Anda di halaman 1dari 6

American Economic Association

Credit, Money, and Aggregate Demand

Author(s): Ben S. Bernanke and Alan S. Blinder
Source: The American Economic Review, Vol. 78, No. 2, Papers and Proceedings of the One-
Hundredth Annual Meeting of the American Economic Association (May, 1988), pp. 435-439
Published by: American Economic Association
Stable URL:
Accessed: 10/08/2010 06:18

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The
American Economic Review.

Credit, Money, and Aggregate Demand


Most standard models of aggregatede- credit are viewed as perfect substitutesfor

mand, such as the textbookIS/LM model, auction-marketcredit("bonds"),and finan-
treat bank assets and bank liabilitiesasym- cial marketsclearonly by price.Modelswith
metrically.Money,the bankliability,is given a distinct role for creditarisewhen eitherof
a special role in the determinationof aggre- these assumptionsis abandoned.
gate demand. In contrast, bank loans are Following James Tobin (1970) and Karl
lumped togetherwith otherdebt instruments Brunnerand Allan Meltzer(1972),we choose
in a "bond market,"which is then conve- to abandon the perfect substitutabilityas-
niently suppressedby Walras'Law. sumption and ignore credit rationing.'Our
Much recent researchprovidesreasonsto model has three assets: money, bonds, and
question this imbalance.A growingtheoreti- loans. Only the loan marketneeds explana-
cal literature,based on models with asym- tion. We assume that both borrowersand
metric information,stressesthe importance lenders choose between bonds and loans
of intermediariesin the provisionof credit according to the interest rates on the two
and the specialnatureof bankloans. Empiri- credit instruments.If p is the interestrateon
cally, the instabilityof econometricmoney- loans and i is the interest rate on bonds,
demandequationshas been accompaniedby then loan demand is: Ld = L(p, i, y). The
new interest in the credit-GNP relation- dependenceon GNP (y) capturesth+etrans-
ship (see especially the work of Benjamin actions demandfor credit,whichmightarise,
Friedman). for example,from workingcapitalor liquid-
We have developedseveralmodels of ag- ity considerations.
gregate demand which allow roles for both To understandthe genesisof loan supply,
money and "credit"(bank loans). We pre- consider a simplified bank balance sheet
sent a particularlysimple one, a variantof (which ignores net worth) with assets: re-
the textbook IS/LM model,in this paper. serves, R; bonds, Bb; loans, LS; and liabili-
Thoughit has a simplegraphicalrepresen- ties: deposits, D. Since reservesconsist of
tation like IS/LM, this model permitsus to requiredreserves, TD, plus excess reserves,
pose a richer array of questionsthan does E, the banks' adding-upconstraintis: Bb +
the traditionalmoney-onlyframework. LS + E = D(1- T). Assuming that desired
portfolio proportionsdependon ratesof re-
I. The Model turn on the availableassets (zero for excess
reserves), we have LS = A(p, i)D(l - ), with
The LM curve is a portfolio-balancecon- similar equations for the+sharesof B b and
dition for a two-assetworld: asset holders E. Thus the conditionfor clearingthe loan
choose between money and bonds. Tacitly, marketis
loans and other forms of customer-market
(1) L(p, i, y) =X(p, i)D(1- T).

tDiscussants: Charles Freedman, Bank of Canada;

Charles I. Plosser, University of Rochester; Robert H.
Rasche, Michigan State University.
*Princeton University, Princeton, NJ 08544. We are 1Blinder (1987) offers a model in which there is
grateful to the NSF for supporting this research. rationing and no substitute for bank credit.

The money market is described by a con-

ventional LM curve. Suppose banks hold
excess reserves equal to e(i)D(1 - Tr).2 Then
the supply of deposits (we ignore cash) is
equal to bank reserves, R, times the money
multiplier, m(i) = [E(i)(1- _) + T)]'-. The
demand for deposits arises from the transac-
tions motive and depends on the interest
rate, income, and total wealth, which is con-
stant and therefore suppressed: D(i, y).
Equating the two gives

(2) D(i,y) = m()R. L c

Implicitly, D(i, y) and L(p, i, y) define the
nonbank public's demand function for bonds y
since money demand plus bond demand
minus loan demand must equal total finan- FIGURE 1
cial wealth.
The remaining market is the goods market,
which we summarize in a conventional IS we call the CC curve (for "commodities and
curve, written generically as3 credit"). It is easy to see that the CC curve is
negatively sloped like an IS curve, and for
(3) y=Y(i,P). much the same reasons. However, it is shifted
by monetary policy (R) and by credit-market
II. GraphicalRepresentation shocks that affect either the L(-) or X(-)
functions, while the IS curve is not. The
Use (2) to replace D(1- T) on the right- CC and LM curves are shown together in
hand side of (1) by (1- T)m(i)R. Then (1) Figure 1.
can be solved for p as a function of i, y, and Our CC curve reduces to the IS curve if
R :4 loans and bonds are assumed to be perfect
substitutes either to borrowers (Lp -* - oo)
(4) p =O(il+y, R). or to lenders (Xp -* oo), or if commodity
demand is insensitive to the loan rate (Yp
Finally, substitute (4) into (3) to get = 0)-which would make the loan market
irrelevant to IS/LM. This clarifies the spe-
(5) y=Y(i, (i,y,R)) cial assumptions implicit in the money-only
which, in deference to Don Patinkin (1956), The opposite extreme, or credit-only view,
would arise if money and bonds were perfect
substitutes (Di -* - xo), which would make
the LM curve horizontal. Keynes' explana-
tion for the liquidity trap is, of course, well
known. We think of high substitutability as
2For simplicity we assume that only i, not p, in-
fluences the demand for excess reserves. more likely to arise from financial innova-
3The interest rates in (3) should be real rates. But a tions which create new money substitutes.
model of aggregate demand takes both the price level However, even with a liquidity trap, mone-
and inflation as given; so we take the expected inflation tary policy still matters because it influences
rate to be constant and suppress it.
4p is an increasing function of i as long as the
the CC curve.
interest elasticity of the money multiplier is not too Now let us turn to the intermediate cases
large. represented by Figure 1.

III. ComparativeStatics5 the model, such a shock should reduce credit,

GNP, and the interest rate on government
Most conventional shocks work in our bonds while raising the interest rate on loans.
model just as they do in IS/LM. For exam- Another notable example with the same pre-
ple, an expenditure shock shifts the CC curve dicted effects is the credit controls of
along a fixed LM curve, and a money- March-July 1980. In this instance " tight
demand shock shifts the LM curve along money" should, and apparently did, reduce
a fixed CC curve. The effects are familiar interest rates on government bonds.
and need not be discussed. The only note-
worthy difference is that a rise in bank re- IV. Implicationsfor MonetaryPolicy
serves might conceivably raise the rate of
interest in the credit model. Graphically, the We turn next to the traditional target and
ambiguity arises because an increase in R indicator issues of monetary policy. The so-
shifts both the CC and LM curves outward. called monetary indicator problem arises if
Economically, the credit channel makes the central bank sees its impact on aggregate
monetary policy more expansionary than in demand only with a lag but sees its impacts
IS/LM and therefore raises the transactions on financial-sector variables like interest
demand for money by more than in the rates, money, and credit more promptly.
conventional model. What does our model say about the suitabil-
Greater interest attaches to issues that ity of money or credit as indicators?
elude the IS/LM model. An upward shift in Table 1 shows the qualitative responses of
the credit supply function, X(.) (which might GNP, money, credit, and bond interest rates
correspond, for example, to a decrease in the to a wide variety of shocks, assuming that
perceived riskiness of loans) shifts the CC bank reserves is the policy instrument. Col-
curve outward along a fixed LM curve, umns 1 and 2 display a conclusion familiar
thereby raising i and y. The interest rate on from IS/LM: money is a good qualitative
loans, p, falls, however. An upward shift in indicator of future GNP movements except
the credit demand function, L (.), which when money demand shocks are empirically
might correspond to a greater need for work- important. Columns 1 and 3 offer the corre-
ing capital, has precisely the opposite effects. sponding conclusion for credit: credit is a
We find it difficult to think of or identify good qualitative indicator except when there
major shocks to credit demand, that is, sharp are important shocks to credit demand. If
increases or decreases in the demand for money demand shocks were indeed more
loans at given interest rates and GNP. But important than credit demand shocks in the
shocks to credit supply are easy to con- 1980's, credit would have been a better indi-
ceptualize and to find in actual history. For cator than money.
example, Bernanke's (1983) explanation for What about the target question, that is,
the length of the Great Depression can be about the choice between stabilizing money
thought of as a downward shock to credit vs. stabilizing credit? Rather than try to con-
supply stemming from the increased riski- duct a complete Poole-style (1970) analysis,
ness of loans and banks' concern for liquid- we simply ask whether policymakers would
ity in the face of possible runs. According to respond "correctly" (i.e., in a stabilizing way)
to various shocks if they were targeting mon-
ey or targeting credit.
Consider first an expansionary IS shock.
Table 1 (line 5) shows that both money and
5 Most comparative statics results require no assump- credit would rise if bank reserves were un-
tions other than the ones we have already made. But, in changed. Hence a central bank trying to
a few cases, we encounter theoretical ambiguities that
can be resolved by invoking certain elasticity assump-
stabilize either money or credit would con-
tions spelled out in a longer version of this paper. If tract bank reserves, which is the correct
output is fixed on the supply side, y would be replaced stabilizing response. Either policy works, at
by P in Figure 1 and in the text discussion that follows. least qualitatively. A similar analysis applies


(1) (2) (3) Interest Period With Money With Credit
Rise in: Income Money Credit Ratea
1953:1-1973:4 .51,37 .17,11
Bank Reserves + + +
1974:1-1979:3 .50,.54 .50,.51
Money Demand - + - +
+ + 1979:4-1985:4 .11,34 .38,.47
Credit Supply + +
Credit Demand - - +
Commodity Demand + + + + aGrowth rates are first differences of natural loga-
aOn bonds. bCorrelations in nominal terms come first; correla-
tions in real terms come second.

to shocks to the supply of credit or to the

money multiplier.
But suppose the demand for money in- Table 2 shows the simple correlations be-
creases (line 2), which sends a contractionary tween GNP growth and growth of the two
impulse to GNP. Since this shock raises M, financial aggregates during three periods.
a monetarist central bank would contract Money was obviously much more highly cor-
reserves in an effort to stabilize money, which related with income than was credit during
would destabilize GNP. This, of course, is the period of stable money demand, 1953-73.
the familiar Achilles heel of monetarism. But the two financial aggregates were on a
Notice, however, that this same shock would more equal footing during 1974:1-1979:3.
make credit contract. So a central bank try- Further changes came during the period of
ing to stabilize credit would expand reserves. unstable money demand, 1979:4-1985:4;
In this case, a credit-based policy is superior money-GNP correlations dropped sharply
to a money-based policy. while money-credit correlations fell only
The opposite is true, however, when there slightly, giving a clear edge to credit.7
are credit-demand shocks. Line 4 tells us More direct evidence on the relative
that a contractionary (for GNP) credit- magnitudes of money-demand and credit-
demand shock lowers the money supply but demand shocks was obtained by comparing
raises credit. Hence a monetarist central bank the residuals from estimated structural mon-
would turn expansionary, as it should, while ey-demand and credit-demand functions like
a creditist central bank would turn contrac- D(*) and L(*) in our model. We used the
tionary, which it should not. logarithmic partial adjustment model, with
We therefore reach a conclusion similar to adjustment in nominal terms, which we are
that reached in discussing indicators: If not eager to defend but which was designed
money-demand shocks are more important to fit money demand. Hence, our procedure
than credit-demand shocks, then a policy of seems clearly biased toward finding rela-
targeting credit is probably better than a tively larger credit shocks than money shocks.
policy of targeting money. Unsurprisingly, estimates for the entire
1953-85 period rejected parameter stability
V. Empirical Evidence across a 1973:4-1974:1 break, so we con-
centrated on the latter period.8 Much to our
The foregoing discussion suggests that the
case for credit turns on whether credit de-
mand is, or is becoming, relatively more from Flow-of-Funds data). For details and analysis of
the latter, see Blinder (1985).
stable than money demand. We conclude 7Similar findings emerged when we controlled for
with some evidence that this is true, at least many variables via a vector-autoregression and looked
since 1979.6 at correlations between VAR residuals.
8Estimation was by instrumental variables. Instru-
ments were current, once, and twice lagged logs of real
6 In what follows, " money" is MI, "credit" is an government purchases, real exports, bank reserves, and
aggregate invented by one of us: the sum of intermedi- a supply shock variable which is a weighted average of
ated borrowing by households and businesses (derived the relative prices of energy and agricultural products.

amazement, we estimated moderately sensi- measure, the variance of money-demand

ble money and credit demand equations for shocks was much smaller than that of
the 1974:1-1985:4 period on the first try credit-demand shocks during the first sub-
(standard errors are in parentheses): period but much larger during the second.
The evidence thus supports the idea that
log M=- .06 + .939 logM1- .222i money-demand shocks became much more
(.34) (.059) (.089) important relative to credit-demand shocks
in the 1980's. But that does not mean we
+ .083logP+ .012logy should start ignoring money and focusing on
(.052) (.059) credit. After all, it is perfectly conceivable
that the relative sizes of money-demand and
SEE = .00811 D W = 2.04, credit-demand shocks will revert once again
to what they were earlier. Rather, the mes-
logC=- 1.75 + .885logC1 - .424p sage of this paper is that a more symmetric
(0.63) (.076) (.285) treatment of money and credit is feasible
and appears warranted.
+ .514i+ .075logP+ .292log y
(.389) (.086) (.107)
SEE =.00797, DW= 2.44.
Bernanke,Ben S., "Nonmonetary Effects of
Here y is real GNP, P is the GNP deflator, the Financial Crisis in the Propagation of
p is the bank prime rate, and i is the three- the Great Depression," American Eco-
month Treasury bill rate. Although the inter- nomic Review, June 1983, 73, 257-76.
est rate coefficients in the credit equation are Blinder,Alan S., "Credit Rationing and Effec-
individually insignificant, they are jointly tive Supply Failures," Economic Journal,
significant, have the correct signs, and are June 1987, 97, 327-52.
almost equal in absolute value-suggesting a , "The Stylized Facts About Credit
specification in which the spread between p Aggregates," mimeo., Princeton Univer-
and i determines credit demand. Notice that sity, June 1985.
the residual variances in the two equations Brunner, Karl and Meltzer, Alan H., "Mon-
are about equal. ey, Debt, and Economic Activity," Journal
Since the sample was too short to test of Political Economy, September/October
reliably for parameter stability, we examined 1972, 80, 951-77.
the residuals from the two equations over Patinkin, Don, Money, Interest, and Prices,
two subperiods with these results: New York: Harper and Row, 1956.
Poole, William," Optimal Choice of Monetary
variance of variance of Policy Instruments in a Simple Stochastic
money credit Macro Model," QuarterlyJournal of Eco-
period residual residual nomics, May 1970, 2, 197-216.
1974:1-1979:3 .265x 10-4 .687x 10-4 Tobin, James, "A General Equilibrium Ap-
1979:4-1985:4 .888x 10-4 .435X 10-4 proach to Monetary Theory," Journal of
Money, Credit and Banking, November
The differences are striking. By this crude 1970, 2, 461-72.