The Macroeconomics of the Great Depression: A Comparative Approach Author(s): Ben S.

Bernanke Source: Journal of Money, Credit and Banking, Vol. 27, No. 1 (Feb., 1995), pp. 1-28 Published by: Ohio State University Press Stable URL: http://www.jstor.org/stable/2077848 . Accessed: 16/08/2011 20:13
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp 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 support@jstor.org.

Ohio State University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Money, Credit and Banking.

http://www.jstor.org

MONEY, CREDIT, AND BANKING LECTURE

The Macroeconomics the GreatDepression: of A Comparative Approach

BEN S. BERNANKE

To UNDERSTAND GREAT THE DEPRESSION Holy Grail is the of macroeconomics.Not only did the Depressiongive birthto macroeconomicsas a distinct field of study, but also to an extent that is not always fully appreciatedthe experienceof the 1930s continuesto influencemacroeconomists' beliefs, policy recommendations,andresearchagendas. And, practicalitiesaside, findingan explanationfor the worldwideeconomic collapse of the 1930s remainsa fascinatingintellectual challenge. We do not yet have our hands on the Grail by any means, but during the past fifteen years or so substantialprogresstowardthe goal of understanding Depresthe sion has been made. This progress has a numberof sources, including improvements in our theoreticalframeworkandpainstaking historicalanalysis. To my mind, however, the most significantrecent developmenthas been a change in the focus of Depressionresearch,from a traditional emphasison events in the United States to a more comparativeapproachthatexaminesthe experiencesof manycountriessimultaneously. This broadeningof focus is importantfor two reasons:First, though in the end we may agree with Romer(1993) thatshocks to the domesticU.S. economy were a primarycause of both the Americanand world depressions, no account of the Great Depression would be complete without an explanationof the worldwide natureof the event, and of the channels throughwhich deflationaryforces spread among countries. Second, by effectively expandingthe data set from one observation to twenty, thirty,or more, the shift to a comparativeperspectivesubstantially
The author thanks BaxTyEichengreen for his comments and Ilian Mihov for excellent research assistance.

Journal of Money, Credit, and Banking, Vol. 27, No. 1 (February1995) Copyright 1995 by The Ohio State UniversityPress

: MONEY, CREDIT,AND BANKING

improves our ability to identify in the strict econometric sense the forces responsible for the world depression. Because of its potentialto bring the profession toward agreementon the causes of the Depression and perhaps, in consequence, macroeconomics I conto greaterconsensus on the centralissues of contemporary sider the improved identificationprovided by comparativeanalysis to be a particularly importantbenefit of that approach. of In this lecture I provide a selective survey of our currentunderstanding the GreatDepression, with emphasis on insights drawnfrom comparativeresearch(by both myself and others). For reasons of space, and because I am a macroeconomist ratherthana historian,my focus will be on broadeconomic issues ratherthanhistorical details. For readerswishing to delve into those details, Eichengreen(1992) proof treatment the monetaryand econonic historyof the vides a recent, authoritative have drawnheavily on Eichengreen'sbook (and his earlierwork) interwarperiod. I in in preparingthis lecture, particularly section 1 below. To review the state of knowledge about the Depression, it is convenientto make the textbook distinction between factors affecting aggregatedemand and those affecting aggregatesupply. I argue in section 1 that the factors that depressedaggregate demand aroundthe world in the 1930s are now well understood,at least in the broadterms. In particular, evidence that monetaryshocks played a majorrole in aroundthe world prithe GreatContraction,and that these shocks were transmitted marily throughthe workingsof the gold standard,is quite compelling. source of the Of course, the conclusion that monetaryshocks were an important Depression raises a central question in macroeconomics, which is why nominal shocks should have real effects. Section 2 of this lecture discusses what we know economies. aboutthe impactsof falling money supplies and price levels on interwar financialcrisis and I considertwo principalchannelsof effect: (1) deflation-induced levels, broughtabout by the in(2) increases in real wages above market-clearing of complete adjustment nominalwages to price changes. Empiricalevidence drawn from a range of countriesseems to provide supportfor both of these mechanisms. (in However, it seems that, of the two channels, slow nominal-wageadjustment the face of massive unemployment)is especially difficultto reconcile with the postulate the of economic rationality.We cannotclaim to understand Depressionuntil we can provide a rationalefor this paradoxicalbehaviorof wages. I concludethe paperwith some thoughtson how the comparativeapproachmay help us make progresson this importantremainingissue.
DEMAND: THE GOLD STANDARD AND WORLDMONEY SUPPLIES 1. AGGREGATE

During the Depression years, changes in outputand in the price level exhibiteda role for strongpositive correlationin almost every country,suggesting an important aggregate demand shocks. Although there is no doubt that many factors affected aggregate demand in various countriesat various times, my focus here will be on the crucial role played by monetaryshocks.

BEN S. BERNANKE : 3

For many years, the principaldebate aboutthe causes of the GreatDepression in the United States was over the importance be ascribedto monetaryfactors. It was to easily observedthatthe money supply,output,and prices all fell precipitouslyin the contractionand rose rapidly in the recovery; the difficulty lay in establishing the causal links among these variables. In theirclassic study of U.S. monetaryhistory, Friedmanand Schwartz(1963) presenteda monetaristinterpretation these obserof vations, arguingthat the main lines of causationran from monetarycontractionthe result of poor policy-makingand continuing crisis in the banking system to declining prices and output. OpposingFriedmanand Schwartz, Temin(1976) contended that much of the monetarycontractionin fact reflecteda passive responseof money to output;and that the main sources of the Depressionlay on the real side of the economy (for example, the famous autonomousdrop in consumptionin 1930). To some extent the proponentsof these two views arguedpast each other, with monetaristsstressingthe monetarysources of the latterstages of the GreatContraction (from late 1930 or early 1931 until 1933), and antimonetarists emphasizingthe likely importance of nonmonetaryfactors in the initial downturn. A reasonable compromise position, adopted by many economists, was that both monetary and nonmonetaryforces were operative at various stages (Gordon and Wilcox 1981). Nevertheless, conclusive resolution of the importanceof money in the Depression was hamperedby the heavy concentrationof the disputantson the U.S. case on one data point, as it were. l Since the early 1980s, however, a new body of researchon the Depression has emerged which focuses on the operationof the international gold standardduring the interwarperiod (Choudhriand Kochin 1980; Eichengreen 1984; Eichengreen and Sachs 1985; Hamilton 1988; Temin 1989; Bernankeand James 1991; Eichengreen 1992). Methodologically,as a naturalconsequenceof their concern with international factors, authors working in this area brought a strong comparative perspective into research on the Depression; as I suggested in the introduction,I consider this developmentto be a majorcontribution,with implicationsthat extend beyond the questionof the role of the gold standard.Substantively in markedcontrast to the inconclusive state of affairs that prevailed in the late 1970s the new gold-standard researchallows us to assert with considerableconfidencethat monetary factorsplayedan important causalrole,both in the worldwidedecline in prices and output and in their eventualrecovery.Two well-documentedobservationssupport this conclusion:2 First, exhaustive analysis of the operation of the interwargold standardhas shown that much of the worldwidemonetarycontractionof the early 1930s was not a passive response to declining output, but instead the largely unintendedresult of
1. That both sides considered only the U.S. case is not strictly true; both Friedmanand Schwartz (1963) and Temin (1976) made useful comparisonsto Canada, for example. Nevertheless, the Depression experiencesof countriesother than the United States were not systematicallyconsidered. 2. More detailed discussions of these points may be found in Eichengreenand Sachs (1985), Temin (1989), Bernankeand James (1991), and Eichengreen(1992). An importantearly precursoris Nurkse (1944).

: MONEY, CREDIT,AND BANKING

an interactionof poorly designed institutions,shortsighted policy-making,andunfavorable political and economic preconditions.Hence the correlationof money and price declines with output declines that was observed in almost every country is most reasonably interpretedas reflecting primarilythe influence of money on the real economy, ratherthan vice versa. Second, for reasonsthat were largely historical,political, and philosophicalrather than purely economic, some governmentsrespondedto the crises of the early 1930s by quickly abandoningthe gold standard,while others chose to remain on gold despite adverse conditions. Countriesthat left gold were able to reflate their money supplies and price levels, and did so after some delay; countriesremaining on gold were forced into furtherdeflation.To an overwhelmingdegree, the evidence shows that countriesthat left the gold standard recoveredfrom the Depressionmore quickly than countriesthat remainedon gold. Indeed, no countryexhibited significant economic recovery while remainingon the gold standard.The strong dependence of the rate of recovery on the choice of exchange-rateregime is further, powerful evidence for the importanceof monetaryfactors. Section 1.1 briefly discusses the first of these two observations, and section 1.2 considers the second. 1.1. The Sources of MonetaryContraction:MultipleMonetaryEquilibria? Despite the focus of the earliermonetaristdebate on the U.S. monetarycontraction of the early 1930s, this country was hardly unique in that respect: The same phenomenon occurred in most market-orientedindustrialized countries, and in many developing nationsas well. As the recentresearchhas emphasized,what most countries experiencing monetarycontractionhad in common was adherenceto the international gold standard. Suspended at the beginning of World War I, the gold standardhad been laboriously reconstructedafter the war: The United Kingdom returnedto gold at the prewarparity in 1925, Francecompleted its returnby 1928, and by 1929 the gold standard was virtuallyuniversalamong marketeconomies. (The shortlist of exceptions includedSpain, whose internalpolitical turmoilpreventeda returnto gold, and some Latin Americanand Asian countrieson the silver standard.)The reconstruction of the gold standard was hailed as a majordiplomaticachievement,an essential step towardrestoringmonetaryand financialconditions- which were turbulent during the 1920s to the relative tranquilitythat characterizedthe classical (18701913) gold-standardperiod. Unfortunately,the hoped-for benefits of gold did not materialize:Insteadof a new era of stability,by 1931 financialpanics and exchangerate crises were rampant,and a majorityof countriesleft gold in that year. A complete collapse of the system occurredin 1936, when Franceand the otherremaining "Gold Bloc" countriesdevaluedor otherwise abandonedthe strictgold standard. As noted, a strikingaspect of the short-livedinterwargold standard was the tendency of the nations that adhered to it to suffer sharp declines in inside money stocks. To understand general terms why these declines happened,it is useful to in

BEN S. BERNANKE : S

consider a simple identitythatrelatesthe inside money stock (say, M1) of a country on the gold standardto its reserves of monetarygold: M1 = (M1IBASE)x (BASEIRES)x (RESIGOLD)x PGOLD x QGOLD where M1 = M1 money supply (money and notes in circulationplus commercialbank deposits), BASE = monetarybase (money and notes in circulationplus reservesof commercial banks), RES = internationalreserves of the central bank (foreign assets plus gold reserves), valued in domestic currency, GOLD = gold reserves of the central bank, valued in domestic currency = PGOLD x QGOLD, PGOLD = the official domestic-currency price of gold, and QGOLD = the physical quantity(for example, in metric tons) of gold reserves. Equation(1) makes the familiarpoints that, underthe gold standard,a country's money supply is affected both by its physical quantityof gold reserves (QGOLD) and the price at which its centralbank standsready to buy and sell gold (PGOLD). In particular,ceteris paribus, an inflow of gold (an increase in QGOLD)or a devaluation (a rise in PGOLD) raises the money supply. However, equation (1) also indicates three additionaldeterminantsof the inside money supply under the gold standard: (1) The "moneymultiplier,"M1IBASE. In fractional-reserve banking systems, the total money supply (including bank deposits) is largerthan the monetarybase. As is familiarfrom textbooktreatments,the so-called money multiplier,M1IBASE, is a decreasing function of the currency-deposit ratio chosen by the public and the reserve-depositratio chosen by commercialbanks. At the beginning of the 1930s, M1IBASEwas relatively low (not much above one) in countriesin which banking was less developed, or in which people retaineda preferencefor currencyin transactions. In contrast, in the financially well-developed United States this ratio was close to four in 1929. (2) The inverse of the gold backingratio, BASEIRES.Because centralbankswere typically allowed to hold domestic assets as well as international reserves, the ratio BASEIRES the inverse of the gold backingratio(also called the coverageratio)exceeded one. Statutoryrequirements usually set a minimumbackingratio (such as the Federal Reserve's 40 percent requirement),implying a maximum value for BASEIRES(for example, 2.5 in the United States). However,therewas typically no statutoryminimumfor BASEIRES,an important asymmetry.In particular, sterilization of gold inflows by surpluscountriesreducedaveragevalues of BASEIRES. (3) The ratio of international reserves to gold, RESIGOLD. Under the gold(1)

: MONEY, CREDIT,AND BANKING

exchange standardof the interwarperiod, foreign exchange convertible into gold reserves, on a one-to-one basis with gold itself.3 could be counted as international Hence, except for a few "reservecurrency"countries, the ratio RESIGOLDalso usually exceeded one. Because the ratio of inside money to monetarybase, the ratioof base to reserves, and the ratio of reserves to monetarygold were all typically greaterthan one, the countries- far from equallingthe value of monemoney supplies of gold-standard tary gold, as might be suggested by a naive view of the gold standard were often large multiples of the value of gold reserves. Totalstocks of monetarygold continued to grow throughthe 1930s; hence, the observed sharpdeclines in inside money entirelyto contractionsin the averagemoney-goldratio. supplies must be attributed Why did the world money-goldratio decline? In the early partof the Depression period, priorto 1931, the consciously chosen policies of some majorcentralbanks role (see, for example, Hamilton 1987). For example, it is now played an important in ratherwidely acceptedthatFederalReservepolicy turnedcontractionary 1928, in an attemptto curb stock marketspeculation.In termsof quantitiesdefined in equafell tion (1), the ratio of the U.S. monetarybase to U.S. reserves (BASEIRES) from 1.871 in June 1928, to 1.759 in June 1929, to 1.626 in June 1930, reflectingboth conscious monetarytightening and sterilizationof induced gold inflows.4 Because of this decline, the U.S. monetarybase fell about6 percentbetween June 1928 and June 1930, despite a more-than-10percentincreasein U.S. gold reservesduringthe same period. This flow of gold into the United States, like a similarly large inflow into Francefollowing the Poincare'stabilization,drainedthe reservesof othergoldstandardcountriesand forced them into paralleltight-moneypolicies.5 However, in 1931 and subsequently,the large declines in the money-gold ratio that occurredaroundthe world did not reflect anyone's consciously chosen policy. The proximate causes of these declines were the waves of banking panics and the exchange-ratecrises thatfollowed the failureof the Kreditanstalt, largestbankin Austria, in May 1931. These developmentsaffectedeach of the componentsof the and money-gold ratio: First, by leading to rises in aggregatecurrency-deposit bank reserve-depositratios, bankingpanics typically led to sharpdeclines in the money multiplier,M1IBASE (Friedmanand Schwartz 1963; Bernankeand James 1991). Second, exchange-rate crises and the associated fears of devaluation led central banks to substitute gold for foreign exchange reserves; this flight from foreignexchange reserves reducedthe ratio of total reserves to gold, RESIGOLD.Finally, in the wake of these crises, central banks attemptedto increase gold reserves and coverage ratios as securityagainst futureattackson their currencies;in many counat 3. The gold-exchange standardwas proposedby participants the Genoa Conferenceof 1922, as a were not means of avertinga feared shortageof monetarygold. Although the Genoa recommendations the formally adopted, as the gold standardwas reconstructed reliance on foreign exchange reserves increased significantlyrelative to the prewarpractice. are 4. U.S. monetarydata in this paragraph from Friedmanand Schwartz(1963). Sumner(1991) suggests the use of the coverage ratio as an indicatorof the stance of monetarypolicy undera gold standard. 5. The gold flow into Francewas exacerbatedby a 1928 law that induceda systematicconversionof foreign exchange reserves into gold by the Bank of France;see Nurkse (1944).

BEN S. BERNANKE : 7

tries, the resulting "scramblefor gold" induced continuing declines in the ratio

BASEIRES. 6
A particularly destabilizingaspect of this process was the tendencyof fears about the soundness of banks and expectationsof exchange-ratedevaluationto reinforce each other (Bernankeand James 1991; Temin 1993). An element that the two types of crises had in common was the so-called "hotmoney,"short-term deposits held by foreigners in domestic banks. On one hand, expectations of devaluationinduced outflows of the hot-moneydeposits (as well as flightby domestic depositors),which threatenedto triggergeneral bankruns. On the otherhand, a fall in confidence in a domestic banking system (arising, for example, from the failure of a majorbank) often led to a flight of short-termcapital from the country,draininginternational reserves and threateningconvertibility.Otherthanabandoning parityaltogether, the centralbanks could do little in the face of combinedbankingand exchange-ratecrises, as the former seemed to demandeasy money policies while the latterrequired monetarytightening. From a theoreticalperspective, the sharpdeclines in the money-goldratio during the early 1930s have an interestingimplication:namely,thatunderthe gold standard as it operatedduringthis period, there appeared be multiple to potential equilibrium valuesof the moneysupply.7 Broadly speaking, when financialinvestors and other membersof the public were "optimistic,"believing that the bankingsystem would remainstable and gold paritieswould be defended, the money-goldratio and hence the money stock itself remained "high." More precisely, confidence in the banks allowed the ratio of inside money to base to remainhigh, while confidence in the exchange rate made centralbanks willing to hold foreign exchange reserves and to keep relativelylow coverage ratios. In contrast,when investorsandthe generalpublic became "pessimistic,"anticipating bankrunsanddevaluation,these expectations were to some degree self-confirmingandresultedin "low"values of the money-gold ratio and the money stock. In its vulnerabilityto self-confirmingexpectations, the gold standard appearsto have bornea stronganalogyto a fractional-reserve banking system in the absence of deposit insurance:For example, Diamond and Dybvig (1983) have shown that in such a system there may be two Nash equilibria,one in which depositorconfidence ensures that there will be no run on the bank, the other in which the fears of a run (and the resulting liquidation of the bank) are selfconurmlng.

An interpretation the monetarycollapse of the interwarperiod as a jump from of one expectationalequilibriumto anotherone fits neatly with Eichengreen's(1992) comparison of the classical and interwargold-standardperiods [see also Eichengreen (forthcoming)].Accordingto Eichengreen,in the classical period, high levels of centralbankcredibilityand international cooperationgeneratedstabilizingexpectations, for example, speculators'activities tendedto reverseratherthanexacerbate
6. Declines in BASE/RES also reflected sterilization of gold inflows by gold-surplus countries concerned about inflation; and, more benignly, the revaluationof gold reserves following currency devaluations. 7. I am investigatingthis possibility more formally in ongoing work with Ilian Mihov.

: MONEY, CREDIT,AND BANKING

movementsof currencyvalues away fromofficialexchangerates.In contrast,Eichengreen argues, in the interwarperiod central banks' credibility was significantly reduced by the lack of effective internationalcooperation(the result of lingering animosities and the lack of effective leadership)and by changingdomestic political equilibria- notably, the growing power of the labor movement, which reducedthe perceived likelihood that the exchange rate would be defended at the cost of higher unemployment. Banking conditions also changed significantlybetween the earlier and andlaterperiods, as war, reconstruction, the financialandeconomic problemsof the 1920s left the banksof manycountriesin a muchweakerfinancialcondition, and thus more crisis-prone.For these reasons, destabilizingexpectationsand a resulting low-level equilibriumfor the money supply seemed muchmorelikely in the interwar environment. Table 1 illustratesequation (1) with data from six representativecountries. The first three countriesin the table were membersof the Gold Bloc, who remainedon until relativelylate in the Depression(Franceand Polandleft gold the gold standard in 1936, Belgium in 1935). The remainingthree countries in the table abandoned gold earlier:the United Kingdom and Sweden in 1931, the United States in 1933. [Throughoutthis lecture I follow Bernankeand James(1991) in treatingany major departurefrom gold-standardrules, including devaluationor the imposition of exchange controls, as "leaving gold."] Of course, the gold leavers gained autonomy for their domestic monetarypolicies; but as these countriescontinuedto hold gold reserves and set an official gold price, the componentsof equation(1) could still be calculatedfor those countries. Several useful points may be gleaned from Table 1: First, observe the strongcormembershipand falling M1 money supplies (a respondencebetween gold-standard minorexception is Poland, which manageda small growth in nominalM1 between 1932 and 1936). Second, note the sharpdeclines in M1IBASEandRESIGOLD,reflecting (respectively)the bankingcrises and exchangecrises (both of which peaked in 1931). Third, the table shows the tendency of gold-surpluscountriesto sterilize (that is, BASEIREStends to fall in countriesexperiencingincreases in gold stocks, QGOLD). A striking case shown in Table 1 is that of Belgium: Although that countrywas the beneficiary of large gold inflows early in the Depression, the combinationof declines in M1IBASE (reflecting banking panics), RESIGOLD(reflecting liquida(the resultof conscious sterilization of foreign-exchangereserves), andBASEIRES tion early in the period, and of attempts to defend the exchange rate against speculative attacklater in the period) induced sharpdeclines in the Belgian money stock. Similarly,because of falls in M1IBASEandRESIGOLD,Franceexperienced almost no nominal growth in M1 between 1930 and 1934, despite a more than 50 percentincrease in gold reserves. The otherGold Bloc countryin the table, Poland, experiencedmonetarycontractionprincipallybecause of loss of gold reserves. Another interestingphenomenonshown in Table 1 is the tendency of countries devaluing or leaving the gold standardto attractgold away from countries still on the gold standard.In the table, the United Kingdom, Sweden, and the United States

(pJnu?luoo)

1Z S b91 S-Ibl Z-6bl 1-8 1 8 Z L6 8 86 a7090

8b 8b 8b 8b 8b 8b 8b 8b
a7OOd

Z Z Z Z Z Z Z Z

SS Z ZbS Z SLS Z
SOZ-Z

Z0'1 6ZO' I
101 ' I ZOZ' I

IIZ, 89Z I IIZ I

901-1

60 8Z 819 Z80

Z I Z Z

ObL' I 19'Z Z80'1 08Z'I

SS#ISS88

Lf I ZZS I 80S I 86b 1

SS8gI I S

a7091SSX

LSSI S I SOZI S801 bOO I IZOI 00 1 886

I rll IS

96 1 S6 1 b6 1 61 Z6 1 161 06 1 6Z61

(161 sqtusxdos ; plBpUP}S

O L6ZZ Z S9bl 1 80bl b 961
Z LL8

; Pl

p3pu3dsns SSLI S9SI 6bbl 80bl Z91 6ZZI 191 8Z I

I rll

NHaHMS

8 88 8-0801 8 6901 a7090

991 991 991 991 991 991 991 991

a7OOd

-0 0 0 0 0 -0 0 0

0 0 0 0 0 0 0 0 a70o1sa

1 1 1 1 1 1 1 1

16Z' S19'b 06S'b S6'b Z8'9 ZSb'9 669' S SZ8'S

SS#ISS88

OOLI b69 1 Zb9 1 089 1 L99 1 6LS I 819 1 O9S I

SS8gI I S

96 1 S61 b6 1 61 Z6 1 161 061 6Z61

(I 61 lJqtU3ld3S 9-19S 8-0ZS O-bZS 6-1LS l-bS b-S l-L8Z 6 SbZ a7090 61 61 06 06 06 06 06 06 CZ Z Z Z Z Z 6Z I 8L I
99Z I

plBpU}S

PloWpapuadsns)NOaONIN [ aINn L9 1 6LS 1 VN bSL I S08 1 LbO Z 9 Z bOS-Z

xsw/ IS

Z8Z I
S9Z I

860- 1 90'1 1 1' I bl'1 S6'1

99Z' I

8S- I LOL I Z6b 1

a70o1saa

L69'1 6b6' 1
SS#ISS88

blb 9S66 VN Z80b 6b1b 98bb OZb9b 88LZb

IS

96 1 S6 1 b6 1 61 Z6 1 161 06 1 6Z61

a7OOd

99 6-bL 6 b8 08 L b8 101 6 b6 811 a7090

Z6 S Z6 S Z6-S Z6 S Z6-S Z6-S Z6 S Z6 S

a7OOd

9LO-I
190 1 9SO I

S81-1 LZ I SS I SL I OSL I
a70o1sa

b9' SSI' 69'Z 96b'Z

LLI 'Z

888' 1 60L'I 06'1

S2XISS88

(S61 t1oleWpanleAsp) Ob 1 6SOZ LLZI L681 10-1 1981 08Z 1 Z081 SLZI LLI L9Z 1 Sb61 8Z 1 ZIZZ 6 1 b8ZZ
SS88/IS 1S

NnID1gS

96 1 S61 b6 1 61 Z61 161 06 1 6Z61

(961 3q}3O 8-199Z 1-806 Z-lb8b 6-bbib 6 68b b 6SOb b 8SI 9SbZ a7090 89 ZZ 96-91 96 91 96 91 96 91 96-91 96-91 96 91
a7OOd

panleAsp'961 llJdY 1 ouoo aSueqoxaposodtul) aNY1Od LSS'I 86Z'I 860' 1 9SI'I

010' 1 101 ' I 901 'I

bZO I
OZO I

Z10 1 S10 1
bSO-I

LO 1 68b 1 Z9 1
a70o1sa

601'1
SS#ISS88

81Z 1 OZI bbZ I b9Z1 9Z 1 6Z 1 SZ 1 bS 1

asw/ IS

L6ZLII 600801 ISb11 98bll 61SIZI 8bLZZI OZLIII Z9S101

IS

96 1 S61 b6 1 61 Z6 1 161 06 1 6Z61

(961lJqOlDo pJnlBA3p) H3NYXA 961-6Z61 S]INnOD

XIS Nl A14dnS AHNOW AHI A0 SINYNINXHIHa

I H1geS

10 : MONEY, CREDIT,AND BANKING

TABLE UNITED

(COntinUed) STATES (SUSPended gO1d Standard MarCh 1933)

M1
1929 1930 1931 1932 1933 1934 1935 1936 26434 24922 21894 20341 19759 22774 27032 30852

M 1iBASE
3.788 3.498 2.831 2.534 2.380 2.396 2.335 2.327

BASEIRES
1.746 1.655 1.854 1.900 2.057 1.154 1.144 1.178

GOLD RESI
1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0

PGOLD
0.6646 0.6646 0.6646 0.6646 0.6646 1.1253 1.1253 1.1253

QGOLD
6014.0 6478.9 6278.8 6358.6 6072.7 7320.9 8997.8 10004.7

NOTES. The table sllustrates the identlty, equation ( I ), for slx countrleS. Where possible, values are end-of-year. the Appendix. Definitions are as follows.

Data sources are given in

M I = Money and notes in circulatlon plus commercial bank deposits; in local currency (milllons). BASE = Money and notes in circulation plus commercial bank reserves; in local currency. RES = Internatlonal reserves (gold plus foreign assets); valued in local currency. GOLD = Gold reserves; valued in local currency at the official gold price = PGOLD X QGOLD. PGOLD = Official gold price (units of local currency per gram); for countries not on the gold standard, a legal fiction rather than a market
prlce.

QGOLD = Physlcal quantlty of gold reserves, in metric tons.

all experienced significant gold inflows startingin 1933. This seemingly perverse result reflected the greaterconfidenceof speculatorsin alreadydepreciatedcurrencies, relative to the clearly overvalued currenciesof the Gold Bloc. This flow of gold away from some importantGold Bloc countrieswSs the final nail in the gold standard'scoffin. 1.2. The MacroeconomicImplicationsof the Choice of Exchange-rateRegime gold standardsuffered We have seen that countriesadheringto the international largely unintendedand unanticipateddeclines in their inside money stocks in the late 1920s and early 1930s. These declines in inside money stocks, particularlyin 1931 and later, were naturallyinfluencedby macroeconomicconditions;but they were hardly continuous, passive responses to changes in output. Instead, money supplies evolved discontinuouslyin response to financialand exchange-ratecrises, crises whose roots in turnlay primarilyin the political and economic conditionsof the 1920s and in the institutionalstructureas rebuilt after the war. Thus, to a first approximation,it seems reasonableto characterizethese monetaryshocks as exogoutput, suggesting a significantcausal role enous with respect to contemporaneous for monetaryforces in the world depression. However, even strongerevidence for the role of nominalfactorsin the Depression is providedby a comparisonof the experiencesof countriesthatcontinuedto adhere to the gold standardwith those that did not. Although, as has been mentioned, the to greatmajorityof countrieshad returned gold by the late 1920s, therewas considerable variation in the strengthof national allegiances to gold during the 1930s: Many countries left gold following the crises of 1931, notably the "sterlingbloc" (the United Kingdom and its tradingpartners).Othercountriesheld out a few years more before capitulating(for example, the United States in 1933, Italy in 1934). Finally, the diehardGold Bloc nations, led by France, remainedon gold until the

BEN S. BERNANKE : 11

final collapse of the system in late 1936. Because countriesleaving gold effectively removed the externalconstrainton monetaryreflation, to the extent that they took advantageof this freedom we should observe these countriesenjoying earlier and strongerrecoveries than the countriesremainingon the gold standard. That a clear divergence between the two groups of countriesdid occur was first noticed in a pathbreaking paper by Choudhriand Kochin (1980), who considered the relative performancesof Spain (which as mentionedneverjoined the gold standardclub), three Scandinaviancountries(which left gold following the sterlingcrisis in September1931), and four countriesthatremainedpartof the Gold Bloc (the Netherlands,Belgium, Italy, and Poland).Choudhri Kochinfoundthatthe goldand standardcountries suffered substantiallymore severe contractionsin output and prices than did Spain and the three Scandinavian nations. In anotherimportant paper, Eichengreenand Sachs (1985) examineda numberof macrovariablesin a sample of ten major countries over the period 1929-1935; they found that by 1935 countriesthat had left gold relatively early had largely recoveredfrom the Depression, while the Gold Bloc countriesremainedat low levels of output and employment. Bernanke and James (1991) confirmed the general findings of the earlier authorsfor a broadersample of twenty-four(mostly industrialized)countries, and Campa(1990) did the same for a sample of Latin Americancountries. If choices of exchange-rateregime were random,these results would leave little doubt as to the importanceof nominal factors in determiningreal outcomes in the Depression. Of course, in practicethe decision aboutwhetherto leave the gold standardwas endogenousto a degree, and so we must be concernedwith the possibility that the results of the literatureare spurious, that is, that some underlying factor accounted for both the choice of exchange-rateregime and the subsequentdifferences in economic performance.In fact, these results are very unlikely to be spurious, for two general reasons: First, as has been documentedin detail by Eichengreen(1992) and others, for most countries the decision to remain on or leave the gold standardwas strongly influencedby internaland externalpolitical factors and by prevailingeconomic and philosophicalbeliefs. For example, the Frenchdecision to stay with gold reflected, among other things, a desire to preserve at any cost the benefits of the Poincare' stabilizationand the associated distributional bargainsamong domestic groups;an overwhelmingly dominant economic view (shared even by the Communists)that sound money and fiscal austeritywere the best long-runantidotesto the Depression; and what can only be describedas a strongassociationof nationalpridewith maintenanceof the gold standard.8 Indeed, as BernankeandJames(1991) point out, economic conditions in 1929 and 1930 were on averagequite similarin those countries
8. The differencesin world views were most apparent the ill-fated 1933 LondonEconomic Conferat ence, in which Gold Bloc delegates decriedlack of sound money as the root of all evil, while representatives of the sterlingbloc stressedthe imperativesof reflationand economic expansion(Eichengreenand Uzan 1993). The persistenceof these attitudesacross decades is fascinating;note the attachmentof the Frenchto thefranc fort in the recent troublesof the EMS, and the contrastingwillingness of the British (as in September 1931) to abandonthe fixed exchange rate in the pursuitof domestic macroeconomic objectives.

12 : MONEY, CREDIT,AND BANKING

that were to leave gold in 1931 and those that would not; thus it is difficultto view this choice as being simply a reflection of cross-sectional differences in macroeconomlc perrormance. Second, and perhaps even more compelling, is that any bias created by endogeneity of the decision to leave gold would appearto go the wrong way, as it were, to explain the facts: The presumptionis that economically weaker countries, or those suffering the deepest depressions, would be the first to devalue or abandon gold. Yet the evidence is that countries leaving gold recovered substantiallymore rapidly and vigorously than those who did not. Hence, any correctionfor endogeneity bias in the choice of exchange-rateregime should tend to strengthenthe association of economic expansion and the abandonment gold. of Tables2 and 3 below extendthe resultsof BernankeandJames(1991) on the links between exchange-rateregime and macroeconomicperformance,using a data set similar to theirs. Both tables employ annualdata on thirteenmacroeconomicvariables for up to twenty-sixcountries,dependingon availability(see the Appendixfor a list of countries, data sources, and data availabilities).Following similartables in Bernankeand James, Table2 shows averagevalues of the log-changes of each variable (except for nominal and real interestrates, which are measuredin percentage points) for all countriesin the sample, and for the subsetsof countrieson and off the gold standardin each year.9 Averages for the whole sample are reportedfor each year from 1930 to 1936; because almost all countries were on gold in 1930 and almost all had left gold by 1936, averagesfor the subsamplesare shown for 19311935 only. The statistical significance of the divergences between gold and nongold countries is assessed in Table3. Lines marked"a"in Table3 presentthe resultsof paneldata regressions of each of the macroeconomicvariablesin Table 2 against a constant, yearly time dummies, and a dummy variablefor gold-standardmembership (ONGOLD). (Lines in Table 3 marked"b" should be ignored for now.) For each country-yearobservation, the variable ONGOLDindicates the fractionof the year that the countrywas on the gold standard(the numberof months on the gold standarddivided by twelve). The regressionsuse data for 1931-1935 inclusive, but the results are not sensitive to addingdata from 1930 or 1936 or to dropping1931. Because each regressioncontains a full set of annualtime dummies, the estimatedcoefficients of ONGOLD in each regression may be interpretedas reflecting purely cross-sectionaldifferencesbetween countrieson andoff gold, holdingconstantaverage macroeconomicconditions. Absolute values of t-statistics,given undereach estimatedcoefficient, indicatethe significanceof the between-groupdifferences. Tables 2 and 3 are generallyquite consistent with the conclusions that (1) moner

9. As noted earlier,we treata countryas leaving gold if it deviates seriouslyfrom gold-standard rules, for example, by imposing comprehensivecontrols or devaluing, as well as if it formally renounces the gold standard.Dates of changes in gold-standard policies for twenty-fourof our countriesare given by Bernankeand James, Table2.1. In addition, we take Argentinaand Switzerlandas leaving gold on their official devaluationdates (December 1929 and October 1936, respectively). Reportedvalues are simple within-groupaverages of the data; however, weighting the results by gold reserves held or relative to 1929 productionlevels (available in League of Nations 1945) did not qualitativelychange the results.

TABLE 2 AVERAGEBEHAVIOR SELECTED OF MACROVARIABLES FORCOUNTRIES AND OFF THE GOLD ON STANDARD, 1930-1936
1930 1931 1932 1933 1934 1935 1936

1. Manufacturing production(log-change) Average -.066 -.116 -.090 ON -.117 -.173 OFF -.113 -.057 2. Wholesale prices (log-change) Average -.116 -.122 -.045 ON -.140 -.133 OFF -.084 -.011 3. M1 money supply (log-change) Average .016 -.088 -.068 ON -.094 -.088 OFF -.076 -.060 4. M1-currencyratio (log-change) Average .030 -.129 -.006 ON -.142 -.052 OFF -.102 .014 5. Nominal wages (log-change) Average .004 -.030 -.053 ON -.027 -.070 OFF -.039 -.045 6. Real wages (log-change) Average .122 .094 .007 ON .110 .064 OFF .059 -.020 7. Employment(log-change) Average -.066 -.117 -.074 ON -.113 -.137 OFF -.127 -.047 8. Nominal interest rate (percentagepoints) Average 5.31 5.43 5.29 ON 5.22 4.20 OFF 5.90 5.68 9. Ex-post real interest rate (percentagepoints) Average 16.89 9.39 6.51 ON 10.38 9.41 OFF 7.16 5.47 10. Relative price of exports (log-change) Average -.033 -.011 -.047 ON .003 -.019 OFF -.040 -.058 11. Real exports (log-change) Average -.073 -.179 -.222 ON -.193 -.292 OFF -.146 -.192 12. Real imports (log-change) Average -.071 -.211 -.264 ON -.159 -.250 OFF -.315 -.271 13. Real share prices (lKog-change) Average -.107 -.186 -.214 ON -.181 -.219 OFF -.198 -.211

.076 .068 .078 017 065 002 006 045 .007 024
009

.100 .025 .120 .018 -.037 .033 .019 -.013 .028 -.002 - .016 .002 - .002 -.031 .007 -.023
.005

.074
001

.072

.008 .024 038 .036 .027 067 .046

01 1

.048

.074

030 030 033 029

009

037 006
001

-.011

022 .004 022 .016 031 .064 016 .083 89 05 86

.19

.031

.032 025
.050

-.018

-.032 .096 .028 .113 3.97 3.26 4.13

1.11

.006 .065 37 69 56 78 94 64 .076 .134 .058 .014 008 .021 .004 006 .008 .133 .139 . 130

.068

3.79

3.35 0.61 .084 .140 .070 .056 .015 .067 .038 -.067 .070 .060 -.028 .092

92 62 067 112 058 .021 024 .030 .020 012 .027 .091 .062 .098

-8.93

.039

.072

.049

.115

NOTES: For each variableand year, the table presentsthe overall average value of the variable,and the averagefor countrieson and off the gold standardin thatyear (see Bernankeand James 1991). As most countrieswere on the gold standard 1930 andoff the gold standard in in 1936, disaggregateddatafor those years are not presented.Data are annualand for up to twenty-sixcountries,dependingon dataavailability (see the Appendix). Real wages, real share prices, and the ex post real rate of interestare computedusing the wholesale price index. If a countryis on the gold standardfor a fractionf of a particular year, the values of its variablesfor the whole year are counted with the gold standardcountrieswith weightf and with non-gold-standard countrieswith weight 1-f for that year. The proportion country-months of "on gold" in each year are as follows: 0.676 (1931), 0.282 (1932), 0.237 (1933), 0.205 (1934), 0.160 (1935).

TABLE

REGRESSIONS DUMMIES,

OF SELECTED
1931-1935

MACRO

VARIABLES

AGAINST

GOLD

STANDARD

AND

BANKING

PANIC

Dependent variable
Manufacturing production ( lb) (la)

ONGOLD
-.0704 (4.04) -.0496 (2.80)

PANIC

AdjustedR2
0.601

-.0926 (3-50)

0.634

Wholesale prices

(2a) (2b)

-.0914 (8.20) -.0885 (7.47) -.0129 (0-73)

0.622 0.620

Money (M 1)

supply

(3a) (3b)

-.0534 (3.26) -.0344 (2.06) -.0846 (3.40)

0.297 0.352

Ml-currency ratio

(4a) (4b)

-.0329 (1.91) -.0176 (0.99) -.0680 (2.55)

0.263 0.294

Nominal wages

(5a) (Sb)

-.0204 (2.62) -.0145 (1.78) -.0262 (2.16)

0.196 0.219

Real

wages

(6a) (6b)

.0605 (5.84) .0656 (5-99) -.0230 (1.41)

0.466 0.470

Employment

(7a) (7b)

-.0610 (4.38) -.0507 (3.48) -.0458 (2.10)

0.557 0.569

Nominal est rate

inter-

(8a) (8b)

-1.22 (2.83) -1 00 -0 97

0.109 0.116

(2.20) Ex-post interest real rate (9b) (9a) 2.70 (2.07) 2.16 (1.56) Relative of price (loa) ( lOb) .0464 (1.70) .0288 (1.00) Real exports (1 la) (l lb) -.0745 (2.08) -.0523 (1 Real imports (12a) (12b) 39)

(1.43) 0.264 2.39 (1.16) 0.198 .0783 (1.83) 0.323 -.0990 (1.76) 0.416 -.1036 (2.25) 0.354 -0.413 (0.97) 0.354 0-435 0.334 0.213 0.266

exports

-.0000

(0.00)
.0232 (0.75) Real share (13a) (13b) -.0299 (1.12) -.0206 (0.72)

prices

NOTES: Entries are estimatedcoefficients from regressionsof the dependentvariablesagalnstdummies for adherenceto the gold standard (ONGOLD) and for the presence of a bankingpanic (PANIC). Absolute values of t-statisticsare in parentheses.Dependentvarlablesare measuredin log-changes, except for the nominal and ex post real interestrates, whlch are sn percentagepolnts (levels). Data are annual, 1931 to 1935 Incluslve, and for up to twenty-six countries, dependlngon data avallabllity(see the Appendlx) Each regresslonIncludesa complete set of year dummies. ONGOLDand PANIC are measuredas the numberof monthsduringthe year In which the countrywas on gold or experlencing a bankingpanlc (see text), dlvided by twelve

BEN S. BERNANKE : 15

tary contractionwas an important source of the Depressionin all countries;(2) subsequent to 1931 or 1932, there was a sharp divergence between countries which remainedon the gold standardand those that left it; and (3) this divergence arose because countries leaving the gold standardhad greaterfreedom to initiate expansionary monetarypolicies. Turningfirst to the behaviorof money supplies, we can see from Table2 (line 3) thatthe inside money stocks of all countriescontractedsharplyin 1931 and 1932. In an arithmeticsense, much of this contractioncan be attributedto declines in the ratioof M1 to currency(line 4), which in turnprimarilyreflectedthe effects of banking crises (note the concentration this effect in 1931).l Duringthe period 1933of 1935, however, Table 2 shows that the money supplies of gold-standard countries continuedto contract, while those of countriesnot on the gold standardexpanded. Table 3 (line 3a) indicates that, over the 1931-1935 period, the growthrate of M1 (line 3a) in countrieson gold averagedabout 5 percentagepoints per year less than in countriesoff gold, with an absolute t-value of 3.26. The behavior of price levels correspondedclosely to the behavior of money stocks. Table 2 (line 2) shows that, althougha sharpdeflationoccurredin all countries through 1931, in countries leaving gold wholesale prices stabilized in 19321933 and began, on average, to rise in 1934. l l Countriesremainingon gold experienced continuingdeflationthrough 1935, leading to a cumulativedifferencein log price levels over 1932-1935 of .329. Accordingto Table3 (line 2a), over the 19311935 period wholesale price inflationwas about9 percentagepoints per year lower (absolutet-value = 8.20) in countrieson gold. Declines in output and employment were strongly correlatedwith money and price declines: Manufacturing production(Table2, line 1) and employment (Table 2, line 7) fell in all countriesin 1930-1931 but afterward began to diverge between the two groups. Over the period 1932-1935, the cumulativedifferencein log output levels was .310, and the cumulativedifferencein log employmentlevels was .301, in favorof countriesnot on gold. The corresponding absolutet-values (Table3, lines la and 7a, for the 1931-1935 sample) were 4.04 and 4.38 for outputand employment, respectively.These are highly significantdifferences, both economically and statistically. The behaviorof other macro variables shown in Tables 2 and 3 are also generally consistent with the monetary-shocksstory. For example, a standardMundellFleming analysis of a small gold-standard economy (Eichengreenand Sachs 1986) would predict that monetarycontractionabroadwould depress domestic aggregate demandby raising the domestic real interestrate. It also would predictan increase
10. The preferredmeasure,M1IBASE,is not used owing to lack of dataon commercialbankreserves for many countries in the sample. Note from Table 3, line 4a, that the fall in the M1-currencyratio is greater on average in gold-standardcountries (and the difference is statisticallysignificant at approximately the S percent level), consistent with our earlier observationthat bankingproblems were more severe in gold-standard countries. 11. Thus price-level stabilizationprecededmonetarystabilizationin the typical countryleaving gold. A possible explanation is that devaluationraised expectationsof future inflation, lowering money demand and raising currentprices.

16 : MONEY, CREDIT,AND BANKING

in the domestic real exchange rate (price of exports), relative to countries not on gold, and an accompanyingdecline in real exports. Table 2 (line 9) shows that ex post real interest rates were universally high in 1930, coming down graduallyin both gold and nongold countries, but being consistently lower in countriesnot on gold. 12 Table 3 (line 9a) confirmsthat, on average, ex post real interestrates were 2.7 percentage points higher in gold-standardcountries (t = 2.07). The real excountries(line 10a of Table3, measuredrelativeto the change rate in gold-standard United States) grew on averageclose to S percentagepoints per year relativeto that real exof nongold countries(but with a t-value of only 1.70), and correspondingly countriesfell between 7 and 8 percentage ports (Table3, line 1la) of gold-standard points per year more quickly (absolutet-value = 2.08). There was no differencein the growthrates of importsbetween gold and nongold countries(Table3, line 12a), presumablyreflectingthe offsettingeffects in Gold Bloc countriesof lower domestic income and improvedterms of trade. index deflatedby the wholeInterestingly,real shareprices (a nominalshare-price countries, falling sale price index) did not fare that much worse in gold-standard about 3 percentagepoints a year faster (absolutet-value = 1.12). There are significant differencesbetween gold and nongold countriesin the behaviorof nominaland real wages, but as these variablesare most closely linked to issues of aggregatesupply, we defer discussion of them until the next section.
2. AGGREGATESUPPLY:THE FAILUREOF NOMINALADJUSTMENT

Although the consensus view of the causes of the GreatDepression has long included a role for monetaryshocks, we have seen in section 1 that recent research taking a comparativeperspective has greatly strengthenedthe empirical case for the money as a majordriving force. Further, effects of monetarycontractionon real economic variablesappearedto be persistentas well as large. Explainingthis persistent non-neutralityis particularlychallenging to contemporarymacroeconomists, since currenttheories of non-neutrality(such as those based on menu costs or the confusion of relative and absoluteprice levels) typically predictthat the real effects of monetaryshocks will be transitory. On the aggregatesupply side, then, we still have a puzzle: Why did the process of adjustmentto nominal shocks appearto take so long in interwareconomies? In this
countriesof course does not 12. A findingthatex post real interestrates were higherin gold-standard settle whetherex ante real interestrates were higher;that depends on whetherdeflationwas anticipated. For the U.S. case, Cecchetti (1992) finds evidence for, and Hamilton(1992) finds evidence against, the propositionthat people anticipatedthe declines in the price level. (I do not know of any studies of this issue for countries other than the United States.) This debate bears less on the question of whetherthe channel of transmission:If deflation was initiating shocks were monetarythan it does on the particular anticipated,so that the ex ante real interestrate was high, then the channelof monetarytransmissionwas throughconventional IS curve effects. If deflation was unanticipated,as both Cecchetti and Hamilton note, then one must rely more on a debt-deflationmechanism(see section 2). The behaviorof nominal lower in goldinterestrates, which remainedwell above zero in most countriesand were not substantially countries(Table2, line 8), suggests to me that much of the deflation standardthan in non-gold-standard horizon. Evans and Wachtel( 1993) drawa similarconcluwas not expected, at least at the medium-term sion based on U.S. nominal interestrate behavior.

BEN S. BERNANKE :

17

section I will discuss the evidence for two leading explanationsof how monetary shocks may have had long-lived effects: inducedfinancialcrisis and sticky nominal wages. 2.1. Deflation and the Financial System If one thinks aboutimportant of contractsin the economy thatare set in nomsets inal terms, and which are unlikely to be implicitly insuredor indexed againstunanticipated price-level changes, financial contracts(such as debt instruments)come immediately to mind. In my 1983 paper I argued that nonindexationof financial contractsmay have provided a mechanismthroughwhich declining money stocks and price levels could have had real effects on the U.S. economy of the 1930s. I discussed two relatedchannels, one operatingthrough"debt-deflation" the othand er throughbank capital and stability. The idea of debt-deflationgoes back to IrvingFisher (1933). Fisher envisioned a dynamic process in which falling asset and commodity prices createdpressureon nominal debtors, forcing them into distress sales of assets, which in turnled to further price declines and financialdifficulties.13 His diagnosis led him to urge President Roosevelt to subordinate exchange-rate considerations the need for reflation, to advice that (ultimately) FDR followed. Fisher's idea was less influential in academic circles, though, because of the counterargument that debt-deflationrepresented no more than a redistribution from one group(debtors)to another(creditors). Absent implausibly large differences in marginalspendingpropensitiesamong the groups, it was suggested, pure redistributionsshould have no significant macroeconomic effects. However, the debt-deflationidea has recently experienceda revival, which has drawn its inspirationfrom the burgeoningliteratureon imperfect informationand agency costs in capital markets.l4 According to the agency approach, which has come to dominate modern corporatefinance, the structureof balance sheets provides an importantmechanism for aligning the incentives of the borrower (the agent) and the lender (the principal).One centralfeatureof the balance sheet is the borrower's net worth, defined to be the borrower'sown ("internal")funds plus the collateralvalue of his illiquid assets. Many simple principal-agent models imply that a decline in the borrower'snet worth increasesthe deadweightagency costs of lending, and thus the net cost of financing the borrower'sproposed investments. Intuitively,if a borrowercan contributerelativelylittle to his or her own projectand hence must rely primarilyon externalfinance, then the borrower'sincentivesto take actions that are not in the lender's interestmay be relativelyhigh; the result is both deadweightlosses (for example, inefficientlyhigh risk-takingor low effort) and the necessity of costly informationprovision and monitoring. If the borrower's net worth falls below a thresholdlevel, he or she may not be able to obtainfunds at all.
13. Kiyotaki and Moore (1993) provide a formal analysis that capturessome of Fisher's intuition. 14. An importantearly paper that applied this approachto consumer spending in the Depression is Mishkin (1978). Bernankeand Gertler(1990) provide a theoreticalanalysis of debt-deflation.See Calomiris (1993) for a recent survey of the role of financialfactors in the Depression.

18 : MONEY, CREDIT,AND BANKING

From the agency perspective, a debt-deflationthat unexpectedly redistributes neutralevent:To the extent wealth away from borrowersis not a macroeconomically that potential borrowershave unique or lower-cost access to particularinvestment projects or spending opportunities,the loss of borrowernet worth effectively cuts off these opportunitiesfrom the economy. Thus, for example, a financially distressed firm may not be able to obtain workingcapitalnecessaryto expandproduction, or to fund a project that would be viable under better financial conditions. Similarly,a household whose currentnominalincome has fallen relativeto its debts may be barredfrom purchasinga new home, even thoughpurchaseis justified in a permanent-incomesense. By inducing financial distress in borrower firms and households, debt-deflationcan have real effects on the economy. If the extent of debt-deflationis sufficientlysevere, it can also threatenthe health of banks and other financial intermediaries(the second channel). Banks typically have both nominal assets and nominal liabilities and so over a certain range are hedged against deflation. However, as the distress of banks' borrowersincreases, the banks' nominalclaims are replacedby claims on real assets (for example, collateral); from that point, deflation squeezes the banks as well. 15Actual and potential impairbankcapitalandhurtbanks'economic loan losses arisingfrom debt-deflation in efficiency in several ways: First, particularly a system withoutdeposit insurance, depositorruns and withdrawalsdeprivebanksof funds for lending;to the extent that these loans are not easily rebank lending is specialized or information-intensive, placed by nonbankforms of credit. Second, the threatof runsalso induces banksto increase the liquidity and safety of their assets, furtherreducingnormallending activity. (The most severely decapitalizedbanks, however, may have incentives to make very risky loans, in a gambling strategy.)Finally, bank and branchclosures may destroylocal informationcapitaland reducethe provisionof financialservices. How macroeconomicallysignificantwere financialeffects in the interwarperiod? My 1983 paper, which consideredonly the U.S. case, showed that measuresof the liabilities of failing commercialfirms and the deposits of failing banks helped predict monthly changes in industrialproduction, in an equation that also included lagged values of money and prices. However, this evidence is not really conclusive: For example, as Green and Whiteman(1992) pointedout, the spikes in commercial and bankingfailures in 1931 and 1932 could well be functioningas a dummy variable, picking up whateverforces- financialor otherwise caused the U. S . Depression to take a sharpsecond dip duringthatperiod. As with the debateon the role of money, the problemis the reliance on what amountsto one datapoint. research, Bernanke However, in the comparativespirit of the new gold standard and James(1991) studiedthe macroeconomiceffects of financialcrises in a panel of twenty-fourcountries. The expansionof the sample broughtwith it datalimitations: Bernankeand James used annualratherthan monthly data, and lack of data on indebtednessand financialdistress forced them to confine their analysis to the effects
15. Banks in universalbanking systems, such as those of centralEurope, held a mixtureof real and nominal assets (for example, they held equity as well as debt). Universal banks were thus subject to pressureeven earlierin the deflationaryprocess.

BEN S. BERNANKE : 19

of bankingpanics. Further,not having a consistentquantitative measureof banking instability,they chose to use dummy variablesto indicateperiods of bankingcrisis (as suggested by theirreadingof historicalsources). Offsettingthese disadvantages, expandingthe sample made it possible to comparethe U.S. case with both countries thatalso sufferedsevere bankingproblemsand countriesin which bankingremained stable despite the Depression. In particular,Bernankeand Jamesarguedthat crossnational differences in vulnerabilityto bankingcrises had more to do with institutional and policy differences than macroeconomic conditions, strengtheningthe case that bankingpanics had an independentmacroeconomiceffect (as opposed to being a purely passive response to the generaleconomic downturn). 16 As a measure of banking instability,Bernankeand James constructeda dummy variable called PANIC, which they defined as the numberof months during each year that countriesin their sample sufferedbankingcrises. 17 In regressionscontrolling for a varietyof factors, includingthe rateof change of prices, wages, and money stocks, the growth rate of exports, and discountratepolicy, Bernankeand James found an economically large and highly statistically significant effect of banking panics on industrialproduction. A reduced-formsummaryof the effects of PANICon our list of macrovariablesis given in the rows of Table3 marked"b," which reportsestimatedcoefficients from regressions of each macro variable against PANIC, the dummy for gold standard membership(ONGOLD),and time dummies for each year. For these estimates we have divided the Bernanke-James PANICvariableby twelve, so that its estimated coefficients may be interpreted annualizedeffects. as The results suggest important macroeconomiceffects of bankpanics that are both independentof gold-standard effects and consistentwith theoreticalpredictions:On the real side of the economy, PANICis found to have economically large and statistically significant effects on manufacturing production(line lb) and employment (line 7b). In particular, with gold-standard membership controlledfor, the effect of a year of bankingpanic on the log-changeof manufacturing productionis estimatedto be-.0926 with an absolute t-value of 3.50; and the effect on the log-change of employment is-.0456, with a t-value of 2.10. Banking panics are also found to reduce both real and nominal wages (lines 6b and Sb), hurt competitiveness and exports (lines 10b and 1lb), raise the ex post real interestrate (line 9b), and reduce
16. Factorscited by Bernankeand Jamesas contributing bankingpanics includedbankingstructure to ("universal"banking systems and systems with many small banks were more vulnerable);reliance on short-termforeign liabilities; and the country'sfinancialand economic experiencesand bankingpolicies duringthe 1920s. See Grossman(1993) for a more detailedand generallycomplementary analysis of the causes of interwarbankingpanics. 17. Bernankeand Jamesdated periods of crisis as startingfrom the first severe bankingproblems, as determinedfrom a readingof primaryand secondarysources. If therewas some cleardemarcation point, such as the U.S. banking holiday of March 1933, that point was used as the ending data of the crisis; otherwise, they arbitrarily assumed that the effects of the crisis would last for one year after its most intense point. Countrieswith nonzerovalues of PANIC includedAustria,Belgium, Estonia, France,Germany, Hungary,Italy, Latvia, Poland, Rumania, and the United States. Results presentedhere add data for Argentinaand Switzerlandto the Bernanke-James sample;consistentwith the Bernanke-James banking crisis chronology, we treat Switzerland(July 1931-November 1933) as a crisis country.Grossman (1993) includes all of these countriesas "crisis"countriesin his studybut differsin countingNorway as a crisis countryas well.

20

: MONEY, CREDIT,AND BANKING

real share prices (line 13b), although estimated coefficients are not always statistically significant. On the nominalside of the economy, bankingpanics significantlylower the money multiplier(proxied in line 4b of Table 3 by the ratio of M1 to currency),as expected. We also find (line 3b) that bankingpanics in a countrysignificantlyreduce the M1 money stock. This effect on the money supply is actuallyinconsistentwith a simple Mundell-Fleming model of a small open economy on the gold standard: With worldwide conditions held constant(by the time dummies), a small country's money stock is determinedby domestic money demand, so that any declines in the money multipliershould be offset by endogenousinflows of gold reserves. Possible reconciliationsof the empiricalresult with the model are that bankingpanics lowered domestic M1 money demandor raisedthe probabilityof exchange-ratedevaluation (either would induce an outflow of reserves); our finding above that panics raised the real interestrate fit with the latterpossibility. A findingthat is consistent with the Mundell-Flemingmodel is that, once gold-standardmembershipis controlled for, bankingpanics had no effect on wholesale prices (line 2b). This last result is important,because it suggests that the observed effects of panics on output and other real variables are operatinglargely throughnonmonetarychannels, for example, the disruptionof credit flows. As with the earlierdebate about the role of monetaryshocks, moving from a focus on the U.S. case to a comparativeinternationalperspective provides much strongerevidence on the potentialrole of bankingcrises in the Depression. Ideally, we should like to extend this evidence to the broaderdebt-deflationstory as well. Indeed, the strongpresumptionis thatdebt-deflation effects were much more pervasive than banking crises, which were relatively more localized in space and time. Unfortunately,consistent internationaldata on types and amounts of inside debt, and on various indicatorsof financialdistress, are not generally available.18 2.2. DeJqation Nominal Wages and Induced financial crisis is a relatively novel proposal for solving the aggregate supply puzzle of the Depression. The more traditional explanationof monetarynonneutralityin the 1930s, as in macroeconomicsmore generally,is thatnominalwages and/or prices were slow to adjust in the face of monetaryshocks. In fact, widely availableprice indexes, such as wholesale and consumerprice indexes, show relatively little nominal inertia during this period (admittedly,the same is not true for many individual prices, such as industrialprices). Hence in contradistinction to contemporarymacroeconomics, which has come to emphasize price over wage rigidity research on the interwarperiod has focused on the slow adjustmentof nominalwages as a source of nonneutrality. Following thatlead, in this subsectionI
18. Eichengreenand Grossman(1994) attemptto measuredebt-deflation an indirectindicator,the by spread between the central bank discount rate and the interestrate on commercialpaper. As they note, this indicatoris not wholly satisfactoryand they obtain mixed results.

BEN S. BERNANKE : 21

discuss the comparativeempirical evidence for sticky wages in the Depression. I defer for the momentthe deeperquestionof how wages could have failed to adjust, conditionsof the Depressionera. given the extreme labor-market The link between nominal wage adjustmentand aggregatesupply is straightforward:If nominalwages adjustimperfectly,then falling price levels raise real wages; employers respondby cutting their workforces.l9 Similarly,in a countryexperiencing monetaryreflation,real wages should fall, permittingreemployment.Although the cyclicality of real wages has been much debated in the postwar context, these two implicationsof the sticky-wagehypothesisare clearly borneout by the comparative interwardata, as can be seen in Tables2 and 3: First, during the worldwide deflationof 1930 and 1931, nominal wages worldwide fell much less slowly than (wholesale) prices, leading to significantincreases in the ratio of nominal wages to prices (Table2, lines 2, 5, and 6). Associated with this sharpincrease in real wages were declines in employmentand output(Table2, lines 7 and 1).2 Second, from about 1932 on, therewas a markeddivergencein real-wagebehavior between countries on and off the gold standard(Table 2, line 6): In countries leaving gold, prices rose more quickly than nominalwages (indeed, the lattercontinued to fall for a while), so thatreal wages fell; simultaneously,employmentrose sharply.In countries remainingon gold, real wages rose or stabilizedand employment remainedstagnant.Table3 (line 6a) indicatesa differencein real wage growth between countries on and off the gold standardequivalent to about 6 percentage points per year, with a t-value of 5.84. This latterresult, thatreal-wagebehaviorvariedwidely between countriesin and out of the Gold Bloc, was first pointed out in the previously cited article by Eichengreenand Sachs (1985). Using data from ten Europeancountriesfor 1935, Eichengreen and Sachs showed that Gold Bloc countries systematicallyhad high output,while countriesnot on gold had much real wages and low levels of industrial lower real wages and higher levels of production(all variableswere measuredrelative to 1929). In a recent paper, Bernankeand Carey (1994) extended the Eichengreen-Sachs analysis in a numberof ways: First, they expandedthe sample from ten to twentytwo countries, and they employed annualdata for 1931-1936 ratherthan for 1935 to only. Second, to avoid the spuriousattribution real wages of price effects operat19. In the standardanalysis, increases in the real wage lead to declines in employmentbecause empossible channelis ployers move northwestalong their neoclassical labordemandcurves. An alternative thathigher wage paymentsdeplete firms'liquidity,leadingto reducedoutputand investmentfor the types of financialreasons discussed above (my thanksto MarkGertlerand BruceGreenwaldfor independently making this suggestion). This latterchannel might be tested by observing whethersmalleror less liquid firms respondedto real-wage increasesby cutting employmentmore severely than did large, financially more robust firms. 20. The wholesale price index is not the ideal deflatorfor nominal wages; to find the productwage, which is relevantto labor demanddecisions, one should deflate by an index of outputprices. The very limited internationaldata on product wages are less supportiveof the sticky-wage hypothesis than the evidence given here; see Eichengreen and Hatton (1988) or Bernanke and James (1991) for further
. .

lscusslon.

22

: MONEY, CREDIT,AND BANKING

the ing throughnonwage channels,2l in regressionsthey separated real wage into its nominal-wageand price-level components. Third, they controlledfor factors other variablestechniquesto thanwages affectingaggregatesupply and used instrumental With these modcorrect for simultaneitybias in outputand wage determination.22 equation describing output supply in ifications, Bernankeand Carey's "preferred" their sample was (theirTable4, line 9):
q = -.600

(3.84) where

w + .673 p + .540 q_, (7.66) (5.10)

.144 PANIC- .69-05 STRIKE

(2)

(5.79)

(3.60)

production(in logs), = currentand lagged manufacturing w = nominal wage index (in logs), p = wholesale price index (in logs), PANIC = number of months in each year of banking panic [see the text or (1991)], divided by 12, and Bernanke-James STRIKE= working days lost to labor disputes (per thousandemployees).
q, q_,

Absolute values of t-statistics are shown in parentheses.The regression pooled cross-sectional data for 1931-1936 and included time dummies and fixed country effects. A consistent estimate of within-countryfirst-orderserial correlation of -.066 was obtainedby applicationof nonlinearleast squares. and The equationindicatesthatbankingpanics (PANIC) work stoppages(STRIKIC) had large and statisticallysignificanteffects on the supply of output,23and the coefficient on lagged output indicates that output adjustedabout half-way to its "tarthe get" level in any given year. Most importantly, coefficient on nominal wages is equal and opposite in magnitudeto the coeffihighly significantand approximately cient on the price level, as suggestedby the sticky-wagehypothesis.24In particular, equation (2) indicates that countries in which nominal wages adjusted relatively slowly towardchanging price levels experiencedthe sharpestdeclines in manufacturingoutput. To illustratethis last point in a very simple way, Figure 1 shows 1935 outputsand nominalwages for five Gold Bloc countries(Belgium, France,the Netherlands,Po21. Suppose that deflationaffects outputthrougha nonwage channel, such as inducedfinancialcrisis, and that nominal-wagedata are relatively noisy (for example, they reflect official wage rates ratherthan rates actually paid). Then we might well observe an inverse relationshipbetween measuredreal wages and output, even though wages are not partof the transmissionchannel. 22. Instrumentsused in the equation to follow included, as aggregate demand shifters, a tradeweighted importprice index and the discountratefor Gold Bloc countries,and M1 for countriesoff gold. Additionally,the bankingpanic and strikevariables, and lagged values of the nominalwage and output, were treatedas predetermined. 23. The coefficienton PANICimplies thatone year of bankingcrisis reducedoutputby approximately is 14 percent. The coefficient on STRIKE aboutwhat one would expect if outputlosses due to strikesare proportional hours of work lost. See Bernankeand Carey (1994) for furtherdiscussion. to 24. That the coefficients on wages and prices are equal and opposite is easily accepted at standard significance levels (p = .573).

BEN S. BERNANKE : 23

0.95 Nethertands
0.9 Z

_0.85 -*Pofand
o " Bel,alum

a) 0.8
-

ll

!0.75 X
o

France

0.7

0.65 SVYitZerland

0.6 0.7

0.75

r 0.8

0.85

0.9

0.95

1Z05

Nominalwage (1929=1.0)
FIG. 1. Outputand Wages in the Gold BlocS 1935

the throughout land, and Switzerland).As they shareda common monetarystandard period these countries had similar wholesale price levels in 1935 but nominal wages differedamong the countries. As Figure 1 indicates, Franceand Switzerland had significantlyhigher nominalwages thanthe otherthreecountries(indeed, those countries had shown almost no nominal wage adjustmentsince 1929); these two countriesalso had sigrlificantlylower outputlevels. A regressionfor just these five datapoints of the log of outputon a constantand the log of the nominalwage yields a coefficient on the nominalwage of-.628 with a t-statisticof-1.49. Although Bernanke and Carey (1994) found cross-sectional evidence for the sticky-wage hypothesis, they emphasized that the time-series evidence is much weaker(recall thattheirregressionincludedyearlytime dummies so thatthe results are based entirely on cross-countrycomparisons).Broadly,the problemwith stscky wages as an explanationof the time-seriesbehaviorof outputin the Depressionis as follows: Although real wages rose sharplyaroundthe world duringthe 1929-1931 downturn in most countries real wages didn't decline much during the recovery phase of the Depression;indeed, some countries(such as the United States)enjoyed strongrecoveries despite rising real wages. Bernankeand Careyreportthat, for the twenty-twocountriesin their sample averageoutputin 1936 was nearly 10 percent above 1929 levels, even though real wages in 1936 remained nearly 20 percent higher than in 1929.25 One possible reconciliationof the cross-section and timeseries results is that actual wages paid kll relativeto reportedor official wage rates
25. In principlethis result could be explained by secular increases in capacity at a given real wage. However, Bernankeand Carey estimate that trend capacity growth of 5.6 percent per year on average would be needed to reconcile the behaviorof outputand real wages.

24

: MONEY, CREDIT,AND BANKING

as the Depressionwore on; and thatthe ratioof actualto reportedwages was similar among the countriesin the sample. in 2.3. Can Failures of Nominal Adjustment the Depression be Explained? I have discussed two generalreasons for the failureof interwareconomies to adjust to the large nominalshocks thathit them in the early 1930s: (1) nonindexeddebt and contracts, throughwhich deflation inducedredistributions financialcrisis; and (2) slow adjustmentof nominal wages (and presumablyother elements of the cost structureas well). From an economic theorist'spoint of view, there is an important which is that-following an distinctionbetween these two sources of nonneutrality, unanticipateddeflation-there are incentives for the partiesto renegotiatenominal wage (or price) agreements, but not nominal debt contracts. In particular,if the nominal wage is "too high"relative to labormarketequilibrium,both the employer and the worker(who otherwisewould be unemployed)shouldbe willing to accept a lower wage, or to take other measuresto achieve an efficient level of employment effects of (Barro 1977). In contrast, there is no presumptionthat the redistributive unanticipateddeflation operating through debt contracts will be undone by some sort of implicit indexing or renegotiationex post, since large net creditorsdo gain from deflation and have no incentive to give up those gains.26Hence the failure of nominal wages (and, similarly,prices) to adjustseems inconsistentwith the postufinancialcrisis does not (given late of economic rationality,while deflation-induced that nonindexedfinancialcontractsexist in the firstplace27). One interestingpossibility for reconciling wage-price stickiness with economic rationalityis that the nonindexationof financialcontracts, and the associated debtdeflation, might in some way have been a source of the slow adjustmentof wages and otherprices. Such a link would most likely arise for political reasons:As deflation proceeded, both the growing threat of financial crisis and the complaints of debtorsincreasedpressureon governmentsto intervenein the economy in ways that inhibit adjustment.In the case of France, for example (which, note from Figure 1, seemed a particularlyslow adjuster),a historianreported: merchants, indusand shopkeepers, declined, farmers, as and as pricesbroke incomes basis,to buildupa complex the trialists facedbankruptcy, statebegan,on anempirical with whichinterfered thefreeoperation measures array interventionist of andinchoate (Kemp 1972,p. 101) situations acquises. certain forcesin order preserve to of market Examples of interventionistmeasuresby the Frenchgovernmentincluded tough agriculturalimport restrictionsand minimumgrain prices, intendedto supportthe nominal incomes of farmers(a politically powerfulgroup of debtors);government(1990), typically predict that debt26. Formal models in the literature, such as Bernanke-Gertler deflationlowers aggregateoutputand investmentbut does not lead to a situationthatis Pareto-inefficient (given the informationconstraints).Thus there is no incentive for renegotiationbetween creditorsand model were enhanced by assuming productionor aggregate demand debtors. If the Bernanke-Gertler but externalities, then debt-deflationcould imply Pareto-inefficiency, not of the sort that can easily be remediedby bilateralrenegotiation. 27. Nonindexationof financialcontractsmight be rationalizedas an attemptto minimize transactions costs ex ante. This strategyis reasonableif the monetaryauthorityis expected to keep inflationstableof assumptiongiven the restoration the gold standard. an understandable

BEN S. BERNANKE : 25

supportedcartelizationof industry,as well as importprotection, with the goal of increasing prices and profits; and measures to reduce labor supply, including repatriationof foreign workers and the shorteningof workweeks.28These measures (comparable to New Deal-era actions in the United States) tended to block the downwardadjustmentof wages and prices. Other links from debt-deflationto wage-price behavior operatedthrough more strictly economic channels. For example, in France, heavy industriessuch as iron and steel expandedextensively duringthe 1920s, which left them with heavy debt burdens.In response to the financialdistresscaused by deflation, firmsacted singly and in combinationto try to restrictoutput,raise prices, and maintainprofitmargins (Kemp 1972, pp. 89ff.) Such behavioris predictedby modernindustrialorganization theory and evidence (see, for example, Chevalierand Scharfstein1994). A variety of other factors no doubt contributedto incomplete nominal adjustment. In some countries, many wages and prices were either directlycontrolledby the government(so that change involved administrative legislative action, with or the usual lags), or were highly politicized. Legislatively set taxes, fees, and tariffs were an additionalsource of nominal rigidity [see Crucini(1994) on tariffs]. Complex, decentralizedeconomies also no doubt faced serious problems of coordination, both internallyand with other economies, an issue thathas been the subjectof recent theoreticalwork (see, for example, Cooper 1990). I believe that, as with other issues relating to the Depression, the comparative internationalapproachholds the most promise for improvingour understanding of the sources of incompletenominaladjustment.In this case, though, the comparative analysis will need to include political and institutionalvariables,such as the proportion of workerscovered by unions;the extent of representation workers,farmers, of industrialists,etc., in the legislature;the share of the workforceemployed by the government, and so on. More qualitatively,historical and case-study comparisons of the political response to deflationin differentcountriesmay help explain the differing degrees of economic damage inflictedby falling prices.
3. CONCLUSION

Methodologically,the main contribution recentresearchon the Depressionhas of been to expand the sample to include many countriesother than the United States. Comparativestudies of a large set of countrieshave greatly improvedour ability to identify the forces that drove the world into depressionin the 1930s. In particular, the evidence for monetarycontractionas an important cause of the Depression, and for monetary reflation as a leading component of recovery, has been greatly strengthened. On the aggregatesupply side of the economy, we have learnedand will continue
28. Of course, the most obvious interventions would have been to stop the deflationby devaluingor to mandatea writedownof all nominalclaims. As we have seen, however, in Francedevaluationwas widely considered as heraldinga plunge into chaos; while the writedownof debts and other claims, besides being administratively complex, would have been considereda politically unacceptableviolation of the sanctity of contracts.

26

: MONEY, CREDIT,AND BANKING

to learn a great deal from the interwarperiod. One key result is that wealth redistributionsmay have aggregateeffects, if they are of the form to induce systematic financial distress. Empiricalevidence has also been found for incomplete adjustUnderstandment of nominal wages as a factor leading to monetarynonneutrality. ing this latterphenomenonwill probablyrequirea broadperspectivethat takes into account political as well as economic factors.
APPENDIX:DATA SOURCES

productiondata are from League of Nations (1945). Wages and Manufacturing LabourOrganization,YearBook of Labor employment data are from International Statistics, various issues. Data on commercialbank reserves, used in constructing monetary base measures, are taken from League of Nations, II.A Economic and FinancialSeries:Money and Banking, variousissues. Monetarydata for the United States are from Friedmanand Schwartz(1963) and Boardof Governors(1943). Other data are from League of Nations, Statistical YearBook and MonthlyBulletin of Statistics, various issues. All data are annual and were collected for as many of the following twenty-six countries as possible: Australia,Argentina,Austria, Belgium, Canada,Czechoslovakia, Denmark, Estonia, Finland, France, Germany,Greece, Hungary,Italy, Japan, Latvia, the Netherlands,Norway, New Zealand, Poland, Rumania, Sweden, Spain, Switzerland,the United Kingdom, and the United States. Data availabilityby variableis describedbelow. Inclusionof countriesin the data set was based on the availabilityof data for key variables, particularlyoutput and
prices.

Data Availability production:All countries, except Spain for 1936. IndustrialproManufacturing duction used for Argentina,from Thorp(1984). Wholesale prices: All countries. Money and notes in circulation:All countries. Commercialbank deposits:All countries, except Greece and Spain for 1936. Nominal wages: All countries, except Finland, Greece, Rumania,Spain. Employment:All countries, except Austria, Belgium, Czechoslovakia, Greece, Spain, and Denmarkfor 1930. Discount rate:All countries, except Argentinaand Switzerland. Exchange rates (relativeto Frenchfranc):All countries. Exports:All countries, except Argentinaand Spain for 1936. Imports:All countries, except Estonia, Finland, Greece, and Spain for 1936. Shareprice index:Availablefor Austria,Belgium, Canada,Czechoslovakia,Denmark, France, Germany,Hungary,Italy, the Netherlands, Norway, Sweden, Spain, Switzerland,the United Kingdom, and the United States.

BEN S. BERNANKE : 27

CITED LITERATURE

Barro, Robert J. "Long-termContracting,Sticky Prices, and MonetaryPolicy." Journal of MonetaryEconomics 3 (1977), 305- 16. of Effects of the FinancialCrisis in the Propagation the Great Bernanke,Ben. "Nonmonetary Depression." AmericanEconomicReview 73 (June 1983), 257-76. Bernanke, Ben, and Kevin Carey. "Nominal Wage Stickiness and Aggregate Supply in the GreatDepression." Unpublished,PrincetonUniversity,January1994. QuarBernanke, Ben, and Mark Gertler."FinancialFragilityand Economic Performartee." terlyJournal of Economics 105 (February1990), 87- 114. Bernanke, Ben, and Harold James. "The Gold Standard,Deflation, and FinancialCrisis in Comparison."In R. G. Hubbard,ed., Financial the Great Depression: An International Marketsand Financial Crises, Chicago:Universityof Chicago Press, 1991. Board of Governorsof the FederalReserve System. Bankingand MonetaryStatistics. Washington: NationalCapita Press, 1943. Calomiris, CharlesW. "FinancialFactorsin the Depression."Journalof Economic Perspectives 7 (Spring 1993), 6186. Campa, Jose Manuel. "ExchangeRates and Economic Recovery in the 1930s: An Extension to Latin America."Journal of EconomicHistory 50 (September1990), 677-82. Cecchetti, Stephen G. "Pricesduringthe GreatDepression:Wasthe Deflationof 1930-1932 AmericanEconomicReview 82 (March 1992), 141-56. Really Unanticipated?" Chevalier,JudithA., and David S. Scharfstein."CapitalMarketImperfectionsand CounterTheoryand Evidence."NationalBureauof EconomicResearchworking cyclical Markups: paperno. 4614, January1994. TransChoudhri,Eh$anU., and Levis A. Kochin. "TheExchangeRate and the International mission of Business Cycle Disturbances:Some Evidence from the Great Depression." Journal of Money, Credit, and Banking 12 ( I980), 565-74. GovernWagesand Prices and the Impactof Expansionary Cooper, Russell. "Predetermined ment Policy." Review of EconomicStudies 57 (1990), 205-14. in Crucini, Mario. "Sourcesof Variation Real TariffRates:The United States, 1900-1940." AmericanEconomic Review 84 (June 1994), 732-43. Diamond, Douglas W., and Philip H. Dybvig. "BankRuns, Deposit Insurance,and Liquidity." Journal of Political Economy91 ( June 1983), 401- 19. ExploraGold Standard." BankCooperationunderthe Interwar Eichengreen,Barry."Central tions in Economic History 21 (1984), 64-87. . Golden Fetters: The Gold Standardand the Great Depression, 1919-1939. New York:Oxford University Press, 1992. . "CentralBank Cooperationand Exchange Rate Comnlitments:The Classical and Financial HistoryReview (forthcoming). Compared." InterwarGold Standards Eichengreen, Barry,and RichardGrossman."Debt Deflation and FinancialInstability:Two Historical Explorations."Unpublished, University of Californiaat Berkeley, June 1994. PerUnemploymentin International Eichengreen, Barry,and Timothy J. Hatton. "Interwar spective: An Overview."In Eichengreen,B. and Hatton,T., eds., InterwarUnemployment in InternationalPerspective, London:KluwerAcademic Publishers, 1988. Eichengreen, Barry, and Jeffrey Sachs. "Exchange Rates and Economic Recovery in the 1930s." Journal of Economic History 45 (1985), 925-46. . "CompetitiveDevaluationand the GreatDepression:A TheoreticalReassessment.t' Economics Letters 22 (1986), 67-71.

28

: MONEY, CREDIT,AND BANKING

Eichengreen, Barry,and Marc Uzan. "The 1933 WorldEconomic Conferenceas an Instance of Failed International Cooperation."In Peter B. Evans et al., eds., Double-EdgedDiplomacy, Berkeley, Calif.: University of CaliforniaPress, 1993. Evans, Martin, and Paul Wachtel."WerePrice Changesduringthe GreatDepressionAnticipated? Evidence from Nominal InterestRates." Journal of MonetaryEconomics 32 (August 1993), 3-34. Fisher, Irving. "TheDebt-DeflationTheoryof GreatDepressions."Econometrica1 (October 1933), 337-57. Friedman,Milton, and Anna J. Schwartz.A MonetaryHistory of the United States, 18671960. Princeton:PrincetonUniversityPress, 1963. Gordon, RobertJ., and JamesWilcox. "Monetarist Interpretations the GreatDepression: of Evaluationand Critique."In Karl Brunner,ed., The GreatDepression Revisited. Boston: MartinusNijhoff, 1981, 49-107. Green, Susan J., and CharlesH. Whiteman."A New Look at Old Evidence:On 'NonmonetaryEffects of the FinancialCrisis in the Propagation the GreatDepression'."University of of Iowa, April 1992. Grossman,RichardS. "The Shoe That Didn't Drop:ExplainingBankingStabilityduringthe GreatDepression."WesleyanUniversity,March 1993. Hamilton, JamesD. "MonetaryFactorsin the GreatDepression."Journal of MonetaryEconomics 19 (1987), 145-69. . "The Role of the International Gold Standard Propagating GreatDepression." in the Contemporary Policy lssues 6 (1988), 67-89. . "Was the Deflation during the Great Depression Anticipated?Evidence from the CommodityFuturesMarket. AmericanEconomicReview 82 (March 1992), 157-78. " Kemp, Tom. The French Economy, 1913-39: The History of a Decline. London:Longman Group, 1972. Kiyotaki, Nobu, andJohnH. Moore. "CreditCycles." Unpublished,Universityof Minnesota and LSE, March 1993. League of Nations. lndustrializationand Foreign Trade. Geneva, 1945. Mishkin, FredericS. "The Household Balance Sheet and the GreatDepression."Journal of EconomicHistory 38 (December 1978), 918-37. Nurkse, Ragnar(with W. A. Brown, Jr.). lnternationalCurrencyExperience:Lessons of the lnter-WarPeriod. Princeton, N.J.: PrincetonUniversity Press for the League of Nations, 1944. Romer, Christina."The Nation in Depression."Journal of Economic Perspectives7 (Spring 1993), 19-40. Sumner,Scott. "TheEquilibriumApproachto DiscretionaryMonetaryPolicy underan International Gold Standard:1926-1932." The ManchesterSchool of Economic Studies (December 1991), 378-94. Temin, Peter. Did MonetaryForces Cause the GreatDepression? New York:W. W. Norton, 1976. . Lessonsfrom the GreatDepression. Cambridge,Mass.: MIT Press, 1989. . "Transmissionof the Great Depression." Journal of Economic Perspectives 7 (Spring 1993), 87-102. Thorp, Rosemary, ed. Latin America in the 1930s: The Role of the Peripheryin the World Crisis. New York:St. Martin'sPress, 1984.