this made the crash much bigger once the bad news hit. CDS is
a kind of insurance market for bond defaults, but instead of
cushioning losses, it made them much worse because often the
buyers of the bonds did not buy the insurance, they sold it.26
One might mistakenly think that CDS should just wash out.
In other words, for every dollar lost on the insurance, there
should be a dollar gained by the recipient. But the optimistic
writers of insurance are very different from the pessimistic
buyers of insurance. When the bad news hits, the former lose
and must reduce their purchases of assets; the latter gain, but
still will not buy the assets. Writers of CDS insurance expose
the economy to the same problems of excessive leverage I
described earlier.
This brings us to the question of just how much leverage one
could actually obtain via the CDS market. Imagine a bond with
$100 face value that is trading for $98, and imagine a CDS
insurance contract promising $1 for every $1 the bond defaults.
The $98 price suggests expected losses to the insurance writer of
$2. If the bond rises to $99, the seller of insurance effectively
makes a dollar and if the bond price falls to $97, the insurance
writer effectively loses $1. Thus, writing insurance is tantamount
to owning the bond. One can therefore compare the collateral a
writer of CDS insurance had to put up with the down payment a
buyer of the bond would have had to make to see where leverage
was higher. It now appears that leverage was higher with CDS.
Many firms, like AIG, were allowed to write CDS insurance with
little or no initial margin. If enough collateral had been put up by
AIG, there would have been no reason to bail it (or more to the
point, all its counterparties) out.
The failure of some buyers of CDS insurance to insist on
proper collateral from the writers of the insurance was made far
worse because the gains and losses from CDS are not netted.
A Firm B that was neutral, betting one way against Firm A on
tranche BBB, and betting the opposite way on the same tranche
against Firm C, could come out a loser anyway. If Firm A
defaults on its insurance payment, then B will be unpaid by A
but still on the hook for paying C. So instead of just one Firm A
going bankrupt and another Firm C going unpaid in the
absence of collateral, as would happen with netting, another
Firm B might also go bankrupt, closing shop, firing workers,
and creating other social costs.
Losses by leveraged buyers of assets can cause a chain
reaction when a margin call forces a leveraged buyer to sell,
which might lower the price and force another leveraged buyer
to sell and so on. But with uncollateralized CDS, the chain
26
Of course, there were undoubtedly some hedge funds that bought bonds they
thought were undervalued, and bought insurance on similar bonds in order to
hedge their position against the risk of a market downturn. These are the
leveraged buyers that survived the crisis without a bailout. AIG is a classic
example of a writer of CDS insurance on mortgages that also held mortgage
securities (see Congressional Oversight Panel Report, June 10, 2010).
114
Solving the Present Crisis and Managing the Leverage Cycle
reaction is more direct: Firm B loses the money irrespective
of market prices. The implication I draw later is that there are
benefits from CDS being traded on an exchange instead of
in bilateral contracts, both to ensure that collateral is always
posted by the writer of the insurance and to make sure losses
are netted.
Another benefit of putting CDS on an exchange would be
the ease with which size and leverage could be monitored by
regulators. In traditional insurance law, as I understand it,
there is a prohibition against overinsuring by taking out
insurance for more than the underlying asset, precisely because
of the moral hazard such practices entail. Similar prohibitions
could be adopted for CDS.27
4.4 Counterparty Risk
In bilateral CDS contracts, it was often the case that the insurer
did not post enough initial margin collateral to guarantee
payment after a big move in default probabilities. This CDS
problem illustrates a more general flaw in the whole system of
contracting on Wall Street. These contracts to a great degree
were written in such a way that only one side of every
transaction was presumed liable to default, so that only the
other side needed protecting. For example, in the repo market,
a hedge fund borrower gets a loan from an investment bank,
and puts up collateral at the bank worth more than the loan.
The investment bank is protected against the potential default
of the hedge fund, because in that event the collateral can be
sold to recover the loan amount. But the contract does not
contemplate the bankruptcy of the investment bank. What
recourse does the hedge fund have if the investment bank goes
out of business, shutting its doors and swallowing the collateral
security? Following the Lehman bankruptcy, traders who never
before had to give a second thought to these counterparty risk
questions suddenly had to reevaluate all their contracts, with
disastrous effects on liquidity and price discovery.
Now, this unplanned-for counterparty risk has become the
primary rationale for the government’s seemingly unending
commitment to inject capital into “too-big-to-fail” institutions.
“We can’t afford another Lehman,” is the common refrain;
we had to intervene with AIG not because it was so vital, but
because if it defaulted a chain reaction might ensue.
The prospective solution to the counterparty risk problem
is to ensure that both sides put up enough collateral. Of
course, people are now more alert to their counterparty
vulnerability than they were before, and thus pressure will
grow, for example, on repo lenders to warehouse the
27
See “A Daring Trade Has Wall Street Seething,” Wall Street Journal, June 12,
2009, about a writer of CDS insurance who found a way to make the bond pay
off to avoid paying the overinsurance.
collateral at a third site that would not be compromised by the
bankruptcy of the lender. This raises questions about whether
there is enough collateral in the economy to back all the
promises people want to make, which I discuss at length in
Geanakoplos (1997) and Geanakoplos and Zame (2009). But
I believe there could be a government initiative to move as
many bilateral contracts onto exchanges as possible; agents
trading with the exchange will be required to put up
collateral, and the netting through the exchange will
economize on the collateral. As for any finance-related
bilateral contracting so particular that it could not be moved
to an exchange, the parties could either accept strict
disclosure requirements and limits on how much of this
contracting they could engage in or accept doing without the
instruments altogether.
4.5 Government Laxity, Deregulation, and
Implicit Guarantees Increased Leverage
The mildness and shortness of the crisis stage of the last two
leverage cycles, in 1994 and 1998, may have led many people,
perhaps including the regulators, to ignore leverage altogether.
The abrupt tightening of margins in 1998 was explained by the
supposed irrationality of lenders, who it was said reacted by
raising margins after the fact, that is, after the fall in prices had
already occurred. It appears that virtually no lenders lost
money on loans against mortgage securities in that crisis. The
run-up in asset prices and home prices during the current cycle
was attributed mostly to irrational exuberance, instead of being
understood, first and foremost, as an inevitable consequence of
the increase in leverage. Partly as a result of this faulty narrative,
government authorities did nothing to curtail the dramatic
growth in homeowner leverage, or consumer leverage more
generally, or corporate leverage, or securities leverage. Banks
were allowed to move assets off their balance sheets and thus
avoid capital requirements, further increasing their leverage.
Not only did the Fed (and everyone else) react passively to
the rising leverage pervading the system, it encouraged the
deregulation that unleashed the leverage inherent in outsized
credit default swaps. As I mentioned earlier, outsized CDS
contracts seem on their face to be either gambling or writing
insurance in excess of the value of the property being insured.
Under either interpretation, they would have run afoul of state
laws prohibiting gambling or overinsurance. Thus, it took a
positive act of Congress to pass legislation in the Commodity
Futures Modernization Act of 2000 exempting CDS from those
limitations.
Perhaps the most important and unwitting government
stimulus to the increased leverage was the implicit government
guarantees for entities that were considered too-big-to-fail.
Fannie Mae and Freddie Mac grew bigger and bigger. The
presumed government guarantee on their promises enabled
them to leverage their assets to 30 or more, and still issue debt
just above Treasury rates. Without this implicit government
backing, they would never have been able to borrow so much
with such little capital.
Many investment banks were allowed to write CDS without
collateralizing their implicit promises, as I observed before. It
seems virtually inexplicable that Wall Street overlooked this
counterparty risk; more likely, many counterparties assumed
that these firms were implicitly backstopped by the Fed or the
Treasury. And indeed, despite some doubts when Lehman
collapsed, that expectation proved correct.28
4.6 The Rating Agencies Effectively
Increased Leverage
The expansion of the mortgage market into less creditworthy
households made it more likely that a shock would someday be
“big and bad and scary,” creating more uncertainty and more
disagreement. The anticipation of that, however remote the
possibility seemed, should have made lenders nervous and
caused them to put a brake on leverage. This rational concern
was dramatically reduced by a faith many investors had in the
rating agencies and their default models, which were widely
relied upon by market participants (and the rating agencies
themselves), but which failed to account adequately for the
probability that defaults in certain circumstances would be
highly correlated. Some investors forgot the incentives of the
rating agencies and the incentives of many market actors to
downplay seriously the probability of highly correlated
defaults. In the face of a long history of low defaults and with
billions of dollars of deals waiting on the blessing of a small
handful of rating agency actors, it would have been astonishing
if ratings had been as tough as they should have been. The same
lesson applies to the mortgage brokers who were able to collect
fees for signing up borrowers without facing any losses
themselves if the borrowers defaulted.
4.7 Global Imbalances Increased Leverage
Caballero, Farhi, and Gourinchas (2008), Caballero (2010),
and others have suggested that the enormous savings glut
28
Bear Stearns was sold to J.P. Morgan, which took on Bear’s obligations, but
only after the government guaranteed $29 billion of Bear’s assets. Many other
investment banks, like Goldman Sachs, were given emergency injections of
$10 billion of Troubled Asset Relief Program (TARP) money.
FRBNY Economic Policy Review / August 2010
115
coming from Asia increased the demand for safe assets. This
presented a profit opportunity to American financiers, who
were thus stimulated to engineer the securitizations that
created apparently safe bonds out of risky assets. It is hard to
assess how important this factor is, but surely a gigantic
demand for safe bonds would indeed give a big incentive to
create those bonds and thus inevitably to concentrate more risk
in other bonds. However, that leaves unexplained why
investors were willing to buy those other bonds, or why
investors bought so much of the new, “safe” AAA-rated bonds
even when their yields revealed that the market did not think
they were perfectly safe. The Chinese, for example, did not buy
these bonds and they did not lose money when they
subsequently defaulted. The global-imbalances hypothesis
relies on an additional mechanism for its power: global
imbalances created lower, truly safe rates, which led American
investors pursuing absolute yields to leverage more, for
example, by buying the new, “safe” bonds with borrowed
money to leverage their tiny excess spreads. Thus, we come
back to leverage.
4.8 All Upside Down
The upshot of the huge credit boom and the plunging prices
was that an extraordinary number of households, businesses,
and banks ended up upside down or underwater, that is, with
debt exceeding their assets. According to First American
CoreLogic, about 13 million of the 55 million mortgage holders
were underwater in early 2010. According to Lender Processing
Services, about 2 million families have lost their homes since
2007, 2.5 million more are in foreclosure, and another 3 million
are not currently paying their mortgages.
The government has assumed trillions of dollars of
mortgage debt through its guarantee of Fannie Mae and
Freddie Mac and through its Federal Housing Authority (FHA)
loans, and has invested hundreds of billions of dollars
supporting banks and firms like AIG; in addition, on account
of the huge number of failing banks, the Federal Deposit
Insurance Corporation is on the verge of borrowing from the
Treasury. A problem of too much private debt has morphed
into a problem of too much government debt.
4.9 Why Didn’t Wall Street Risk Managers
Anticipate the Collapse?
Having discussed many of the factors that exacerbated the crisis
of 2007-09, I am now in a position to assess the widely held
view that nobody saw it coming.
116
Solving the Present Crisis and Managing the Leverage Cycle
Nobody doubts that Wall Street understood that there was
considerable risk in subprime mortgage pools. That is why they
were tranched into different tiers, called AAA, AA, and down to
BBB. And these bonds were all senior to residual pieces and
overcollateralization, which together provided another 8 percent of protection. So, the question is really not whether Wall
Street overlooked the risk, but rather how did it come to be that
Wall Street so badly underestimated the size of the risk?
The answer, I believe, is that it was nearly impossible to
foresee the devastating consequences of the multiple feedbacks
between securities and houses embodied in the double leverage
cycle. Complex adaptive systems are notoriously hard to
predict. Contrary to the myth that nobody imagined that
housing prices could go down as well as up, I suspect that
virtually every large bank and hedge fund considered a scenario
in which housing prices went down at least 10 percent. Recall
that if 25 percent of the loans result in homeowners being
thrown out of their houses, with 25 percent losses on each
foreclosed home, that amounts to losses of just 6.25 percent =
.25 x .25 for the pool as a whole, which would leave the rated
bonds unscathed. Better still, if 70 percent of the homeowners
refinanced according to historical patterns, then even with
50 percent defaults and 50 percent losses on the remaining
30 percent of the loans, losses would come only to 6.75 percent
= 30 percent x .5 x .5. But how many anticipated that at the
same time as housing prices went down mortgage down
payments would rise to the point that subprime refinancing
virtually stopped, dropping from 70 percent to zero? Or that
subprime mortgage originations would cease, causing further
house declines? And that at the same time servicers and banks
would refuse to write down principal, leading to more
foreclosures and further house declines? And that in the face of
so much homeowner misery and the destruction of so much
property, the government would wait until March 2009—more
than two years after the crash of the subprime ABX index in
January 2007—to launch its Home Affordable Modification
Program (HAMP)?
5. The Solution to the Crisis:
A Multi-Pronged Approach
Once the economy is plunged into circumstances as dangerous
as we saw last year, the government has no choice but to act
boldly. The correct course of action is to reverse the final stages
of the crisis and thus stop the panic. At the outset of this crisis,
I recommended the three-pronged approach I present here—
a thematic solution to the crisis that addresses in order of
importance all aspects of the final stages of the leverage.29
29
See Geanakoplos (2008).
As I explained above, all leverage cycles end with 1) bad
news creating uncertainty and disagreement, 2) sharply
increasing collateral rates, and 3) losses and bankruptcies
among the leveraged optimists. These three factors reinforce
and feed back on each other. In particular, what begins as
uncertainty about exogenous events creates uncertainty about
endogenous events, like how far prices will fall or who will go
bankrupt, which leads to further tightening of collateral, and
thus further price declines and so on. In the aftermath of the
crisis, we always see depressed asset prices, reduced economic
activity, and a collection of agents that are not yet bankrupt but
hovering near insolvency. How long the aftermath persists
depends on the depth of the crisis and the quality of the
government’s response. Whether we find ourselves in a similar
crisis in the future depends on whether, understanding how
leverage got us here, we adopt reforms that require supervisors
to monitor and regulate leverage in good times. First, I take up
what government actions could have been taken, and in what
order, to address the final stage of the double leverage cycle that
the government was called on to address in 2007.
The thematic solution once the crisis has started is to reverse
the three symptoms of the crisis: contain the bad news,
intervene to bring down margins, and carefully inject
“optimistic” equity back into the system. To be successful, any
government plan must respect all three remedial prongs, and
should be explainable and explained to the public in terms that
it can understand. Without public confidence, which can only
flow from public understanding, any federal government
(hereafter, “government”) plan undermines its own objectives
and limits its prospects for success. The government’s actions
thus far have not addressed all three prongs adequately and
policymakers have thus far largely failed to explain how their
various solutions are tied to the roots of the crisis we face.
Unfortunately, the TARP, the government’s first
intervention plan to buy distressed assets, was not clearly
thought through and neither it, the ostensible solution, nor
the problem that required a solution were clearly explained.
After its announcement, asset prices fell further. But even now,
after the panic has subsided, we must ask who or what is the
government trying to save? Many in the public have come to
believe it is merely trying to save banks, or some big banks,
from failure because somehow their failure would signal a
catastrophe for the American brand, to be prevented at all
costs.30 The confusion about the government’s goals has
created its own set of problems, which we can ill afford.
Clarifying the government’s goals will be harder now, but it
remains an indispensible step.
30
“Sixty-seven percent (67 percent) of adults believe Wall Street will benefit
more from the new bank bailout plan than the average U.S. taxpayer.”
Rasmussen Reports, February 2009/56.
5.1 Step One—Addressing the Precipitating
Cause of the Crisis: “Scary Bad” News
(Massive Uncertainty) about Housing
and the Assets Built on Housing
To foster recovery from the dramatic final stage of a leverage
cycle as large as the one we have just experienced, the
government must address the cause of the uncertainty that
triggered the end stage. Without that, the efforts taken thus far
to bring margins down and recapitalize banks, even had they
been perfectly implemented, would not be enough to reverse
the cycle and restore the economy to health. In this crisis, with
its roots in housing, that means doing something for housing
prices and homeowners. This makes undeniable sense in this
crisis, not just because addressing the cause of the uncertainty
and disagreement (the scary bad news) is critical to reverse any
leverage cycle, but because the biggest social losses will
probably come from the displaced homeowners. And, of
course, the biggest reason for the tumbling mortgage security
prices, and the resulting insolvency of the banking sector, is
fear that housing prices will keep falling.
Saving the Homeowners: Stemming the Tsunami
of Foreclosures to Come
One of the saddest stories in this financial meltdown is that
millions of homeowners are being thrown out of their homes
for defaulting on their mortgages. Throwing somebody out of
his home is tragic for the homeowner, but also very expensive
for the lender. One of the shocking aspects of the foreclosure
crisis is how low the recoveries have become on foreclosed
properties, after expenses. (Interestingly, the mortgage bond
index markets anticipated these bad recoveries.) Nobody gains
when the homeowners are thrown out and the banks and/or
investors collect pennies on the dollar for the money they
loaned. Yet, as we saw, 2 million homeowners have already
been evicted, another 2.5 million are seriously delinquent and
almost surely will be evicted in the near future, and at least
another 3 million will eventually default and be evicted if trends
continue. Without much bolder action than has thus far been
taken by the government, the stream of evictions and bad
recoveries for lenders will continue and accelerate, becoming a
torrent that will further depress housing prices and impede
economic recovery.
Negative equity is a key driver of mortgage defaults. When
faced with an income shock, borrowers who are in positive
equity have the option to sell the house rather than default.
Borrowers who are underwater (in negative equity) may
choose to default even in the absence of an income shock.
FRBNY Economic Policy Review / August 2010
117
The connection between LTV and default is illustrated in
Chart 8. For each mortgage in the First American CoreLogic
LoanPerformance Data Base, the current LTV is estimated by
taking the appraisal value of the house at the moment the first
loan was given, and then assuming thereafter that the house
changed in value according to the Case-Shiller index for houses
with the same Zip code.
As the chart shows, homeowners who have positive equity
in their homes default infrequently. But for homeowners with
negative equity, the rate of default is staggering. For subprime
borrowers with a 160 percent loan-to-value ratio (that is, the
ratio of all the mortgages on the home divided by the current
home price), the default rate is 8 percent per month.
These findings seemed surprising when I first presented
them in a New York Times editorial written with Susan Koniak
on March 5, 2009 (Geanakoplos and Koniak 2009). But
nowadays, many other researchers are reaching the same
conclusion.31 The conclusion is an inescapable matter of
incentives. It may not be economically rational for a
homeowner to continue to pay off a $160,000 loan when his
house is only worth $100,000.32 Mortgage loans have turned
out to be no-recourse—after seizing the house, the lender
almost never comes after the borrower for more payments.
Besides the ability to live in the house, the only other thing the
homeowner loses by defaulting is his credit rating, but
especially for a nonprime borrower with a low credit rating to
begin with, how much can that be worth? Finally, a choice
today by a negative equity borrower to default may be moving
up in time a necessity to default at some point in the future. In
this case, the borrower’s credit rating will likely be damaged
anyway.
31
Haughwout, Peach, and Tracy (2008) stress the importance of negative
equity as a determinant of early defaults among nonprime borrowers. The
Congressional Oversight Panel cited negative equity as the single greatest
predictor of default in its report of March 6, 2009. It included the data I provide
here as evidence of this fact, data that I supplied to the Panel in advance of its
report, as well as data from an array of government agencies, all of which
corroborated the Ellington Capital Management data presented here. That is
not to say that joblessness is not now having a significant effect on default rates.
It is. But even now, negative equity is the best predictor of default and many
Americans with jobs are defaulting, and will continue to default, not just the
unemployed. See generally the Congressional Oversight Panel’s Report of
October 9, 2009, on the continuing foreclosure problem and the unimpressive
results from government foreclosure prevention efforts taken thus far. Finally,
to the extent that job loss has become (it was not at the start of this crisis) a
significant cause of defaults, strong effective measures to eliminate the scary
bad news—that is, efforts to stabilize the housing market—will help the
economy recover faster and thus help the employment rate.
32
The implication of this statement is that the HAMP plan of reducing interest
rates to lower mortgage payments to homeowners who are underwater is, at
least for those seriously underwater, an invitation or encouragement to act in a
manner that may make no or little economic sense, that is, stretching to make
mortgage payments, albeit lowered from their highs, on homes those people
will never own when many of them might be able to rent more cheaply.
118
Solving the Present Crisis and Managing the Leverage Cycle
Chart 8
Monthly Net Flow (Excluding Modifications) from
Less than Sixty Days to Sixty or More Days Delinquent
Based on Performance from November 2008 to January 2009
for All Deals Issued in 2006
Monthly default rate (percent)
8
Option
adjustable-rate
mortgage
7
6
Subprime
5
alt-A
4
Prime
3
2
1
0
0
50
100
150
200
250
BHPA-adjusted CLTV (percent)
Note: Circles indicate median combined-loan-to-value (CLTV) ratios
by product.
Foreclosures are horribly expensive for the lender. At the
present time, subprime lenders collect about 25 cents per dollar
of loan when they foreclose. For example, if the loan is for
$160,000 and the house has fallen in value to $100,000 and the
homeowner defaults and is evicted, the lender can expect to get
back $40,000. It takes eighteen months on average to evict a
homeowner, and during that time he does not pay his mortgage
or his taxes, the house is often left empty and vandalized, a
realtor must be hired to sell the house, and so on. Of course, the
main reason the average recoveries are so low is that the
defaulters are the homeowners who are furthest underwater
(see Chart 8).
In a rational world, many foreclosure losses would never
happen. The lenders would renegotiate the loans by reducing
the principal so the homeowners could pay less and stay in their
homes, and the lenders would actually get more by avoiding the
losses from legal fees and bad home price sales. If the above
loan were written down to $80,000, the homeowner would
likely find a way to pay it, or else fix up the house and sell it for
$100,000. Either way, the lender would get $80,000 instead of
$40,000. That would have the further benefit of keeping many
homes off the market and thereby aid in the stabilization of
home prices.
The Home Affordable Modification Program pays servicers
to temporarily reduce interest payments and to extend the term
of the mortgage in order to reduce the monthly payments on
the mortgage, but does not incentivize servicers to cut principal. Cutting monthly payments by half will temporarily reduce
the homeowner’s payments by the same amount that cutting
principal by half would. But under the government’s plan, the
cut is temporary, not permanent, and thus is likely to lead to
many more defaults in the long run than cutting principal
would as soon as the interest rate goes back up.33 In fact, since
the homeowner will still be underwater, he will not in any
meaningful sense own his house. He will be less likely to make
repairs, he will not be able to give the house to his children, he
will not be able to sell it if he gets a job in another city.34 In
short, there is every reason to think he will likely default even
before the interest rate goes back up. For loan modifications
where there is no principal reduction, the redefault rate is
above 50 percent within nine months.35 Indeed, because the
government’s present plan allows servicers to increase
principal while cutting interest by adding fees and other costs
to the old principal amount, the plan is likely to leave more
homeowners underwater than there would be absent the plan
and others more deeply underwater—that is, with even less
chance of ever owning their homes and thus less incentive to
keep up with mortgage payments—than they would have
without this government “rescue” plan.
HAMP started off slowly and only recently is beginning to be
able to process a larger flow of mortgages. In the first six months
of the plan, according to the Congressional Oversight Panel’s
October 2009 report, only 85,000 mortgages had been modified,
and of those only 1,711 were “permanent” modifications (that is,
permanent/temporary, since interest rate reductions under the
plan are designed to end in a few years), and of those only 5
involved principal reductions.36 As of May 2010, HAMP had
started trial modifications on 1,244,184 loans, of which 429,696
had been canceled and 340,459 had been converted into
permanent modifications. Again, virtually none of the
permanent modifications involved principal reduction. Of the
5.7 million loans that were delinquent sixty or more days in May,
only 1.7 million were eligible for HAMP modifications.
33
Haughwout, Okah, and Tracy (2009) find in a sample of pre-HAMP
subprime mortgage modifications that reducing principal is twice as effective
as cutting the interest rate in terms of reducing the post-modification redefault
rate.
34
See Gyourko and Saiz (2004).
35
See “OCC/OTS Mortgage Metrics Report,” 2Q 2009.
36
To be clear, my criticism of HAMP is not based on the number of the timelimited “permanent” modifications completed, but rather is centered on the
near-exclusive concentration on interest reduction and, as I explain in the text
below, on leaving the servicers in charge of the modification decision. I could
find no updated information in the report on how many, if any, of the trial or
permanent modifications involved principal reduction as opposed to interest
reduction, and I have no reason to assume that the percentage of modifications
with principal reductions has increased. It is also worth noting that in the
Congressional Oversight Panel’s Report of October 2009 (p. 127), the Panel
notes that the apparent rise in modifications due to the administration’s plan
might be overstated, as there was some evidence of a “substitution effect,” that
is, the number of “voluntary” modifications by servicers (or modifications
made outside of the administration’s plan) went down in the first six months
of the plan, suggesting that the gross number of modifications attributable to
the plan itself might be exaggerated. The new report by the government does
not provide data from which one can assess any substitution effect.
The design of any modification program must recognize
that the servicers have incentives that at times put them at odds
with bondholders and homeowners, so that they may actually
prevent modifications that would help lenders and homeowners but hurt servicers. In the case of many nonprime
borrowers, the loans have been pooled in a trust, and their
principal has been tranched into many different bonds, each
held by a different investor. The lenders are the bondholders,
but they are numerous and dispersed and by contract have
given up the legal right to renegotiate with homeowners,
delegating that right to an agent.37 That agent is the servicer,
who has a fiduciary responsibility to act in the interests of the
bondholders in the trust.38 In “normal” times, this
arrangement worked tolerably enough. But in this crisis, with
so many mortgages in or near default, it has failed miserably for
at least four reasons, all traceable to a misalignment of interests
between servicers and those whose interests they are supposed
to protect, which has now ruptured with terrible effects.
First, modifying loans is a time-consuming and expensive
operation. The servicers who have the legal right to make
modifications do not get paid directly for improving the cash
flows to loans. It is generally cheaper for them to move into
foreclosure. In particular, they have no incentive to set up the
huge infrastructure and to hire and train the extra staff
required to make sensible modifications on a grand scale.
Second, modifying the loans has different effects on different
bondholders. It has proved difficult to modify loans in a way that
pleases everyone. The servicers say they are terrified of lawsuits
from the bondholders if their modifications help most
bondholders but hurt others. For example, writing down
principal immediately may make more money for the trust as a
whole, but it would immediately wipe out the BBB bonds and
possibly other lower level bondholders. Letting the borrowers
remain in their houses without paying during the foreclosure
process means that during all that time all the bondholders,
including the BBB, get their coupons paid in full from servicer
advances. The servicers then recoup their advances, at the expense
of the trust, when the house is finally sold.39 In reality, servicers
37
It should be noted that this right was given up to avoid the collective action
problems inherent when the lenders are numerous and dispersed, and thus was
given to a third party (the servicer) to be exercised on the lender’s behalf, the
servicer acting as a fiduciary for the lenders. It was not given to the servicer to
be used to benefit the servicer’s interests at the expense of the principals (the
lenders), and using the discretion to modify or foreclose that way is self-dealing
on the part of servicers and a breach of their obligation to the lenders.
38
See Alan Kronovet, “An Overview of Commercial Mortgage Backed
Securitization: The Devil Is in the Details,” 1 N.C. Banking Inst. 288, 311
(1997), explaining fiduciary duties of servicers. Section 1403 of the new
housing bill that was signed into law on July 30, 2008 (HR 3221, the Housing
and Economic Recovery Act of 2008, P.L. 110-289), lays out the fiduciary
responsibilities of servicers of pooled mortgages.
39
This requires that the servicers have access to capital to finance the coupon
payments until the foreclosure process is concluded.
FRBNY Economic Policy Review / August 2010
119
were not deterred only by potential lawsuits. That was revealed
when Congress passed legislation that freed servicers from
lawsuits by bondholders.40 Principal reduction modifications
did not follow. To put it all another way, there is a complex
negotiation that is not taking place, and the government needs
to intervene to break an impasse for the public good.
Third, now that HAMP, which is based on interest
reductions, has given the servicers cover to reduce interest
instead of principal, they can be counted on to do the former
and eschew the latter. Cutting the principal by half, for
example, immediately reduces the servicer’s fee by half (since
the fee is computed as a percentage of principal), while cutting
interest does not. Moreover, cutting principal increases the
likelihood that the homeowner will sell or refinance, which
would cause the servicer to lose his fee entirely.
Fourth, the biggest servicers happen to be owned by the
biggest banks, which in turn own a huge number of second-lien
loans. Cutting principal on first loans almost implies cutting the
principal drastically, if not to zero, on second loans. But that
would mean that the banks could no longer hold the secondliens on their books at potentially inflated prices. The banks want
desperately to postpone the write-down of those second-liens,
which is to say, they have yet another powerful motive not to do
what is in the interest of lenders, homeowners, and the economy
as a whole: reduce principal on the first-lien loans they are
servicing. By contrast, cutting interest on first-lien loans makes it
easier to justify carrying the second-liens on bank balance sheets
at higher values for the near term (which is what matters to the
banks), as homeowners are more likely to be able to make the
lower monthly payments (from lower interest rates) than their
original payments, at least in the short run.41
Another indication that servicers have bad incentives is that
when the big banks hold the same kind of loans in their private
portfolios, they do reduce principal. During the second quarter
of 2009, 30 percent of all modifications done to loans directly
held in bank portfolios involved some principal reduction.
During that same quarter, the servicers reduced principal on
0 percent of their loan modifications, as did the governmentowned agencies Fannie Mae and Freddie Mac.42
Loans that have not been securitized and are held entirely
by banks (whole loans) are also not being written down fast or
40
See Section 201 of the Helping Families Save Their Homes Act of 2009,
preventing lenders/bondholders from suing servicers who modify mortgages
under a qualified mortgage modification plan, which is defined in the Act
broadly enough to include all economically sensible modifications, that is,
those with a reasonable prospect of returning more money to the lenders than
a foreclosure.
41
Cutting the monthly payments will also push the likely default further into
the future. Under current accounting rules, this reduces the loss reserves that
the banks have to hold against these loans.
42
See OCC/OTS Mortgage Metrics Report, Q2 2009.
120
Solving the Present Crisis and Managing the Leverage Cycle
far enough.43 The pathology this time is, if anything, more
distressing. It appears that the banks, abetted by the suspension
of mark-to-market rules, are unwilling to fully recognize the
losses that have occurred on their residential mortgages.44 They
may prefer to keep a mortgage on their books at $160,000, even
though it will eventually bring them only $40,000, than to
reduce the principal to $80,000 and mark the loan at this value
today. The suspension of mark-to-market rules has also fed the
pathology discussed above on second-liens. Perpetuating a
conflict between the economic value and the accounting value
of an asset is bad government policy when it leads to actions
that further reduce the asset’s value. This conflict is also
obscuring the value of bank assets, many of which are being
guaranteed by the government, and thus in turn obscuring the
value of mortgage assets now owned by the government. In my
terms, this only ensures the continuation of “scary bad” news
(uncertainty), when the goal should be for government plans to
clarify the situation (the value of assets) that keeps leverage
severely constricted.
Insuring that economically efficient mortgage
modifications are made for borrowers can be greatly facilitated
by placing the decisions with impartial agents. In October 2008,
Susan Koniak and I urged the government to take the
reworking process out of the hands of the servicers and put the
decision into the hands of government-hired trustees. In our
approach, the government-hired trustees would be told only
about the homeowners, and would be blind to the bonds built
atop the loans. Their job would be to choose modifications or
foreclosure, whichever they judged would lead to the greatest
recovery for the lenders on the original loan. They would thus
be carrying out the duties of the servicers exactly as they were
intended, but free from the conflicts of interest and perverse
incentives that have prevented the servicers from carrying out
their mission.45
If there is a second-lien loan, the government trustees would
make the same calculation, deciding what modification, if any,
would maximize total revenue. If this involved reducing
principal, then the second-loan principal would be reduced to
43
At first, it appeared that they were not being written down at any greater rate
than securitized loans, although the data are not perfect on this. Foote et al.
(2009) argue that this showed there was no real incentive to write down loans.
Now, again based on imperfect data, there seems to be some evidence that
principal on whole loans, at least at some banks, is being written down more
often than principal on securitized loans (which effectively never see
reductions in principal), although reductions in principal on whole loans are
still much less frequent and much less widespread than one would expect to see
given the economics of the situation, that is, that reducing principal for many
underwater homeowners will yield much more money than foreclosure or
(over the long term) interest reductions.
44
Banks may also still be holding out for some more direct government subsidy
for their failing whole loans, either through government assumption of the
mortgage risk or some other form of direct payment for anticipated whole loan
losses.
zero. The second-loan holder could still receive some cash,
however. I would recommend distributing the same percentage
of the monthly payments to the second loan as it was getting
before principal was reduced for a period of, say, two years.
After that, the second loan would be completely extinguished
and all cash flows would flow to the first-loan holder.
For a vast number of homeowners now upside-down in
their mortgages, that is, owing more than their home is
presently worth, this process would likely result in a reduction
of principal. Why? Because reducing principal rather than
cutting interest rates would be more effective at preventing
defaults and would yield investors/lenders more money than
foreclosing, as we have seen.46
If the government handled this correctly, most homeowners
who were unable to pay the original loan but were willing and able
to pay a modestly lesser amount would get to stay in their homes,
the bondholders collectively would get more payments than they
are currently expecting (though some tranches would be hurt),
and the government would not have to invest any capital.
This plan is not the same as “cramdown” in bankruptcy,
which Congress has thus far rejected and which entails costs
and creates some perverse incentives that my plan avoids.
Giving reductions in principal through bankruptcy (assuming
the law were changed to allow that) would encourage
homeowners now current on their mortgages but underwater
and thus likely to default sometime in the future to default
immediately to support their petition for bankruptcy relief.
However, my plan, as originally conceived, does not build in
any incentives for the borrower to default in order to increase
the chance that the mortgage will be modified. Principal
reductions would be done first for homeowners who have not
defaulted yet, and only later for homeowners who have
defaulted under some special hardship. It would give
underwater homeowners now holding on for the short term a
continued incentive to keep paying until the government
trustees could evaluate their loans and circumstances for a
reduction in principal. Second, my plan differs from
bankruptcy in that it does not subject homeowners to the
shame and devastating harm to future credit and thus to their
economic circumstance that a bankruptcy proceeding entails.
Third, my plan contemplates putting local housing market
45
See Geanakoplos and Koniak (2008). Under this plan, the servicers would
still collect the servicing fees they do now. They would continue their duties of
sending letters to homeowners, collecting the monthly payments and
distributing them to bondholders, evicting homeowners who did not pay,
selling their homes, and so on. The only change is that the mortgage loan
modification would be taken out of their hands and put into the hands of the
government trustees. This reassignment of a particular duty in the contract is
not a “takings” from the servicer, among other reasons because the servicers
have failed to carry out their fiduciary obligations to the bondholders who
employ them to get the most possible value out of the loans. See Dana
(forthcoming).
46
See Haughwout et al. (2009) for evidence based on subprime modifications.
experts and community bankers in place as government
trustees, not bankruptcy judges who are neither numerous
enough to handle the number of defaulting homeowners who
should justifiably qualify for principal reduction nor as
knowledgeable as the personnel I would put in charge.47 If my
plan were indeed up and running, bankruptcy might be
something worth considering as a true last resort for those
already deeply in default. Finally, bankruptcy involves all kinds
of hidden costs, like lawyer fees and trustee expenses (on top of
the costs associated with the experts required to advise the
bankruptcy judges) that are unnecessary and wasteful for the
vast majority of homeowners and lenders who should be able
to make a win-win deal without incurring those costs.48
My original plan called for legislation to cut through the
agency-problem mess in securitized pools of mortgages by
eliminating contract provisions in pooling arrangements that
now enable servicers to act contrary to the interests of the
investors that the provisions were originally designed to
protect. Thus, I envisioned that the government trustees would
only be empowered to modify securitized mortgages. This
would leave unsolved the problem of whole loans that banks
are still refusing to modify sensibly, by writing down principal
for underwater homeowners.
I believe, however, that once a government program of
modifications for securitized loans proved its worth by
resulting in more recovery for investors, banks would be likely
to adopt similar standards to modify whole loans. Nonetheless,
a solid government plan to force sensible principal reductions
for securitized loans would, I believe, go a long way toward
convincing the banks that no better deal from the government
was forthcoming, particularly if the government clearly
articulated that this was so, and would exert discipline on the
valuation of the whole loans and second loans on the banks’
balance sheets. Obliging the banks to mark to market would, of
course, also push them to get the most value out of their loans
by writing down principal for underwater homes.
Finally, what if home prices vastly appreciate by the time the
homeowner sells his home? To prevent unwarranted windfall
profits to homeowners, the government plan could easily
require the homeowner to share 50/50 with the lenders any
appreciation in home price up to the full amortized value of the
original mortgage, and the plan might even provide that, for
houses sold for more than the original loan price, lenders
receive a greater percentage of the appreciation.
47
Indeed, it is highly doubtful that our bankruptcy courts could handle the job
Congress would be giving them if so-called cramdown legislation were
adopted, at least not if it were adopted without first having a plan like the one
I propose up and running to handle the vast majority of underwater
homeowners.
48
My plan envisions the government paying for the trustees (community
bankers) to decide on whether principal modification would bring in more for
bondholders than foreclosure, but I estimate that government expenditure
should come to less than $5 billion.
FRBNY Economic Policy Review / August 2010
121
A Floor to Housing Prices and Restarting Private
Lending on Mortgages: Government Equity Stake
in Homes
There are at least four reasons to support housing prices
directly, in addition to doing so through effective foreclosure
relief. First, if housing prices held firm, fewer homeowners
would be underwater; thus, more would have an incentive to
make their payments. That would keep them in their homes.
Second, firm housing prices would staunch the losses on
mortgage securities even if there were foreclosures. Third, once
there is a floor to housing prices, pessimistic lenders would be
relieved of the disaster scenario for many mortgage securities,
and margins on mortgage securities would come down
significantly, enabling optimistic buyers to purchase them
using leverage, pushing up the price of mortgage securities.49
Fourth, the leverage cycle is less severe for housing than
for mortgage securities, so it can be fixed more easily by
government intervention, because home buyers generally lock
in their loans and leverage for the duration of time they live in
the house. Only new buyers of homes, and those who want to
change homes, need to confront the tougher margins. Existing
homeowners cannot be forced to put more money down,
whereas mortgage security holders who borrowed on one-day
repos have found that they now face tougher margin requirements that involve putting more money down. Thus, there are
fewer homes in play than there are mortgage securities.
The government has recognized the need to try and support
housing prices. A concern is that the measures taken will
expose the government to the risk of billions of dollars of future
losses, in addition to substantial current costs, while leaving
private mortgage lending dead in the water. We simply cannot
sustain a situation where all mortgage lending is done by the
government. The plan I propose helps to stabilize housing
prices and to reinvigorate private lending. And in the long run,
it may cost the government much less, possibly even making
money.
Current government FHA policy is to make mortgage loans
with as little as 3.5 percent down. In addition, borrowers can
finance some of their closing costs as well as the up-front
mortgage insurance premium. As a result, the effective LTV on
new FHA mortgages can exceed 100. These homeowners start
with little incentive to continue making payments, particularly
in rough economic times. Given the transaction costs of selling
a house, absent a rise in housing prices these borrowers will
remain underwater and thus create a new source of future
defaults. This policy is a repetition (albeit on a smaller scale) of
49
As I discuss below, margins must in the future be monitored by the Federal
Reserve to assure that they do not once again get excessively low, precipitating
another massive and dangerous leverage cycle.
122
Solving the Present Crisis and Managing the Leverage Cycle
the low down payment lending practices that got us here. It
exposes the government to a huge risk of default, and does
nothing to stimulate private mortgage lending.50
The government has also tried to stabilize housing prices
through its efforts to keep mortgage interest rates low and
thereby encourage purchases and refinancing. To this end, the
Federal Reserve’s Large-Scale Asset Purchase program has
purchased $1.25 trillion of agency mortgage securities. Like the
HAMP modification program, this choice reflects once again
a concentration on interest rates rather than on the collateral
(leverage) effects that are at the core of my argument. The
Large-Scale Asset Purchase program appears to have lowered
mortgage interest rates, but surprisingly few homeowners were
able to take advantage of the lower rates by refinancing because
they could not come up with a down payment and/or their
credit had deteriorated.51 One might worry that as the
purchases wind down, mortgage rates may go back up.
A third government initiative is to give an $8,000 tax credit
to buyers of homes. This tax credit does appear to have been
more successful at stimulating home purchases. But the tax
credit has no upside for taxpayers and it does nothing to
reinvigorate private lending since most of the new mortgages
were guaranteed by the FHA. If $8,000 were spent on 7 million
homes, the cost to taxpayers would come to $56 billion. By
contrast, the equity stake plan I propose below is a purchase of
value for value; in the long run, it may cost nothing and actually
have upside for taxpayers. It should also stimulate demand, and
it would reinvigorate private mortgage lending.
As I observed earlier, toughening margins have affected
housing prices, because many homeowners can no longer put
up the cash payment needed to buy new homes. New
homeowners are being asked to put as much as 30 to 40 percent
down if they cannot get a government loan. The government
could stimulate demand for new purchases, and also mitigate
the margin problem, by offering to buy a 20 percent equity
stake in any new home purchase (under some maximum price,
as with agency conforming loans). Thus, suppose a house is
purchased for $100. The government pays $20 and gets a
20 percent equity piece, which it collects whenever the
homeowner sells. If down the line, the house sells for $200,
the government gets $40. The government is thus earning the
home price appreciation on its piece, without having to bear
the expense of maintaining the house. The homeowner gains
50
For more on FHA risk, see Aragon et al. (2010).
See Caplin, Freeman, and Tracy (1997) for a discussion of down payment
constraints on refinancing and Peristiani et al. (1997) for a discussion of credit
constraints. To address this concern, the administration started the Home
Affordable Refinance Program, which allows borrowers with prime mortgages
to refinance with current LTVs as high as 125. In addition, the FHA introduced
a “streamline refinance” program for borrowers with high-LTV FHA loans to
refinance to a new FHA loan.
51
because he gets to live in the whole house while paying for only
80 percent of it. If the home buyer needs a loan to get the house,
the government equity piece reduces the down payment the
buyer must make, and the ongoing mortgage payments he
must make. And if we make the government’s equity piece
the second loss piece, it leaves the lenders in a very, very safe
position, encouraging lending. In effect, it lowers the margin
to the borrower, and raises the margin of safety to the lender.
Here is how it works.52
Under the plan, the home buyer who wanted a loan to
purchase the house would be allowed to borrow at most
80 percent of the $80 of the house he bought, or $64. He would
have to put up 20 percent x $80 = $16 of his own cash. The
homeowner would then have a big incentive to make his
payments. If he walks away from his debt, he can save $64, but
he has to give up living in a $100 house on which he had an $80
ownership share. But if the borrower does default, and if the
lender has to foreclose, the lender would be able to collect his
debt out of the house sale proceeds ahead of the government
equity piece. The government would collect next, and lastly the
buyer would get any leftover cash. If the house sold in
foreclosure (net of expenses) for $82, the lender would get his
$64, the government would get $18, and the homeowner
nothing. The effective margin for the homeowner is thus
16 percent on the asset price of $100, but the margin of safety
for the lender is 36 percent. This should make the lender feel
very safe and encourage private lending on mortgages. The
homeowner’s down payment of 16 percent on the total home
price is about half the down payment many nongovernment
lenders are demanding now. On top of that, the new buyer’s
mortgage payments would be 20 percent lower than before,
because he would be paying on a loan of $64 instead of $80.
What about the costs of my plan? Last year, there were
5.5 million new home purchases, down from a high of 7 million. Even if the government had to buy the equity in the entire
7 million, at an average home price of $200,000, it would cost
$280 billion. But the government would own equity, and be
protected by the homeowner’s down payment. Housing prices
would need to fall another 16 percent before the government
lost equity value. As housing prices stabilized, the government
would gradually phase out the program, in all likelihood in a
year, at most two, after adoption. To lower the government’s
overall equity investment, the program could be limited to
first-time home buyers.
52
Equity sharing arrangements could also form with private investors. For a
discussion, see Caplin et al. (1997).
5.2 Step Two—A Fed Lending Facility
to Help Restore Reasonable Leverage
The most easily implementable step and the second priority,
after addressing the source of the uncertainty (the scary bad
news), in responding to the final stage of any leverage cycle
could be government action to decrease astronomical collateral
rates. Thus, in October 2008 I suggested that the most
immediate step the Federal Reserve could take was to lend
money using the so-called troubled assets (those that suddenly
became nearly impossible to use as collateral, as I explained
earlier) as no-recourse collateral. I suggested 50 percent
margins on average, a reasonable halfway level between the
5 percent margins required at the peak of the leverage bubble
and the 70-90 percent margin rate demanded in 2008. The
Asset-Backed Securities Loan Facility (TALF) and the PublicPrivate Investment Program (PPIP), announced in early 2009
at what turned out to be the bottom of the price cycle, embody
the spirit of my recommendation. Indeed, the PPIP did lend on
these bonds at exactly 50 percent margins. The turnaround of
prime mortgage security prices (displayed in Chart 2) after
these programs were announced seems to me to be some
evidence for the wisdom of the intervention. But in terms of
some important details, those programs did not go as I would
have recommended. In any case, it now appears that having
achieved their purpose, they have been drastically attenuated.
Lending with smaller margins (haircuts) than the market is
willing to offer to borrowers who might not repay is a great
departure from the traditional role of the Federal Reserve. The
orthodox view is that the Fed injects liquidity into the system
by lending money to banks and others with impeccable
reputations for repaying so as to reduce the riskless rate of
interest on very short-term loans. The banks would then
presumably turn around and relend that money to investors, at
a lower interest rate than would have obtained absent the Fed’s
intervention. However, the great bulk of lending in the
investment world is not based on the reputation of the
borrower but based instead on the value of the collateral. The
lesson of the leverage cycle is that when lenders demand too
much collateral for their loans, liquidity dries up. The Fed
cannot undo this by making riskless loans at a lower interest
rate than the market, because in liquidity crises it is not the
interest rate the banks charge that impedes investor borrowing
but rather the amount of collateral they require. The Fed needs
to step around the banks and make risky loans directly to
investors with smaller haircuts than the market demands, if
it is to have the desired effect.
The mechanics of such a massive lending program require
some careful thought, but nothing compared with the
FRBNY Economic Policy Review / August 2010
123
difficulties of directly buying. The Fed could simply announce
that any arm’s-length buyer of any designated security could, at
the moment of purchase, take that security to the Fed and
receive a five-year loan of 50 percent of the price in exchange
for putting the security up as collateral. The Fed would not
need to price the security itself. The market would have just
done the pricing. With a 50 percent margin, the government
money is still quite safe. Remember, the 50 percent loan is
against the price the securities will be traded at, not against the
original price when issued. The government could thereafter
monitor prices, periodically demanding more cash from the
borrower to maintain its 50 percent margin, which would make
the government lending safer and more responsible.53
Monitoring the collateral price is a much easier job than
deciding the price to buy, since there is a 50 percent margin of
error: the price monitoring only has to be half right. And the
government could consider charging a slightly higher interest
rate than the fed funds rate or discount rate, thereby potentially
making a profit for taxpayers. That would also make the
program easier for the public and politicians to accept.
Needless to say, the 50 percent margin cannot be applied
to all bonds. Some bonds have such high volatility in their
cash flows that even a 50 percent margin is unsafe. Other
bonds can safely be leveraged much more. The Fed must
exercise its own expertise in setting these margins, as I discuss
later. But in a crisis, they should be set at levels substantially
more generous than the market is offering, and significantly
less generous than the market had been offering in the
ebullient stage before the crisis.
The five-year term can also be chosen flexibly. But it is
important that there is a longish term commitment to
borrowers that the loan will not be pulled from under them.
The last thing a buyer wants to do in a crisis is leverage to buy
and then have his financing pulled, or his margins increased. Of
course, the Fed needs to worry about its exit strategy; if it lends
too much money long term, it will not be able to reel it all back
in should inflation pick up. However, by lending at margins
and interest rates that are favorable in the crisis but that
borrowers will find onerous once markets pick up, and by
making margin calls, the Fed can count on most borrowers
refinancing their loans privately once the market heats up.
The government might even arrange all this lending without
having to come up with the money. Under this alternative, the
government could loan slightly less, say, 40 percent, and give
up the right to make margin calls. The loan could then be
securitized, guaranteed by the government, and sold off to the
private sector. With the government guarantee, the money
would easily be raised. Or even more directly, for some bonds
where this makes sense, the government could simply
53
Even if the securities gradually lost all their value, the Fed would still not lose
any money if it made frequent margin calls.
124
Solving the Present Crisis and Managing the Leverage Cycle
guarantee a certain percentage of the principal payments.
Private lenders could then lend this much without any risk of
default. Of course, on some securities the government might be
able to lend much more than 40 percent and still regard the
money as safe.
At 50 percent margins, buyers would be able to purchase
securities using only half the cash they need to put up at the
bottom of a cycle when margins might become 100 percent.
Aside from allowing investors’ own cash to go further, this
borrowing allows investors to earn leveraged returns. If they
think the security trading for 60 might only rise to 66 in the near
future, they can buy it with 30 down and earn a return of
20 percent when it rises to 66 instead of a return of 10 percent.
Buying will be stimulated and the depressed prices at the
bottom of the leverage cycle will be pushed back up. Again, with
this potential for private profit, the program would make more
political sense if a somewhat higher interest rate for the loans
were charged, thus building in a real chance for taxpayer profit.
Lending is better than the government’s first (and quickly
shelved) idea, as proposed by former Treasury Secretary Henry
Paulson, of buying up the “troubled assets.” As I explained in
October 2008, lending against collateral does not require the
government to choose what prices to pay, as it would have to if
the Treasury directly bought securities. Moreover, lending,
unlike buying, is direct action to restore leverage and restoring
leverage is the thematic solution to the leverage cycle crisis. It is
not some stop-gap band-aid invented only under the pressures
of the moment.
Further, lending puts taxpayer money at far less risk than
buying does. Assuming the Fed lends at 50 percent margins,
every dollar the government lends using the targeted assets as
collateral will necessarily be matched by money the investor
spends on those assets. The government can say its money is
being leveraged. The investors who avail themselves of the
government lending will still have their money at risk. Because
these investors, and not the government, will do the buying,
there is little, if any, chance that this action will push prices
to outrageous levels and enrich undeserving sellers.
The Fed has boldly gone a long way in this direction, further
than any previous Fed. Through the TALF and the PPIP, the
Fed and the Treasury, respectively, have indeed embodied
many of these ideas. The PPIP lends at 50 percent margins
on troubled mortgage securities, just as I recommended. Its
announcement, I believe, played a pivotal role in starting what
is now more than a year-long rebound in security prices. Given
the condition of the asset markets in early 2009, the rebound in
prices seems almost miraculous, and in many ways one must
judge the TALF/PPIP a resounding success.
Nevertheless, I believe that the Fed-Treasury leverage
intervention would have been better if it had been
implemented somewhat differently. This difference is
important to bear in mind not just for this crisis, but also in
case there is another crisis in which prices do not rebound as
quickly after a leverage intervention. In my opinion, the two
programs did not encompass a wide enough set of assets or a
wide enough set of borrowers, they took too long to get going,
and in some cases TALF actually took leverage up almost to the
crazy levels it had been before. Had TALF started earlier, and
had it lent on more assets, it would not have been forced to give
such high leverage on the narrow band of assets it did lend
against.
In the emergency stages of the leverage cycle, the Fed should
have extended lending on more kinds of collateral. TALF
restricted leverage mostly to new securities, or to securities that
were still AAA-rated. As more and more mortgage securities
get downgraded below investment-grade status, they lose their
ability to be used as collateral even in the private sector.
Lending against the most toxic securities is actually necessary to
maintain their value.54
The TALF program made government loans on new credit
cards, auto loans, college loans, and other securitizations at 20
to 1 leverage. In my opinion, this repeats the error of the FHA
mortgage program, lending at the same inflated leverage that
got us into trouble in the first place. The Fed has rightly
observed that propping up new security values is more
important than propping up legacy security values, because
new securities represent new activities. When new prices go
down, new securities are not issued and the underlying activity
for which the securities would be issued (students going to
school, cars being purchased, new houses being built,
consumers buying with credit cards) stops. However, as I argue
more formally in Geanakoplos (2010), in the depths of the
leverage cycle, the Fed could raise the price of new securities
further by leveraging them less, if it would also leverage the
legacy securities to modest levels. The reason is that potential
buyers of these new securities are tempted instead to put all
their capital into the depressed legacy assets where they are
nearly sure of a high return. This indeed is one of the main
reasons banks stop lending to businesses or homeowners: they
can get better returns by buying depressed legacy assets. Given
the depressed legacy security prices, the only way TALF could
redirect this private money into new securities was by giving
54
Again, such lending would be much less risky if the government had adopted
a sensible plan to staunch foreclosures and stabilize housing prices, such as I
have just outlined, because such a plan would reduce the toxicity of the
securities at issue. And the quicker the government moves to do that, the less
risky such lending will become, not to mention the good it would do for the
value of the toxic securities the government now owns through one program
or another or now guarantees, representing continuing and enormous
government money still at considerable risk. This point is why I stress the
importance of understanding the nature of the crisis in crafting sensible
solutions and how failing to address one part of the problem, in our case the
failure to adequately address housing, limits the good that otherwise sensible
programs might make.
leverage on the new securities at astronomical 20:1 ratios. If
instead the Fed would give much lower and safer 2 to 1 leverage
on the legacy assets, it would raise the legacy asset prices, and
thus even the new security prices, because it would remove the
bargains investors are seeking in the legacy assets.55 The new
assets would not need so much leverage, and the risk to the
taxpayers would be reduced. This would also go a long way to
solving the bank lending problem. As I show again in
Geanakoplos (2010) (in a stylized example, to be sure), despite
lending on a much larger scale, by allowing leverage at 2 to 1 on
a wide array of assets rather than at 20 to 1 on a narrow set of
assets, the Fed could actually reduce its expected defaults while
increasing the prices of all the securities. A year later, it now
appears that the Fed will not face significant losses on these
TALF loans, and private leverage is also returning. But had
things gone worse, the Fed might have been stuck with some
dangerous loans.
In the crisis stage, the Fed needs to go around the banks and
lend directly to more investors. In theory, the Fed could make
no-recourse loans only to a few banks, who would turn around
and relend to everyone else. But the banks are nervous about
showing too much lending on their books, they ask for too
much collateral, and now the Fed is giving them more
profitable ways to make money than by lending; so the Fed
must reach out directly to more borrowers. Curiously, the PPIP
has been restricted to ten potential borrowers/investors,
making its scope and size in the end less than what was
anticipated. Also, with only ten investors taking government
money, the potential for conflicts of interest seems very high,
as I discuss later.
The TALF and PPIP programs took too long to get up and
running. Hopefully, at the bottom of the next leverage cycle, or
even earlier, similar programs could be implemented sooner.
I recommend that the Fed keep a standing, permanent lending
facility up and running. In normal times, it would lend a little
bit across a wide range of assets, to be ready to spring into
action if private collateral rates became too high. This facility
could be administered directly by the Fed, by people it hired, or
it could be run through the repo desks of the Wall Street banks.
In the latter case, it would be wise to insist that the banks put
some of their capital at risk along with the Fed money. The
advantage of using repo desks is that they are already staffed
with trained personnel, who have great expertise in making
margin calls. Duplicating that expertise would be expensive.56
The advantage of a permanent facility is that the Fed would be
ready to quickly lend on a grand scale, on many securities, and
to many lenders, in the next crisis.
55
Another reason why it actually could raise new security prices is that by
leveraging the legacy securities at 2 to 1, it will free some investor equity to put
into the new securities.
56
I presented this proposal for a lending facility to the Liquidity Working
Group at the Federal Reserve Bank of New York in early 2009.
FRBNY Economic Policy Review / August 2010
125
5.3 Step Three—Restoring “Optimistic”
Capital
Lending will not by itself bring the prices of assets to their old
levels (which is okay, given that “old” values were inflated by
excessive leverage, as I have explained). But that means that the
most optimistic buyers, unfortunately including some of the
biggest and most prominent financial institutions in America,
have irretrievably lost a huge amount of capital. Not only is
their capital no longer available to spend on these securities,
but similarly the money they borrowed to spend on these
securities has also disappeared.
The most obvious thing the government could do, it did:
inject money into financial firms. The idea was that then the
firms would continue to function as optimistic buyers and their
workers would not join the ranks of the unemployed. But the
main problem with the way the government injected capital is
that this injection of capital was not coordinated with vigorous
programs to address the two other prongs of the end of any
leverage cycle: the source of the scary bad news (here, housing)
and the precipitous drop in leverage, which I have just
addressed in my discussion of Fed lending.
In the absence of vigorous programs to address the first two
prongs of any leverage crisis, injecting capital does nothing but
push an ultimate reckoning down the road. Without steps one
and two, the true financial status of our financial institutions
is unknown and unknowable because there is no reliable way
to price many of the assets they hold. The danger is that the
injection of new capital keeps the banks from failing immediately, but it is not enough to restore their previous activities,
leaving them in a kind of limbo and actually creating more
uncertainty in the system about whether they will survive. As
long as no one knows whether and to what extent our biggest
financial institutions are sound, our economy cannot recover.
Bailouts with Punishment
After a double leverage cycle as outsized as we have just been
through, it is likely that even with a lending facility established,
and capital injected properly into the system, some, maybe
many, firms would still fail. In general, that is what we should
want. The government cannot afford to make good
everybody’s debt. Some debtholders must lose when a financial
system is allowed to become bloated by artificially high prices
maintained by excess leverage from the ebullient stage of the
leverage cycle. In the ebullient phase of this cycle, too many
people were drawn into the financial sector by the resultant
artificial profits. Failures will remove many of these excesses.
But what if those institutions are seen by the government as,
in current jargon, systemically important? For those firms, the
126
Solving the Present Crisis and Managing the Leverage Cycle
Treasury might want to intervene, as the Fed did last year, on a
case-by-case basis. But, if that approach is used, important
issues are the degree to which the shareholders have to give up
their shares and the bondholders lose their value, and whether
new management should be put in place. Even in cases where
old management is not that old, that is, cannot be reasonably
charged with responsibility for all the excess, replacing
management may be wise, if only to help bolster public support
for the government’s actions and expenditures of taxpayer
funds. It is also imperative that the government decide as
quickly as possible after a crisis presents itself (and on grounds
that can be explained as fair and objective), who it will let fail,
and then coordinate an orderly liquidation. Quite possibly the
biggest public relations risk the government runs in the bottom
of the leverage cycle is to appear to be bailing out ailing firms
on too generous terms.
If instead of injecting funds into an ailing firm the
government takes it over, it must quickly decide what it will do
with the creditors. Once it guarantees their debts, there is no
turning back when the full extent of the firm’s asset value
becomes clarified. In the case of AIG, it now appears that the
government will lose much less money than was initially feared.
But in the case of Fannie Mae and Freddie Mac, where the
stakes are orders of magnitude bigger, we still do not know
what the government losses will be. It is conceivable they may
approach $1 trillion, though that does not seem likely at the
moment. This is another reason why steps one and two are
urgently needed at the very outset of the crisis to clarify prices.
Government Purchases of Assets
The government could replace the lost optimistic capital by
buying distressed securities directly. In effect, the Treasury
would take conservative and pessimistic taxpayers’ money that
would never be invested in these securities, and invest it there,
assuming, of course, that it did so with the expertise necessary
to make reasonably sound judgments on which securities to
buy and how much to pay for them. This was the plan that
Secretary Paulson originally proposed.
Government buying plans are a risky approach—riskier
than the steps I have laid out above—and thus, if ever used,
must be implemented with extreme care. An argument that is
often blithely made for government buying is that when
security prices are terribly depressed in “fire sales,” the
government might make some good investments. It is likely,
the argument goes, that the general taxpayer is too
conservative, and by transforming pessimistic capital into
optimistic capital, the government might even be directly
helping the taxpayer, while at the same time staunching the
collapse of security prices.
Forcing natural pessimists into purchases they fear, however
much potential financial upside, may well undermine public
confidence in government, especially if the investments start to
go bad. But even if taxpayers were on board, caution should be
the watchword. The lending mentioned earlier (a much more
direct approach to restoring leverage) would probably raise
security prices, so the government purchases would not be at
rock-bottom prices. Private investors (naturally more agile and
quicker than the government), knowing that the government
would be buying, would rush to buy first, reducing potential
government profits. Of course, that, in some sense, would be
what the government would want to happen because it would
mean that security prices would rise more quickly. But it might
also result in taxpayers getting stuck with the worst assets,
causing public outrage and charges of foul play.
The biggest obstacle and the one that apparently stopped
Secretary Paulson’s original plan to buy the troubled assets is
the enormous challenge of deciding what to buy, and at what
price. We must not forget that the downward swing in the
leverage cycle is always triggered by genuine bad news, which I
call scary because it creates more uncertainty. Private investors
hold back for fear of “catching a falling knife”; the government
has far less expertise than these private investors. Since the
distressed mortgages are very heterogeneous, it is not at all clear
how the government acting alone could figure out what prices
to pay. Indeed, since Secretary Paulson’s call for government
purchases of distressed securities, a large number of them
(including most CDOs) have continued to lose value, with
some even going to zero. In retrospect, a program of
indiscriminate buying might have been a disaster. But how
could the government decide what to buy, and at what prices?57
The dangers of government buying look so profound that in
October 2008, I recommended that if the government were to
buy at all, it would be better for the government to invest
through professional money managers, again piggy-backing on
the choices they make to invest their own capital.58 To help
ensure that money managers had the right incentives, I also
recommended dividing the government money up among a
large number of private managers and making the investments
57
One suggestion that was made is by reverse auction. The government would
divide the securities into different categories, and then buy from each category
those securities that the current asset holders are willing to sell for the lowest
price. But how would the government decide what the categories are and how
much to spend on each? And how would it be protected from sellers’ efforts to
unload the worst securities in each category? If the purchases were to be made
by an auction mechanism, I would have suggested a variation in which private
bidders were allowed to enter the auction, not just private sellers. I would have
recommended that the government commit to buying half the winners’
purchases, at their winning prices. That way, the government could ride on the
expertise of the private buyers. Still, even that solution could be gamed,
particularly given that some private buyers might hold other positions—I am
thinking of CDS here—that made it worthwhile for them to overbid in a
manner that might not be easy to deter or discover.
58
See Geanakoplos (2008).
and returns of these companies very public. These managers
would then be competing with each other on a world stage to
see how their investments performed. A more conventional
incentive device would be to say that a manager gets no fees
until the return on the assets passes some hurdle. Only after the
taxpayers make money would the managers earn any fees.
The PPIP embodies a number of the same principles I
advocated. Under the PPIP plan, the government has set up
accounts with professional money managers in which each
government equity dollar is invested side-by-side in the same
securities with a dollar of investor capital. (This is in addition
to the money loaned to the managers.)
Should another crisis arise, the government must be aware
of the pitfalls of a large government buying program. The
government cannot appear to the public as enriching the
managers it entrusts with its money with fees that are too high.
However, they must be given incentives to perform well.
Otherwise, they might be tempted to spend taxpayer money
buying portfolios sold by the failing companies of their cronies,
in exchange for favors later on. Or they might pay less attention
to the government investments than to the investments of their
fee-paying clients. Or they might buy for the government with
an eye toward benefiting their private clients by raising prices
of assets the clients hold, or in some other way. These conflicts
of interest become more acute to the extent that the number of
managers is small and to the extent that they each have a huge
amount of government money to wield. For example, a big
enough buyer with government money could conceivably offer
to rid a bank of toxic assets, at favorable prices, in exchange for
favors like easier credit later on.
Another potential pitfall in government buying is the
perverse incentives it might set up among sellers eager to get
their securities purchased. For example, it may be that the
banks were waiting for the government purchase not just of
securities, but shaky whole loans too, and that hope may have
contributed to their failure to modify whole loans in a rational
manner.
Thus, even with all the advice I have offered about how the
government should buy if it must, buying may still not be a
wise policy, particularly not as a substitute for an adequate
lending program, such as I described above.
6. Moral Hazard
It is often said that with every bailout comes a moral hazard
that leads to a bigger problem the next time. The problem
would be that bailing people out in this crisis would lead to
higher leverage in the next cycle. There really is only one
reliable antidote to that, and that is regulation of leverage.
FRBNY Economic Policy Review / August 2010
127
One observation, which appeared in Geanakoplos and
Kubler (2005), is that general systemwide interventions, like
restoring sane leverage, in the crisis do not always create
deleterious incentives in the long run. Surviving a crisis means
tremendous profit opportunities in the good phase of the next
cycle. If a systemic intervention gives prudent firms a chance to
survive, rather than everyone going under, those firms will
have an increased incentive to be prudent. Bailouts that rescue
firms, no matter how imprudent they have been (in fact,
precisely because they in particular were imprudent), are the
source of moral hazard.
Some have suggested that writing down principal on
mortgage loans will also cause moral hazard. They say it will
encourage homeowners to behave badly, and the government
to intervene in too many markets, and threaten the sanctity of
contracts. I disagree, because the writing down of principal
could be done as a function of the decline in some index of
housing prices. The index is beyond the control of the
homeowner, so it does not distort homeowner incentives.
Moreover, it could be done first for homeowners who have not
defaulted yet, and only later for homeowners who have
defaulted under some special hardship. It could only be done,
as I have said, if it promises to bring more money to the lenders.
A good test of whether it is a good idea is whether it would be
written into the contract in the first place if people had thought
of the possibility of this much home price decline. I agree with
Shiller (2008), who suggests that just these kinds of mortgages,
with principal automatically reduced if some housing index
falls enough, could and will likely become the standard
mortgages of the future.
7. Managing the Ebullient Stage
of the Leverage Cycle
After this crisis passes, we must prepare for the next leverage
cycle. The first step is to constantly monitor leverage at the
securities level, at the investor level, and at the CDS level.
Every newspaper prints the interest rates every day, but
none of them mentions what margins are. The Federal Reserve
needs to settle on a menu of different security classes, monitor
their haircuts daily by talking to all the big lenders and
borrowers, and then make averages public on a regular
schedule, say every month or quarter.
The leverage of money managers could also be public.
Moreover, legislation and regulations could contain strong and
clear prohibitions against misleading the public or regulators
on the degree of leverage.
128
Solving the Present Crisis and Managing the Leverage Cycle
I discussed at great length in Sections 3 and 4 how CDS
contracts provide an opportunity to leverage, so these must
be monitored as well. Putting them on an exchange would
facilitate monitoring, as well as netting and ensuring enough
collateral is posted. All too often CDS insurance buyers allowed
the writers of insurance to get away without actually putting up
the collateral. Repo lending too must be reorganized so that
borrowers are protected in case the lenders go bankrupt and
swallow up the borrower’s collateral.
Transparency about actual leverage could bring a great deal
of discipline to the market, and warn investors of impending
trouble. In my earlier leverage charts, one can see the
tremendous spikes in margins during the crisis stages of the last
two cycles. One can also see a drift down in haircuts in the
ebullient stage of the last cycle.
But transparency alone is not enough. Some investors will
not curtail their leverage, no matter how much scrutiny by the
public, and how far out of line with recent practice they
become. Put bluntly, the market alone will not take care of
outsized leverage. It is thus imperative that the Fed put outside
limits on leverage. It will still be necessary to regulate leverage.
The lesson of the leverage cycle is that there are many
externalities (eight that I listed), and we should always expect
cycles of too much leverage followed by too little leverage.
The most direct way to regulate leverage might be by
empowering a “leverage supervisor” who could simply forbid
loans at too high leverage in ebullient times, setting different
leverage limits for different security classes. Banks would
simply not be allowed to lend 97 percent of the value of the
house, and repo lenders would not be allowed to reduce
haircuts too far.
Many people have argued that setting margin limits is
difficult because securities are so heterogeneous. But I believe
this problem will eventually be solved once the haircut data
history becomes more public. It was not obvious how to
manage interest rates either. But little by little, the Fed has
gotten better at it. The same will be true with leverage. The
combination of security leverage data, investor leverage data,
CDS leverage data, and asset price data could give the Fed
tremendous information for managing future leverage cycles
that it did not have, or chose to ignore, in this and in past
leverage cycles. The critical thing is that with the data in hand,
the Fed will be able to monitor dramatic changes in leverage
and asset prices, and therefore will easily recognize when we are
reaching either end of the cycle.
Another way of controlling leverage is to tax firms that
borrow excessively, or that borrow excessively on their
collateral, or that lend excessively on collateral. (The tax rate
again would have to differ depending on the kind of
borrowing.) A very small tax might go a long way to discourage
excessive leverage, and might also change the maturity
structure, inducing longer term loans, if it were designed
properly. Another advantage of the leverage tax is that revenues
from it could be used to finance the lending facility the Fed
would need to keep at the ready in anticipation of the downside
of future leverage cycles.
Yet another way of controlling leverage is by mandating that
lenders can only tighten their security margins very slowly.
Knowing they cannot immediately adapt if conditions get more
dangerous, lenders will be led to keep tighter margins in good,
safe times.
Leverage constraints have been proposed at the investor
level for selected financial firms. Congress is considering a hard
cap on bank leverage of 15. There are six potential advantages,
however, to limiting leverage at the securities level instead of at
the investor level. The first is that many people can leverage;
limiting leverage at banks or at a few other financial institutions
might just induce leveraged purchases to move somewhere
else. Second, the leverage of an investor is often a meaningless
number, at least as an indicator of credit tightness, since just
when things are getting bad, and margins on securities are
tightening and the whole economy is being forced to
deleverage, many firms will appear to be more leveraged
because their equity will be disappearing. (It has become
fashionable nowadays to say that leverage regulation should be
countercyclical, by which people mean that investor leverage
should be allowed to go up in bad times and down in good
times. Enforcing a hard cap on investor leverage would
paradoxically exacerbate the leverage cycle by forcing firms to
sell at the bottom of the cycle, even if they had long-term loans
that did not require rolling over.) Third, different securities
include different amounts of “embedded leverage.” Thus, it
makes sense to mandate different leverage numbers for
different securities. Setting an absolute leverage limit like 15,
independent of the portfolio mix, might induce banks to shift
their investments into securities with higher embedded
leverage. Fourth, a focus on securities leverage would lead to
derivatives such as CDS becoming part of the leverage
numbers. As we saw, writing CDS insurance is like owning the
underlying bond, so taking the ratio of the collateral required
on the CDS to the cash price of the bond gives a good measure
of the CDS leverage. Fifth, it is harder to hide securities leverage
than investor leverage; for one thing, there is a counterparty to
each security transaction reporting the same number that can
be used by regulators as a check on reported numbers. Finally,
a leverage supervisor managing securities leverage numbers
might be less vulnerable to political pressure because his
mandate would be more technical.
8. Conclusion
The leverage cycle brought us to the edge of a cliff. We have
moved back from the precipice, but unless we understand the
features of the leverage cycle and design our responses to
address the specific problems that characterize the end stage
of an outsized leverage cycle, we are left hoping for a miracle
to restore our financial prosperity. Marking time and waiting
for the miracle of things getting better appear to be part of the
current government policy, at least as it relates to housing and
foreclosures. That miracle, if it comes, will be nothing more
than the start of another cycle, maybe one even worse than
the one we have just experienced. My recommendations for
solving the present crisis and managing the leverage cycle
in its ebullient stage might prevent such an outcome.
FRBNY Economic Policy Review / August 2010
129
References
Adrian, T., and H. Shin. Forthcoming. “Liquidity and Leverage.”
Journal of Financial Intermediation.
Ferreira, F., J. Gyourko, and J. Tracy. Forthcoming. “Housing Busts
and Household Mobility.” Journal of Urban Economics.
Aragon, D., A. Caplin, S. Chopra, J. V. Leahy, Y. LeCun, M. Scoffier, and
J. Tracy. 2010. “Reassessing FHA Risk.” NBER Working Paper
no. 15802, March.
Foote, C. L., K. S. Gerardi, L. Goette, and P. S. Willen. 2009. “Reducing
Foreclosures.” Federal Reserve Bank of Boston Public Policy
Discussion Papers, no. 09-02.
Backman, A., and G. O’Connor. 1995. “Rating Cash Flow Transactions
Backed by Corporate Debt: 1995 Update.” Moody’s.
Fostel, A., and J. Geanakoplos. 2008. “Leverage Cycles and the Anxious
Economy.” American Economic Review 98, no. 4 (September):
1211-44.
Bernanke, B., M. Gertler, and S. Gilchrist. 1996. “The Financial
Accelerator and the Flight to Quality.” Review of Economics and
Statistics 78, no. 1 (February): 1-15.
———. 1999. “The Financial Accelerator in a Quantitative Business
Cycle Framework.” In J. B. Taylor and M. Woodford, eds.,
Handbook of Macroeconomics, vol. 1C, 1341-93. Elsevier.
Brunnermeier, M. K., and L. H. Pedersen. 2009. “Market Liquidity and
Funding Liquidity.” Review of Financial Studies 22, no. 6
(June): 2201-38.
Caballero, R. 2010. “The ‘Other’ Imbalance and the Financial Crisis.”
VOX, January 14.
Caballero, R., E. Farhi, and P.-O. Gourinchas. 2008. “An Equilibrium
Model of ‘Global Imbalances’ and Low Interest Rates.” American
Economic Review 98, no. 1 (March): 358-93.
Caplin, A., S. Chan, C. Freeman, and J. Tracy. 1997. Housing
Partnerships: A New Approach to a Market at a
Crossroads. Cambridge, Mass.: MIT Press.
Caplin, A., C. Freeman, and J. Tracy. 1997. “Collateral Damage:
Refinancing Constraints and Regional Recessions.” Journal
of Money, Credit, and Banking 29, no. 4, part 1 (November):
496-516.
Dana, D. Forthcoming. “The Foreclosure Crisis and the AntiFragmentation Principle in State Property Law.” University of
Chicago Law Review.
Farhi, E., and J. Tirole. 2009. “Collective Moral Hazard, Maturity
Mismatch, and Systemic Bailouts.” NBER Working Paper
no. 15138, July.
130
Solving the Present Crisis and Managing the Leverage Cycle
Geanakoplos, J. 1997. “Promises, Promises.” In W. B. Arthur,
S. Durlauf, and D. Lane, eds., The Economy as an Evolving
Complex System, vol. 2, 285-320. Reading, Mass.: AddisonWesley.
———. 2003. “Liquidity, Default, and Crashes: Endogenous
Contracts in General Equilibrium.” In M. Dewatripont, L. P.
Hansen, and S. J. Turnovsky, eds., Advances in Economics and
Econometrics: Theory and Applications, vol. 2, 170-205.
Econometric Society Monographs: Eighth World Congress.
Cambridge University Press.
———. 2008. “Solving the Present Crisis and Managing the Leverage
Cycle.” Remarks delivered at the Board of Governors of the Federal
Reserve System, Washington, D.C., October 3.
———. 2010. “The Leverage Cycle.” In D. Acemoglu, K. Rogoff, and
M. Woodford, eds., NBER Macroeconomics Annual 2009,
vol. 24, 1-65. Chicago: University of Chicago Press.
Geanakoplos, J., and S. Koniak. 2008. “Mortgage Justice Is Blind.”
Editorial. New York Times, October 30.
———. 2009. “Principal Matters.” Editorial. New York Times,
March 5.
Geanakoplos, J., and F. Kubler. 2005. “Leverage, Incomplete Markets,
and Pareto Improving Regulation.” Unpublished paper.
Geanakoplos, J., and H. Polemarchakis. 1986. “Existence, Regularity,
and Constrained Suboptimality of Competitive Allocations when
the Asset Market Is Incomplete.” In W. P. Heller, R. M. Starr, and
D. A. Starrett, eds., Social Choice and Public Decision
Making: Essays in Honor of Kenneth J. Arrow, 65-96.
Cambridge University Press.
References (Continued)
Geanakoplos, J., and W. Zame. 2009. “Collateralized Security
Markets.” Unpublished paper.
Kiyotaki, N., and J. Moore. 1997. “Credit Cycles.” Journal of
Political Economy 105, no. 2 (April): 211-48.
Gromb, D., and D. Vayanos. 2002. “Equilibrium and Welfare in
Markets with Financially Constrained Arbitrageurs.” Journal
of Financial Economics 66, nos. 2-3 (November-December):
361-407.
Minsky, H. P. 1986. Stabilizing an Unstable Economy.
New Haven, Conn.: Yale University Press.
Gyourko, J., and A. Saiz. 2004. “Reinvestment in the Housing Stock:
The Role of Construction Costs and the Supply Side.” Journal
of Urban Economics 55, no. 2 (March): 238-56.
Haughwout, A., E. Okah, and J. Tracy. 2009. “Second Chances:
Subprime Mortgage Modification and Re-Default.” Federal
Reserve Bank of New York Staff Reports, no. 417, December.
Haughwout, A., R. Peach, and J. Tracy. 2008. “Juvenile Delinquent
Mortgages: Bad Credit or Bad Economy?” Journal of Urban
Economics 64, no. 2 (September): 246-57.
Myers, S. C. 1977. “Determinants of Corporate Borrowing.” Journal
of Financial Economics 5, no. 2 (November): 147-75.
Peristiani, S., P. Bennett, G. Monsen, R. Peach, and J. Raiff. 1997.
“Effects of Household Creditworthiness on Mortgage
Refinancing.” Journal of Fixed Income 7, no. 3
(December): 7-21.
Shiller, R. 2008. The Subprime Solution: How Today’s Global
Financial Crisis Happened and What to Do About It.
Princeton, N.J.: Princeton University Press.
Tobin, J., and S. S. Golub. 1998. Money, Credit, and Capital.
New York: McGraw-Hill/Irwin.
Holmstrom, B., and J. Tirole. 1997. “Financial Intermediation,
Loanable Funds, and the Real Sector.” Quarterly Journal
of Economics 112, no. 3 (August): 663-91.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York
or the Federal Reserve System. The Federal Reserve Bank of New York provides no warranty, express or implied, as to the
accuracy, timeliness, completeness, merchantability, or fitness for any particular purpose of any information contained in
documents produced and provided by the Federal Reserve Bank of New York in any form or manner whatsoever.
FRBNY Economic Policy Review / August 2010
131
@FedFRASER