Full text of Business Review (Federal Reserve Bank of Philadelphia) : Second Quarter 2010

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

De-Leveraging and the Financial Accelerator:
How Wall Street Can Shock Main Street*
BY SATYAJIT CHATTERJEE

T

he severity of the recent economic downturn
raises questions about the role of financial
markets in modern market economies. Why
did rising defaults in a relatively small portion
of the U.S. housing market cause a financial crisis?
Why do financial crises have outsized adverse effects
on the rest of the economy? As a general rule, a decline
in economic activity in the nonfinancial sector, such as
occurs during a typical recession, induces greater restraint
on the part of the financial sector and that restraint —
manifested usually in a pullback of credit and funding
— in turn causes further setbacks to the nonfinancial
sector. In the academic literature, this feedback effect is
called the financial accelerator. In this article, Satyajit
Chatterjee looks at what underlay the financial shock
that emanated from Wall Street in the fall of 2007. Then
he focuses on the channels through which the financial
accelerator works and how the accelerator can turn a
financial market disruption into a deep recession.

In the first quarter of 2006, when
delinquencies on subprime mortgages

Satyajit
Chatterjee is a
senior economic
advisor and
economist in
the Research
Department of
the Philadelphia
Fed. This article
is available free of
charge at www.philadelphiafed.org/researchand-data/publications/.
www.philadelphiafed.org

first began their sustained rise, the
unemployment rate in the United
States stood at 4.75 percent. Under
the impact of the ensuing financial
crisis, the U.S. economy fell into
recession in December 2007, and the
unemployment rate shot up to 9.5

*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.

percent within a year and a half. What
began as a problem in the subprime
segment of the U.S. mortgage market
snowballed into a full-blown financial
crisis and one of the worst recessions of
the postwar era.
The severity of the current
downturn raises questions about the
role of financial markets in modern
market economies. Why did rising
defaults in a relatively small portion
of the U.S. housing market cause a
financial crisis? Why do financial
crises have outsized adverse effects on
the rest of the economy?
As a general rule, a decline in
economic activity in the nonfinancial
sector, such as occurs during a typical
recession, induces greater restraint on
the part of the financial sector and
that restraint — manifested usually in
a pullback of credit and funding — in
turn causes further setbacks to the
nonfinancial sector. In the academic
literature, this feedback effect is
called the financial accelerator. The
terminology alludes to the fact that
greater financial restraint can cause a
downturn to gather additional speed
or lesser financial restraint can cause
an upturn to do the same. When the
initial shock is a shock to the financial
sector itself, the financial accelerator
can combine with the shock to
produce a particularly steep decline in
economic activity.
First we’ll look at what underlay
the financial shock that emanated
from Wall Street in the fall of
2007, and then we’ll focus on the
channels through which the financial
accelerator works and how the
accelerator can turn a financial market
disruption into a deep recession.
Business Review Q2 2010 1

SOME BACKGROUND ON THE
FINANCIAL CRISIS
The financial crisis that erupted
in the fall of 2007 has its origins in
subprime mortgages, that is, loans
made to risky borrowers for the
purposes of buying a house. The
subprime segment of the U.S. housing
market is relatively small, so it is
puzzling that default on these loans
could become the source of a major
financial crisis.
One reason is leverage. The
financial firms (commercial banks,
investment banks, and hedge funds)
that bought the risky mortgages
funded these purchases by borrowing
from other financial market
participants. Thus, when these
mortgages began failing, it was not just
the financial firms that had bought
the mortgages that got into trouble, so
did the entities that had lent money
to the financial firms. These entities,
typically other financial firms, in turn
had borrowed money to fund their
loans; so the creditors of these other
financial firms also got into trouble.
Leverage is the reason that a relatively
small pool of failing assets can cause a
systemic problem. Leverage makes the
insolvency of one financial institution
a trigger for the insolvency of other
financial institutions.
But leverage alone can hardly
be the culprit for the financial crisis.
Leverage is at the heart of efficient
financial intermediation and has been
a fact of life in industrial economies for
centuries. A more important proximate
cause of the crisis was the manner in
which financial firms leveraged their
purchase of risky mortgages. They
funded their purchases by borrowing
short term. They promised their
investors that they could have their
funds back within a short period of
time. Since the mortgages bought
would not mature until many years
into the future, the cash flow from

2 Q2 2010 Business Review

the investment was insufficient to pay
off the maturing debt. The financial
firms made up the shortfall by issuing
new short-term debt. In most cases,
the new debt was absorbed by existing
investors. In other words, the financial
firms were relying on their investors
to “roll over” their loans as the loans
matured. The mode of operation of
financial firms was to fund purchases

same time, it becomes impossible for
the bank to meet its obligations.
Rising defaults on subprime
mortgages in 2006 led investors to
reassess the risks inherent in assets
based on subprime mortgages. As the
market value of these assets declined,
investors became worried that future
investors might refuse to issue new
loans against these suspect assets.

Leverage alone can hardly be the
FXOSULW IRU WKH ¿QDQFLDO FULVLV /HYHUDJH
LV DW WKH KHDUW RI HI¿FLHQW ¿QDQFLDO
LQWHUPHGLDWLRQ DQG KDV EHHQ D IDFW RI OLIH
LQ LQGXVWULDO HFRQRPLHV IRU FHQWXULHV
of long-term assets (risky mortgages)
with a sequence of short-term debt.
That is, they engaged in maturity
transformation.
While maturity transformation
is part of a well-functioning financial
system and financial firms will engage
in it (as well as in leverage) to generate
value for their investors, maturity
transformation entails some risks.
The danger is that if investors become
nervous about a firm’s solvency, they
can refuse to renew their loans to
the firm and thereby put the firm in
a real bind. This is called a rollover
crisis. Bank runs are a famous example
of this sort of crisis. In a bank run,
depositors rush to withdraw their
deposits from the bank because they
fear the bank will fail. Banks are
subject to rollover crises because they
engage in maturity transformation.
They borrow very short term (in effect
they tell their depositors they can
withdraw funds at any time), but they
use the deposits to purchase assets
that pay off gradually over time. If all
depositors (or a good many of them)
attempt to withdraw deposits at the

If that happened, the firm would be
unable to pay off the new loan it was
issued today. This lack of confidence
led to a rollover crisis in which current
investors refused to renew their loans
to financial firms. Of course, as
investors refused to renew their loans,
financial firms holding suspect assets
began to experience great difficulty in
meeting their short-term obligations.
In some cases, the firms simply went
bankrupt. In other cases, the firms
suffered huge losses in equity, since
they had used their own funds to
service their short-term obligations.
Leverage and maturity
transformation were the main
proximate causes of the financial crisis.
But it is important to understand
that these are proximate causes. The
reasons why large financial firms
engaged in this type of leveraging
remain a matter of controversy. As
researchers and analysts probe into the
ultimate causes of the crisis, they will
uncover some of the deeper reasons
as to how and why financial firms got
themselves into this bind. Also, it is
important to remember that leverage

www.philadelphiafed.org

and maturity transformation are not
problems per se. Perhaps the catalyst
that turned these well-known forms of
financial intermediation into a recipe
for financial ruin was the unwelcome
concentration of financial risk within a
handful of very large financial firms. It
appears that the considerable default
risk of subprime mortgages was not
passed on to ultimate investors such
as households, corporations, pension
funds, and insurance companies but
was instead absorbed by a few large
financial firms. Although the subprime
segment of the U.S. housing market is
relatively small, the concentration of
the default risk of subprime mortgages
in a few large financial firms ended
up causing a problem for the entire
financial system.
Now let’s turn to a discussion of
the main channels through which
a loss in equity in the financial
sector retards economic activity in
the nonfinancial sector. The goal is
to provide some perspective on the
severity of the downturn that followed
in the wake of the financial crisis.
DE-LEVERAGING AND THE
CREDIT CRUNCH
There are several channels
through which a loss in equity of
financial firms has adverse effects on
the nonfinancial sector. Some of these
channels involve direct effects and
others, indirect effects. Among the
direct effects is a pullback in the supply
of credit that results from de-leveraging
by financial firms.1
To understand the role of deleveraging, we need to understand
the balance sheet of a financial firm.
Table 1 gives a simple example of an
investment bank’s balance sheet. On
the asset side of the balance sheet

are loans made by the investment
bank. Typically, these loans are made
to the nonfinancial sector, which
includes businesses, households, and
the government. In the example, the
investment bank has bought mortgages
from households worth $100 and has
business-sector loans worth $200. On
the liability side of the balance sheet,
the investment bank has debt worth
$250 in the form of commercial paper
and equity worth $300-$250, or $50.2
What the balance sheet says is that
the investment bank has invested $50
of its own money and $250 worth of
borrowed money to purchase $300
worth of assets.
An important aspect of the
balance sheet is the leverage ratio,
defined as the ratio of the value of
assets to equity; in the example, the
leverage ratio is 6 (300÷50). An
investment bank likes to maintain a
target leverage ratio that is low enough
so as to assure investors who lend it
money that there is a high probability
their loans will be repaid. In the
example, the investment bank can
sustain losses of up to $50, or one-sixth
of the value of its assets, and still be

2

Commercial paper is an unsecured promissory
note issued by large banks or corporations with
a maturity date of less than one year. Since the
loan is not backed by collateral, only highly
reputable firms can issue commercial paper.

in a position to pay off its creditors.
The higher the leverage ratio, the
less capacity the investment bank
has to absorb losses without affecting
its creditors.3 It stands to reason
that an investment bank’s target
leverage ratio will ultimately depend
on investors’ perception of risk in the
financial system. During periods of
low perceived risk, the target can be
expected to rise, and during periods of
high perceived risk, the target can be
expected to fall.
The important point is that an
adverse shock to the market value
of assets causes the leverage ratio to
rise above its target. To understand
why, suppose the market value of the
mortgages held by the bank declines
by 20 percent. In dollar terms, this
is a loss of $20. Generally accepted
accounting principles (GAAP) require
the investment bank to record the
value of mortgages on its books at
current market value. Thus, the bank
is required to mark down the value
of mortgages on its books from $100
to $80. This means that the bank’s
equity (which is simply the difference

3
A target leverage ratio is not directly observable (since financial firms do not announce it),
but its value can be inferred from the observed
behavior of financial firms. See the article by
Allen Berger, Robert DeYoung, Ozde Oztekin,
and David Lee for evidence in support of target
leverage ratios.

TABLE 1
Assets

Liabilities

Mortgages

$100

Commercial Paper

$250

Business Loans

$200

Equity

$50

Total

$300

Total

$300

1

The role of de-leveraging in the pullback of
credit is discussed at length in the article by
Adrian Tobias and Hyun Song Shin.

www.philadelphiafed.org

Business Review Q2 2010 3

between the value of its assets and the
value of its liabilities) drops from $50
to $30 ($280-$250) and its leverage
ratio rises to 8¹/³. If the bank’s target
leverage ratio was 6 to begin with,
the loss in the market value of assets
results in a leverage ratio higher than
the target.
An increase in the leverage ratio
above its target value makes investors
less eager to lend to the investment
bank. Investors will demand a higher
interest rate on any new commercial
paper issued by the investment bank
(to compensate for the higher risk
of loss) or stop buying the bank’s
commercial paper altogether. Thus,
market forces make it hard for the
bank to maintain the same level of
short-maturity debt as before.4 At this
point, the investment bank must either
raise more equity or reduce its assets
(both of which will lower its leverage
ratio). Raising equity is usually not
much of an option for banks in the
midst of a financial crisis, although
some financial firms did raise equity
in the early phase of the current crisis
and have returned to equity markets
in recent months. Typically, in a crisis,
the adjustment in the leverage ratio is
accomplished by reducing assets. For
instance, in the example above, the
bank might bring its leverage ratio
back to 6 by reducing its loans to the
business sector from $200 to $100,
with a corresponding $100 reduction

4

It is worth pointing out that the leverage ratio
is closely related to a bank’s capital ratio, a ratio
that plays an important role in bank regulation.
The capital ratio is simply the ratio of bank equity (capital) to a risk-weighted sum of bank assets. As such, it is closely related to the inverse
of the leverage ratio. A rise in the leverage ratio
will be accompanied by a decline in the bank’s
capital ratio. If a bank’s capital ratio falls below
the level determined by regulation, the bank is
required by law to take steps to increase its capital ratio. Thus, an increase in a bank’s leverage
ratio resulting from a drop in asset values may
force a bank to take steps to lower its leverage
ratio for regulatory reasons.

4 Q2 2010 Business Review

in commercial paper (from $250 to
$150). The bank’s balance sheet after
this adjustment will be as shown in
Table 2.
The amount by which the
investment bank must reduce its assets
is closely related to the leverage ratio
it would like to maintain, in this case
6. For every dollar decline in equity,
the bank must reduce assets by $6.
Therefore, a $20 decline in equity
requires the bank to shrink its assets
by $120. Taking into account the fact
that the $20 decline in equity was
triggered by a $20 decline in the value
of mortgages, the bank must reduce its
assets by an additional $100 ($120$20).
This process of reducing the
leverage ratio by reducing assets in
the wake of a loss in equity is called
de-leveraging. The important point to
note is that since the leverage ratio
is a number quite a bit greater than
one, de-leveraging can convert any
given decline in equity into a much
larger decline in investment bank
assets. Since a bank’s assets are mostly
loans to the nonfinancial sector,
de-leveraging results in a constriction
in the flow of credit to nonfinancial
firms. A reduction in the supply of
credit, in turn, raises the firm’s cost
of credit, thereby reducing firms’
demand for investment goods and
consumer spending by households and,
ultimately, lowers employment.

Two additional points are worth
making. First, how much assets have
to fall because of de-leveraging also
depends on what happens to the target
leverage ratio following the initial
shock to equity. At the start of the
current crisis, the rise in uncertainty
caused investors to look for lower
leverage ratios than was customary
in the recent past. This, in turn, led
investment banks to lower their target
leverage ratio and that became an
additional factor in the de-leveraging
engaged in by financial firms. To
continue with our example, suppose
that the new target leverage ratio is
4 instead of 6. Then, starting from
the position shown in Table 2, the
new balance-sheet position might
look like the one in Table 3. To get its
leverage ratio down to 4, given equity
of $30, the firm must reduce its asset
holdings to $120 ($30 times 4). Thus, it
must reduce its assets by $60 ($180$120). In the example in Table 3, the
reduction is accomplished by reducing
mortgages and business loans by $30
each. On the liability side, the bank’s
commercial paper declines by $60
(from $150 to $90).
Second, a reduction in investment
bank debt (in the example, commercial
paper) goes hand-in-hand with the deleveraging. Given this, it is important
to ask: What happens to the funds
that investors were formerly lending to
this bank? In a crisis, the funds end

TABLE 2
Assets
Mortgages

Liabilities
$80

Commercial Paper

$150

Business Loans

$100

Equity

$30

Total

$180

Total

$180

www.philadelphiafed.org

up in the hands of entities that can
borrow with very low risk of default.
Economists refer to this re-allocation
of funds from risky borrowers to safe
borrowers as a flight to quality.5 Of
course, the safe assets are bought from
existing holders of these assets, who
are likely to use the proceeds from
their sale to obtain other safe assets,
such as deposits at commercial banks.
So the process of de-leveraging during
a crisis is likely to increase deposits at
commercial banks, and if commercial
banks do not lend out this new inflow
of funds, it will also increase the
reserves that the banks hold with the
Fed. Thus, during a crisis, the process
of de-leveraging tends to move funds
out of circulation and into reserves,
which also puts downward pressure
on the inflation rate as money in
circulation tends to fall (or grow more
slowly).6
FALLING PROPERTY VALUES,
DEBT CAPACITY, AND THE
FINANCIAL ACCELERATOR
The decline in the flow of credit
resulting from de-leveraging also has
adverse consequences for property
values. The reason is that the value
of many properties, such as residential
homes and office buildings, is sensitive
to the free flow of credit. When credit
is not easily available, people and
businesses cannot easily buy houses
and commercial property. This causes
a drop-off in the demand for such
property and results in a fall in their
market price. For instance, when a

5
See the article by Evan Gatev and Philip
Strahan and the article by William Lang and
Leonard Nakamura for evidence on the “flight
to quality” during earlier contractionary episodes.

TABLE 3
Assets

Liabilities

Mortgages

$50

Commercial Paper

$90

Business Loans

$70

Equity

$30

$120

Total

$120

Total

homeowner who wishes to sell his or
her house cannot find many buyers
who can get financing (to buy the
house), he or she may be tempted to
drop the asking price. The same is true
for commercial properties.
A decline in the value of
residential and commercial properties
has further consequences for the
level of aggregate spending. The
reason is that a decline in property
values reduces the debt capacity of
businesses and households — which
is the maximum amount they are
permitted to borrow — and thereby
reduces business investment and
consumer expenditures (and,
ultimately, aggregate output). Thus,
falling property values lead to a
decrease in credit offered to the
nonfinancial sector. This effect is
what economists call the financial
accelerator. To understand how the
financial accelerator works, we need
to understand why there is a debt
capacity and why it declines with
property values.7
Commercial and residential
properties often serve as collateral in
business and household borrowing.

6

A consequence of de-leveraging is that safe
borrowers get to borrow at a lower interest rate.
This could potentially mitigate the contractionary
effects of de-leveraging except that the primary
beneficiary of the flight to quality tends to be the
government, not the private sector.

www.philadelphiafed.org

7

This discussion draws upon the ideas in the
article by Ben Bernanke and Mark Gertler
and the article by Nobuhiro Kiyotaki and John
Moore, especially the latter.

For instance, a business that wishes to
expand its operations could finance
the expansion by taking out a loan
from a bank using its property as
collateral for the loan. What this
means is that if the business cannot
repay the loan, the bank (the lender)
takes ownership of the property offered
as collateral against the loan. Banks
typically only make loans against
collateral because doing so makes
the loans less risky and encourages
borrowers to spend the borrowed funds
wisely (poor use of the funds results in
the loss of the collateral). Naturally,
there is a close connection between
the value of the collateral and the size
of the loan offered against it. Banks
are typically willing to lend up to some
fraction of the value of the property
offered as collateral. The maximum
amount that banks are willing to lend
against the borrower’s property is the
borrower’s debt capacity. When there
is a decline in the value of property
that can be offered as collateral, there
is a decline in the debt capacity of the
nonfinancial sector.
The reduction in debt capacity
reduces the flow of credit to firms
with productive uses for funds. Even
in the midst of a severe downturn,
there will be businesses that can put
funds to good use. There will also be
some financial intermediaries (the
ones unscathed by the crisis, perhaps)

Business Review Q2 2010 5

that will be eager to lend. But when
the debt capacity of businesses is
lowered by a decline in property prices,
businesses with good uses of funds
cannot borrow as much as they would
like. This financial constraint curtails
business investment and eventually
leads to an output level that is lower
than it would be in the absence of a
decline in debt capacity.
A decline in debt capacity is also
the reason a decline in home equity
depresses consumer spending. As
has been remarked upon many times
during the current crisis, households
borrow against their home equity
to pay for all kinds of consumer
expenditures. These expenditures
go well beyond home improvement
projects (which remain a main
motivation for home equity loans) and
encompass expenditures for which it
would be hard to get a loan directly.
A decline in residential property
prices reduces how much households
can borrow because the property (the
house in this case) being offered as
collateral is worth less.8 Once again,
even in the midst of a crisis, there
will be households that would like to
borrow more than their (reduced) debt
capacity, and these households will
have to reduce their spending. To the
extent that the decline in household
debt capacity constrains business
investments by small businesses that
rely on the owners’ assets to get loans,
the decline will have deleterious effects
on future output and employment as
well.
Every decline in property prices
reduces the debt capacity of the

8

If the household has a mortgage against the
property already, only the value of the house
in excess of the outstanding mortgage — the
owner’s home equity — can be offered as collateral against the new loan. A decline in the
value of residential property reduces home
equity dollar for dollar.

6 Q2 2010 Business Review

nonfinancial sector and thereby adds
fuel to the financial accelerator.
The downturn gathers further speed
and feedback effects kick in: As
unemployment rises and economic
activity declines, property prices
decline even more, which leads to
further decreases in debt capacity and
further decreases in expenditures,
output, and employment.
Eventually, this downward spiral
in property prices and economic
activity comes to a halt, in part
because the financial accelerator
begins to lose its potency. Recall
that the accelerator works through
a reduction in debt capacity, which
is the maximum level of debt the
nonfinancial sector can borrow

OTHER FACTORS FEEDING
THE ACCELERATOR
The previous section began with
the observation that the credit crunch
adversely affected property prices
because property prices are sensitive
to the free flow of credit. Given the
magnitude of the financial shock and
the consequent de-leveraging, the
credit crunch is probably an important
factor in the decline in property values.
However, other factors have played a
role in the decline in property values
as well and have therefore fed the
financial accelerator. We discuss the
more important channels here.
As already noted, declines in
residential house prices result in
declines in home equity and therefore

(YHU\ GHFOLQH LQ SURSHUW\ SULFHV UHGXFHV WKH
GHEW FDSDFLW\ RI WKH QRQ¿QDQFLDO VHFWRU DQG
WKHUHE\ DGGV IXHO WR WKH ¿QDQFLDO DFFHOHUDWRU
(given existing property prices). But
when economic activity is quite low,
debt capacity is not what constrains
investment on the part of firms
and households. They reduce their
investment simply because investment
is not remunerative when the general
level of economic activity is low. When
that stage is reached, a further decline
in property prices and debt capacity
does not cause additional reductions in
expenditure and output because there
are very few entities (businesses or
households) that would want to borrow
more than their debt capacity allows.
The downward spiral is also arrested
in part because of policy actions.
Accommodative monetary and fiscal
policies shore up expenditures and
therefore offset, to some extent, the
decline in expenditures that stems
from the operation of the financial
accelerator.

a decline in debt capacity. If the
household already has an existing
mortgage, the decline in house values
can lead to home equity becoming
negative. That is, the value of the
debt owed becomes larger than
the value of the property. In this
situation the homeowner may choose
to default on his or her mortgage.
In the run-up to the crisis, many
families bought homes with very low
down payments. Consequently, the
decline in house prices has resulted in
many families having negative home
equity. The result has been a huge
rise in foreclosures.9 Foreclosures, in
turn, depress house prices. Foreclosed
properties are sold at a heavy discount
because lenders (banks) that end up

9

For an excellent discussion of the connection
between negative home equity and mortgage
defaults, see the article by Ronel Elul.

www.philadelphiafed.org

owning them find it costly to hold
on to the houses. As the number
of foreclosures rises, the increased
presence of sellers willing to sell homes
at low prices puts downward pressure
on the price of all properties, including
those not in foreclosure.
In addition, foreclosures reduce
the demand for housing space
because families who lose homes
in a foreclosure end up renting less
space than they owned. They end
up renting because they cannot get a
new mortgage after defaulting on the
previous one, and they rent less space
than they owned because renting
does not have the tax advantages that
homeownership does. Thus, the overall
demand for housing space declines
with foreclosures. This puts further
downward pressure on house prices.10
The decline in house prices also
reduces the household net worth of
families who do not go into foreclosure.
Household net worth is the difference
in the value of all household assets and
all household liabilities. It is a measure
of household wealth. As house prices
decline, the value of household assets
declines. But there is no immediate
change in household liabilities (if the
household does not choose to default
on the mortgage), and therefore, there
is a decline in household wealth.
Lower wealth translates into lower
spending because families feel poorer
and spend less. Economists refer to
this as the wealth effect. The negative
wealth effect of a decline in household
net worth lowers consumer spending,
which lowers output and employment
and further depresses property prices.
Increased uncertainty about the
future also plays a role in reducing
current output and depressing

10

The interaction between the foreclosures, the
tax code, and the demand for housing space
was investigated in my recent paper with Burcu
Eyigungor.

www.philadelphiafed.org

property prices. Greater uncertainty
(higher probability of both good
and bad outcomes) makes firms and
households delay decisions that cannot
be easily reversed. Most investment
decisions fall into this category so that
uncertainty reduces expenditures on
business fixed investment.11 Greater

move out of circulation and into bank
reserves.
In sum, there are a host of factors
working to reduce property prices in
the wake of the crisis. The severity and
speed of the current downturn reflects,
in part, the operation of the financial
accelerator. The good news is that the

,QFUHDVHG XQFHUWDLQW\ DERXW WKH IXWXUH
DOVR SOD\V D UROH LQ UHGXFLQJ FXUUHQW RXWSXW
DQG GHSUHVVLQJ SURSHUW\ SULFHV *UHDWHU
XQFHUWDLQW\ PDNHV ¿UPV DQG KRXVHKROGV GHOD\
GHFLVLRQV WKDW FDQQRW EH HDVLO\ UHYHUVHG
uncertainty (especially uncertainty
regarding future earnings prospects)
increases a household’s desire for
precautionary savings (a rainy-day
fund), which reduces aggregate
consumer spending. Overall,
uncertainty can be a potent force
for lowering business and household
investment and is undoubtedly an
important factor in the current
downturn.12 It should be noted that the
increase in the precautionary savings
on the part of households and firms
(firms that delay investments park
their funds in safe financial assets) also
means that these entities allocate more
of their funds to safe assets, and this
is another factor putting downward
pressure on the yield on safe assets
and, ultimately, causing funds to

11
The role of uncertainty in delaying investment is discussed in the article by Ben Bernanke
and in the book by Avinash Dixit and Robert
Pindyck.
12
A discussion of how uncertainty affects
consumer spending can be found in my recent
Business Review article. It is worth noting that
there has been a sharp increase in the personal
savings rate in the U.S. since the crisis began.
The savings rate averaged less than 2 percent in
2007 but rose to around 4.5 percent in 2009 and
remains elevated.

financial accelerator works both ways.
As economic activity begins to revive,
perhaps because of accommodative
government policy or some good shock,
desired investment on the part of
firms begins to rise. At first, firms that
desire to borrow less than their debt
capacity (unconstrained firms) are the
ones that can get funding to undertake
their investment. But new investment
(and rising economic activity more
generally) puts upward pressure on
property prices. As property prices
begin to recover, constrained firms
(firms that would like to borrow more
than their debt capacity) can borrow
more as well because the increase
in property prices increases their
debt capacity. Of course, additional
investment increases aggregate output
in the short run and, eventually, in the
long run, as well. Thus increases in
property prices provide extra impetus
to the rise in economic activity in the
same way that declines in property
prices provided extra impetus to the
decline in economic activity at the
start of the downturn.
CONCLUSION
In the wake of the financial crisis,
the U.S. economy has suffered one

Business Review Q2 2010 7

of the worst recessions of the postWorld War II era. There is little doubt
that this episode will be the focus of
research and analysis for a long time
to come and our understanding of the
origins of the crisis and its aftermath
will evolve over time. At this point,
we can only give provisional answers to
the question: Why did rising defaults
in the subprime mortgage market
cause a financial crisis that led to such
a severe downturn?
This article suggests that leverage

and maturity transformation are the
proximate reasons as to why defaults in
a relatively small segment of the U.S.
housing market led to a financial crisis.
And the severity of the downturn is
most likely the result of an interaction
between declining property prices —
brought on by de-leveraging in the
financial sector — and the consequent
decline in the debt capacity of the
nonfinancial sector. This interaction,
called the financial accelerator in the
academic literature, has the potential

to feed on itself and cause a large
— and more or less simultaneous
— decline in property prices and
economic activity. This diagnosis
has implications for the future: If the
financial accelerator played a role in
making the downturn steep and quick,
we may expect it to play a role in the
recovery as well. When the recovery
takes root, the workings of the
accelerator will tend to make it sharp
and rapid. BR

Chatterjee, Satyajit, and Burcu
Eyigungor. “House Price Dynamics and
Foreclosures: A Quantitative Analysis of
the Mortgage Crisis and the Foreclosure
Prevention Policy,” Federal Reserve Bank
of Philadelphia Working Paper 09-22
(October 2009).

Kiyotaki, Nobuhiro, and John Moore.
“Credit Cycles,” Journal of Political
Economy, 105:2 (1997).

REFERENCES

Berger, Allen, Robert DeYoung, Ozde
Oztekin, and David Lee. “Why Do Large
Banking Organizations Hold So Much
Capital?” Journal of Financial Research,
34:2/3 (2008), pp. 123-50.
Bernanke, Ben. “Irreversibility,
Uncertainty, and the Cyclical Investment,”
Quarterly Journal of Economics, 98:1 (1983),
pp. 85-106.
Bernanke, Ben, and Mark Gertler.
“Agency Costs, Net Worth, and Business
Fluctuations,” American Economic Review,
79:1 (1989), pp. 14-31.
Chatterjee, Satyajit. “The Peopling of
Macroeconomics,” Federal Reserve Bank
of Philadelphia Business Review (First
Quarter 2009).

8 Q2 2010 Business Review

Dixit, Avinash, and Robert Pindyck.
Investment Under Uncertainty. Princeton:
Princeton University Press, 1994.
Elul, Ronel. “Residential Mortgage
Defaults,” Federal Reserve Bank of
Philadelphia Business Review (Third
Quarter, 2006).

Lang, William, and Leonard Nakamura.
“Flight to Quality in Banking and
Economic Activity,” Journal of Monetary
Economics, 36 (1995), pp. 145-64.
Tobias, Adrian, and Hyun Song Shin.
“Liquidity and Leverage,” Federal Reserve
Bank of New York Staff Reports No. 328
(January 2009).

Gatev, Evan, and Philip Strahan.
“Banks’ Advantage in Hedging Liquidity
Risk: Theory and Evidence from the
Commercial Paper Market,” Journal of
Finance, 61:2 (2008), pp. 867-92.

www.philadelphiafed.org

Monetary Policy in a Liquidity Trap*
BY MICHAEL DOTSEY

I

n the United States, the Federal Reserve sets
monetary policy by targeting the federal funds
rate. This process usually involves lowering
short-term interest rates when economic
growth is weak and raising them when economic growth
is strong. A wide class of economic models has shown
that, in theory, conducting policy in this way allows the
economy to employ resources efficiently. In addition, many
empirical studies have shown that most central banks
actually behave in this manner. In normal times, it is fairly
easy for the central bank to conduct policy in this fashion.
But there is one instance when conducting policy in this
manner becomes problematic: when the economy finds
itself in a “liquidity trap,” a situation in which the shortterm nominal interest rate is zero or very close to zero. In
this article, Mike Dotsey analyzes the difficulties a central
bank faces in such circumstances and discusses the tools
available to monetary policymakers. Policy as usual is not
an option, and the central bank’s framework for
conducting policy must change.
Monetary policy typically operates
by targeting a short-term interest
rate. For example, in the United

Mike Dotsey is a
vice president and
senior economic
policy advisor
in the Research
Department of
the Philadelphia
Fed. This article
is available free
of charge at www.
philadelphiafed.org/research-and-data/
publications/.
www.philadelphiafed.org

States, the Federal Reserve targets the
federal funds rate. In order to conduct
monetary policy, central banks
generally vary the short-term interest
rate target in response to economic
conditions. They do so because
setting the short-term interest rate
at a level consistent with economic

*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.

fundamentals generally attains both
the most efficient level of output1 and
an inflation rate consistent with longrun inflation objectives.
This process usually involves
lowering short-term interest rates when
economic growth is weak or inflation
or expected inflation is below some
desired rate and raising them when the
economy is growing strongly or when
inflation or expectations of inflation
are high. It has been theoretically
shown in a wide class of economic
models that conducting policy in this
way allows the economy to employ
resources efficiently. Low and stable
inflation is a desirable feature of a wellmanaged economy, and setting the
interest rate in a pro-cyclical manner is
consistent with economic efficiency.
This way of conducting monetary
policy is not just theoretically
sound. Many empirical studies have
shown that most central banks
actually behave in this manner. This
description of monetary policy —
varying the interest rate in response to
inflation and economic activity — is
called a Taylor rule or a Taylor-type
rule, named after John Taylor, who first
described these types of policies.
In normal times it is fairly easy
for the central bank to conduct policy
according to a Taylor-type rule. But
there is one instance when conducting
policy in this manner becomes
problematic: when the economy finds
itself in a “liquidity trap,” which is
defined as a situation in which the

1

The efficient level of output is the output
that would occur if all prices and wages were
continuously adjusted in response to changes in
economic conditions.

Business Review Q2 2010 9

short-term nominal interest rate is zero
or very close to zero.
Because the nominal interest
rate is generally bounded below by
zero, the central bank cannot lower
interest rates further even if it would
be desirable to do so, as it would be if
the economy were in a deep recession.
Furthermore, as I’ll discuss below, in
this situation, trying to stimulate the
economy by injecting more money
or liquidity through open market
operations may have little or no effect
on output. Therefore, it may appear
that monetary policy is impotent under
these conditions.
This article analyzes the
difficulties a central bank faces in such
circumstances and discusses the tools
available to monetary policymakers.
Policy as usual is not an option,
and the central bank’s framework
for conducting policy must change.
Importantly, it must change in ways
that alter individuals’ expectations of
what policy will be like when the zero
lower bound on interest rates is no
longer binding.
Thus, the conduct of monetary
policy becomes quite subtle and
depends on the credibility of proposed
future actions. Further, economists
have been concerned about the design
of appropriate monetary policy in a
liquidity trap for quite some time, and
in what follows, I will draw heavily
on the work of Gauti Eggertsson and
Michael Woodford; Alan Auerbach
and Maurice Obstfeld; and Paul
Krugman.
ECONOMIC PROBLEMS
ASSOCIATED WITH A
LIQUIDITY TRAP
To understand the economic
problems that ensue when an economy
is in a liquidity trap, we must first
understand the concept of the real
interest rate and its role in efficiently
allocating economic resources. What

10 Q2 2010 Business Review

follows will be a fairly abbreviated
analysis.2
The real interest rate, defined as
the nominal interest rate less expected
inflation, plays an important role in
determining what fraction of output
is consumed and what fraction is invested. In a perfectly competitive economy, the movement of the real interest
rate in response to economic shocks is
consistent with the optimal allocation
of economic resources. That is, the

is very weak, the real interest rate may
even become negative. A negative real
interest rate is sometimes observed
during recessions.
Generating a Liquidity Trap. If
the economy is sufficiently weak that a
real interest rate below zero is desirable, it is possible for the economy to
enter a liquidity trap. As indicated
above, the real interest rate is defined
as the nominal interest rate minus the
expected rate of inflation. But this

7KH UHDO LQWHUHVW UDWH GH¿QHG DV WKH
QRPLQDO LQWHUHVW UDWH OHVV H[SHFWHG LQÀDWLRQ
SOD\V DQ LPSRUWDQW UROH LQ GHWHUPLQLQJ ZKDW
IUDFWLRQ RI RXWSXW LV FRQVXPHG DQG
ZKDW IUDFWLRQ LV LQYHVWHG
real rate responds in such a way that
the level of output and its allocation
between consumption and investment
is the one that provides the highest
level of economic welfare. This interest
rate, which is associated with perfect
competition, is generally referred to
as the economy’s natural interest rate.
For example, strong economic growth
is associated with an opportune time
to make investments, especially if
that growth is generated by increased
productivity. At such times, consumers are also wealthier and hence desire
more consumption. In order to induce
enough saving for financing the optimal quantity of investment, the real
interest rate rises. Thus, resources are
allocated toward increasing the capital
stock, which, in turn, results in higher
future output, higher future consumption, and higher wages. Analogously,
when the economy is weak, the real interest rate falls, and when the economy

2

For a more detailed discussion, see my 2004
Business Review article.

means that the nominal interest rate
is the sum of two components: the real
interest rate and the expected rate of
inflation. This relationship is known
as the Fisher equation. Importantly,
the nominal interest rate cannot be
negative because no one would lend at
a negative rate. If they did, they would
get less money back than they lent,
and they would be better off putting
their money in their mattress. Thus,
in a liquidity trap, when the nominal
interest rate is zero and a negative
real interest rate is also desirable, the
Fisher equation implies that expected
inflation must be equal and of opposite
sign to this negative real interest rate.
Therefore, the desirability of a negative
real interest rate implies the desirability of positive expected inflation.
If features of the economy prevent
prices from adjusting flexibly, expected
inflation may, in the end, not be
high enough to generate a sufficiently
low real interest rate. The monetary
authority is also unable to lower the
nominal rate below zero. Thus, in addition to the economic shocks that are

www.philadelphiafed.org

responsible for the recession, interest
rates cannot adjust in an optimal way.
The presence of the liquidity trap
places the economy in even greater
jeopardy. Furthermore, because money
and bonds are now perfect substitutes — each is earning a zero rate of
interest — the inflation rate is not a
current monetary policy phenomenon.
The fact that both assets are now
earning the same zero rate of interest
implies that the public is indifferent
between the relative amounts of money
and bonds in its portfolio.3 Therefore,
current open market operations that
alter the amount of bonds and money
in public hands have no impact on
inflation. Second, with no opportunity
cost for holding money, the public is
willing to hold just about any amount
of money the central bank supplies.
Thus, current injections of money
have little effect on prices or inflation.
This is why the occurrence of a zero
nominal interest rate is called a liquidity trap.
However, future monetary policy
can prove effective in the current
environment, but understanding the
subtle and indirect way in which that
happens requires an understanding of
how monetary policy affects prices in
more normal times.
Controlling the Price Level and
Inflation. In normal times, standard
economic models suggest that a central
bank should adjust the short-term
nominal interest rate one-for-one with
perceived movements in the real interest rate. This type of policy engenders
an efficient economic response to the
various types of disturbances that

3

Currency earns a zero rate of interest and other
types of money, such as bank reserves, have,
until quite recently, earned a zero rate of interest. When short-term bonds, such as Treasury
bills, earn a positive rate of interest, holding
money incurs an opportunity cost in terms of
forgone interest.

www.philadelphiafed.org

affect economic activity. Moreover,
this type of policy is consistent with
a policy of low and stable inflation.
Only policy changes that move the
real interest rate by larger amounts
than dictated by underlying economic
fundamentals have a substantive effect
on inflation and economic activity. For
example, a severe tightening of policy
raises the short-term real interest rate
above its efficient or natural level, temporarily choking off consumption and
investment. The tightening of policy
also brings down inflation. A good
example is the disinflation during the
tenure of Fed Chairman Paul Volcker,

rate to the natural rate would require a
nominal rate of -2.0 percent, which is
impossible.
This higher-than-natural real
rate will serve to choke off aggregate
demand beyond what occurs due
to economic disturbances, and the
economy will be in for a deeper recession than it otherwise would be. This
is the situation in which the liquidity
trap has severe consequences and why
all central banks endeavor to keep the
economy out of these circumstances.
As discussed, this is also the situation in which the nominal rate cannot
be lowered further, and standard mon-

,Q QRUPDO WLPHV VWDQGDUG HFRQRPLF PRGHOV
VXJJHVW WKDW D FHQWUDO EDQN VKRXOG DGMXVW WKH
VKRUWWHUP QRPLQDO LQWHUHVW UDWH RQHIRURQH
ZLWK SHUFHLYHG PRYHPHQWV LQ WKH UHDO
LQWHUHVW UDWH
when the Fed maintained very high
nominal and real interest rates. This
policy contributed to the two ensuing
recessions and a significant lowering of
the inflation rate. The opposite occurs
when the central bank reduces the real
interest rate below its natural rate. The
result is temporarily higher output and
an increase in the inflation rate.
However, a liquidity trap is a time
when the central bank would like to
bring the real rate down. Therefore, in
theory, the central bank should desire
an increase in near-term inflation that
makes the real interest rate negative
enough so that the economy is able to
best cope with the fundamental factors
that have reduced output growth. For
example, suppose the natural real interest rate is -3.0 percent and inflation
expectations are 1.0 percent. The zero
lower bound on nominal interest rates
implies that the real interest rate in
financial markets can, at best, be lowered to -1.0 percent. To lower the real

etary policy that relies on simple Taylor-style interest rate rules is helpless
in ameliorating the weakness in the
economy. Unfortunately for Japan in
the 1990s and the U.S. economy today,
this is where we find ourselves. Fortunately, there are policies the central
bank can follow that will mitigate the
effects of the liquidity trap, but policies
in this situation involve departing from
normal operating procedures and the
rules that normally govern monetary
policy. As a result, these alternative
policies may be difficult to communicate, and because liquidity traps are
rare events, these policies may not be
deemed fully credible since the public
has little experience with these situations, as well.
MONETARY POLICY IN THE
LIQUIDITY TRAP
Credibility is an essential feature
of the simple policy I will discuss
and a feature of any successful

Business Review Q2 2010 11

monetary policy during a liquidity
trap, and it may be an even more
important ingredient than when the
economy is functioning under normal
circumstances.4 The reason is that
the central bank must depart from
its normal behavior, and the public,
having little experience with a liquidity
trap, may not believe that policy has
actually changed. Absent perfect
credibility, the policies described below
would lead to very different and much
less beneficial economic outcomes.
If the economy is in a liquidity
trap and the weakness in the economy
is significant, it may be desirable
to generate an increase in inflation
expectations. In our previous example,
lowering the financial real interest
rate to a desirable -3.0 percent requires
inflation expectations to increase to
3.0 percent. However, doing so requires
the public to believe that future
inflation will indeed reach 3.0 percent.
The success of altering future
policy also requires that the economy
not be in the liquidity trap forever.
Historically, all instances of actual
liquidity traps have been temporary.
The current crisis appears to be
temporary as well, and it appears
that the public believes this to be the
case. That inference is based on the
fact that long-term interest rates are
currently positive. Because longterm interest rates are an average of
current and future short-term interest
rates, a positive long-term interest
rate implies that at some point in the
future short-term interest rates will be
positive as well. Hence, the evidence
from long-term bond markets indicates
that the zero lower bound will not
last indefinitely. Liquidity traps,

4

For a discussion of the importance of credibility in general, see my 2008 article and the
Federal Reserve Bank of Philadelphia’s 2007
annual report.

12 Q2 2010 Business Review

fortunately, appear to be temporary
phenomena.
Role of Nominal Interest
Rate in a Liquidity Trap. We have
emphasized that there is nothing
current monetary policy can
accomplish while the economy is in a
liquidity trap. However, once economic
activity recovers to the point at which
the nominal interest rate is positive,
monetary policy can influence the
level of economic activity. So at some

economic activity more than offsets
the cost of somewhat higher inflation.
But because the commitment pertains
to future actions, it will have an effect
only if the policy is believed. This
feature is an important component
of the influential work of Gauti
Eggertsson and Michael Woodford,
who have analyzed the liquidity
trap in great depth. An important
theme resonating throughout their
analysis is policy’s ability to influence

,Q D VWDQGDUG WKHRUHWLFDO PRGHO D
commitment by the central bank to
WHPSRUDULO\ LQFUHDVH IXWXUH LQÀDWLRQ DERYH
ZKDW LW ZRXOG EH LQ WKH DEVHQFH RI D
OLTXLGLW\ WUDS LV D EHQH¿FLDO SROLF\ UHVSRQVH
ZKHQ WKH HFRQRP\ LV LQ D OLTXLGLW\ WUDS
point in the future, a lower-thannormal future short-term nominal
interest rate will stimulate future
economic activity.
Generating increased output
growth in the future can have
consequences for current output.
Investment now becomes more
attractive, and firms may be reluctant
to lay off as many workers if they
are confident that higher than
normal output is around the corner.
Expectations of better times ahead will
also stimulate current consumption.
The cost of the future monetary
stimulus will be that future inflation
would be higher than it otherwise
would have been.
Thus, in a standard theoretical
model a commitment by the central
bank to temporarily increase future
inflation above what it would be
in the absence of a liquidity trap is
a beneficial policy response when
the economy is in a liquidity trap.
The central bank makes such a
commitment because the gain in

expectations and, importantly,
inflationary expectations over long
horizons. By doing so, the monetary
authority influences the term structure
of real interest rates and thereby
influences current aggregate demand.5
So, even in an environment where
both prices and inflation respond
slowly to economic shocks and
monetary policy, the policies prescribed
by Eggertsson and Woodford have
substantial effects.
In their work, Eggertsson and
Woodford show that the zero bound
can cause a significant problem for
monetary policy in the case in which
the interest rate rule does not change
when the economy exits the liquidity
trap. That is, a Taylor-type rule that
works fine in normal times may not
work so well when there is a zero lower
bound problem.

5

The term structure of interest rates describes
the relationship between interest rates on bonds
of varying maturities.

www.philadelphiafed.org

A particularly important result
of their analysis is that many policies advocated in the popular press
when the economy is in a liquidity
trap with zero nominal interest rates
are not useful. In particular, in their
framework, not only are current open
market operations, which exchange
short-term bonds for money, irrelevant,
but temporarily providing additional
bank reserves through increased open
market operations will have no effect
on the economy, irrespective of the
types of assets the monetary authority
purchases.
This last result occurs because efficient pricing of, say, long-term bonds
that are currently yielding a positive
interest rate can have an effect on
behavior only if those purchases imply
a change in the path of short-term
rates. This is because, as mentioned,
long-term rates are merely an average
of short-term rates.6 Thus, any policy
response today that does not also reflect a change in future policy will not
affect future economic activity. Therefore, it will not affect future short-term
interest rates and hence should not
affect the long-term bond rate in any
meaningful way.
Two features of their model are
responsible for the ineffectiveness of
large-scale increases in central bank
liabilities, often called quantitative
easing: (1) any increases in money at
the zero bound is done through open
market operations and, therefore,
does not affect the value of government liabilities, and (2) any increase
in money, even if it is accomplished
via government transfers, is transitory. Thus, as in the analysis by Alan
Auerbach and Maurice Obstfeld, for
increases in money to be beneficial,

6

Eggertsson and Woodford’s argument is in fact
more general and encompasses the government’s purchase of any asset.

www.philadelphiafed.org

the increase must be permanent. By
necessity, the underlying interest rate
rule must change once the economy
escapes from the zero lower bound. If
policy returns to a normal interest rate
rule, the money injected during the liquidity trap will have to be withdrawn
to ward off an increase in the inflation
rate. But this action would be inconsis-

greater output growth in the future.
The increase in future output growth
implies greater output growth in the
present, when the zero lower bound is
binding, and implies that the natural
interest rate is somewhat higher in the
current environment than it would be
absent the promise of future inflation.
Thus, Eggertsson and Woodford show

$ SROLF\ WKDW SHUPDQHQWO\ FKDQJHV WKH
PRQHWDU\ EDVH WRGD\ PXVW DOVR EH DVVRFLDWHG
ZLWK D FKDQJH LQ WKH LQWHUHVW UDWH UXOH LI LW
LV WR KDYH HIIHFWV ,W LV QRW MXVW WKH FXUUHQW
VHWWLQJ RI WKH LQWHUHVW UDWH WKDW LV LPSRUWDQW
EXW WKH SDWK WKDW SROLF\ VHWV IRU IXWXUH
VKRUWWHUP LQWHUHVW UDWHV PDWWHUV DV ZHOO
tent with the higher inflation promised
while the economy was in the liquidity trap. Hence, if the public believes
that upon exiting the liquidity trap the
central bank would immediately return
to normal policy, the promise of additional near-term inflation would not
have been believed in the first place.
Thus, a policy that permanently
changes the monetary base today must
also be associated with a change in the
interest rate rule if it is to have effects.
It is not just the current setting of the
interest rate that is important, but the
path that policy sets for future shortterm interest rates matters as well. This
is analogous to saying that the systematic component of policy is important
and that more importance should be
attached to what will be done in the
future than what is done today.
But there is an additional subtlety
here. As mentioned, a change in future
policy implies that the central bank
must tolerate additional inflation in
the future even after the zero bound is
no longer a problem. This policy leads
to less deflation at the zero bound and

that the economic losses associated
with a real interest rate that is too low
can be reduced.
A SPECIFIC POLICY
Eggertsson and Woodford provide
specific policy advice for the central
bank when a liquidity trap occurs.
The specifics of their proposal are
complex and particular to their
model. However, they suggest that
dealing with the public’s expectations
when the economy is in a liquidity
trap will take some skill on the part
of any central bank. Interestingly,
in their framework, a simple pricelevel targeting rule comes very close
to achieving the best outcome, and
such a policy should be relatively
easy to communicate. Rather than
targeting inflation per se, as is typical
of most central bank behavior, in a
liquidity trap, the central bank should
actually target the path of prices.7 The

7

For a detailed discussion of price-level targeting, see the article by Alexander Wolman.

Business Review Q2 2010 13

important distinction is that a price
path implies that should inflation
be relatively low today so that the
price level is below its target, future
inflation must increase to get the price
level back on track. Therefore, the
occurrence of deflation would require
higher future inflation, and as we have
seen, somewhat higher than normal
inflation is a useful mechanism for
ameliorating the adverse effects of a
liquidity trap.8
A price-level target is a way of
formalizing that policy prescription.
Because no central bank employs a
price-level target, that could make
credibility for this option problematic.
The proposal could be couched as a
time-varying inflation target, whereby
the targeted inflation rate would be
the rate that would get prices back to
the price-level path. But, again, the
public has little experience with such
a rule. Thus, establishing credibility
for future expansionary policy is an
essential, but perhaps difficult, feature
of successful policy at the zero lower
bound.
Thus, a central message of
Eggertsson and Woodford’s research is
that the monetary authority must be
able to commit to expansionary policy
once the zero-lower-bound problem
is alleviated. In particular, it must
commit to higher inflation than would
otherwise occur if the zero bound
had not been reached. A proposal of
raising the price of long-term debt
or, equivalently, lowering long-term
interest rates is consistent with the
optimal lower future path of short-term
rates. It could, therefore, be useful for

a central bank to buy large quantities
of long-term debt as a way of signaling
its intention to increase near-term
inflation and inflation expectations. In
this case, not carrying through on its
implied promise would result in a fall
in bond prices and a capital loss for the
central bank.9
A LARGE INCREASE IN THE
FED’S BALANCE SHEET
However, a potential challenge
from the standpoint of the monetary
authority is that once higher shortterm inflation is realized, the public
will alter its expectations of inflation
and the central bank will now be
facing an inflation scare and the
problems that accompany a departure
of inflation expectations from target.

A lack of perfect
FUHGLELOLW\ ZKLFK PD\
be an unavoidable
UHDOLW\ DFWV DV D WZR
HGJHG VZRUG WKDW
PDNHV GHDOLQJ ZLWK D
OLTXLGLW\ WUDS GLI¿FXOW
Problems such as these have been well
documented in Marvin Goodfriend’s
study and discussed in my essay with
Charles Plosser.
Thus, a lack of perfect credibility,
which may be an unavoidable reality,
acts as a two-edged sword that makes
dealing with a liquidity trap difficult.

Without full credibility, it is hard
to generate an increase in inflation
beyond what the public would
normally expect, and if that inflation
is generated, it subsequently may be
difficult to return expectations of
inflation to ones that are consistent
with price stability. As discussions in
the media suggest, the current large
increase in the Federal Reserve’s
balance sheet could represent such a
threat to the credibility of the Fed’s
long-run inflation target.10
The concern being expressed is
that if it becomes difficult to unwind
some of the assets currently on the
balance sheet, the future money supply
could be permanently higher. However,
with interest rates returning to normal
levels, the demand for money will not
be permanently higher. A permanent
increase in the money supply without a
permanent increase in money demand
can only lead to higher prices and
higher inflation.
Currently, there is every
expectation that the Fed will
successfully reduce its balance sheet
as the banking system recovers, and
survey data on inflation expectations
confirm this belief. Managing that
expectation is thus an important part
of policy, as evidenced in a number
of speeches by Federal Reserve
policymakers, including Philadelphia
Fed President Charles Plosser.11 It has
become increasingly important for
the Federal Open Market Committee
(FOMC) to articulate an exit strategy
and to indicate to the public that
it will follow an exit strategy that
does not ignite future inflation.

9

8

In other models, such as the one in the study
by Andrew Levin, David Lopez-Salido, Edward
Nelson, and Tack Yun, a price-level target does
not duplicate optimal policy nearly as well.
Their model calls for even more aggressive
policy, which leads to a permanent increase in
the price-level path.

14 Q2 2010 Business Review

Alternatively, as Lars Svensson has suggested,
the central bank could deflate the value of the
currency using an exchange-rate peg. Doing so
would require purchasing foreign assets, and
this policy may also be useful in establishing
credibility for higher inflation. If higher inflation is not forthcoming, the home currency
would appreciate, and the foreign assets on the
central bank’s balance sheet would depreciate,
resulting in a capital loss for the central bank.

10

The size of the Federal Reserve’s balance sheet
has more than doubled from $954 billion on
September 17, 2008, to slightly more than $2
trillion as of August 26, 2009.

11

See, for example, the speech by Charles
Plosser.

www.philadelphiafed.org

Indeed, the FOMC has been quite
explicit concerning its intentions for
maintaining long-run price stability.
SUMMARY
This article describes the
difficulties of conducting monetary
policy when there is a liquidity trap.
A very weak economy can require
negative real interest rates, and rates
that are sufficiently negative can be
hard to achieve when the short-term
nominal interest rate is bounded
below by zero. Although, in theory,
generating increased expectations of

future inflation is helpful, this may be
difficult to achieve because standard
monetary policy that targets a nominal
interest rate is ineffective once
nominal interest rates have reached
zero. Achieving the necessary increase
in expected inflation falls on promises
of future policy, but successfully
accomplishing this goal may require
credibility for temporarily deviating
from the central bank’s long-run
inflation target. Furthermore, that
deviation, if successful, could result
in the public’s no longer believing in
the long-run target. The central bank

could face a future destabilization
of inflation expectations and all the
problems that ensue when that occurs.
Thus, a liquidity trap is a
perilous place for the economy
and a central bank. Successfully
navigating a liquidity trap requires
open communication and transparency
because it requires the public to
understand not only current policy but
future policy as well. BR

Eggertsson, Gauti B., and Michael
Woodford. “The Zero Bound on Interest
Rates and Optimal Monetary Policy,”
Brookings Papers on Economic Activity, 1
(2003), pp. 139-233.

Plosser, Charles I. “The Economic Outlook
and Some Challenges Facing the Federal
Reserve,” speech at the 2009 Economic
Outlook Panel, University of Delaware,
January 14, 2009.

Goodfriend, Marvin S. “Interest Rate
Policy and the Inflation Scare Problem:
1979-1992,” Federal Reserve Bank of
Richmond Economic Quarterly, 79:1
(Winter 1993), pp 1-23.

Svensson, Lars E. O. “The Zero Bound in
an Open Economy: A Foolproof Way of
Escaping from a Liquidity Trap?” Monetary
and Economic Studies (Special Edition)
(February 2001), pp. 277-312

Krugman, Paul R. “It’s Baaack: Japan’s
Slump and the Return of the Liquidity
Trap,” Brookings Papers on Economic
Activity, 2 (1998), pp. 137-205.

Wolman, Alexander L. “Staggered Price
Setting and the Zero Bound on Nominal
Interest Rates,” Federal Reserve Bank
of Richmond Economic Quarterly, 84:4
(1998), pp. 1-24.

REFERENCES

Auerbach, Alan, and Maurice Obstfeld.
“The Case for Open Market Purchases in a
Liquidity Trap,” American Economic Review
(March 2005), pp. 110-37.
Bernanke, Ben S. “Four Questions about
the Financial Crisis,” speech, Morehouse
College, April 14, 2009.
Dotsey, Michael. “How the Fed Affects
the Economy: A Look at Systematic
Monetary Policy,” Federal Reserve Bank of
Philadelphia Business Review (First Quarter
2004), pp. 6-15.
Dotsey, Michael. “Commitment Versus
Discretion in Monetary Policy,” Federal
Reserve Bank of Philadelphia Business
Review (Fourth Quarter 2008), pp. 1-8.
Dotsey, Michael, and Charles I. Plosser.
“Commitment versus Discretion in
Monetary Policy,” Federal Reserve Bank
of Philadelphia 2007 Annual Report, pp.
4-17.

www.philadelphiafed.org

Levin, Andrew, David Lopez-Salido,
Edward Nelson, and Tack Yun.
“Limitations on the Effectiveness of
Forward Guidance at the Zero-Lower
Bound,” International Journal of Central
Banking, 6:1 (March 2010).

Business Review Q2 2010 15

Hiring, Job Loss, and the
Severity of Recessions*
BY R. JASON FABERMAN

T

he hiring and firing decisions of individual
businesses are one of the drivers behind
movements in the unemployment rate during
expansions and recessions. Whether a
recession is driven by large job losses or weak hiring will
greatly affect the composition and consequences of the
unemployed and can have important policy implications.
The extent to which recessions are times of weak hiring
or high job loss depends in large part on the severity of
the downturn. A recession is a time when the fraction of
businesses that are expanding goes down and the fraction
of businesses that are contracting goes up. A severe
recession is one in which the shift in this distribution is
more dramatic. In this article, Jason Faberman discusses
how the severity of a recession determines whether high
job loss or weak hiring will be the more important source
of declining employment and rising unemployment
through disproportionate changes in the distribution of
business-level employment growth.

What drives movements in the
unemployment rate during expansions
and recessions? Obviously, much of it

Jason Faberman is
a senior economist
in the Philadelphia
Fed’s Research
Department.
This article is
available free of
charge at www.
philadelphiafed.org/
research-and-data/publications/.
16 Q2 2010 Business Review

is driven by the hiring and firing decisions of individual businesses. When
businesses hire more workers than
they lose (whether those workers leave
voluntarily or involuntarily), employment expands and the unemployment
rate tends to go down. When businesses lose more workers than they

*The views expressed here are those of the
author and do not necessarily represent
the views of the Federal Reserve Bank of
Philadelphia or the Federal Reserve System.

hire, employment contracts and the
unemployment rate rises. This does
not mean, though, that boom times are
driven entirely by hiring and recessions are driven entirely by job losses.
For example, if firms cut back sharply
on their hiring with little change in
the number of workers they lose, the
unemployment rate would rise because
people would find it harder to find new
work.
Whether a recession is driven
by large job losses or weak hiring will
greatly affect the composition and
consequences of the unemployed and
can therefore have important policy
implications. Laid-off workers can
come from a variety of backgrounds.
Oftentimes, these workers lose valuable human capital in the process,
especially if the laid-off employees
are older, more experienced workers
with a lot of job-specific skills. Weak
hiring affects all individuals looking
for work: those who were recently laid
off, those just entering the workforce
(e.g., recent graduates), and those who
are currently employed but want a new
job. Weak hiring implies that there are
fewer jobs to apply for, which makes it
more difficult for the unemployed to
find work.
The recessions of the 1970s and
1980s, as well as the most recent
downturn, saw steep declines in
employment and sharp increases in
unemployment. At the same time,
the pace of layoffs was very high but
relatively short-lived. In comparison,
the fall in employment and the rise in
unemployment during the 1990-91 and
2001 recessions were much less severe.
During these recessions, there was a
moderate rise in job losses but a relawww.philadelphiafed.org

tively steep drop in hiring, particularly
during the 2001 recession. Furthermore, the 1990-91 and 2001 recessions
had declines that persisted well after
the official end of the recession.1
In academic circles, the contrast
in behavior has led to two diverging
views on recessions and the labor market. Some economists, such as Robert
Hall and Robert Shimer, focus on the
more recent downturns and take the
view that rising unemployment during
recessions is driven by weak hiring and
hence a low probability that the unemployed will find a job. Others, such
as Shigeru Fujita and Garey Ramey,
and Michael Elsby, Ryan Michaels, and
Gary Solon, cite the historical evidence and argue that rising unemployment is driven by high rates of job loss.
In reality, the extent to which
recessions are times of weak hiring or
high job loss depends on the severity
of the downturn. Severe recessions
are typically characterized by a sharp
drop in output and large amounts of
job loss, while moderate recessions
are characterized by smaller declines
in output and relatively weak hiring.
These results come about because, at
any point in time, there is a distribution of businesses that are expanding,
contracting, or keeping their employment steady. A recession is a time
when the fraction of businesses that
are expanding goes down and the fraction of businesses that are contracting
goes up. A severe recession is one in
which the shift in this distribution is
more dramatic. Furthermore, when
businesses expand or contract by a
certain amount, they tend to do so
with a fairly consistent mix of hiring,
quits (voluntary worker separations),
and layoffs (involuntary worker separa-

1

Here, “official” dates refer to the business cycle
peaks and troughs as designated by the National
Bureau of Economic Research (NBER).

www.philadelphiafed.org

tions). Fast-growing businesses tend
to have mostly hires, fast-declining
businesses tend to have mostly layoffs,
and businesses with smaller employment changes tend to have a mix of
hiring, quits, and layoffs that occur
simultaneously. During a severe recession, the number of businesses with
large contractions increases sharply. As
a result, the layoff rate at the national
level increases drastically. In contrast,
a mild recession generally has a smaller
increase in the number of contracting
businesses, so the resulting drop in hiring at the national level can outweigh
the more modest rise in the layoff rate.
HIRES, SEPARATIONS, AND
BUSINESS GROWTH
The Difference Between Gross
and Net Employment Changes. To
understand how the above findings
come about, we need to start with
the basic fact that the net change in
employment that we observe from the
Employment Situation Report of the
Bureau of Labor Statistics (BLS) each
month is the result of literally millions
of workers either starting or leaving a
job at thousands of businesses.2
We can examine gross changes in
employment in two ways: by tracking
the movements of the workers or by
tracking the employment behavior
of the businesses that employ them.
Shigeru Fujita details the first approach in an earlier Business Review
article, and he shows that following the
flow of workers between employment,
unemployment, and nonparticipation in the labor force provides much
more information on the state of the

2

The statistics in the Employment Situation
Report come from two surveys: a monthly
payroll survey, Current Employment Statistics,
which surveys businesses about their
employment, and a monthly household survey,
the Current Population Survey, which queries
households about the employment behavior of
their members.

labor market than looking at, say, the
unemployment rate or employment
growth alone.
The second approach provides
more insights as well, and it turns out
to be more useful for our purposes. Using it allows us to relate what are often
called worker flows, which are the gross
amount of hires or separations occurring in the economy, to the employment growth (or decline) at individual
businesses. Separations are the sum of
all quits, layoffs, and any other type of
separation, such as a retirement, and
the change in a business’s employment is simply the difference between
its total hires and total separations.
For example, if a business hired three
people and had one separation, its employment will have expanded by two
jobs. Given that businesses can have
hires while contracting and separations while expanding, one can have
complex interactions between worker
flows and business-level employment
growth.
Movements in National-Level
Worker Flows over Time. Next,
we need to know what the nationallevel patterns of the worker flows look
like. The Job Openings and Labor
Turnover Survey (JOLTS) of the BLS
reports the total amount of hiring,
quits, and layoffs at all businesses in
the economy each period. The data
measure the monthly rates of total hiring and total separations as a percent
of total employment, with the latter
broken out into quits (those who leave
their jobs voluntarily), layoffs (those
who are separated involuntarily), and
other separations (e.g., retirements).3
The JOLTS time series begins only in
December 2000 but now covers two
recessions.

3

Other separations are a very small fraction of
total separations and vary little with the business cycle, so I ignore them in this article.

Business Review Q2 2010 17

18 Q2 2010 Business Review

FIGURE 1
Hiring, Quits and Layoffs, 2000-2009,
JOLTS Data
Percent of Employment
5.0

Hiring Rate
4.0

3.0
Quit Rate
2.0
Layoff Rate

1.0

Dec-09

Jun-09

Dec-08

Jun-08

Dec-07

Jun-07

Dec-06

Jun-06

Dec-05

Jun-05

Dec-04

Jun-04

Dec-03

Jun-03

Dec-02

Jun-02

Dec-01

Jun-01

0.0
Dec-00

Figure 1 illustrates how the JOLTS
aggregate estimates behave over time.
The 2001 recession officially started
in March and ended in November of
that year, but employment losses (as
measured by the BLS payroll survey)
continued until August 2003. The
current recession began in December
2007. Figure 1 shows a clear decline
in both hiring and quits during these
downturns, suggesting that these two
flows are procyclical; that is, they rise
and fall in sync with economic activity.
It also shows a very modest rise in
layoffs during the 2001-03 period and
a more noticeable increase in the
2008-09 period, suggesting that layoffs
are at least somewhat countercyclical:
Layoffs go up when economic growth
goes down.
Figure 2 illustrates that other
measures of job loss, such as the job
destruction rate (a summary measure
of employment losses at all contracting
businesses) for manufacturing
employment reported in the BLS
Business Employment Dynamics
(BED) data, and the Department of
Labor’s data on initial unemployment
insurance (UI) claims by the recently
laid-off, provide stronger evidence
of the countercyclicality of job loss.
They also show that the rate of job
loss spikes sharply during the deep
recessions of the 1970s, 1980s, and the
current downturn relative to the rises
in the 1990-91 and 2001-03 periods.
How Business-Level Changes
Relate to the National-Level Data.
Finally, we need to know how the
hires and separations at the business
level aggregate to the national-level
statistics observed in Figures 1 and 2.
To do so, it is useful to think of the
national-level worker flow statistics in
Figure 1 as weighted averages of each
worker flow rate across individual
businesses. The key insight from the
weighted average approach will be
that movements in worker flows at the

Source: Author’s tabulations from published JOLTS statistics

FIGURE 2
Quarterly Job Destruction and Unemployment
Insurance Initial Claims, 1967-2009
10.0
9.0
8.0
Job Destruction Rate
(through 2009Q1)

7.0
6.0
5.0
4.0

UI Initial Claims Rate
(through 2009Q4)

3.0
2.0
1967

1972

1977

1982

1987

1992

1997

2002

2007

Source: Job destruction rates are estimates for manufacturing from my working paper, updated
through 2009 with published BED data. The UI claims rate is total weekly claims (in all sectors)
during the quarter as a percent of total employment, from published UI claims statistics.

national level can come from one of
two sources: changes in business-level
worker flow rates or changes in the

distribution of business-level activity.
For our purposes, we want to
relate the worker flows to the business-

www.philadelphiafed.org

level employment growth rates, and
the example below illustrates the
relationship. It splits all businesses into
contracting, stable, and expanding
businesses and then calculates
the average worker flow rates and
employment shares for each group.
Suppose that, for a given period,
estimates from the business-level
micro-data provided us with the
employment shares and worker flow
rates shown in Table 1.
In Table 1, 25 percent of businesses are losing workers on net, 45 percent
have no change in their employment,
and 30 percent are adding workers
on net. All three groups have some
amount of both hiring and separations
(defined as the sum of quits and layoffs
here). At the contracting businesses,
the average separation rate must be
higher than the average hiring rate
(both measured as percentages of the
businesses’ employment); otherwise,
they would not be contracting. The
opposite is true of the expanding
businesses. At the stable businesses,
the hiring and separation rates exactly
offset each other. As we will see below,
the numbers in this example are similar to what we observe in an average
month in the U.S. data. Stable businesses have the lowest average hiring
and separation rates because many of
them have no employment changes at
all in a given month.
Putting the data in our example
together, we get the following formulas
for deriving what the national-level
hiring and separation rates will be in
this case:
1DWLRQDO/HYHO
5DWH

§ Share of ·§ Avg. Rate at ·
¸
¸¨
¨
¨ Contracting ¸¨ Contracting ¸ 
¨ Businesses ¸¨ Businesses ¸
¹
¹©
©
§ Share of ·§ Avg. Rate at ·
¸
¸¨
¨
¨ Stable ¸¨ Stable ¸ 
¨ Businesses ¸¨ Businesses ¸
¹
¹©
©
§ Share of ·§ Avg. Rate at ·
¸
¸¨
¨
¨ Expanding ¸¨ Expanding ¸
¨ Businesses ¸¨ Businesses ¸
¹
¹©
©

www.philadelphiafed.org

National-Level Hiring Rate =
(0.25)(2.0)+(0.45)(1.0)+(0.30)(12.0)
= 4.550 percent
National-Level Quit Rate =
(0.25)(4.0)+(0.45)(0.5)+(0.30)(2.0)
= 1.825 percent
National-Level Layoff Rate =
(0.25)(7.0)+(0.45)(0.5)+(0.30)(1.0)
= 2.275 percent
In each case, we see that the nationallevel estimates average across the
hiring or separation rates of the
three groups using their share of
total employment as a weight. The
difference between the national-level
hiring rate (4.55 percent) and the
national-level quit and layoff rates
(1.825+2.275 = 4.10 percent) implies
that total employment grew, on net,
by 0.45 percent. Just as it is in the
actual JOLTS data, this is a much
smaller number than the 4.55 percent
of workers who were just hired this
month.
Now, if we were to expand our
example to include finer growth rate
intervals (e.g., businesses that grow
or contract less than 1 percent, 1
to 2 percent, etc.), we would get the
following formula:
WFt

¦s
g

gt wf gt

,

where WFt is the national-level
worker flow rate (i.e., hiring, quits,
or layoffs) in period t, sgt is the share
of employment at businesses with a
growth rate of g in period t, and wfgt
is the average worker flow rate for
businesses with a growth rate of g in
period t. Thus, the weighted average
expression shows that movements in
worker flows at the national level can
come from either changes in businesslevel worker flows (i.e., changes in
wfgt) or changes in the distribution of
business-level employment growth (i.e.,
changes in the business-level weights,
sgt).
THE EVIDENCE ON
BUSINESS-LEVEL
EMPLOYMENT BEHAVIOR
Figures 3 and 4 illustrate what
the “real-world” equivalents of the
business-level worker flow rates,
the wfgt, look like. The figures show
estimates of the average hiring, quit,
and layoff rates as a function of the
business-level employment growth
rate built from the JOLTS businesslevel micro-data in my paper with
Steven Davis and John Haltiwanger.
Figure 3 shows that the hiring rate
rises proportionately with the growth
rate when growth is positive and

TABLE 1
Contracting
Businesses

Stable
Businesses

Expanding
Businesses

Share of Employment
(Employment in Group/
Total Employment)

25.0

45.0

30.0

Average Hiring Rate
(Hires/Employment)

2.0

1.0

12.0

Average Quit Rate
(Quits/Employment)

4.0

0.5

2.0

Average Layoff Rate
(Layoffs/Employment)

7.0

0.5

1.0

Business Review Q2 2010 19

is essentially flat when growth is
negative. Figure 4 shows that the layoff
rate increases proportionately with
the size of a contraction when growth
is negative but layoffs are essentially
flat when growth is positive. It also
shows that the quit rate increases
when a contraction is relatively small
and that quits are essentially constant
(albeit at a higher rate) during larger
contractions. Like the layoff rate, the
quit rate is essentially flat when growth
is positive. Comparing Figures 3 and 4,
we see that the hiring and layoff rates
at the business level exhibit opposing
“hockey-stick” patterned relationships
to business-level growth.
The two figures tell us that when
a business expands employment by,
say, 10 percent, it tends to do so with
a hiring rate of 13.5 percent because,
on average, 2.5 percent of its workforce
will quit and another 1 percent will
be either laid off or discharged as it
tries to expand. Similarly, when a
business wants to contract by, say, 10
percent, it will lay off only 5 percent
of its workforce because, on average,
7.9 percent will leave, of which the
business will replace 2.9 percent, on
average, to counteract some of the
turnover.
Figure 5 shows how the shares
of employment at businesses with
different growth rates, the sgt terms,
change over time by showing the
business-level employment growth
rate distribution at two points:
one for a period of high nationallevel employment growth (i.e., an
expansion) and one for a period of
low national-level employment growth
(i.e., a recession). As the economy
moves from expansion to recession,
the distribution shifts to the left,
meaning that the sgt shares for growing
establishments go down and the sgt
shares for contracting establishments
go up. While the shift may appear
subtle, the statistics listed in the figure

20 Q2 2010 Business Review

FIGURE 3
Hiring vs. Business-Level Growth
Monthly Rate, Percent of Employment
120.0

100.0

80.0

60.0

40.0

20.0

0.0
-100.0

-80.0

-60.0

-40.0

-20.0

0.0

20.0

40.0

60.0

80.0

100.0

Monthly Employment Growth Rate, Percent

Source: Estimates from my study with Steven Davis and John Haltiwanger, which uses establishment micro-data from JOLTS pooled over 2001-2006. The dashed line represents a 45-degree
line emanating from the origin.

FIGURE 4
Separations (by Type) vs. Business-Level Growth
Monthly Rate, Percent of Employment
120.0

100.0

80.0
Layoff Rate
60.0

40.0

20.0

0.0
-100.0

Quit Rate

-80.0

-60.0

-40.0

-20.0

0.0

20.0

40.0

60.0

80.0

100.0

Monthly Employment Growth Rate, Percent

Source: Estimates from my study with Steven Davis and John Haltiwanger, which uses establishment micro-data from JOLTS pooled over 2001-2006.

www.philadelphiafed.org

show that the changes for businesses
with high growth or large contractions
are substantial. Moving from an
expansion period to a recession period
reduces the share of employment at
businesses with high growth (greater
than 10 percent of employment) from
18.7 percent to 14.1 percent. This
reduction corresponds to changes that
affect roughly 6.2 million workers.
Figure 5 also shows that the shift
in the distribution is asymmetric:
The shift skews the distribution of
employment away from a small range
of expanding businesses and toward a
broad range of contracting businesses.
Figure 5 shows that when moving from
expansion to recession, the fraction of
employment at high-growth businesses
falls 4.6 percent, while the fraction of
employment at businesses with a large
contraction rises 6.3 percent.
Finally, it turns out that the
worker flow rates depicted in Figures
3 and 4 barely change over time, as
my research with Steven Davis and
John Haltiwanger shows.4 Therefore,
the movements in the national-level
worker flows observed in Figure 1
occur primarily through the shifts in
the growth rate distribution depicted
in Figure 5.

percent, and a layoff rate of 2.275
percent at the national level. With the
new employment shares, national-level
hiring and separation rates are now:

ing rate but a considerable increase in
the national-level separation rate.
Suppose the economy from the
previous example falls into recession,
causing the growth rate distribution to
shift to the left. Assume that the shift
is asymmetric, just as it is in Figure 5.
In the example in Table 2, the
fraction of employment at declining
businesses rises by 10 percentage
points, while the fraction of
employment at growing businesses falls
by 5 percentage points. The difference
is made up by a 5-percentage-point
fall in the fraction of employment at
stable businesses. We assume that the
business-level hiring and separation
rates are the same as before, consistent
with what we find in the data. Recall
that the previous shares of employment
at contracting, stable, and expanding
businesses produced a hiring rate
of 4.55 percent, a quit rate of 1.825

National-Level Hiring Rate =
(0.35)(2.0)+(0. 40)(1.0)+(0.25)(12.0)
= 4.10 percent
National-Level Quit Rate =
(0.35)(4.0)+(0.40)(0.5)+(0.25)(2.0)
= 2.10 percent
National-Level Layoff Rate =
(0.35)(7.0)+(0.40)(0.5)+(0.25)(1.0)
= 2.90 percent
As the economy moves from
expansion to recession, the hiring rate
falls from 4.55 to 4.10 percent and the
separation rate rises from 4.10 to 5.00
percent. As a result, the national-level
employment growth rate moves from
+0.45 to -0.90 percent. The labor
market is now contracting rather

FIGURE 5
The Distribution of Business-Level
Employment Growth
Percent of Employment
18.0

IMPLICATIONS FOR CYCLICAL
EMPLOYMENT CHANGES
The fact that the growth rate
distribution tends to have an asymmetric shift when moving into a recession
is an important reason some recessions
are driven by relatively high job loss,
while others are driven by relatively
weak hiring. The example in Table 2
shows how an asymmetric shift toward
contracting businesses can produce a
modest drop in the national-level hir-

18.0

16.0

16.0

14.0
12.0

Employment at Large
Contractions (< –10%):
13.2% during high-growth qtrs.
19.5% during low-growth qtrs.

High-Growth Quarters

14.0
12.0

Employment at Large
Expansions (> 10%):
18.7% during high-growth qtrs.
14.1% during low-growth qtrs.

10.0
8.0
Low-Growth Quarters

6.0

10.0
8.0
6.0

4.0

4.0

2.0

2.0

0.0
-30.0

-25.0

-20.0

-15.0

-10.0

-5.0

0.0

5.0

10.0

15.0

20.0

25.0

0.0
30.0

Quarterly Employment Growth Rate, Percent
4

In the data, the exception is the quit rate
relationship in Figure 4, which shifts down
during recessions.

www.philadelphiafed.org

Source: Estimates from my study with Steven Davis, John Haltiwanger, and Ian Rucker, which
uses quarterly establishment-level growth rates from BED micro-data from 2001-2006.

Business Review Q2 2010 21

TABLE 2
Contracting
Businesses

Stable
Businesses

Expanding
Businesses

Old Share of
Employment
(Expansion)

25.0

45.0

30.0

New Share of
Employment
(Recession)

35.0

40.0

25.0

sharply. Moreover, the change is driven
more by the rise in the separation rate
(+0.90 percent), particularly the layoff
rate (+0.63 percent), than by the fall
in the hiring rate (-0.45 percent).5
Thus, our example produces the same
result we find in the data: Severe
recessions have relatively high layoff
rates, more so than low hiring rates, at
the national level.
Besides the asymmetric shift,
the other reason our example is able
to generate large layoffs during a
deep recession is that it assumes that
the hiring and layoff rates exhibit
the “hockey-stick” relationships we
observe in Figures 3 and 4. Since
the layoff rate rises sharply with the
size of a business’s contraction, larger
leftward shifts in the growth rate
distribution, that is, larger increases
in the share of businesses experiencing a large contraction, will drive the
national-level layoff rate even higher.
Figure 6 illustrates this phenomenon

5
The quit rate rises in this example, contradicting its behavior in Figure 1, because, for simplicity, I have assumed away the fact that the quit
relationship in Figure 4 is the only one of the
three that changes (by shifting down) during a
recession. This does not affect the main point of
the example, though.

22 Q2 2010 Business Review

with a hypothetical interaction of the
business-level layoff rate with movements in the growth rate distribution.
The further the growth rate distribution shifts to the left, the greater is the
share of employment at businesses with
very high layoff rates. This causes the
national-level layoff rate to increase
sharply. Since the shift in the growth
rate distribution is asymmetric, the rise
in the layoff rate is greater than the
decline in the hiring rate. In contrast,
a mild recession has a relatively small
shift to the left, meaning that there is
only a small increase in the share of
businesses with very high layoff rates,
and consequently, the asymmetry plays
less of a role. In this case a rise in the
national-level layoff rate may be similar
to, or even smaller than, the decline in
the hiring rate.
Intuitively, a mild recession means
that there is a relatively large share
of businesses cutting their workforces
modestly. Figure 4 shows that such
businesses generally do so with an
equal mix of quits and layoffs. Since
the contraction is small, a business
can use regular attrition to shrink
its employment and will have to lay
off only a few additional workers, on
average. At the same time, however,
these businesses are not hiring, so

those workers that do lose their jobs
find it difficult to find new work and
remain unemployed for some time. A
deep recession involves an increase
in the share of businesses undergoing
large contractions. Figure 4 shows that,
in these cases, attrition is not enough
to get businesses to their new desired
employment levels, so they must let
sizable fractions of their workforces
go, adding to the unemployment rolls
through these layoffs.
The recessions of the 1970s and
1980s had sharp, deep declines in
employment. While we do not have
data on business-level growth distributions that go back that far, our exercise
and the large spikes in job destruction
and UI claims observed during these
periods (Figure 2) suggest that these
periods likely involved large leftward
shifts of the distribution.6 The rise in
unemployment during these periods
was driven by the large number of
workers who lost their jobs as a result
of these layoffs. The 1990-91 recession
and the 2001 recession had relatively
modest declines in employment, suggesting that these periods involved
much smaller shifts in the growth rate
distribution. As our exercise would
imply, these periods saw only modest rises in the layoff rate. In relative
terms, there was a decline in hiring
that was just as important during
these periods. Without a large spike in
layoffs, the unemployment rate did not
rise as much as it did in the 1970s and
1980s. Both recessions, however, were
followed by “jobless recoveries,” during which hiring remained depressed
for an extended period. During this
time, it was difficult for those who did
lose their jobs to find new work, and
consequently, the unemployment rate

6

The two studies by Steven Davis and John
Haltiwanger find similar spikes in job destruction during these periods.

www.philadelphiafed.org

remained elevated for some time.
By historical standards, the current recession is very deep, and the
pace of layoffs is comparable to that
seen in the 1970s and 1980s. Consequently, it likely represents a large leftward shift of the growth rate distribution. Through the end of 2009, a high
layoff rate led to a sharp increase in
the unemployment rate, but as Figure
1 shows, there has also been a sharp
drop in hiring.
The exercise in this article,
though, speaks only to the severity of
a recession, not to its length, which is
generally determined by the nature
of the macroeconomic shocks to
the economy that cause a recession.
Historically, deep recessions have been
relatively brief (i.e., have a “V-shaped”
recovery), implying that the growth
rate distribution shifts to the left for a
short period of time and then quickly
begins shifting back toward the right,
while the more shallow recessions have
extended periods of job loss (i.e., have
an “L-shaped” recovery), implying that
the distribution shifts to the left and
remains there for a while.
As of this writing, the current
recession could have either a V-shaped
or L-shaped recovery. Under the first
scenario, the growth rate distribution would shift sharply to the left but
then revert relatively quickly, creating a large but brief spike in layoffs
and a subsequent sharp, but similarly
brief, rise in the unemployment rate.
Under the second and more troubling
scenario, the growth rate distribution
would shift to the left and remain
there for some time. Layoff rates

www.philadelphiafed.org

FIGURE 6
An Illustration of the Interaction Between Growth
Distribution and Employer Flow Functions
Percent of Employment

Layoff Rate
(as a function of growth)
Employment Growth
Rate Distributions
(black = high-growth period,
grey = low-growth period)

-30

-20

-10

0

10

20

30

Employment Growth Rate, Percent

Note: Figure is a hypothetical illustration of the interaction between the business-level employment growth rate distribution (Figure 5) and the layoff rate (Figure 4). The shift in the distribution is exaggerated for illustrative purposes.

would remain high and hiring would
remain depressed, leading to very high
unemployment rates that persist for
some time.
CONCLUSION
A weak labor market is the outcome of two different types of employment adjustment: weaker hiring and
greater job loss. This article has shown
that the severity of a recession determines whether high job loss or weak
hiring will be the more important
source of declining employment and

rising unemployment through asymmetric shifts in the distribution of business-level employment growth. These
shifts interact with kinked “hockey
stick” relationships between hiring,
layoffs, and business-level growth to
generate this result. Therefore, an
important part of understanding the
behavior of employment and the primary causes of unemployment during
an economic downturn is understanding how the employment behavior of
individual businesses changes over the
business cycle. BR

Business Review Q2 2010 23

REFERENCES

Davis, Steven J., and John Haltiwanger.
“Gross Job Creation and Destruction:
Microeconomic Evidence and Macroeconomic Implications,” in Olivier
Blanchard and Stanley Fischer, eds., NBER
Macroeconomics Annual 1990. Cambridge,
MA: MIT Press, 1990, pp. 123-68.
Davis, Steven J., and John Haltiwanger.
“Gross Job Creation, Gross Job Destruction
and Employment Reallocation,” Quarterly
Journal of Economics, 107:3 (1992), pp.
819-63.
Davis, Steven J., R. Jason Faberman, and
John Haltiwanger. “The Flow Approach
to Labor Markets: New Evidence and
Micro-Macro Links,” Journal of Economic
Perspectives, 20:3 (2006), pp. 3-24.

Davis, Steven J., R. Jason Faberman,
John Haltiwanger, and Ian Rucker.
“Adjusted Estimates of Worker Flows and
Job Openings in JOLTS,” in Katharine
Abraham, Michael Harper, and James
Spletzer, eds., Labor in the New Economy,
Chicago: University of Chicago Press
(forthcoming).
Elsby, Michael, Ryan Michaels, and Gary
Solon. “The Ins and Outs of Cyclical
Unemployment,” American Economic
Journal: Macroeconomics, 1:1 (2009), pp.
84-110.
Faberman, R. Jason. “Job Flows, Jobless
Recoveries, and the Great Moderation,”
Federal Reserve Bank of Philadelphia
Working Paper 08-11 (2008).

Fujita, Shigeru. “What Do Worker Flows
Tell Us About Cyclical Fluctuations in
Employment?” Federal Reserve Bank of
Philadelphia Business Review (Second
Quarter 2007), pp. 1-10.
Fujita, Shigeru, and Garey Ramey. “The
Cyclicality of Separation and Job Finding
Rates,” International Economic Review, 50:2
(2009), pp. 415-30.
Hall, Robert E. “Job Loss, Job Finding,
and Unemployment in the U.S. Economy
over the Past Fifty Years,” in Mark Gertler
and Kenneth Rogoff, eds., 2005 NBER
Macroeconomics Annual. Cambridge, MA:
National Bureau of Economic Research,
2005, pp. 101-37.
Shimer, Robert. “Reassessing the Ins and
Outs of Unemployment,” University of
Chicago, mimeo (2007).

24 Q2 2010 Business Review

www.philadelphiafed.org

RESEARCH RAP

Abstracts of
research papers
produced by the
economists at
the Philadelphia
Fed

You can find more Research Rap abstracts on our website at: www.philadelphiafed.org/research-and-data/
publications/research-rap/. Or view our working papers at: www.philadelphiafed.org/research-and-data/
publications/.

ESTIMATING THE ELASTICITY
OF LABOR SUPPLY: HOME
PRODUCTION, DEMOGRAPHICS,
AND HOUSEHOLD
CHARACTERISTICS
This paper revisits the argument, posed
by Rupert, Rogerson, and Wright (2000),
that estimates of the intertemporal elasticity
of labor supply that do not account for
home production are biased downward.
The author uses the American Time Use
Survey, a richer and more comprehensive
data source than those used previously,
to replicate their analysis, but he also
explores how other factors interact with
household and market work hours to
affect the elasticity of labor supply. An
exact replication of their analysis yields
an elasticity of about 0.4, somewhat larger
than previously estimated. Once the author
accounts for demographics and household
characteristics, particularly the number of
children in the household, the estimate
is essentially zero. This is true even when
accommodating extensive-margin labor
adjustments. Households’ biological inability
to smooth childbearing over the life cycle
and the resulting income effect on market
work hours drive this result.
Working Paper 10-3, “Revisiting the Role
of Home Production in Life-Cycle Labor
Supply,” R. Jason Faberman, Federal Reserve
Bank of Philadelphia

25 Q2 2010 Business Review

EXPLAINING OUTPUT
VARIABILITY ACROSS COUNTRIES
Inference about common international
stochastic trends and interest rates is
gained using a small open-economy model,
data from seven developed countries, and
Bayesian methods. Shocks to these common
factors explain up to 17 percent of the
variability of output in several economies.
Country-specific preference and premium
disturbances account for the bulk of the
volatility observed in the data. There is
substantial heterogeneity in the estimated
structural parameters as well as stochastic
processes for the countries in the sample.
This diversity translates into a rich array
of impulse responses across countries.
According to the model, the recent low
international interest rates might have
initially deepened the decline of GDP in
several developed economies.
Working Paper 10-4, “Common Factors
in Small Open Economies: Inference and
Consequences,” Pablo Guerron-Quintana,
Federal Reserve Bank of Philadelphia
MONITORING MACROECONOMIC
ACTIVITY IN REAL TIME
The authors sketch a framework for
monitoring macroeconomic activity in
real time and push it in new directions.
In particular, they focus not only on
real activity, which has received the

www.philadelphiafed.org

most attention to date, but also on inflation and
its interaction with real activity. As for the recent
recession, the authors find that (1) it likely ended
around July 2009; (2) its most extreme aspects concern
a real activity decline that was unusually long but
less unusually deep, and an inflation decline that was
unusually deep but brief; and (3) its real activity and
inflation interactions were strongly positive, consistent
with an adverse demand shock.
Working Paper 10-5, “Real-Time Macroeconomic
Monitoring: Real Activity, Inflation, and Interactions,” S.
%RUDùDQ $UXRED 8QLYHUVLW\ RI 0DU\ODQG DQG 9LVLWLQJ
Scholar, Federal Reserve Bank of Philadelphia, and Francis
; 'LHEROG 8QLYHUVLW\ RI 3HQQV\OYDQLD DQG 9LVLWLQJ
Scholar, Federal Reserve Bank of Philadelphia
EXPECTATIONS AND MACROECONOMIC
FLUCTUATIONS
Using survey-based measures of future U.S. economic activity from the Livingston Survey and the
Survey of Professional Forecasters, the authors study
how changes in expectations, and their interaction with
monetary policy, contribute to fluctuations in macroeconomic aggregates. They find that changes in expected future economic activity are a quantitatively important driver of economic fluctuations: a perception that
good times are ahead typically leads to a significant rise
in current measures of economic activity and inflation.
The authors also find that the short-term interest rate
rises in response to expectations of good times as monetary policy tightens. Their results provide quantitative
evidence on the importance of expectations-driven
business cycles and on the role that monetary policy
plays in shaping them.
Working Paper 10-6, “Expectations and Economic
Fluctuations: An Analysis Using Survey Data,” Sylvain
Leduc, Federal Reserve Bank of San Francisco, and Keith
Sill, Federal Reserve Bank of Philadelphia
PROVIDING INCENTIVES TO DETER FRAUD
Social and private insurance schemes rely on legal
action to deter fraud and tax evasion. This observation
guides the authors to introduce a random state verification technology in a dynamic economy with private
information. With some probability, an agent’s skill
level becomes known to the planner, who prescribes a
punishment if the agent is caught misreporting. The
authors show how deferring consumption can ease the

26 Q2 2010 Business Review

provision of incentives. As a result, the marginal benefit
may be below the marginal cost of investment in the
constrained-efficient allocation, suggesting a subsidy
on savings. They characterize conditions such that the
intertemporal wedge is negative in finite horizon economies. In an infinite horizon economy, the authors find
that the constrained-efficient allocation converges to a
high level of consumption, full insurance, and no labor
distortions for any probability of state verification.
Working Paper 10-7, “Fraud Deterrence in Dynamic
Mirrleesian Economies,” Roc Armenter, Federal Reserve
Bank of Philadelphia, and Thomas M. Mertens, New York
University
TRANSFERRING RISK THROUGH LOAN SALE
AND SECURITIZATION
Depository institutions may use information advantages along dimensions not observed or considered
by outside parties to “cream-skim,” meaning to transfer
risk to naïve, uninformed, or unconcerned investors
through the sale or securitization process. This paper
examines whether “cream-skimming” behavior was
common practice in the subprime mortgage securitization market prior to its collapse in 2007. Using Home
Mortgage Disclosure Act data merged with data on
subprime loan delinquency by ZIP code, the authors
examine the bank decision to sell (securitize) subprime
mortgages originated in 2005 and 2006. They find that
the likelihood of sale increases with risk along dimensions observable to banks but not likely observed or
considered by investors. Thus, in the context of the
subprime lending boom, the evidence supports the
cream-skimming view.
Working Paper 10-8, “’Cream-Skimming’ in Subprime
Mortgage Securitizations: Which Subprime Mortgage
Loans Were Sold by Depository Institutions Prior to the
Crisis of 2007?” Paul Calem, Federal Reserve Board of
Governors; Christopher Henderson, Federal Reserve Bank
of Philadelphia; and Jonathan Liles, Freddie Mac
AN ALTERNATIVE APPROACH TO
MEASURING BANK COMPETITION
Measuring banking competition using the HHI,
Lerner index, or H-statistic can give conflicting results.
Borrowing from frontier analysis, the authors provide
an alternative approach and apply it to Spain over 19922005. Controlling for differences in asset composition,
productivity, scale economies, risk, and business cycle

www.philadelphiafed.org

influences, they find no differences in competition
between commercial and savings banks nor between
large and small institutions, but the authors conclude
that competition weakened after 2000. This appears
related to strong loan demand where real loan-deposit
rate spreads rose and fees were stable for activities
where scale economies should have been realized.
Working Paper 10-9, “A Revenue-Based Frontier
Measure of Banking Competition,” Santiago Carbó,
University of Granada, Spain; David Humphrey, Florida
6WDWH 8QLYHUVLW\ DQG 9LVLWLQJ 6FKRODU )HGHUDO 5HVHUYH
Bank of Philadelphia; and Francisco Rodriguez, University
of Granada, Spain
CHANGE IN ICEBERG COSTS AND THE RISE
IN U.S MANUFACTURING EXPORTS
The authors study the rise in U.S. manufacturing
exports from 1987 to 2002 through the lens of a
monopolistically competitive model with heterogeneous
producers and sunk costs of exporting. Using the
model, they infer that iceberg costs fell nearly 27
percent in this period. Given this change in iceberg
costs, the authors use the model to calculate the
predicted increase in trade. Contrary to the findings in
Yi (2003), they find that the exports should have grown
an additional 70 percent (78.7 vs. 46.4). The model
overpredicts export growth partly because it misses the
shift in manufacturing to relatively small establishments
that did not invest in becoming exporters. Contrary
to the theory, employment was largely reallocated
from very large establishments, those with more than
2,500 employees, toward very small manufacturing
establishments, those with fewer than 100 employees.
The authors also find that very little of the contraction
in U.S. manufacturing employment can be attributed to
trade.
Working Paper 10-10, “Do Falling Iceberg Costs
Explain Recent U.S. Export Growth?” George Alessandria,
Federal Reserve Bank of Philadelphia, and Horag Choi,
University of Auckland
OPTIMAL CAPITAL INCOME TAXATION
This paper quantitatively investigates the optimal
capital income taxation in the general equilibrium
overlapping generations model, which incorporates
characteristics of housing and the U.S. preferential
tax treatment for owner-occupied housing. Housing
tax policy is found to have a substantial effect on

www.philadelphiafed.org

how capital income should be taxed. Given the U.S.
preferential tax treatment for owner-occupied housing,
the optimal capital income tax rate is close to zero,
contrary to the high optimal capital income tax rate
implied by models without housing. A lower capital
income tax rate implies a narrowed tax wedge between
housing and non-housing capital, which indirectly
nullifies the subsidies (taxes) for homeowners (renters)
and corrects the over-investment to housing.
Working Paper 10-11, “Optimal Capital Income
Taxation with Housing,” Makoto Nakajima, Federal
Reserve Bank of Philadelphia
A NEW APPROACH TO MODELING
LONG-TERM DEBT
In this paper, the authors present a new approach
to incorporating long-term debt into equilibrium
models of unsecured debt and default. They make three
sets of contributions. First, the authors advance the
theory of sovereign debt begun in Eaton and Gersovitz
(1981) by proving the existence of an equilibrium price
function with the property that the interest rate on
debt is increasing in the amount borrowed. Second,
using Argentina as a test case, they show that unlike a
one-period debt model, their model of long-term debt is
capable of accounting for the average external debt-tooutput ratio, average spread on external debt, and the
standard deviation of spreads for the 1993-2001 period,
without any deterioration in the model’s ability to
account for Argentina’s other cyclical facts. Third, the
authors propose a new and very accurate method for
solving the model.
Working Paper 10-12, “Maturity, Indebtedness, and
Default Risk,” Satyajit Chatterjee, Federal Reserve Bank of
Philadelphia, and Burcu Eyigungor, Koç University
MORTGAGE DEFAULT: ASSESSING THE ROLE
OF NEGATIVE EQUITY AND ILLIQUIDITY
This paper assesses the relative importance of two
key drivers of mortgage default: negative equity and
illiquidity. To do so, the authors combine loan-level
mortgage data with detailed credit bureau information
about the borrower’s broader balance sheet. This gives
them a direct way to measure illiquid borrowers: those
with high credit card utilization rates. The authors
find that both negative equity and illiquidity are
significantly associated with mortgage default, with
comparably sized marginal effects. Moreover, these two

Business Review Q2 2010 27

factors interact with each other: The effect of illiquidity
on default generally increases with high combined loanto-value ratios (CLTV), though it is significant even for
low CLTV. County-level unemployment shocks are also
associated with higher default risk (though less so than
high utilization) and strongly interact with CLTV. In
addition, having a second mortgage implies significantly
higher default risk, particularly for borrowers who have
a first-mortgage LTV approaching 100 percent.
Working Paper 10-13, “What ‘Triggers’ Mortgage
Default?” Ronel Elul, Federal Reserve Bank of
Philadelphia; Nicholas S. Souleles, University of
Pennsylvania; Souphala Chomsisengphet, Office of the
Comptroller of the Currency; Dennis Glennon, Office of
the Comptroller of the Currency; and Robert Hunt, Federal
Reserve Bank of Philadelphia
ACCOUNTING FOR TIME-VARYING
VOLATILITY IN U.S. AGGREGATE DATA:
STOCHASTIC VOLATILITY VS. CHANGES IN
MONETARY POLICY
This paper compares the role of stochastic
volatility versus changes in monetary policy rules
in accounting for the time-varying volatility of U.S.
aggregate data. Of special interest to the authors is
understanding the sources of the great moderation
of business cycle fluctuations that the U.S. economy
experienced between 1984 and 2007. To explore this
issue, the authors build a medium-scale dynamic
stochastic general equilibrium (DSGE) model with
both stochastic volatility and parameter drifting in
the Taylor rule and they estimate it non-linearly using
U.S. data and Bayesian methods. Methodologically, the
authors show how to confront such a rich model with
the data by exploiting the structure of the high-order
approximation to the decision rules that characterize

28 Q2 2010 Business Review

the equilibrium of the economy. Their main empirical
findings are: 1) even after controlling for stochastic
volatility (and there is a fair amount of it), there
is overwhelming evidence of changes in monetary
policy during the analyzed period; 2) however, these
changes in monetary policy mattered little for the
great moderation; 3) most of the great performance
of the U.S. economy during the 1990s was a result of
good shocks; and 4) the response of monetary policy
to inflation under Burns, Miller, and Greenspan was
similar, while it was much higher under Volcker.
Working Paper 10-14, “)RUWXQH RU 9LUWXH 7LPH
9DULDQW 9RODWLOLWLHV 9HUVXV 3DUDPHWHU 'ULIWLQJ,” Jesus
)HUQDQGH]9LOODYHUGH 8QLYHUVLW\ RI 3HQQV\OYDQLD Pablo
Guerron-Quintana, Federal Reserve Bank of Philadelphia;
and Juan F. Rubio-Ramirez, Duke University and Federal
Reserve Bank of Atlanta
ESTIMATING A DSGE MODEL TO EXAMINE
RECENT U.S. MONETARY HISTORY
The authors report the results of the estimation of
a rich dynamic stochastic general equilibrium model
of the U.S. economy with both stochastic volatility
and parameter drifting in the Taylor rule. They use
the results of this estimation to examine the recent
monetary history of the U.S. and to interpret, through
this lens, the sources of the rise and fall of the great
American inflation from the late 1960s to the early
1980s and of the great moderation of business cycle
fluctuations between 1984 and 2007.
Working Paper 10-15, “Reading the Recent Monetary
History of the U.S., 1959-2007,” Jesus Fernandez9LOODYHUGH 8QLYHUVLW\ RI 3HQQV\OYDQLD Pablo GuerronQuintana, Federal Reserve Bank of Philadelphia; and Juan
F. Rubio-Ramirez, Duke University and Federal Reserve
Bank of Atlanta

www.philadelphiafed.org