Journal of Economic Perspectives—Volume 24, Number 4—Fall 2010—Pages 21–44

Financial Intermediation and

Macroeconomic Analysis

Michael Woodford

I
ssues relating to financial stability have always been part of the macroeco-
nomics curriculum, but they have often been presented as mainly of historical
interest, or primarily of relevance to emerging markets. However, the recent
financial crisis has made it plain that even in economies like the United States,
significant disruptions of financial intermediation remain a possibility. Under-
standing such phenomena and the possible policy responses requires the use of
a macroeconomic framework in which financial intermediation matters for the
allocation of resources.
In this paper, I first discuss why neither standard macroeconomic models that
abstract from financial intermediation nor traditional models of the “bank lending
channel” are adequate as a basis for understanding the recent crisis. I argue that
instead we need models in which intermediation plays a crucial role, but in which
intermediation is modeled in a way that better conforms to current institutional
realities. In particular, we need models that recognize that a market-based finan-
cial system—one in which intermediaries fund themselves by selling securities in
competitive markets, rather than collecting deposits subject to reserve require-
ments—is not the same as a frictionless system.
I then sketch the basic elements of an approach that allows financial inter-
mediation and credit frictions to be integrated into macroeconomic analysis in a
straightforward way. I show how the model can be used to analyze the macroeco-
nomic consequences of the recent financial crisis and conclude with a discussion of
some implications of the model for the conduct of monetary policy.

■ Michael Woodford is the John Bates Clark Professor of Political Economy, Columbia Univer-
sity, New York City, New York. His e-mail address is 〈[email protected]
[email protected]〉〉.
doi=10.1257/jep.24.4.21
22 Journal of Economic Perspectives

Why a New Framework for Macroeconomic Analysis is Needed

It may be useful first to review why familiar macroeconomic models do not

already incorporate the features needed to make sense of recent economic devel-
opments. I shall argue that it is difficult to understand why either the significant
decline in house prices since 2006 or the substantial losses sustained by financial
firms should have so seriously affected aggregate employment and economic
activity except in the context of a model in which financial intermediaries play a
crucial role and in which their ability to fulfill that function can at some times be
significantly impaired.

Housing Prices and Aggregate Demand

While the severity of the recent financial crisis has been extensively discussed,
some have questioned whether it was really the primary cause of the Great Recession.
For example, Baker (2010) argues that a substantial reduction in aggregate demand
can be explained as a wealth effect on consumer expenditure, given the decline
in U.S. households’ housing wealth by several trillion dollars. In this analysis, “the
problem is not first and foremost a financial crisis.” But as Buiter (2010) points out,
there is no aggregate wealth effect of a decline in housing prices, since the house-
hold sector in aggregate is both the owner of the housing stock and the consumer
of the services supplied by it. A fall in house prices reduces the value of an asset,
but also reduces the cost of buying the stream of housing services that people were
planning to purchase by exactly the same amount.
It is possible to have a nonzero effect on aggregate expenditure on other goods
(when other prices remain unchanged), even without financial frictions, owing to
redistribution of wealth between households with a net “long” position in housing
and those with net “short” positions, if the average marginal propensities to consume
out of wealth are different between the two types. Nonetheless, because the positive
and negative wealth effects will largely offset one another, the effect on aggregate
demand is likely to be fairly small relative to the size of the aggregate decrease in
housing wealth.
Larger effects are instead possible if one recognizes that the losses resulting
from the collapse of housing prices were disproportionately concentrated in certain
financial institutions which play a role in the allocation of resources that cannot
easily be replaced by those to whom wealth was redistributed. A model of this kind
is sketched below. For a quantitative analysis of the effects of the fall in U.S. housing
prices that stresses such effects, see Greenlaw, Hatzius, Kashyap, and Shin (2008).

Banking and the Money Supply

It is also difficult to understand why large losses by financial institutions on
housing-related bets should have such a significant effect on the real economy
without a model that takes account of credit frictions. According to the well-known
monetarist view, banking crises affect the economy because they reduce the total
supply of money in the economy, since the “money multiplier”—the factor by which
 Michael Woodford 23

the economy’s money supply exceeds the “monetary base” supplied by the central
bank—falls when funds are withdrawn from commercial banks in response to
concerns about their stability. The lower money supply is then only consistent with
money demand to the extent that money demand is also reduced, through some
combination of lower economic activity and deflation. This is the classic account by
Friedman and Schwartz (1963) of how the widespread bank failures in the United
States deepened the Great Depression.
However, such a model, at least as conventionally elaborated, cannot explain
why the recent problems of the financial sector should have caused a sharp reces-
sion, for the Friedman–Schwartz story depends on the monetary base remaining
fixed despite a collapse of the money multiplier. But under contemporary insti-
tutional arrangements, the Fed automatically adjusts the supply of base money as
necessary to maintain its target for the federal funds interest rate; thus, any change
in the money multiplier due to a banking crisis should automatically be offset by a
corresponding increase in the monetary base, neutralizing any effect on interest
rates, inflation, or output.1
Moreover, many of the institutions whose failure or near-failure appeared to
do the most damage in the recent crisis, such as Lehman Brothers, did not issue
liabilities that would count as part of Friedman and Schwartz’s measure of the
money supply. Under a classic monetarist view, the failure of such institutions
should pose no threat to the aggregate economy. (Hence the proposals by some
that finance can remain only lightly regulated, as long as commercial banks are
strictly excluded from the riskier activities.) But the consequences of the failure
of Lehman suggest otherwise.

Models of the Bank Lending Channel

Models that postulate an essential role for banks in financing certain kinds of
expenditure are better able to explain how a financial crisis could have such dire conse-
quences for the real economy as we have observed. However, the kinds of financial
constraints that were emphasized in many past models of this kind assumed specific
institutional forms and regulatory requirements that have become less relevant to the
U.S. financial system over time.
Consider, for example, traditional accounts of the “bank lending channel” of
the transmission of monetary policy. This argument emphasized the indispensable
role of commercial banks as sources of credit for certain kinds of borrowers, in
particular those without direct access to capital markets. Deposits were in turn held
to be an indispensable source of funding for the lending of commercial banks, and
these were subject to legal reserve requirements. To the extent that reserve require-
ments were typically a binding constraint, a reduction in the supply of reserves by
the Federal Reserve would require the volume of deposits to be reduced, which

1
For example, in the model without credit frictions expounded in Figure 2 below, a banking panic
that reduces the money multiplier will have no effect other than to increase the supply of base money
required to implement the central bank reaction function represented by the schedule MP.
24 Journal of Economic Perspectives

would in turn require less lending by commercial banks. Bernanke and Blinder
(1988) and Kashyap and Stein (1994) offer expositions of this view; Smant (2002)
provides a critical review of the literature.
Clearly, the importance of this channel for effects of monetary policy on
economic activity depended on the validity of each of the links in the proposed
mechanism: that reserve requirements were a binding constraint for many banks;
that commercial banks lacked sources of funding other than deposits; that an
important subset of borrowers lacked sources of credit other than commercial
banks; and that banks lacked opportunities to substitute between other assets and
lending to bank-dependent borrowers. Each of these assumptions was less obviously
defensible after the financial innovations and regulatory changes of the 1980s and
1990s. Adrian and Shin (forthcoming a, b) discuss the changing structure of the
U.S. financial system in more detail.
Nonbank financial intermediaries became increasingly important as sources of
credit, particularly as a result of the growing popularity of securitization. Figure 1A
shows the contributions of several categories of financial institutions to total net
lending in the United States; while commercial banks are clearly still important,
they are far from the only important source of credit. More importantly, both the
recent lending boom and the more recent financial crisis had more to do with
changes in financial flows of several of the other types shown in the figure; for
example, lending by issuers of asset-backed securities surged in the period up until
the summer of 2007 and then crashed, while lending by market-based mutual funds
and other market-based financial intermediaries2 crashed after the fall of 2008.
Nor are deposits the main source of funding for the financial sector, even in
the case of commercial banks. Figure 1B shows the net increase in financial sector
liabilities each quarter from several sources. Checkable deposits are only a small
part of the sector’s financing; moreover, deposits shrank during the years of the
lending boom, but have risen again during the crisis—so that neither the growth in
credit during the boom nor the contraction of credit in 2008–09 can be attributed
to variations in the availability of deposits as a source of financing. Even to the
extent that deposits do matter, one may doubt the extent to which the availability of
such funding is constrained by reserve requirements, as in recent years these have
ceased to be a binding constraint for many banks (for example, see Bennett and
Peristiani, 2002).
In response to skepticism about the relevance of the traditional bank lending
channel, Bernanke and Gertler (1995) have instead stressed the importance of
an alternative “broad credit channel,” in which the balance sheets of ultimate
borrowers constrain the amount that they are able to borrow; models incorpo-
rating such effects include those of Kiyotaki and Moore (1997) and Bernanke,
Gertler, and Gilchrist (1999). However, the recent crisis, at least in its initial phase,

2
This category includes mutual funds, the government-sponsored enterprises (GSEs), GSE-backed mort-
gage pools, finance companies, real-estate investment trusts, broker-dealers, and funding corporations.
The “market-based financial intermediaries” terminology derives from Adrian and Shin (forthcoming a).
 Financial Intermediation and Macroeconomic Analysis 25

Figure 1
Financial Flows over the Most Recent Credit Cycle

A: Contributions to U.S. Total Net Lending from Several Categories of Financial

Institutions
(quarterly, in billions of dollars)
2,000

1,500

1,000
$ billions

500

–500
Commercial banking
Money market mutual funds
–1,000 Asset-backed securities issuers
Other market-based financial institutions
–1,500
20

20

20

20

20

20

20

20

20

20
03

03

04

05

06

06

07

08

09

09
-Q

-Q

-Q

-Q

-Q

-Q

-Q

-Q

-Q

-Q
1

B: Contributions from Several Sources of Funding to the Net Increase in the 4

Liabilities of the U.S. Financial Sector
(quarterly, in billions of dollars)
2,000

1,500

1,000
$ billions

500

–500
Checkable deposits
Money market mutual fund shares
–1,000 Fed funds and repos
Commercial paper
–1,500
20

20

20

20

20

20

20

20

20

20
0

0
3-

3-

4-

5-

6-

6-

7-

8-

9-

9-
Q

Q
1

Source: Federal Reserve Board, Flow of Funds Accounts.

26 Journal of Economic Perspectives

resulted more from obstacles to credit supply, resulting from developments in the
financial sector itself, than from a reduction in credit demand owing to the prob-
lems of ultimate borrowers.
Hence, what is needed is a framework for macroeconomic analysis in which
intermediation plays a crucial role and in which frictions that can impede an
efficient supply of credit are allowed for, a framework which also takes account of
the fact that the U.S. financial sector is now largely market-based. Fortunately, the
development of a new generation of macroeconomic models with these features
is now well underway. Adrian and Shin (forthcoming b) and Gertler and Kiyotaki
(forthcoming) provide surveys of recent work in this area. Here, I sketch a basic
version of such a model, show how it can be used to interpret the recent crisis,
and then discuss some implications of a model of this kind for monetary policy. A
complete monetary dynamic stochastic general equilibrium model based on the
approach sketched here is developed in Cúrdia and Woodford (2009).

Credit and Economic Activity: A Market-Based Approach

The theory sketched here is appropriate to a market-based financial system in

which the most important marginal suppliers of credit are no longer commercial
banks and in which deposits subject to reserve requirements are no longer the most
important marginal source of funding even for commercial banks.

Macroeconomics with a Single Interest Rate

It is useful to begin by recalling how interest-rate policy affects aggregate
activity in a conventional model that abstracts from financial frictions. In the
simplest versions of such models, financial conditions can be summarized by a single
interest rate, the equilibrium value of which is determined in a market for credit.
Figure 2A shows the key equilibrium condition. The loan supply schedule LS shows
the amount of lending L that ultimate savers are willing to finance (by refraining
from expenditure themselves) for each possible value of the interest rate i received
by savers, while the loan demand schedule LD shows the demand for such funds
for each possible value of the interest rate that must be paid by borrowers. Note
that the slopes for the curves LS and LD both reflect the same principle, which is
that a higher interest rate gives both savers and borrowers a reason to defer current
spending to a greater extent. Equilibrium in the credit market then determines
both a market-clearing interest rate and an equilibrium volume of lending, as shown
by i1 and L1 in the figure.
In Figure 2A, the loan supply and demand curves are specified while holding
constant a great many variables other than the current interest rate. In particular,
the curves are shown assuming a particular level of current-period aggregate output
(and hence income) Y.. A higher level of income should increase the supply of loans
at any given interest rate (as not all of the additional income should be consumed
if future income expectations are held fixed); hence an increase in Y should shift
 Michael Woodford 27

Figure 2
Interest-Rate and Output Determination in the Standard Model

A: Effect of an Increase in Aggregate Income on Loan Supply and Demand

i
LS
Interest rate

i1

i2
LD

L1 L2 L
Volume of lending

B: Effect of a Loosening of Monetary Policy on Interest Rates and Output

i
MP
Interest rate

i1

i2
IS

Y1 Y2 Y
Aggregate income

Notes: In panel A, LS is the loan supply schedule and LD is the loan demand schedule, which are specified
holding constant aggregate income, Y. The arrows show how the curves shift with an increase in Y.
Panel B shows an IS schedule, derived by tracing out the equilibrium interest rate for any assumed
level of current income Y, and a monetary policy reaction function (MP ), showing how the central
bank’s interest rate target will vary with the level of economic activity. The MP curve is drawn for a given
inflation rate. The arrow shows the consequence of an exogenous shift in the policy reaction function
that implies a lower interest rate for any given level of economic activity.
28 Journal of Economic Perspectives

the LS curve down and to the right, as shown by the arrow. It should also reduce
the demand for loans, insofar as borrowers have more current income available
out of which to finance current spending needs or opportunities, in which case the
LD curve shifts down and to the left, as also shown in the figure. The vertical shift in
the LD curve is likely to be smaller than the vertical shift of the LS curve, as shown in
Figure 2A, if the expenditure of borrowers is more interest-elastic than the expendi-
ture of savers. The intersection of the grey curves shows the new equilibrium values,
i2 and L2.
Tracing out the equilibrium interest rate for any assumed level of current
income Y,, one obtains the IS schedule plotted in Figure 2B. (Alternatively, for
each possible interest rate i,, the schedule shows the level of national income for
which investment equals savings, as this is equivalent to equality between supply
of and demand for funds.) The monetary policy reaction function of the central
bank, indicating how the central bank’s interest-rate target will vary with the level of
economic activity, is shown by the curve MP in this figure.3
If we suppose that the MP curve is drawn for a given inflation rate, then the
upward slope shown indicates a response of interest rates to the level of output (rela-
tive to trend or to potential), of a kind implied, for example, by the “Taylor rule”
(Taylor, 1993)—that is, higher interest rates when output is high relative to trend or
potential, and lower interest rates when output is low relative to trend or potential.
In this case, the equilibrium level of output determined in Figure 2B depends on
the inflation rate; a graph showing how the equilibrium level of output would vary
with inflation yields an aggregate demand relation in inflation–output space. Plot-
ting that relation along with a Phillips curve (or aggregate supply) relation between
inflation and output, one can then finally determine equilibrium output.4
This kind of model provides a straightforward account of the way in which a
central bank’s interest-rate policy affects the level of economic activity (and also the
inflation rate, once one adjoins a Phillips curve to the model). However, this model
of the credit market—in which ultimate savers lend directly to ultimate borrowers so
that the interest rate received by savers is the same as that paid by borrowers—clearly
omits some important features of actual financial systems. In actual economies, we
observe multiple interest rates that do not move perfectly together. Changes in
spreads between certain of these interest rates have been important indicators of
changing financial conditions, both during the recent housing boom and during
the subsequent crash, as is discussed further below.

3
In the case that monetary policy is assumed to correspond to some fixed supply of money, then the
MP curve becomes simply the Hicksian LM curve. However, an upward-sloping relation of the kind
shown in the figure will exist under many other hypotheses, including ones more descriptive of actual
central bank behavior than the Hicksian construct. On the relation between IS–MP analysis and the older
IS–LM analysis, see, for example, Romer (2000) in this journal.
4
Alternatively, one can substitute the inflation rate implied by the Phillips curve (for a given level of
output) into the central bank reaction function, and plot the resulting relation for i as a function of Y
as the curve MP. In this case, MP slopes upward, as shown, even if the central bank’s reaction function
responds only to inflation; and the equilibrium shown in Figure 2B already takes account of the endo-
geneity of the inflation rate.
 Financial Intermediation and Macroeconomic Analysis 29

Introducing Multiple Interest Rates

Here I illustrate one way to introduce multiple interest rates into this model.
Suppose that instead of directly lending to ultimate borrowers themselves, savers
fund intermediaries, who use these funds to lend to (or acquire financial claims
on) the ultimate borrowers. Then, it becomes necessary to distinguish between the
interest rate i s (the rate paid to savers) at which intermediaries are able to fund
themselves and the interest rate i b (the borrowing or loan rate) at which ultimate
borrowers are able to finance additional current expenditure. We can still think in
terms of the two schedules shown in Figure 2A, but now the LS schedule represents
the supply of funding for intermediaries rather than the supply of loans to ultimate
borrowers, and we must now recognize that the supply of funding and the demand
for loans are functions of two different interest rates. Hence the equilibrium level
of lending L can be at a point other than the one where the two schedules cross, as
shown in Figure 3A.
What determines the equilibrium relation between the two interest rates i s and
b
i ? Given the funding supply and loan demand curves (which means, given the
values of a set of variables that include the current value of income Y ), we can deter-
mine the unique volume of intermediation that is consistent with any given spread
ω between i b and i s. If the funding supply curve LS and the loan demand curve LD
have the slopes shown, then a larger credit spread ω implies a lower equilibrium
volume of intermediated credit L.. This relation between the quantity of intermedi-
ated credit and the credit spread is graphed as the curve XD in Figure 3B, which we
can think of as the “demand for intermediation.”
The demand for intermediation schedule XD indicates the degree to which
borrowers are willing to pay an interest rate higher than the one required in order
to induce savers to supply funds to finance someone else’s expenditure. This repre-
sents a profit opportunity for intermediaries, to the extent that they are able to
arrange for the transfer of funds at sufficiently low cost. The volume of lending that
actually occurs, though, will also depend on the capacity of the financial sector to
supply this service at a margin low enough for the services to be demanded.
The corresponding “supply of intermediation” schedule, indicating the credit
spread required to induce financial institutions to intermediate a certain volume
of credit between savers and ultimate borrowers, is depicted by the curve XS in
Figure 3B. This curve reflects the consequences of profit maximization by inter-
mediaries, where the intermediaries in question need not be understood to consist
solely or even primarily of traditional commercial banks. Both the equilibrium credit
spread and the equilibrium volume of credit are then determined by the intersec-
tion between the XS and XD schedules. And given an equilibrium credit spread ω,
determined in Figure 3A, one can use Figure 3A to determine the two interest rates.

Determinants of the Supply of Intermediation

The structural relationship represented by the supply of intermediation
schedule XS in Figure 3B can be motivated in various ways. One model assumes
that intermediaries have costs of originating and servicing loans, or of managing
30 Journal of Economic Perspectives

Figure 3
Credit Market Equilibrium with Credit Supply Frictions

A: Effect of a Credit Spread ω1 on the Equilibrium Interest Rates for

Borrowers and Savers, and on the Equilibrium Volume of Credit

i LS
Interest rate

ib

ω1
e
is

LD

L1 L
Volume of lending

B: Determination of the Equilibrium Credit Spread

ω XS
(between savers and borrowers)
Interest rate spread

ω1

XD

L1 L
Volume of lending

Notes: i s is the interest rate paid to savers, at which intermediaries are able to fund themselves, and i b is the interest
rate (the borrowing or loan rate) at which ultimate borrowers are able to finance additional current expenditure.
In this figure, LS schedule represents the supply of funding for intermediaries, the LD schedule is the loan demand
schedule, and these schedules are functions of two different interest rates. Hence the equilibrium level of lending L
can be at a point other than the one where the two schedules cross, as shown in Figure 3(a). ω is the spread between
i b and i s. Given the LS and LD curves we can determine the unique volume of intermediation that is consistent with
any given spread ω. This relation between the quantity of intermediated credit and the credit spread is graphed as
the curve XD in panel B, which we can think of as the “demand for intermediation.” The corresponding “supply of
intermediation” schedule XS indicates the credit spread required to induce financial institutions to intermediate a
certain volume of credit between savers and ultimate borrowers.
 Michael Woodford 31

their portfolios, so that in a competitive equilibrium, the rate i b at which they are
willing to lend (or the return that they will require on assets that they purchase) will
exceed their cost of funds i s by a spread that reflects the marginal cost of lending.
This marginal cost may be increasing in the volume of lending by the intermediary
if the production function for loans involves diminishing returns to increases in the
variable factors, owing to the fixity of some factors (such as specialized expertise or
facilities that cannot be expanded quickly).5
Probably a more important determinant of the supply of intermediation derives
from the limited capital of intermediaries—or, more fundamentally, the limited
capital of the “natural buyers” of the debt of the ultimate borrowers—together with
limits on the degree to which these natural buyers are able to leverage their posi-
tions. The market for the debt of the ultimate borrowers may be limited to a narrow
class of “natural buyers” for any of a variety of reasons: special expertise may be
required to evaluate such assets; other costs of market participation may be lower
for certain investors; or the natural buyers may be less risk averse, or less uncertainty
averse, or more optimistic about returns on the particular assets.
Leverage may also be constrained for any of a variety of reasons. The recent
literature has emphasized two broad types of constraints. On one hand, there may
be a limit on the size of the losses that the intermediary would be subject to in
bad states of the world, relative to its capital; such limits may result from regula-
tory capital requirements, or (the case of greatest relevance in the recent crisis)
such limits may be imposed by the intermediary’s creditors, who are unwilling to
supply additional funding if the leverage constraint is exceeded (as in Zigrand,
Shin, and Danielsson, 2010; Adrian, Moench, and Shin, 2010b; Adrian and Shin,
forthcoming b; Beaudry and Lahiri, 2010).6
Alternatively, intermediaries may raise funds by pledging particular assets as
collateral for individual loans, and the amount that they can borrow may be limited
by the value of available collateral. Gârleanu and Pedersen (2010) and Ashcraft,
Gârleanu, and Pedersen (forthcoming) consider the consequences of collateral
constraints in a model where the fraction of each asset’s value that can be borrowed
using that asset as collateral is among the defining characteristics of the asset.
Geanakoplos (1997, 2003, 2010) instead proposes a theory in which margin require-
ments are endogenously determined in competitive markets.
Under these types of theories, the capital of intermediaries becomes a crucial
determinant of the supply of intermediation. For a given quantity of capital, the

5
This is one of two relatively reduced-form models of endogenous credit spreads considered in the
monetary dynamic stochastic general equilibrium model we present in Cúrdia and Woodford (2009).
The device of a “loan production function” is also used in Goodfriend and McCallum (2007) and in
Gerali, Neri, Sessa, and Signoretti (2010).
6
The “value-at-risk constraint” assumed by authors such as Zigrand, Shin, and Danielsson (2010), Adrian,
Moench, and Shin (2010b), and Adrian and Shin (forthcoming b) is an example of a constraint of this
form. Beaudry and Lahiri (2010) impose a similar constraint by simply assuming that intermediaries can
sell only riskless debt. The constraint assumed by Adrian and Shin (forthcoming b) is formally equivalent
to the one assumed by Beaudry and Lahiri (2010), though the former authors prefer to interpret the
constraint as one on value at risk.
32 Journal of Economic Perspectives

supply schedule XS will be upward sloping, as shown in Figure 3B, if the accept-
able leverage ratio is higher when the spread between the expected return on
the assets in which intermediaries can invest and the rate they must pay on their
liabilities is greater. Consider, for example, a value-at-risk constraint, that requires
the future value of the intermediary’s assets to be worth at least some fraction k
of the amount owed on its debt, with at least some probability 1 – p;; and suppose
that the risky asset in which the intermediary invests will pay at least a fraction s of
its expected payoff with probability 1 – p.. Then the value-at-risk constraint is satis-
fied if and only if the intermediary’s leverage ratio (debt as a fraction of the total
value of its assets) is no greater than s/k times the factor (1 + i b)/(1 + i s), where
i b is the expected return on the risky asset and i s is the rate that the intermediary
must pay on its debt. Thus the acceptable leverage ratio, and correspondingly the
maximum value of assets that the intermediary can acquire, will be an increasing
function of the credit spread.

The IS––MP Model with Credit Frictions

The equilibrium credit spread and volume of credit shown in Figure 3A are
determined for a particular value of national income Y ; because the schedules LS
and LD depend on Y,, as shown in Figure 2A, the location of the schedule XD (at
least) in Figure 3B also depends on Y.. For reasons already discussed above, a higher
level of Y should shift LS to the right and LD to the left, and each of these effects
results in a lower equilibrium value of the rate i s paid to savers, for any given position
of the schedule XS.. Hence we can once again derive an IS schedule, indicating the
equilibrium value of i s for any assumed level of income Y,, but now the IS schedule
will also include a given assumption about the supply of intermediation.7
The resulting model makes many of the same qualitative predictions about the
effects of economic disturbances or policy changes as the standard IS–MP model
(which is simply the special case in which the curve XS is assumed to be horizontal
at ω = 0). However, the dependence of the supply of intermediation on the capital
of intermediaries provides a channel for the amplification and propagation of the
effects of economic disturbances. An increase in aggregate economic activity will
generally increase the value of intermediaries’ assets (loans are more likely to be
repaid, land prices increase with increases in income, and so on) and hence their
net worth. This will allow additional borrowing by the intermediaries, and hence a
larger volume of credit for any given credit spread. Thus, the supply of intermedia-
tion schedule XS will shift down and to the right. A reduction in the interest rate

7
In fact, we can now solve for both i s and i b as functions of Y, but it is the relation between i s and Y
that is relevant for the IS–MP diagram, since it is i s—the rate at which intermediaries are able to fund
themselves—that corresponds to the operating target of the central bank. Plotting again the reaction of
the central bank’s target for i s to changes in economic activity as a curve MP, we again have a diagram of
exactly the kind shown in Figure 2B to determine simultaneously the equilibrium values of the interest
rate and output; the only important difference is that now we must clarify that the interest rate on the
vertical axis is the policy rate i s rather than the borrowing rate i b. Once the equilibrium values of Y and
i s have been determined, they can be transferred back to Figures 3A and B to determine the implied
equilibrium values of i b and L as well.
 Financial Intermediation and Macroeconomic Analysis 33

i s at which intermediaries are able to fund themselves can also increase intermedi-
aries’ net worth, if (as is often the case) they fund longer-term assets with short-term
borrowing that they must roll over, and in this case a reduction in i s will shift the XS
curve down and to the right as well.
Each of these effects will make the IS curve flatter (more interest-elastic) than
it would otherwise be.8 This means that a shift in the MP curve—due either to a
change in monetary policy or to a supply-side disturbance that shifts the aggregate-
supply curve—will have a larger effect on output as a consequence of these “financial
accelerator” effects. Bernanke and Gertler (1995) discuss evidence for the impor-
tance of such effects in the case of monetary policy shocks. Moreover, if a disturbance
leads to an increase or decrease in the capital of the intermediary sector, the altered
level of capital is likely to persist for some time. This can result in effects on economic
activity that are more persistent than the initial disturbance.
The presence of an upward-sloping XS curve (representing credit frictions)
essentially makes the IS curve steeper, and consequently acts to dampen the effects
on aggregate output of disturbances that shift the MP curve, to the extent that
the XS schedule is not itself shifted by the disturbances. In fact, however, the XS
schedule may well shift, in which case the net effect may well be to amplify output
fluctuations, rather than to dampen them.

Consequences of Shifts in the Supply of Intermediation

A more important consequence of this extension of the model is the fact that
shifts in the XS schedule—for either exogenous or endogenous reasons—become
an additional source of variations in aggregate demand, and hence in economic
activity and inflation.9 A disruption of the supply of intermediation will shift the
XS schedule up, so that financial intermediaries supply less credit at every level
of the credit spread ω. As shown in Figure 4A an upward shift in XS results in a
higher equilibrium credit spread, and a lower volume of lending, for any given level
of economic activity (reflected in the location of the XD schedule in the figure).
Transferring this larger spread back to Figure 3A, one observes that the implied
value of i s will be smaller, and the implied value of i b higher, for the given value of
Y.. Because this is true for each possible value of Y,, the IS schedule is shifted down
and to the left, as shown in Figure 4B. (Note that IS plots the equilibrium value of
i s rather than that of i b, because the policy reaction function specifies a target for
i s rather than for i b.) In the absence of any change in the monetary policy reaction

8
Of course, these reasons for the IS curve to be flatter must be balanced against the observation that a
steeper XS curve for a given level of capital in the intermediary sector will imply a steeper IS curve. This
is why the degree of amplification from credit frictions that is found in quantitative dynamic stochastic
general equilibrium models is sometimes quite modest.
9
The empirical dynamic stochastic general equilibrium models of Christiano, Motto, and Rostagno
(2010) and Gilchrist, Ortiz, and Zakrajsek (2009) each attribute a substantial fraction of the short-run
variability of real GDP to disturbances that vary the severity of financial frictions.
34 Journal of Economic Perspectives

Figure 4
Effects of a Disruption of Credit Supply

A: Effects on the Equilibrium Credit Spread ω and Volume of Lending L

for a Given Level of Aggregate Income Y

ω
XS
(between savers and borrowers)
Interest rate spread

ω2
ω1

XD

L2 L1 L
Volume of lending

B: Effects on the Equilibrium Policy Rate and Aggregate Income,

Taking into Account the Monetary Policy Reaction

is MP
Interest rate

i1
i2

IS

Y2 Y1 Y
Aggregate income

Notes: XS and XD are the supply and demand for intermediation. ω is the spread between i b, the interest
rate for borrowers, and i s, the interest rate for savers. The IS schedule shows the equilibrium interest
rate for any assumed level of current income Y, and MP is the monetary policy reaction function (MP).
 Michael Woodford 35

function, the upward shift in XS should result in both a decline in the policy interest
rate and a contraction of real activity.10
This prediction matches the consequences observed, for example, when the
Carter administration imposed credit controls in the second quarter of 1980. This
policy was followed by a contraction in real GDP at a rate of minus 8 percent per
year in that quarter, while the federal funds rate also fell from a level over 17 percent
per annum in April to only 9 percent by July 1980. The effects of a policy tightening
of this kind cannot be understood as a shift of the MP curve (or LM curve) in a
conventional IS–MP (or IS–LM)) diagram, but they are easily understood when one
realizes how changes in the supply of intermediation schedule should be expected
to shift the IS curve.
The dependence of the supply of intermediation on the capital of interme-
diaries also introduces an important channel through which additional types of
disturbances can affect aggregate activity. Any disturbance that impairs the capital
of the banking sector will shift the schedule XS up and to the left, with the effects
just discussed. This means that shocks that might seem of only modest significance
for the aggregate economy—in terms, say, of the total value of business losses that
directly result from the shock—can have substantial aggregate effects if the losses
in question happen to be concentrated in highly leveraged intermediaries, who
suffer significant reductions in their capital as a result. This was an important reason
for the dramatic aggregate effects in 2008–2009 of the losses in the U.S. subprime
mortgage market.
The supply of intermediation can also shift as a result of factors other than
a change in the capital of intermediaries; in particular, leverage constraints can
tighten or loosen, as a result of changes in the attitudes of intermediaries’ creditors
regarding the acceptable degree of leverage, or in the margin requirements associ-
ated with borrowing against the securities that intermediaries hold. Gorton and
Metrick (2009), Adrian and Shin (2009), and Geanakoplos (2010) have all stressed
the importance of increases in margin requirements in the overnight repurchase
(or “repo”) market as a factor that contracted the supply of credit in 2008 and 2009.
Even when shocks to the supply of intermediation originate in a tightening of
leverage constraints and/or margin constraints owing to an increased assessment
of the risk associated with intermediaries’ assets, the effects of the shocks will be
amplified by the dependence of the supply of intermediation on the capital of the
intermediary sector. Intermediaries that are forced to sell assets as a result of tight-
ened leverage constraints are likely to suffer losses, and more so to the extent that
many of them are forced to sell similar assets at the same time, or to the extent
that they are the only “natural buyers” of the assets in question. These losses will
then further reduce their capital, further reducing the amount that they are able

10
In this respect the framework sketched here agrees with the one proposed by Bernanke and Blinder
(1988), who refer to the relation that I call the IS curve as the “commodities and credit curve” instead,
precisely because it is shifted by credit-supply shocks in addition to the usual determinants of the IS
curve. The framework proposed here differs from that of Bernanke and Blinder primarily in offering a
different model of the supply of intermediation.
36 Journal of Economic Perspectives

to borrow, and hence requiring further asset sales. The result is a vicious spiral
that under some circumstances can substantially reduce credit supply. The resulting
contraction of aggregate output may result in further losses to the banks, further
reducing their capital, and hence tightening credit supply even more.

The Most Recent U.S. Credit Cycle

Understanding variations in financial conditions over the most recent credit

cycle requires attention to the behavior of multiple interest rates, not just the federal
funds rate that is targeted by the Federal Reserve. As shown in Figure 5, the Fed
Open Market Committee raised its target for the funds rate to a higher level during
the period 2006–07; but financial conditions did not tighten as much as one might
expect from the increase in the funds rate. First of all, spending decisions depend
more on the level of long-term interest rates, which in turn depend on the expected
average level of short rates over the coming decade, rather than the current level
of short rates alone. Since there was good reason to regard the low level of the
federal funds rate in 2003–04 as a temporary anomaly,11 the long rate implied by the
expected average future level of the short rates did not greatly increase as a result of
the increase in the funds rate between 2004 and 2006.
Moreover, yields on long-term Treasury bonds did not rise by even this much.
The term premium, which indicates the amount by which the actual yield on a
long-term bond exceeds the expected average level of short-term interest rates over
the term to maturity of the bond, declined during this period, as Figure 5 illustrates
for the case of a 10-year bond. In turn, the rates at which private parties can borrow
are not those applicable to the U.S. Treasury; the figure also shows, for example,
that the spread between the average yield on Baa-rated corporate bonds and 10-year
Treasuries also fell between 2004 and 2006.12 Hence corporate borrowing costs actu-
ally fell, despite the increase in the federal funds rate, owing to the declines in the
two spreads! In contrast, the increases in the two spreads during the financial crisis
greatly increased the cost of borrowing.
Even in the case of short-term borrowing, the federal funds rate alone is
not always an adequate measure of money market conditions. Figure 5 also plots
the spread between the three-month U.S. dollar London Interbank Offer Rate

11
For example, see the “fitted” long rates implied by the forecasting model of Kim and Wright (2005).
Indeed, the series plotted in Figure 5 is taken from the estimates of Kim and Wright (2005); their series
is updated at ⟨http://www.federalreserve.gov/econresdata/researchdata.htm⟩.
12
The spread between yields on this class of moderately risky corporate bonds and on similar-maturity
Treasury bonds is a commonly watched indicator of disturbances to the market for corporate debt,
which is strongly correlated with variations in economic activity. Gilchrist, Ortiz, and Zakrajsek (2009)
use an index of corporate bond spreads as a measure of the time-varying financial wedge in an estimated
monetary dynamic stochastic general equilibrium model and find that the co-movements of the bond
spreads with other aggregate variables are consistent with this interpretation.
 Financial Intermediation and Macroeconomic Analysis 37

Figure 5
The Federal Funds Rate Target and Some Interest-rate Spreads

6.0
FF target
10-year term premium
5.0 Baa-Treasury spread
LIBOR-OIS spread

4.0
Percentage points

3.0

2.0

1.0

0.0
2003 2004 2005 2006 2007 2008 2009 2010

Sources: The FF target is from the Federal Reserve Board; the 10-year term premium was calculated by
Don H. Kim and Jonathan H. Wright (available at the Federal Reserve Board website); the Baa–Treasury
spread is from the Federal Reserve Bank of St. Louis; the LIBOR-OIS spread is from Bloomberg.
Notes: The “FF target” is the Federal Funds rate target. The “10-year term premium” is the amount by
which the yield on a 10-year bond exceeds the expected average level of short-term interest rates over
the term to maturity of the bond. The “Baa–Treasury spread” is the spread between Baa-rated corporate
bonds and 10-year Treasuries. The “LIBOR–OIS spread” is the spread between the three-month U.S.
dollar London Interbank Offer Rate (LIBOR) and the overnight interest-rate rate swap (OIS) rate.

(LIBOR)13 and the overnight interest-rate rate swap (OIS) rate, which can be viewed
as essentially a market forecast of the average level of the federal funds rate over that
three-month period. The sharp increases in this spread during the crisis indicate
that the short-term borrowing costs of many banks (especially late in 2008) were
considerably higher than would be indicated by the federal funds rate.
It is popular to attribute the credit boom (at least in part) to the Federal Reserve
having kept the federal funds rate “too low for too long,” but comparison of the
path of the funds rate in Figure 5 with the measures of credit growth in Figure 1A
shows that the increase in lending was greatest in 2006 and the first half of 2007,
after the federal funds rate had already returned to a level consistent with normal
benchmarks. Instead, the fact that spreads were unusually low precisely during the
period of strongest growth in lending—as can be seen by comparing the spreads

13
The LIBOR rate is an average of quoted rates at which banks are able to borrow funds for a short term
(3 months, in the case of the series plotted here) on an uncollateralized basis. It is important not only
because it is the cost of additional funds for some banks, but because other lending rates—such as the
interest rate at which commercial and industrial loans are available to firms under existing loan commit-
ments—are often tied to the LIBOR rate. For alternative interpretations of variations in the LIBOR–OIS
spread, see Giavazzi (2008), Sarkar (2009), and Taylor and Williams (2009).
38 Journal of Economic Perspectives

shown in Figure 5 with the quantities in Figure 1—indicates that an outward shift
of the supply of intermediation schedule XS was responsible, rather than a move-
ment along this schedule in response to a loosening of monetary policy. The reason
for the shift seems to have been an increased appetite of investors for purportedly
low-risk short-term liabilities of very highly leveraged financial intermediaries; in
this journal, Brunnermeier (2009) details the changes in financing patterns during
this period.
The effects of such a shift were like those shown in Figure 4, but with the
reverse sign; as a consequence, the Fed’s increase in the funds rate over the period
between 2004 and 2006 did less to restrain demand than would ordinarily have been
expected.14 The increase in the riskless short-term rate did reduce households’ and
firms’ willingness to hold demand deposits, as a conventional money-demand equa-
tion would imply, and checkable deposits declined during this period, as shown in
Figure 1B; but this did not prevent a net increase in the overall liabilities of financial
intermediaries, and hence in credit supply.
The financial crisis that began in summer 2007 also originated in a change in
the supply of intermediation. It began when increased perceptions of risk resulted
in increases in the margin requirements demanded by creditors in short-term
lending collateralized by mortgage-backed securities, creating a liquidity crisis for
issuers of asset-backed commercial paper. The effect of deleveraging in this sector
on the market value of mortgage-backed securities further impaired the capital of
financial intermediaries more broadly, requiring further deleveraging, in a vicious
spiral: again, Brunnermeier (2009) describes this process in detail in this journal.
In terms of the model, the net result of both reductions in the acceptable degree
of leverage and impairment of the capital of the financial sector was a sharp leftward
shift of the supply of intermediation XS.. As illustrated by Figure 4, the result was a
simultaneous contraction of the volume of lending, as shown in Figure 1, and an
increase in spreads, as shown in Figure 5. The resulting leftward shift of the IS curve
(Figure 4) meant a contraction of aggregate demand, despite the substantial cuts
in the federal funds rate shown in Figure 5. The reduction in the riskless short-term
rate caused an increased willingness to hold transactions deposits, and checkable
deposits increased substantially, as seen in Figure 1B. But plentiful deposits were
not enough to restore the flow of credit, for an inability to increase the volume of
deposits was not the relevant constraint on the supply of credit.
Once this process was underway—and given that, for a time, it appeared
that the crisis might spiral out of control—uncertainty about the macroeconomic
environment likely caused a further leftward shift of the IS curve, by increasing
precautionary saving and increasing the option value of deferring investment.
Once the IS curve shifted sufficiently far, it ceased to be possible to maintain

14
Under this analysis, the fact that the Fed did not tighten policy even further can be said to have contrib-
uted to the credit boom. But the problem was not that the Fed failed to conform to the conventional
benchmark provided by the “Taylor rule,” as argued by Taylor (2009), but rather that it followed it too
faithfully, rather than taking account of the change in financial conditions.
 Michael Woodford 39

output near potential through cuts in the federal funds rate alone, owing to the
zero lower bound on nominal interest rates. Of course, the fact that reduced
aggregate demand resulted in lower economic activity and employment, rather
than simply in reductions in wages and prices to the extent needed to maintain
full employment, depended on the stickiness of wages and prices, as described in
standard textbook accounts.

Implications for Monetary Policy

To what extent does this extension of the standard model imply changes to the
conventional conduct of monetary policy?

Taking Account of Financial Conditions

The model’s most obvious implication is that decisions about interest-rate policy
should take account of changes in financial conditions—in particular, of changes in
interest-rate spreads.. Suppose that one’s goal is to set a value of the policy rate at each
point in time that is consistent with output equal to potential (or, more precisely,
the “natural rate of output” in the sense of Friedman, 1968). In the model sketched
above, this interest rate can be determined at any time given two other numbers:
1) the current value of the “natural rate of interest”—the real interest rate required
for output equal to the natural rate, in the absence of financial frictions15—converted
into an equivalent nominal interest rate by adding the current expected inflation
rate, and 2) the current interest-rate spread ω.16
The model therefore suggests that changes in credit spreads should be an
important indicator in setting the federal funds rate; the funds rate target should be
lower than would otherwise be chosen, given other conditions, when credit spreads
are larger. John Taylor (2008) has proposed, in this spirit, that his well-known
rule for setting the federal funds rate target (explained in Taylor, 1993) should
be modified to specify a funds rate target equal to that prescribed by the standard
“Taylor rule” minus the current value of the LIBOR–OIS spread shown in Figure 5.
In Cúrdia and Woodford (2010a), my coauthor and I show, in the context of a
New Keynesian dynamic stochastic general equilibrium model with credit frictions,
that such a modification of the standard Taylor rule can improve the economy’s
response to disturbances to the supply of intermediation.

15
This concept, derived from the ideas of Knut Wicksell, is discussed extensively in Woodford (2003,
chap. 4). One might alternatively define the natural rate as the real rate that would be required for
output equal to the natural rate of output under the assumption of a credit spread equal to some normal
(steady state) level; the important feature of the proposed definition is that it abstracts from the effects
of variations in the size of credit frictions.
16
In Cúrdia and Woodford (2009), we derive an intertemporal version of the “IS curve” in which the
credit spread appears as a shift factor. Gaspar and Kashyap (2006) were perhaps the first to propose such
a relation.
40 Journal of Economic Perspectives

Alternatively, a forecast-targeting approach to monetary policy of the kind I

recommended in this journal in Woodford (2007)—in which the central bank’s
target for the policy rate should be adjusted as necessary in order for its projections
of inflation and real activity to satisfy a quantitative target criterion—will automati-
cally incorporate responses to changes in financial conditions to the extent that
these shift the IS curve, as in the model sketched above. In addition, this alternative
approach has the advantage of not requiring the central bank to focus on a single
interest rate spread when multiple aspects of financial conditions are each relevant
to aggregate demand and supply determination.

“Unconventional” Monetary Policies

The model also implies that traditional interest-rate policy alone will not, in
general, provide a fully adequate response to a disturbance to credit supply, no
matter how large the cut in the policy rate that may be engineered. The reason is
that even if a sufficient reduction in the policy rate can offset the decline in aggre-
gate demand that would otherwise result from the shift in the IS curve, this does not
fully undo the distortions created by the increase in credit spreads. To the extent
that savers would be willing to supply additional funds at an interest rate lower than
the rate at which borrowers would be willing to borrow additional funds, then there
remains a misallocation of expenditure, even if the aggregate level of expenditure is
optimal: on this point, in Cúrdia and Woodford (2009), my coauthor and I provide
an explicit welfare analysis. Thus, to the extent that it is possible for policy to reduce
the size of the credit spread, this is desirable, even when interest-rate policy is able
to maintain output at potential.
But the case for acting to reduce credit spreads becomes even stronger if
the policy rate is constrained by the zero lower bound on nominal interest rates.
In the case of a large enough disturbance to the supply of intermediation, the IS
curve may shift so far down to the left that the point on it corresponding to the
natural rate of output may involve a negative nominal interest rate; for quanti-
tative examples, see Cúrdia and Woodford (2010b). In this case, conventional
monetary policy is unable to achieve the required level of aggregate demand,
because even a massive expansion of the supply of bank reserves cannot drive the
policy rate below zero. (The Federal Reserve found itself in this situation after
December 2008, as shown in Figure 5.) Under such circumstances, a policy that
can reduce credit spreads can further increase aggregate demand (by shifting
the IS curve to the right), despite the lack of room for any further reduction in
the policy rate.
Broadly speaking, two types of “unconventional” central-bank policies can
reduce credit spreads by shifting the supply of intermediation schedule XS to the
right. One is the extension of credit to intermediaries by the central bank on easier
terms than are available from private creditors; in particular, in the case that the
relevant financing constraint is the existence of too-high margin requirements for
private lending using assets held by the intermediaries as collateral, the central
bank may choose to lend against that collateral with a lower margin requirement.
 Financial Intermediation and Macroeconomic Analysis 41

Ashcraft, Gârleanu, and Pedersen (forthcoming) discuss the Federal Reserve’s Term
Asset-Backed Lending Facility, which provided financing for private purchases of
asset-backed securities, as an example of a policy of this kind and present evidence
of its success at reducing the spreads associated with asset-backed securities eligible
for the program. Such a policy can relax the constraint on the size of intermediary
balance sheets resulting from limited capital in the intermediary sector, by allowing
increased leverage.
Alternatively, the central bank may directly purchase debt claims issued by
private borrowers, so that total credit extended to the private sector can exceed
the size of intermediary balance sheets. Examples of policies of this kind during
the recent crisis include the Fed’s purchases of commercial paper through its
Commercial Paper Funding Facility and its purchases of mortgage-backed secu-
rities and agency debt. On the motivation for and effects of these programs,
see, for example, Adrian, Kimbrough, and Marchioni (2010), Gagnon, Raskin,
Remache, and Sack (2010), and in this journal Kacperczyk and Schnabl (2010).
In this case as well, the supply of intermediation XS is shifted to the right even
though the equilibrium relation between the credit spread and the quantity of
risky assets that can be held on the balance sheets of private intermediaries does
not change.17
It should not be assumed that because it is possible in principle for the central
bank to reduce equilibrium spreads through direct intervention in credit markets,
it is therefore desirable for the central bank to intervene continually to maintain
zero spreads. In Cúrdia and Woodford (2010b), we assume costs of central-bank
lending to the private sector that imply that under normal circumstances, it
will not be optimal for the central bank to hold assets other than highly liquid
Treasury securities on its balance sheet; but even so, central-bank lending to the
private sector can be justified on welfare grounds in the case of a large enough
disruption of credit supply. Gertler and Karadi (2010) reach a similar conclusion
using a related model.

Monetary Policy and Financial Stability

Finally, the fact that a reduction in the capital of intermediaries has an adverse
effect on the supply of intermediation—which in turn can seriously disturb both aggre-
gate demand and the composition of expenditure—implies that it is desirable to reduce
how frequently such crises occur. The role that monetary policy can or should play in
this regard remains controversial. However, a crisis that sharply reduces intermediary
capital can more easily occur, in the sense that the size of the required exogenous
disturbance is smaller, when intermediaries are highly leveraged. Thus, while the
increased volume of lending that a relaxation of leverage constraints makes possible

17
Note that on this analysis, the effects of targeted central bank asset purchases have nothing to do with
“quantitative easing,” as the effects do not depend on the purchases being financed by an increase in
bank reserves, nor do conditions in the market for bank reserves play any role in our analysis. See Cúrdia
and Woodford (2010b) for further discussion.
42 Journal of Economic Perspectives

can improve the short-run allocation of resources, this benefit must be weighed against
the increased risk of occurrence of a crisis that will (if it occurs) increase distortions in
the future, in ways that monetary policy will not then be able to counteract fully.
The model sketched here implies that increased leverage in the financial sector
is a natural consequence of looser monetary policy because of the effects of higher
incomes on loan demand and supply, shown in Figure 2A. Other, more complex
mechanisms, such as the model of misperception of risk by the funders of interme-
diaries proposed by Dubecq, Mojon, and Ragot (2009) can make this effect even
stronger. Given this, the consequences of policy for financial stability need to be
considered in making interest-rate decisions, alongside the consequences of policy
for aggregate economic activity and inflation.
The nature of this consideration should not be completely symmetrical:
marginal adjustments of interest rates always have consequences for output and
inflation, while they will have nonnegligible consequences for the risk of financial
instability only at certain times when the leverage is extreme enough for even small
changes in asset values to have substantial effects on intermediary capital. Improved
regulation and/or macroprudential supervision could further reduce the range of
circumstances in which this consideration would matter for monetary policy deci-
sions; and this would be desirable, if possible, as freeing monetary policy to focus
solely on output and inflation stabilization should allow those goals to be more
effectively achieved. But in the absence of a complete solution of that kind, it is
difficult to defend the view that financial stability can be ignored in monetary policy
decisions; and the development of practical real-time indicators of risks to financial
stability is accordingly an important challenge of the present moment.

■ I would like to thank Tobias Adrian, Bill Brainard, Vasco Cúrdia, Jamie McAndrews,
Benoit Mojon, Tommaso Monacelli, Julio Rotemberg, and Argia Sbordone for helpful
discussions, Luminita Stevens for research assistance, and the editors of this journal, David
Autor, Chad Jones, and Timothy Taylor, for many useful comments on earlier drafts. I would
also like to thank the National Science Foundation for research support under grant number
SES-0820438.

References

Adrian, Tobias, Karin Kimbrough, and Dina Intermediary Balance Sheet Quantities.” Federal
Marchioni. 2010. “The Federal Reserve’s Commer- Reserve Bank of New York Staff Report 428.
cial Paper Funding Facility.” Federal Reserve Bank Adrian, Tobias, and Hyun Song Shin. 2009.
of New York Staff Report 423. “Prices and Quantities in the Monetary Policy
Adrian, Tobias, Emanuel Moench, and Hyun Transmission Mechanism.” International Journal of
Song Shin. 2010a. “Financial Intermediation, Asset Central Banking, 5(4): 131–42.
Prices and Macroeconomic Dynamics.” Federal Adrian, Tobias, and Hyun Song Shin. Forth-
Reserve Bank of New York Staff Report 422. coming (a). “The Changing Nature of Financial
Adrian, Tobias, Emanuel Moench, and Hyun Intermediation and the Financial Crisis of
Song Shin. 2010b. “Macro Risk Premium and 2007–09.” Annual Review of Economics.
 Michael Woodford 43

Adrian, Tobias, and Hyun Song Shin. Forthcom- Dubecq, Simon, Benoit Mojon, and Xavier
ing (b). “Financial Intermediaries and Monetary Ragot. 2009. “Fuzzy Capital Requirements, Risk
Economics.” In Handbook of Monetary Economics, Shifting, and the Risk-Taking Channel of Monetary
vol. 3, ed. B. M. Friedman and M. Woodford. Policy.” Banque de France Working Paper 254.
Amsterdam: Elsevier. http://www.banquedefrance.fr/gb/publications
Ashcraft, Adam, Nicolae Gârleanu, and Lasse /telechar/ner/DT254.pdf.
Heje Pedersen. Forthcoming. “Two Monetary Friedman, Milton. 1968. “The Role of Monetary
Tools: Interest Rates and Haircuts.” In NBER Policy.” American Economic Review, 58(1): 1–17.
Macroeconomics Annual 2010, ed. D. Acemoglu Friedman, Milton, and Anna J. Schwartz. 1963.
and M. Woodford. Chicago: University of Chicago A Monetary History of the United States, 1867–1960.
Press. Princeton: Princeton University Press.
Baker, Dean. 2010. “Blame It on the Bubble.” Gagnon, Joseph, Matthew Raskin, Julie
The Guardian, March 8. Remache, and Brian Sack. 2010. “Large-Scale
Beaudry, Paul, and Amartya Lahiri. 2010. “Risk Asset Purchases by the Federal Reserve: Did They
Allocation, Debt Fueled Expansion, and Financial Work?” Federal Reserve Bank of New York Staff
Crisis.” June. http://faculty.arts.ubc.ca/alahiri Report 441.
/BeaudryLahiri.pdf. Gârleanu, Nicolae, and Lasse Heje Pedersen.
Bennett, Paul, and Stavros Peristiani. 2002. “Are 2010. “Margin-Based Asset Pricing and Deviations
U.S. Reserve Requirements Still Binding?” Federal from the Law of One Price.” http://pages.stern.nyu
Reserve Bank of New York Economic Policy Review, .edu/~lpederse/papers/MarginPricingLoOP.pdf.
8(1): 53–68. Gaspar, Vitor, and Anil K. Kashyap. 2006.
Bernanke, Ben S., and Alan S. Blinder. 1988. “Stability First: Reflections Inspired by Otmar
“Credit, Money, and Aggregate Demand.” American Issing’s Success as the ECB’s Chief Economist.”
Economic Review, 78(2): 435–39. In Monetary Policy: A Journey from Theory to Practice,
Bernanke, Ben S., and Mark Gertler. 1995. 86–118. Frankfurt: European Central Bank.
“Inside the Black Box: The Credit Channel of Geanakoplos, John. 1997. “Promises, Promises.”
Monetary Policy Transmission.” Journal of Economic In The Economy as an Evolving Complex System II,
Perspectives, 9(4): 27–48. ed. W. B. Arthur, S. N. Durlauf. and D. A. Lane,
Bernanke, Ben S., Mark Gertler, and Simon 285–320. Reading, MA: Addison Wesley.
Gilchrist. 1999. “The Financial Accelerator in a Geanakoplos, John. 2003. “Liquidity, Default
Quantitative Business Cycle Framework.” Chap. and Crashes: Endogenous Contracts in General
21 in Handbook of Macroeconomics, vol. 1C, ed. J. B. Equilibrium.” Chap. 5 in Advances in Economics
Taylor and M. Woodford. Amsterdam: Elsevier. and Econometrics, Theory and Applications II, ed. M.
Brunnermeier, Markus. 2009. “Deciphering the Dewatripont, L. P. Hansen, and S. J. Turnovsky.
Liquidity and Credit Crunch 2007–2008.” Journal of Cambridge, Cambridge University Press.
Economic Perspectives, 23(1): 77–100. Geanakoplos, John. 2010. “The Leverage
Buiter, Willem H. 2010. “Housing Wealth Isn’t Cycle.” In NBER Macroeconomics Annual 2009, ed.
Wealth.” Economics: The Open-Access, Open Assess- D. Acemoglu, K. Rogoff, and M. Woodford, 1–65.
ment E-Journal, vol. 4, no. 2010-22. http://www Chicago: University of Chicago Press.
.economics-ejournal.org/economics/journal Gerali, Andrea, Stefano Neri, Luca Sessa, and
articles/2010-22. Federico M. Signoretti. 2010. “Credit and Banking
Christiano, Lawrence, Roberto Motto, and in a DSGE Model of the Euro Area.” Journal of
Massimo Rostagno. 2010. Financial Factors in Money, Credit and Banking, 42(S1): 107–141.
Economic Fluctuations. European Central Bank Gertler, Mark, and Peter Karadi. 2010. “A Model
Working Paper 1192. http://www.ecb.europa.eu of Unconventional Monetary Policy.” http://www
/pub/pdf/scpwps/ecbwp1192.pdf. .carnegie-rochester.rochester.edu/april10-pdfs
Cúrdia, Vasco, and Michael Woodford. /Gertler%20Karadi.pdf.
2009. “Credit Frictions and Optimal Monetary Gertler, Mark, and Nobuhiro Kiyotaki. Forth-
Policy.” http://www.columbia.edu/~mw2230 coming. “Financial Intermediation and Credit
/credit_0709. Policy in Business Cycle Analysis.” In Handbook of
Cúrdia, Vasco, and Michael Woodford. 2010a. Monetary Economics, vol. 3, ed. B. M. Friedman and
“Credit Spreads and Monetary Policy,” Journal of M. Woodford. Amsterdam: Elsevier.
Money, Credit and Banking, 42(6, Supp.): 3–35. Giavazzi, Francesco. 2008. “Why Does the
Cúrdia, Vasco, and Michael Woodford. Spread between LIBOR and Expected Future
2010b.“The Central-Bank Balance Sheet as an Policy Rates Persist, and Should Central Banks Do
Instrument of Monetary Policy.” NBER Working Something About It?” VoxEU.org, posted June 2.
Paper 16208. Gilchrist, Simon, Alberto Ortiz, and Egon
44 Journal of Economic Perspectives

Zakrajsek. 2009. “Credit Risk and the Macroecon- economics without the LM Curve.” Journal of
omy: Evidence from an Estimated DSGE Model.” Economic Perspectives, 14(2): 149–69.
http://sws.bu.edu/aortizb/Credit%20Risk%20 Sarkar, Asani. 2009. “Liquidity Risk, Credit Risk,
and%20the%20Macroeconomy%20Gilchrist%20 and the Federal Reserve’s Responses to the Crisis.”
Ortiz%20%20Zakrajsek.pdf. Financial Markets and Portfolio Management, 23(4):
Goodfriend, Marvin, and Bennett T. McCallum. 335–48.
2007. “Banking and Interest Rates in Monetary Smant, David. 2002. “Bank Credit in the
Policy Analysis: A Quantitative Exploration.” Transmission of Monetary Policy: A Critical Review
Journal of Monetary Economics, 54(5): 1480–1507. of the Issues and Evidence.” MRPA paper 19816,
Gorton, Gary B., and Andrew Metrick. 2009. Munich Personal RePEc Archive.
“Haircuts.” NBER Working Paper 15273. Taylor, John B. 1993. “Discretion versus Policy
Greenlaw, David, Jan Hatzius, Anil K. Kashyap, Rules in Practice.” Carnegie-Rochester Conference Series
and Hyun Song Shin. 2008. Leveraged Losses: Lessons on Public Policy, vol. 39, pp 195–214.
from the Mortgage Market Meltdown. U.S. Monetary Taylor, John B. 2008. “Monetary Policy and
Policy Forum Report No. 2. Chicago: University of the State of the Economy.” Testimony before the
Chicago Booth School of Business. Committee on Financial Services, U.S. House of
Kacperczyk, Marcin, and Philipp Schnabl. 2010. Representatives, February 26.
“When Safe Proved Risky: Commercial Paper Taylor, John B. 2009. Getting Off Track: How
during the Financial Crisis of 2007–2009.” Journal Government Actions and Interventions Caused,
of Economic Perspectives, 24(1): 29–50. Prolonged, and Worsened the Financial Crisis. Stanford:
Kashyap, Anil, and Jeremy Stein. 1994. “Mone- Hoover Press.
tary Policy and Bank Lending.” In Monetary Policy, Taylor, John B., and John C. Williams. 2009.
ed. N. G. Mankiw, 221–56. Chicago: University of “A Black Swan in the Money Market,” American
Chicago Press. Economic Journal: Macroeconomics, 1(1): 58–83.
Kim, Don H., and Jonathan H. Wright. 2005. Woodford, Michael. 2003. Interest and Prices:
“An Arbitrage-Free Three-Factor Term Structure Foundations of a Theory of Monetary Policy. Princeton:
Model and the Recent Behavior of Long-Term Princeton University Press.
Yields and Distant-Horizon Forward Rates.” Woodford, Michael. 2007. “The Case for
Finance and Economics Discussion Series No. Forecast Targeting as a Monetary Policy Strategy.”
2005-33, Federal Reserve Board, August. Journal of Economic Perspectives, 21(4): 3–24.
Kiyotaki, Nobuhiro, and John Moore. 1997. Zigrand, Jean-Pierre, Hyun Song Shin, and
“Credit Cycles.” Journal of Political Economy, 105(2): Jon Danielsson. 2010. “Risk Appetite and Endog-
211–48. enous Risk.” Financial Markets Group Discussion
Romer, David. 2000. “Keynesian Macro- Paper 647.