c11482

Volume Title: Financial Markets and Financial CrisesVolume Author/Editor: R. Glenn Hubbard, editorVolume Publisher: University of Chicago PressVolume ISBN: 0-226-35588-8

Volume URL: http://www.nber.org/books/glen91-1Conference Date: March 22-24,1990Publication Date: January 1991

Chapter Title: The Gold Standard, Deflation, and Financial Crisisin the Great Depression: An International ComparisonChapter Author: Ben Bemanke, Harold JamesChapter URL: http://www.nber.org/chapters/c11482Chapter pages in book: (p. 33 - 68)

The Gold Standard, Deflation,and Financial Crisis inthe Great Depression:An International ComparisonBen Bernanke and Harold James2.1 IntroductionRecent research on the causes of the Great Depression has laid much of theblame for that catastrophe on the doorstep of the international gold standard.In his new book, Temin (1989) argues that structural flaws of the interwar goldstandard, in conjunction with policy responses dictated by the gold standard's\"rules of the game,\" made an international monetary contraction and deflationalmost inevitable. Eichengreen and Sachs (1985) have presented evidence thatcountries which abandoned the gold standard and the associated contraction-ary monetary policies recovered from the Depression more quickly than coun-tries that remained on gold. Research by Hamilton (1987, 1988) supports thepropositions that contractionary monetary policies in France and the UnitedStates initiated the Great Slide, and that the defense of gold standard paritiesadded to the deflationary pressure.1The gold standard-based explanation of the Depression (which we willelaborate in section 2.2) is in most respects compelling. The length and depthof the deflation during the late 1920s and early 1930s strongly suggest a mon-etary origin, and the close correspondence (across both space and time) be-tween deflation and nations' adherence to the gold standard shows the powerof that system to transmit contractionary monetary shocks. There is also ahigh correlation in the data between deflation (falling prices) and depression(falling output), as the previous authors have noted and as we will demonstrateagain below.Ben Bemanke is professor of economics and public affairs at Princeton University and a re-search associate of the National Bureau of Economic Research. Harold James is assistant profes-sor of history at Princeton University.The authors thank David Fernandez, Mark Griffiths, and Holger Wolf for invaluable researchassistance. Support was provided by the National Bureau of Economic Research and the NationalScience Foundation.3334 Ben Bernanke and Harold JamesIf the argument as it has been made so far has a weak link, however, it isprobably the explanation of how the deflation induced by the malfunctioninggold standard caused depression; that is, what was the source of this massivemonetary non-neutrality?2 The goal of our paper is to try to understand betterthe mechanisms by which deflation may have induced depression in the1930s. We consider several channels suggested by earlier work, in particulareffects operating through real wages and through interest rates. Our focus,however, is on a channel of transmission that has been largely ignored by therecent gold standard literature; namely, the disruptive effect of deflation on thefinancial system.Deflation (and the constraints on central bank policy imposed by the goldstandard) was an important cause of banking panics, which occurred in anumber of countries in the early 1930s. As discussed for the case of the UnitedStates by Bernanke (1983), to the extent that bank panics interfere with nor-mal flows of credit, they may affect the performance of the real economy;indeed, it is possible that economic performance may be affected even withoutmajor panics, if the banking system is sufficiently weakened. Because severebanking panics are the form of financial crisis most easily identified empiri-cally, we will focus on their effects in this paper. However, we do not want tolose sight of a second potential effect of falling prices on the financial sector,which is \"debt deflation\" (Fisher 1933; Bernanke 1983; Bernanke and Gertler1990). By increasing the real value of nominal debts and promoting insol-vency of borrowers, deflation creates an environment of financial distress inwhich the incentives of borrowers are distorted and in which it is difficult toextend new credit. Again, this provides a means by which falling prices canhave real effects.To examine these links between deflation and depression, we take a com-parative approach (as did Eichengreen and Sachs). Using an annual data setcovering twenty-four countries, we try to measure (for example) the differ-ences between countries on and off the gold standard, or between countriesexperiencing banking panics and those that did not. A weakness of our ap-proach is that, lacking objective indicators of the seriousness of financialproblems, we are forced to rely on dummy variables to indicate periods ofcrisis. Despite this problem, we generally do find an important role for finan-cial crises—particularly banking panics—in explaining the link between fall-ing prices and falling output. Countries in which, for institutional or historicalreasons, deflation led to panics or other severe banking problems had signifi-cantly worse depressions than countries in which banking was more stable. Inaddition, there may have been a feedback loop through which banking panics,particularly those in the United States, intensified the severity of the world-wide deflation. Because of data problems, we do not provide direct evidenceof the debt-deflation mechanism; however, we do find that much of the appar-ent impact of deflation on output is unaccounted for by the mechanisms we35 Financial Crisis in the Great Depressionexplicitly consider, leaving open the possibility that debt deflation was impor-tant.The rest of the paper is organized as follows. Section 2.2 briefly recapitu-lates the basic case against the interwar gold standard, showing it to have beena source of deflation and depression, and provides some new evidence con-sistent with this view. Section 2.3 takes a preliminary look at some mecha-nisms by which deflation may have been transmitted to depression. In section2.4, we provide an overview of the financial crises that occurred during theinterwar period. Section 2.5 presents and discusses our main empirical resultson the effects of financial crisis in the 1930s, and section 2.6 concludes.2.2 The Gold Standard and DeflationIn this section we discuss, and provide some new evidence for, the claimthat a mismanaged interwar gold standard was responsible for the worldwidedeflation of the late 1920s and early 1930s.The gold standard—generally viewed at the time as an essential source ofthe relative prosperity of the late nineteenth and early twentieth centuries—was suspended at the outbreak of World War I. Wartime suspension of the goldstandard was not in itself unusual; indeed, Bordo and Kydland (1990) haveargued that wartime suspension, followed by a return to gold at prewar pari-ties as soon as possible, should be considered part of the gold standard's nor-mal operation. Bordo and Kydland pointed out that a reputation for returningto gold at the prewar parity, and thus at something close to the prewar pricelevel, would have made it easier for a government to sell nominal bonds andwould have increased attainable seignorage. A credible commitment to thegold standard thus would have had the effect of allowing war spending to befinanced at a lower total cost.Possibly for these reputational reasons, and certainly because of wide-spread unhappiness with the chaotic monetary and financial conditions thatfollowed the war (there were hyperinflations in central Europe and more mod-erate but still serious inflations elsewhere), the desire to return to gold in theearly 1920s was strong. Of much concern however was the perception thatthere was not enough gold available to satisfy world money demands withoutdeflation. The 1922 Economic and Monetary Conference at Genoa addressedthis issue by recommending the adoption of a gold exchange standard, inwhich convertible foreign exchange reserves (principally dollars and pounds)as well as gold would be used to back national money supplies, thus \"econo-mizing\" on gold. Although \"key currencies\" had been used as reserves beforethe war, the Genoa recommendations led to a more widespread and officiallysanctioned use of this practice (Lindert 1969; Eichengreen 1987).During the 1920s the vast majority of the major countries succeeded in re-turning to gold. (The first column of table 2.1 gives the dates of return for the36 Ben Bernanke and Harold Jamescountries in our data set.) Britain returned at the prewar parity in 1925, despiteKeynes's argument that at the old parity the pound would be overvalued. Bythe end of 1925, out of a list of 48 currencies given by the League of Nations(1926), 28 had been pegged to gold. France returned to gold gradually, fol-lowing the Poincare stabilization, although at a new parity widely believed toundervalue the franc. By the end of 1928, except for China and a few smallcountries on the silver standard, only Spain, Portugal, Rumania, and Japanhad not been brought back into the gold standard system. Rumania went backon gold in 1929, Portugal did so in practice also in 1929 (although not offi-cially until 1931), and Japan in December 1930. In the same month the Bankfor International Settlements gave Spain a stabilization loan, but the operationwas frustrated by a revolution in April 1931, carried out by republicans who,as one of the most attractive features of their program, opposed the foreignstabilization credits. Spain thus did not join the otherwise nearly universalmembership of the gold standard club.The classical gold standard of the prewar period functioned reasonablysmoothly and without a major convertibility crisis for more than thirty years.In contrast, the interwar gold standard, established between 1925 and 1928,had substantially broken down by 1931 and disappeared by 1936. An exten-sive literature has analyzed the differences between the classical and interwargold standards. This literature has focused, with varying degrees of emphasis,both on fundamental economic problems that complicated trade and monetaryadjustment in the interwar period and on technical problems of the interwargold standard itself.In terms of \"fundamentals,\" Temin (1989) has emphasized the effects of theGreat War, arguing that, ultimately, the war itself was the shock that initiatedthe Depression. The legacy of the war included—besides physical destruc-tion, which was relatively quickly repaired—new political borders drawn ap-parently without economic rationale; substantial overcapacity in some sectors(such as agriculture and heavy industry) and undercapacity in others, relativeto long-run equilibrium; and reparations claims and international war debtsthat generated fiscal burdens and fiscal uncertainty. Some writers (notablyCharles Kindleberger) have also pointed to the fact that the prewar gold stan-dards was a hegemonic system, with Great Britain the unquestioned center. Incontrast, in the interwar period the relative decline of Britain, the inexperienceand insularity of the new potential hegemon (the United States), and ineffec-tive cooperation among central banks left no one able to take responsibilityfor the system as a whole.The technical problems of the interwar gold standard included the followingthree:1. The asymmetry between surplus and deficit countries in the requiredmonetary response to gold flows. Temin suggests, correctly we believe, thatthis was the most important structural flaw of the gold standard. In theory,under the \"rules of the game,\" central banks of countries experiencing gold37 Financial Crisis in the Great DepressionTable 2.1Dates of Changes in Gold Standard PoliciesSuspension ofGold StandardDecember 1929April 1933—October 1931—September 1931June 1933October 1931——April 1932——December 1931——September 1931September 1931——ForeignExchangeControlOctober 1931——September 1931November 1931November 1931——July 1931September 1931July 1931May 1934July 1932October 1931———April 1936May 1932CountryAustraliaAustriaBelgiumCanadaCzechoslovakiaDenmarkEstoniaFinlandFranceGermanyGreeceHungaryItalyJapanLatviaNetherlandsNorwayNew ZealandPolandRumaniaReturn to GoldApril 1925April 1925October 1926July 1926April 1926January 1927January 1928January 1926August 1926-June 1928September1924May 1928April 1925December1927December1930August 1922April 1925May 1928April 1925October 1927March 1927-February1929April 1924—May 1925June 1919DevaluationMarch 1930September 1931March 1935September 1931February 1934September 1931June 1933October 1931October 1936—April 1932—October 1936December 1931—October 1936September 1931April 1930October 1936—SwedenSpainUnited KingdomUnited StatesSeptember 1931—September 1931March 1933—May 1931—March 1933September 1931—September 1931April 1933Source: League of Nations, Yearbook, various dates; and miscellaneous supplementary sources.inflows were supposed to assist the price-specie flow mechanism by expand-ing domestic money supplies and inflating, while deficit countries were sup-posed to reduce money supplies and deflate. In practice, the need to avoid acomplete loss of reserves and an end to convertibility forced deficit countriesto comply with this rule; but, in contrast, no sanction prevented surplus coun-tries from sterilizing gold inflows and accumulating reserves indefinitely, ifdomestic objectives made that desirable. Thus there was a potential deflation-ary bias in the gold standard's operation.This asymmetry between surplus and deficit countries also existed in theprewar period, but with the important difference that the prewar gold standardcentered around the operations of the Bank of England. The Bank of England38 Ben Bernanke and Harold Jamesof course had to hold enough gold to ensure convertibility, but as a profit-making institution it also had a strong incentive not to hold large stocks ofbarren gold (as opposed to interest-paying assets). Thus the Bank managedthe gold standard (with the assistance of other central banks) so as to avoidboth sustained inflows and sustained outflows of gold; and, indeed, it helpedensure continuous convertibility with a surprisingly low level of gold re-serves. In contrast, the two major gold surplus countries of the interwar pe-riod, the United States and France, had central banks with little or no incentiveto avoid accumulation of gold.The deflationary bias of the asymmetry in required adjustments was mag-nified by statutory fractional reserve requirements imposed on many centralbanks, especially the new central banks, after the war. While Britain, Norway,Finland, and Sweden had a fiduciary issue—a fixed note supply backed onlyby domestic government securities, above which 100% gold backing was re-quired—most countries required instead that minimum gold holdings equal afixed fraction (usually close to the Federal Reserve's 40%) of central bankliabilities. These rules had two potentially harmful effects.First, just as required \"reserves\" for modern commercial banks are notreally available for use as true reserves, a large portion of central bank goldholdings were immobilized by the reserve requirements and could not be usedto settle temporary payments imbalances. For example, in 1929, according tothe League of Nations, for 41 countries with a total gold reserve of $9,378million, only $2,178 million were \"surplus\" reserves, with the rest requiredas cover (League of Nations 1944, 12). In fact, this overstates the quantity oftruly free reserves, because markets and central banks became very worriedwhen reserves fell within 10% of the minimum. The upshot of this is thatdeficit countries could lose very little gold before being forced to reduce theirdomestic money supplies; while, as we have noted, the absence of any maxi-mum reserve limit allowed surplus countries to accept gold inflows withoutinflating.The second and related effect of the fractional reserve requirement has to dowith the relationship between gold outflows and domestic monetary contrac-tion. With fractional reserves, the relationship between gold outflow and thereduction in the money supply was not one for one; with a 40% reserve re-quirement, for example, the impact on the money supply of a gold outflowwas 2.5 times the external loss. So again, loss of gold could lead to an imme-diate and sharp deflationary impact, not balanced by inflation elsewhere.2. The pyramiding of reserves. As we have noted, under the interwar gold-exchange standard, countries other than those with reserve currencies wereencouraged to hold convertible foreign exchange reserves as a partial (or insome cases, as a nearly complete) substitute for gold. But these convertiblereserves were in turn usually only fractionally backed by gold. Thus, just as ashift by the public from fractionally backed deposits to currency would lowerthe total domestic money supply, the gold-exchange system opened up the39 Financial Crisis in the Great Depressionpossibility that a shift of central banks from foreign exchange reserves to goldmight lower the world money supply, adding another deflationary bias to thesystem. Central banks did abandon foreign exchange reserves en masse in theearly 1930s, when the threat of devaluation made foreign exchange assetsquite risky. According to Eichengreen (1987), however, the statistical evi-dence is not very clear on whether central banks after selling their foreignexchange simply lowered their cover ratios, which would have had no directeffect on money supplies, or shifted into gold, which would have been con-tractionary. Even if the central banks responded only by lowering cover ratios,however, this would have increased the sensitivity of their money supplies toany subsequent outflow of reserves.3. Insufficient powers of central banks. An important institutional feature ofthe interwar gold standard is that, for a majority of the important continentalEuropean central banks, open market operations were not permitted or wereseverely restricted. This limitation on central bank powers was usually theresult of the stabilization programs of the early and mid 1920s. By prohibitingcentral banks from holding or dealing in significant quantities of governmentsecurities, and thus making monetization of deficits more difficult, the archi-tects of the stabilizations hoped to prevent future inflation. This forced thecentral banks to rely on discount policy (the terms at which they would makeloans to commercial banks) as the principal means of affecting the domesticmoney supply. However, in a number of countries the major commercialbanks borrowed very infrequently from the central banks, implying that ex-cept in crisis periods the central bank's control over the money supply mightbe quite weak.The loosening of the link between the domestic money supply and centralbank reserves may have been beneficial in some cases during the 1930s, if itmoderated the monetary effect of reserve outflows. However, in at least onevery important case the inability of a central bank to conduct open marketoperations may have been quite destabilizing. As discussed by Eichengreen(1986), the Bank of France, which was the recipient of massive gold inflowsuntil 1932, was one of the banks that was prohibited from conducting openmarket operations. This severely limited the ability of the Bank to translate itsgold inflows into monetary expansion, as should have been done in obedienceto the rules of the game. The failure of France to inflate meant that it contin-ued to attract reserves, thus imposing deflation on the rest of the world.3Given both the fundamental economic problems of the international econ-omy and the structural flaws of the gold standard system, even a relativelyminor deflationary impulse might have had significant repercussions. As ithappened, both of the two major gold surplus countries—France and theUnited States, who at the time together held close to 60% of the world's mon-etary gold—took deflationary paths in 1928-29 (Hamilton 1987).In the French case, as we have already noted, the deflationary shock tookthe form of a largely sterilized gold inflow. For several reasons—including a40 Ben Bernanke and Harold Jamessuccessful stabilization with attendant high real interest rates, a possiblyundervalued franc, the lifting of exchange controls, and the perception thatFrance was a \"safe haven\" for capital—beginning in early 1928 gold floodedinto that country, an inflow that was to last until 1932. In 1928, France con-trolled about 15% of the total monetary gold held by the twenty-four countriesin our data set (Board of Governors 1943); this share, already disproportionateto France's economic importance, increased to 18% in 1929, 22% in 1930,28% in 1931, and 32% in 1932. Since the U.S. share of monetary gold re-mained stable at something greater than 40% of the total, the inflow to Franceimplied significant losses of gold by countries such as Germany, Japan, andthe United Kingdom.With its accumulation of gold. France should have been expected to inflate;but in part because of the restrictions on open market operations discussedabove and in part because of deliberate policy choices, the impact of the goldinflow on French prices was minimal. The French monetary base did increasewith the inflow of reserves, but because economic growth led the demand forfrancs to expand even more quickly, the country actually experienced a whole-sale price deflation of almost 11% between January 1929 and January 1930.Hamilton (1987) also documents the monetary tightening in the UnitedStates in 1928, a contraction motivated in part by the desire to avoid losinggold to the French but perhaps even more by the Federal Reserve's determi-nation to slow down stock market speculation. The U.S. price level fell about4% over the course of 1929. A business cycle peak was reached in the UnitedStates in August 1929, and the stock market crashed in October.The initial contractions in the United States and France were largely self-inflicted wounds; no binding external constraint forced the United States todeflate in 1929, and it would certainly have been possible for the French gov-ernment to grant the Bank of France the power to conduct expansionary openmarket operations. However, Temin (1989) argues that, once these destabiliz-ing policy measures had been taken, little could be done to avert deflation anddepression, given the commitment of central banks to maintenance of the goldstandard. Once the deflationary process had begun, central banks engaged incompetitive deflation and a scramble for gold, hoping by raising cover ratiosto protect their currencies against speculative attack. Attempts by any individ-ual central bank to reflate were met by immediate gold outflows, which forcedthe central bank to raise its discount rate and deflate once again. According toTemin, even the United States, with its large gold reserves, faced this con-straint. Thus Temin disagrees with the suggestion of Friedman and Schwartz(1963) that the Federal Reserve's failure to protect the U.S. money supply wasdue to misunderstanding of the problem or a lack of leadership; instead, heclaims, given the commitment to the gold standard (and, presumably, the ab-sence of effective central bank cooperation), the Fed had little choice but tolet the banks fail and the money supply fall.For our purposes here it does not matter much to what extent central bank41 Financial Crisis in the Great Depressionchoices could have been other than what they were. For the positive questionof what caused the Depression, we need only note that a monetary contractionbegan in the United States and France, and was propagated throughout theworld by the international monetary standard.4If monetary contraction propagated by the gold standard was the source ofthe worldwide deflation and depression, then countries abandoning the goldstandard (or never adopting it) should have avoided much of the deflationarypressure. This seems to have been the case. In an important paper, Choudhriand Kochin (1980) documented that Spain, which never restored the goldstandard and allowed its exchange rate to float, avoided the declines in pricesand output that affected other European countries. Choudhri and Kochin alsoshowed that the Scandinavian countries, which left gold along with the UnitedKingdom in 1931, recovered from the Depression much more quickly thanother small European countries that remained longer on the gold standard.Much of this had been anticipated in an insightful essay by Haberler (1976).Eichengreen and Sachs (1985) similarly focused on the beneficial effects ofcurrency depreciation (i.e., abandonment of the gold standard or devalua-tion). For a sample of ten European countries, they showed that depreciatingcountries enjoyed faster growth of exports and industrial production thancountries which did not depreciate. Depreciating countries also experiencedlower real wages and greater profitability, which presumably helped to in-crease production. Eichengreen and Sachs argued that depreciation, in thiscontext, should not necessarily be thought of as a \"beggar thy neighbor\" pol-icy; because depreciations reduced constraints on the growth of world moneysupplies, they may have conferred benefits abroad as well as at home (al-though a coordinated depreciation presumably would have been better thanthe uncoordinated sequence of depreciations that in fact took place).5Some additional evidence of the effects of maintaining or leaving the goldstandard, much in the spirit of Eichengreen and Sachs but using data from alarger set of countries, is given in our tables 2.2 through 2.4. These tablessummarize the relationships between the decision to adhere to the gold stan-dard and some key macroeconomic variables, including wholesale price infla-tion (table 2.2), some indicators of national monetary policies (table 2.3), andindustrial production growth (table 2.4). To construct these tables, we dividedour sample of twenty-four countries into four categories:6 1) countries not onthe gold standard at all (Spain) or leaving prior to 1931 (Australia and NewZealand); 2) countries abandoning the full gold standard in 1931 (14 coun-tries); 3) countries abandoning the gold standard between 1932 and 1935 (Ru-mania in 1932, the United States in 1933, Italy in 1934, and Belgium in1935); and 4) countries still on the full gold standard as of 1936 (France,Netherlands, Poland).7 Tables 2.2 and 2.4 give the data for each country, aswell as averages for the large cohort of countries abandoning gold in 1931,for the remnant of the gold bloc still on gold in 1936, and (for 1932-35, whenthere were a significant number of countries in each category) for all gold42 Ben Bernanke and Harold Jamesstandard and non-gold standard countries. Since table 2.3 reports data on fourdifferent variables, in order to save space only the averages are shown.8The link between deflation and adherence to the gold standard, shown intable 2.2, seems quite clear. As noted by Choudhri and Kochin (1980),Spain's abstention from the gold standard insulated that country from the gen-eral deflation; New Zealand and Australia, presumably because they retainedlinks to sterling despite early abandonment of the strict gold standard, didhowever experience some deflation. Among countries on the gold standard asof 1931, there is a rather uniform experience of about a 13% deflation in both1930 and 1931. But after 1931 there is a sharp divergence between thosecountries on and those off the gold standard. Price levels in countries off thegold standard have stabilized by 1933 (with one or two exceptions), and thesecountries experience mild inflations in 1934-36. In contrast, the gold standardcountries continue to deflate, although at a slower rate, until the gold stan-dard's dissolution in 1936.With such clearly divergent price behavior between countries on and offgold, one would expect to see similarly divergent behavior in monetary pol-icy. Table 2.3 compares the average behavior of the growth rates of three mon-etary aggregates, called for short MO, Ml, and M2, and of changes in thecentral bank discount rate. MO corresponds to money and notes in circulation,Ml is the sum of MO and commercial bank deposits, and M2 is the sum ofMl and savings bank deposits.9 The expected differences in the monetary po-lices of the gold and non-gold countries seem to be in the data, although some-what less clearly than we had anticipated. In particular, despite the twelvepercentage point difference in rates of deflation between gold and non-goldcountries in 1932, the differences in average money growth in that year be-tween the two classes of countries are minor; possibly, higher inflation expec-tations in the countries abandoning gold reduced money demand and thusbecame self-confirming. From 1933 through 1935, however, the various mon-etary indicators are more consistent with the conclusion stressed by Eichen-green and Sachs (1985), that leaving the gold standard afforded countriesmore latitude to expand their money supplies and thus to escape deflation.The basic proposition of the gold standard-based explanation of theDepression is that, because of its deflationary impact, adherence to the goldstandard had very adverse consequences for real activity. The validity of thisproposition is shown rather clearly by table 2.4, which gives growth rates ofindustrial production for the countries in our sample. While the countrieswhich were to abandon the gold standard in 1931 did slightly worse in 1930and 1931 than the nations of the Gold Bloc, subsequent to leaving gold thesecountries performed much better. Between 1932 and 1935, growth of indus-trial production in countries not on gold averaged about seven percentagepoints a year better than countries remaining on gold, a very substantial effect.In summary, data from our sample of twenty-four countries support theTable 2.2Log-differences of the Wholesale Price Index19301931193219331934193519361. Countries not on gold standard or leaving prior to 1931SpainAustralia (1929)New Zealand (1930)-.00-.12-.03.01-.11-.07-.01-.01-.03-.05-.00.03.03.04.01.01-.00.03.02.05.012. Countries abandoning full gold standard in 1931AustriaCanadaCzechoslovakiaDenmarkEstoniaFinlandGermanyGreeceHungaryJapanLatviaNorwaySwedenUnited KingdomAverage-.11-.10-.12-.15-.14-.09-.10-.10-.14-.19-.16-.08-.14-.17-.13-.07-.18-.10-.13-.11-.07-.12jj-.05-.17-.18-.12-.09-.18-.12.03-.08-.08.02-.09.07-.14.18-.01.05.00.00-.02-.04-.04.01-.03.07.02-.01-.03.12-.14.11-.02-.00-.02.01.00.02.06.02.09.00.01.05-.01.00-.01-.01.02.06.04.02-.00.01.04.02-.01.00.03.02.08.04.05.03.02.04.03-.01.03.00.05.08.02.02.02.03.06.04.05.03.06.04-.013. Countries abandoning gold standard between 1932 and 1935Rumania (1932)United States (1933)Italy (1934)Belgium (1935)-.24-.10-.11-.13-.26-.17-.14-.17-.11-.12-.07-.16-.03.02-.09-.06.00.13-.02-.06.14.07.10.13.13.01.11.094. Countries still on full gold standard as of 1936FranceNetherlandsPolandAverage-.12-.11-.12-.12-.10-.16-.14-.13-.16-.17-.13-.15-.07-.03-.10-.07-.06.00-.06-.04-.11-.02-.05-.06.19.04.02.085. Grand averagesGold standard countriesNon-gold countries-.13-.01.07.00-.04.03-.05.04Note: Data on wholesale prices are from League of Nations, Monthly Bulletin of Statistics and Yearbook,various issues. Dates in parentheses are years in which countries abandoned gold, with \"abandon-ment\" defined to include the imposition of foreign exchange controls or devaluation as well as suspen-sion; see table 2.1.44 Ben Bernanke and Harold JamesTable 2.3Monetary Indicators19301931193219331934193519361. Countries abandoning full gold standard in 1931MO growthMl growthM2 growthDiscount rate change-.04.01.03-0.8-.02-.11-.080.4-.07-.07-.04-0.2.06.02.03-1.2.05.05.05-0.4.05.04.05-0.1.08.08.06-0.12. Countries still on full gold standard as of 1936M0 growthMl growthM2 growthDiscount rate change.03.05.08-1.4.07-.06-.00-0.4-.06-.07-.020.1-.02-.05-.02-0.4.01.01.02-0.4-.03-.06-.030.8.03.08.05-0.33. Grand averages: Countries on goldM0 growthMl growthM2 growthDiscount rate change-.04-.09-.050.2-.03-.04-.01-0.5.01-.01.01-0.4-.02-.06-.020.74. Grand averages: Countries off goldM0 growthMl growthM2 growthDiscount rate change-.07-.06-.03-0.3.05.01.02-1.0.03.04.04-0.4.06.05.05-0.2Note: M0 is money and notes in cirulation. Ml is base money plus commercial bank deposits.M2 is Ml plus savings deposits. Growth rates of monetary aggregates are calculated as log-differences. The discount rate change is in percentage points. The data are from League of Na-tions, Monthly Bulletin of Statistics and Yearbook, various issues.view that there was a strong link between adherence to the gold standard andthe severity of both deflation and depression. The data are also consistent withthe hypothesis that increased freedom to engage in monetary expansion was areason for the better performance of countries leaving the gold standard earlyin the 1930s, although the evidence in this case is a bit less clear-cut.2.3 The Link Between Deflation and DepressionGiven the above discussion and evidence, it seems reasonable to accept theidea that the worldwide deflation of the early 1930s was the result of a mone-tary contraction transmitted through the international gold standard. But this45 Financial Crisis in the Great DepressionTable 2.4 Log-differences of the Industrial Production Index1930 1931 1932 1933 1934 1935 19361. Countries not on gold standard or leaving prior to 1931SpainAustralia (1929)New Zealand (1930)-.01-.11-.25-.06-.07-.14-.05.07.05-.05.10.02.01.09.13.02.09.09NA.07.142. Countries abandoning full gold standard in 1931AustriaCanadaCzechoslovakiaDenmarkEstoniaFinlandGermanyGreeceHungaryJapanLatviaNorwaySwedenUnited Kingdom-.16-.16-.11.08-.02-.10-.15.01-.06-.05.08.01.03-.08-.19-.18-.10-.08-.09-.13-.24.02-.08-.03-.20-.25-.07-.10-.14-.20-.24-.09-.17.19-.24-.08-.06.07-.08.17-.08-.00.03.04-.05.14.05.02.13.10.07.15.31.01.02.05.11.20.10.11.17.03.27.12.12.13.15.04.19.11.13.10.05.07.10.10.16.12.07.10.05.10.11.07.07.10.14.04.10.09.12-.03.10.06.04.09.09.09Average -.05 -.12 -.07 .08 .13 .10 .083. Countries abandoning gold standard between 1932 and 1935Rumania (1932)United States (1933)Italy (1934)Belgium (1935)-.03-.21-.08-.12.05-.17-.17-.09-.14-.24-.15-.16.15.17.10.04.19.04.08.01-.01.13.16.12.06.15-.07.054. Countries still on full gold standard as of 1936FranceNetherlandsPoland-.01.02-.13-.14-.06-.14-.19-.13-.20.12.07.09-.07.02.12-.04-.03.07.07.01.10Average -.04 -.11 -.17 .10 .02 .00 .065. Grand averagesGold standard countries -.18 .09 .03 .01Non-gold countries -.06 .08 .12 .09Note: Data on industrial production are from League of Nations, Monthly Bulletin of Statisticsand Yearbook, various issues, supplemented by League of Nations, Industrialization and ForeignTrade, 1945.46 Ben Bernanke and Harold Jamesraises the more difficult question of what precisely were the channels linkingdeflation (falling prices) and depression (falling output). This section takes apreliminary look at some suggested mechanisms. We first introduce here twoprincipal channels emphasized in recent research, then discuss the alternativeof induced financial crisis.1. Real wages. If wages possess some degree of nominal rigidity, then fall-ing output prices will raise real wages and lower labor demand. Downwardstickiness of wages (or of other input costs) will also lower profitability, poten-tially reducing investment. This channel is stressed by Eichengreen and Sachs(see in particular their 1986 paper) and has also been emphasized by Newelland Symons (1988).Some evidence on the behavior of real wages during the Depression is pre-sented in table 2.5, which is similar in format to tables 2.2-2.4. Note thattable 2.5 uses the wholesale price index (the most widely available price in-dex) as the wage deflator. According to this table, there were indeed large realwage increases in most countries in 1930 and 1931. After 1931, countriesleaving the gold standard experienced a mild decline in real wages, while realwages in gold standard countries exhibited a mild increase. These findings aresimilar to those of Eichengreen and Sachs (1985).The reliance on nominal wage stickiness to explain the real effects of thedeflation is consistent with the Keynesian tradition, but is nevertheless some-what troubling in this context. Given (i) the severity of the unemployment thatwas experienced during that time; (ii) the relative absence of long-term con-tracts and the weakness of unions; and (iii) the presumption that the generalpublic was aware that prices, and hence the cost of living, were falling, it ishard to understand how nominal wages could have been so unresponsive.Wages had fallen quickly in many countries in the contraction of 1921-22. Inthe United States, nominal wages were maintained until the fall of 1931 (pos-sibly by an agreement among large corporations; see O'Brien 1989), but fellsharply after that; in Germany, the government actually tried to depress wagesearly in the Depression. Why then do we see these large real wage increasesin the data?One possibility is measurement problems. There are a number of issues,such as changes in skill and industrial composition, that make measuring thecyclical movement in real wages difficult even today. Bernanke (1986) hasargued, in the U.S. context, that because of sharp reductions in workweeksand the presence of hoarded labor, the measure real wage may have been apoor measure of the marginal cost of labor.Also in the category of measurement issues, Eichengreen and Hatton(1987) correctly point out that nominal wages should be deflated by the rele-vant product prices, not a general price index. Their table of product wageindices (nominal wages relative to manufacturing prices) is reproduced for1929-38 and for the five countries for which data are available as our table2.6. Like table 2.5, this table also shows real wages increasing in the early47 Financial Crisis in the Great DepressionTable 2.5 Log-differences of the Real Wage1930 1931 1932 1933 1934 1935 19361. Countries not on gold standard or leaving prior to 1931SpainAustralia (1929)New Zealand (1930)not available-.04-.05.10.03.01.00-.05-.00.03.01.01-.01-.03.102. Countries abandoning full gold standard in 1931AustriaCanadaCzechoslovakiaDenmarkEstoniaFinlandGermanyGreeceHungaryJapanLatviaNorwaySwedenUnited Kingdom.14.11.14.17.16.12.14.05.20.08.17.17.05.15.11.11.07.06-.00.21.18.08.09.16-.04.00.08-.03.02-.03-.07-.04-.15.02.01.02-.00.05-.06.05.02.04-.07.09-.06.01not available-.00.07not available.09.06-.12.02-.05.01-.02.01-.02.06-.02.03-.03.02-.05-.01.06-.03-.11-.05-.05-.03-.01-.03.06-.01-.00-.04-.03-.02-.00-.05-.02-.02-.02-.03Average .14 .11 -.02 -.03 -.04 -.03 -.023. Countries abandoning gold standard between 1932 and 1935Rumania (1932)United States (1933)Italy (1934)Belgium (1935).20.10.10.19.14.13.07.10-.10-.01.05.07-.05-.03.07.04-.02.04-.01.01-.15-.03-.11-.16-.12.02-.06-.024. Countries still on full gold standard as of 1936FranceNetherlandsPoland.21.12.11.09.14.06.12.09.05.07-.02.00.06-.04.01.09-.01.02-.06-.06-.03Average .15 .10 .09 .02 .01 .03 -.055. Grand averagesGold standard countries .05 .03 .01 .02Non-gold countries -.02 -.03 -.03 -.04Note: The real wage is the nominal hourly wage for males (skilled, if available) divided by thewholesale price index. Wage data are from the International Labour Office, Year Book of LaborStatistics, various issues.48 Ben Bernanke and Harold JamesTable 2.6Year1929193019311932193319341935193619371938Indices of Product WagesUnited Kingdom100.0103.0106.4108.3109.3111.4111.3110.4107.8108.6United States100.0106.1113.0109.6107.9115.8114.3115.9121.9130.0Germany100.0100.4102.296.899.3103.0105.3107.7106.5107.7Japan100.0115.6121.6102.9101.8102.3101.699.287.186.3Sweden100.0116.6129.1130.0127.9119.6119.2116.0101.9115.1Source: Eichengreen and Hatton (1987, 15).1930s, but overall the correlation of real wage increases and depression doesnot appear particularly good. Note that Germany, which had probably theworst unemployment problem of any major country, has almost no increase inreal wages;10 the United Kingdom, which began to recover in 1932, has realwages increasing on a fairly steady trend during its recovery period; and theUnited States has only a small dip in real wages at the beginning of its recov-ery, followed by more real wage growth. The case for nominal wage stickinessas a transmission mechanism thus seems, at this point, somewhat mixed.2. Real interest rates. In a standard IS-LM macro model, a monetary con-traction depresses output by shifting the LM curve leftwards, raising real in-terest rates, and thus reducing spending. However, as Temin (1976) pointedout in his original critique of Friedman and Schwartz, it is real rather thannominal money balances that affect the LM curve; and since prices were fall-ing sharply, real money balances fell little or even rose during the contraction.Even if real money balances are essentially unchanged, however, there isanother means by which deflation can raise ex ante real interest rates: Sincecash pays zero nominal interest, in equilibrium no asset can bear a nominalinterest rate that is lower than its liquidity and risk premia relative to cash.Thus an expected deflation of 10% will impose a real rate of at least 10% onthe economy, even with perfectly flexible prices and wages. In an IS-LM dia-gram drawn with the nominal interest rate on the vertical axis, an increase inexpected deflation amounts to a leftward shift of the IS curve.Whether the deflation of the early 1930s was anticipated has been exten-sively debated (although almost entirely in the United States context). We willadd here two points in favor of the view that the extent of the worldwidedeflation was less than fully anticipated.First, there is the question of whether the nominal interest rate floor was infact binding in the deflating countries (as it should have been if this mecha-nism was to operate). Although interest rates on government debt in theUnited States often approximated zero in the 1930s, it is less clear that this49 Financial Crisis in the Great Depressionwas true for other countries. The yield on French treasury bills, for example,rose from a low of 0.75% in 1932 to 2.06% in 1933, 2.25% in 1934, and3.38% in 1935; during 1933-35 the nominal yield on French treasury billsexceeded that of British treasury bills by several hundred basis points on av-erage.\"Second, the view that deflation was largely anticipated must contend withthe fact that nominal returns on safe assets were very similar whether coun-tries abandoned or stayed on gold. If continuing deflation was anticipated inthe gold standard countries, while inflation was expected in countries leavinggold, the similarity of nominal returns would have implied large expecteddifferences in real returns. Such differences are possible in equilibrium, if theyare counterbalanced by expected real exchange rate changes; nevertheless,differences in expected real returns between countries on and off gold on theorder of 11-12% (the realized difference in returns between the two blocs in1932) seem unlikely.123. Financial crisis. A third mechanism by which deflation can inducedepression, not considered in the recent literature, works through deflation'seffect on the operation of the financial system. The source of the non-neutrality is simply that debt instruments (including deposits) are typically setin money terms. Deflation thus weakens the financial positions of borrowers,both nonfinancial firms and financial intermediaries.Consider first the case of intermediaries (banks).13 Bank liabilities (primar-ily deposits) are fixed almost entirely in nominal terms. On the asset side,depending on the type of banking system (see below), banks hold either pri-marily debt instruments or combinations of debt and equity. Ownership ofdebt and equity is essentially equivalent to direct ownership of capital; in thiscase, therefore, the bank's liabilities are nominal and its assets are real, so thatan unanticipated deflation begins to squeeze the bank's capital position im-mediately. When only debt is held as an asset, the effect of deflation is for awhile neutral or mildly beneficial to the bank. However, when borrowers'equity cushions are exhausted, the bank becomes the owner of its borrowers'real assets, so eventually this type of bank will also be squeezed by deflation.As pressure on the bank's capital grows, according to this argument, itsnormal functioning will be impeded; for example, it may have to call in loansor refuse new ones. Eventually, impending exhaustion of bank capital leads toa depositors' run, which eliminates the bank or drastically curtails its opera-tion. The final result is usually a government takeover of the intermediationprocess. For example, a common scenario during the Depression was for thegovernment to finance an acquisition of a failing bank by issuing its own debt;this debt was held (directly or indirectly) by consumers, in lieu of (vanishing)commercial bank deposits. Thus, effectively, government agencies becamepart of the intermediation chain.14Although the problems of the banks were perhaps the more dramatic in theDepression, the same type of non-neutrality potentially affects nonfinancial50 Ben Bernanke and Harold Jamesfirms and other borrowers. The process of \"debt deflation\crease in the real value of nominal debt obligations brought about by fallingprices, erodes the net worth position of borrowers. A weakening financialposition affects the borrower's actions (e.g., the firm may try to conserve fi-nancial capital by laying off workers or cutting back on investment) and also,by worsening the agency problems in the borrower-lender relationship, im-pairs access to new credit. Thus, as discussed in detail in Bernanke and Ger-tler (1990), \"financial distress\" (such as that induced by debt deflation) can inprinciple impose deadweight losses on an economy, even if firms do notundergo liquidation.Before trying to assess the quantitative impact of these and other channelson output, we briefly discuss the international incidence of financial crisisduring the Depression.2.4 Interwar Banking and Financial CrisesFinancial crises were of course a prominent feature of the interwar period.We focus in this section on the problems of the banking sector and, to a lesserextent, on the problems of domestic debtors in general, as suggested by thediscussion above. Stock market crashes and defaults on external debt werealso important, of course, but for the sake of space will take a subsidiary rolehere.Table 2.7 gives a chronology of some important interwar banking crises.The episodes listed actually cover a considerable range in terms of severity, asthe capsule descriptions should make clear. However the chronology shouldalso show that (i) quite a few different countries experienced significant bank-ing problems during the interwar period; and (ii) these problems reached avery sharp peak between the spring and fall of 1931, following the Creditan-stalt crisis in May 1931 as well as the intensification of banking problems inGermany.A statistical indicator of banking problems, emphasized by Friedman andSchwartz (1963), is the deposit-currency ratio. Data on the changes in thecommercial bank deposit-currency ratio for our panel of countries are pre-sented in table 2.8. It is interesting to compare this table with the chronologyin table 2.7. Most but not all of the major banking crises were associated withsharp drops in the deposit-currency ratio; the most important exception is in1931 in Italy, where the government was able to keep secret much of the bank-ing system's problems until a government takeover was affected. On the otherhand, there were also significant drops in the deposit-currency ratio that werenot associated with panics; restructurings of the banking system and exchangerate difficulties account for some of these episodes.What caused the banking panics? At one level, the panics were an endoge-nous response to deflation and the operation of the gold standard regime.51 Financial Crisis in the Great DepressionTable 2.7A Chronology of Interwar Banking Crises, 1921-36DateJune 19211921-221922April 1923May 1923September 1923 JAPANSeptember 1925 SPAINJuly-September POLAND19261927 NORWAY, ITALYApril 1927 JAPANAugust 1929 GERMANYNovember 1929 AUSTRIANovember 1930 FRANCEDecember 1930 U.S.April 1931(continued)CountrySWEDENNETHERLANDSDENMARKNORWAYAUSTRIAESTONIAITALYARGENTINACrisesBeginning of deposit contraction of 1921-22, leadingto bank restructurings. Government assistanceadministered through Credit Bank of 1922.Bank failures (notably Marx & Co.) andamalgamations.Heavy losses of one of the largest banks, DanskeLandmandsbank, and liquidation of smaller banks.Landmandsbank continues to operate until a restructingin April 1928 under a government guarantee.Failure of Centralbanken for Norge.Difficulties of a major bank, AllgemeineDepositenbank; liquidation in July.In wake of the Tokyo earthquake, bad debts threatenBank of Taiwan and Bank of Chosen, which arerestructured with government help.Failure of Banco de la Union Mineira and BancoVasca.Bank runs cause three large banks to stop payments.The shakeout of banks continues through 1927.Numerous smaller banks in difficulties, but no majorfailures.Thirty-two banks unable to make payments.Restructuring of 15th Bank and Bank of Taiwan.Collapse of Frankfurter Allgemeine Versicherungs AG,followed by failures of smaller banks, and runs onBerlin and Frankfurt savings banks.Bodencreditanstalt, second largest bank, fails and ismerged with Creditanstalt.Failure of Banque Adam, Boulogne-sur-Mer, andOustric Group. Runs on provincial banks.Failure of two medium-sized banks, EstoniaGovernment Bank Tallin and Reval Credit Bank; crisislasts until January.Failure of Bank of the United States.Withdrawals from three largest banks begin. A panicensues in April 1931, followed by a governmentreorganization and takeover of frozen industrial assets.Government deals with banking panic by allowingBanco de Nacion to rediscount commercial paper fromother banks at government-owned Caja de Conversion.52 Ben Bernanke and Harold JamesTable 2.7Date(continued)CountryCrisesMay 1931 AUSTRIA Failure of Creditanstalt and run of foreign depositors.BELGIUM Rumors about imminent failure of Banque deBruxelles, the country's second largest bank, inducewithdrawals from all banks. Later in the year,expectations of devaluation lead to withdrawals offoreign deposits.June 1931 POLAND Run on banks, especially on Warsaw Discount Bank,April-July 1931 GERMANYJuly 1931 HUNGARYAugust 1931September 1931 U.K.LATVIAAUSTRIACZECHOSLOVAKIATURKEYEGYPTSWITZERLANDRUMANIAMEXICOU.S.ESTONIAassociated with Creditanstalt; a spread of the AustrianBank runs, extending difficulties plaguing the bankingsystem since the summer of 1930. After large loss ofdeposits in June and increasing strain on foreignexchanges, many banks are unable to make paymentsand Darmstadter Bank closes. Bank holiday.Run on Budapest banks (especially General CreditBank). Foreign withdrawals followed by a foreigncreditors' standstill agreement. Bank holiday.Run on banks with German connections. Bank ofLibau and International Bank of Riga particularly hardhit.Failure of Vienna Mercur-Bank.Withdrawal of foreign deposits sparks domesticwithdrawals but no general banking panic.Run on branches of Deutsche Bank and collapse ofBanque Turque pour le Commerce et l'lndustrie, inwake of German crisis.Run on Cairo and Alexandria branches of DeutscheOrientbank.Union Financiere de Geneve rescued by takeover byComptoir d'Escompte de Geneve.Collapse of German-controlled Banca Generala a TariiRomanesti. Run on Banca de Credit Roman and BancaRomaneasca.Suspension of payments after run on Credito Espanolde Mexico. Run on Banco Nacional de Mexico.Series of banking panics, with October 1931 the worstmonth. Between August 1931 and January 1932, 1,860banks fail.External drain, combined with rumors of threat toLondon merchant banks with heavy European(particularly Hungarian and German) involvements.General bank run following sterling crisis; secondwave of runs in November.53 Financial Crisis in the Great DepressionTable 2.7DateOctober 1931(continued)CountryRUMANIACrisesFailure of Banca Marmerosch, Blank & Co. Heavybank runs.Collapse of major deposit bank Banque Nationale deCredit (restructured as Banque Nationale pour leCommerce et l'lndustrie). Other bank failures and bankruns.Weakness of one large bank (SkandinaviskaKreditaktiebolaget) as result of collapse of Kreugerindustrial and financial empire, but no general panic.Losses of large investment bank Banque de l'UnionParisienne forces merger with Credit MobilierFrancois.Series of bank failures in Chicago.New wave of bank failures, especially in the Midwestand Far West.General banking panic, leading to state holidays and anationwide bank holiday in March.Restructuring of large bank (Banque Populaire Suisse)after heavy losses.Failure of Banque Beige de Travail develops intogeneral banking and exchange crisis.Bank problems throughout the fall induce government-sponsored merger of four weak banks (Banco Espanoldel Rio de la Plata, Banco el Hogar Argentina, BancoArgentina-Uruguayo, Ernesto Tomquist & Co.).Deposits fall after Italian invasion of Abyssinia.After years of deposit stability, legislation introducinga tax on bank deposits leads to withdrawals (until fall).Anticipation of second devaluation of the crown leadsto deposit withdrawals.FRANCEMarch 1932SWEDENMay 1932FRANCEJune 1932October 1932February 1933November 1933March 1934September 1934U.S.U.S.U.S.SWITZERLANDBELGIUMARGENTINAOctober 1935January 1936October 1936ITALYNORWAYCZECHOSLOVAKIAWhen the peak of the world banking crisis came in 1931, there had alreadybeen almost two years of deflation and accompanying depression. Consistentwith the analysis at the end of the last section, falling prices lowered the nom-inal value of bank assets but not the nominal value of bank liabilities. In ad-dition, the rules of the gold standard severely limited the ability of centralbanks to ameliorate panics by acting as a lender of last resort; indeed, sincebanking panics often coincided with exchange crises (as we discuss furtherbelow), in order to maintain convertibility central banks typically tightenedmonetary policy in the face of panics. Supporting the connection of bankingproblems with deflation and \"rules of the game\" constraints is the observationthat there were virtually no serious banking panics in any country after aban-54 Ben Bernanke and Harold JamesTable 2.8CountryAustraliaAustriaBelgiumCanadaCzechoslovakiaDenmarkEstoniaFinlandFranceGermanyGreeceHungaryItalyJapanLatviaNetherlandsNorwayNew ZealandPolandRumaniaSwedenSpainUnited KingdomUnited StatesLog-differences of Commercial Bank Deposit-Currency Ratio1930-.05.17-.13*.07-.11.08.16.09-.07-.11*.17.07.04.09.03.10.04.04.07.11-.00.00.03.001931-.12*-.40*-.22*-.01-.08-.03-.29*-.05-.12*-.40*.07-.07-.01.03-.57*-.36*-.15*-.11*-.29*-.76*-.00-.24*-.07-.15*1932.05-.06-.10*.03.07.00-.02.14-.01.05-.27*.10.05-.12*.11-.05-.06.03-.02-.05-.02.08.10-.26*1933.01-.20*.07-.05.02-.07-.05-.04-.10*-.09-.03-.03.06-.04-.06-.06-.09.07-.08-.11*-.06.03-.07-.15*1934.05-.07-.13*.00.07.02.10-.06-.07-.01.06-.08.01.03.12-.05-.01.15.10-.28*-.11*.01-.02.141935-.03-.01-.27*.01-.03.02.05-.04-.10-.08-.04-.05-.20*-.00.10-.08.03-.08-.06.10-.08.06.01.051936-.01-.02-.02-.06-.11*-.00.13-.09-.03-.02.02-.03.08.09.45.24-.23*-.32*.10-.16*-.07N.A.-.03.02Note: Entries are the log-differences of the ratio of commercial bank deposits to money and notes incirculation. Data are from League of Nations, Monthly Bulletin of Statistics and Yearbook, various is-sues.*Decline exceeds .10.donment of the gold standard—although it is also true that by time the goldstandard was abandoned, strong financial reform measures had been taken inmost countries.However, while deflation and adherence to the gold standard were neces-sary conditions for panics, they were not sufficient; a number of countriesmade it through the interwar period without significant bank runs or failures,despite being subject to deflationary shocks similar to those experienced bythe countries with banking problems.15 Several factors help to explain whichcountries were the ones to suffer panics.1. Banking structure. The organization of the banking system was an im-portant factor in determining vulnerability to panics. First, countries with\"unit banking,\" that is, with a large number of small and relatively undiversi-fied banks, suffered more severe banking panics. The leading example is ofcourse the United States, where concentration in banking was very low, but ahigh incidence of failures among small banks was also seen in other countries(e.g., France). Canada, with branch banking, suffered no bank failures during55 Financial Crisis in the Great Depressionthe Depression (although many branches were closed). Sweden and theUnited Kingdom also benefited from a greater dispersion of risk throughbranch systems.16Second, where \"universal\" or \"mixed\" banking on the German or Belgianmodel was the norm, it appears that vulnerability to deflation was greater. Incontrast to the Anglo-Saxon model of banking, where at least in theory lend-ing was short term and the relationship between banks and corporations hadan arm's length character, universal banks took long-term and sometimesdominant ownership positions in client firms. Universal bank assets includedboth long-term securities and equity participations; the former tended to be-come illiquid during a crisis, while the latter exposed universal banks (unlikeAnglo-Saxon banks, which held mainly debt instruments) to the effects ofstock market crashes. The most extreme case was probably Austria. By 1931,after a series of mergers, the infamous Creditanstalt was better thought of as avast holding company rather than a bank; at the time of its failure in May1931, the Creditanstalt owned sixty-four companies, amounting to 65% ofAustria's nominal capital (Kindleberger 1984).2. Reliance of banks on short-term foreign liabilities. Some of the mostserious banking problems were experienced in countries in which a substantialfraction of deposits were foreign-owned. The so-called hot money was moresensitive to adverse financial developments than were domestic deposits.Runs by foreign depositors represented not only a loss to the banking systembut also, typically, a loss of reserves; as we have noted, this additional externalthreat restricted the ability of the central bank to respond to the banking situ-ation. Thus, banking crises and exchange rate crises became intertwined.17The resolution of a number of the central European banking crises required\"standstill agreements,\" under which withdrawals by foreign creditors wereblocked pending future negotiation.International linkages were important on the asset side of bank balancesheets as well. Many continental banks were severely affected by the crises inAustria and Germany, in particular.3. Financial and economic experience of the 1920s. It should not be partic-ularly surprising that countries which emerged from the 1920s in relativelyweaker condition were more vulnerable to panics. Austria, Germany, Hun-gary, and Poland all suffered hyperinflation and economic dislocation in the1930s, and all suffered severe banking panics in 1931. While space constraintsdo not permit a full discussion of the point here, it does seem clear that theorigins of the European financial crisis were at least partly independent ofAmerican developments—which argues against a purely American-centeredexplanation of the origins of the Depression.It should also be emphasized, though, that not just the existence of financialdifficulties during the 1920s but also the policy response to those difficultieswas important. Austria is probably the most extreme case of nagging bankingproblems being repeatedly \"papered over.\" That country had banking prob-56 Ben Bernanke and Harold Jameslems throughout the 1920s, which were handled principally by merging fail-ing banks into still-solvent banks. An enforced merger of the Austrian Bod-encreditanstalt with two failing banks in 1927 weakened that institution,which was part of the reason that the Bodencreditanstalt in turn had to beforceably merged with the Creditanstalt in 1929. The insolvency of the Cre-ditanstalt, finally revealed when a director refused to sign an \"optimistic\" fi-nancial statement in May 1931, sparked the most intense phase of the Euro-pean crisis.In contrast, when banking troubles during the earlier part of the 1920s weremet with fundamental reform, performance of the banking sector during theDepression was better. Examples were Sweden, Japan, and the Netherlands,all of which had significant banking problems during the 1920s but respondedby fundamental restructurings and assistance to place banks on a sound foot-ing (and to close the weakest banks). Possibly because of these earlier events,these three countries had limited problems in the 1930s. A large Swedish bank(Skandinaviska Kreditaktiebolaget) suffered heavy losses after the collapse ofthe Kreuger financial empire, and a medium-sized Dutch bank (Amstelbank)failed because of its connection to the Creditanstalt; but there were no wide-spread panics, only isolated failures.A particularly interesting comparison in this regard is between the Nether-lands and neighboring Belgium, where banking problems persisted from 1931to 1935 and where the ultimate devaluation of the Belgian franc was the resultof an attempt to protect banks from further drains. Both countries were heav-ily dependent on foreign trade and both remained on gold, yet the Nether-lands did much better than Belgium in the early part of the Depression (seetable 2.4). This is a bit of evidence for the relevance of banking difficulties tooutput.Overall, while banking crises were surely an endogenous response todepression, the incidence of crisis across countries reflected a variety of insti-tutional factors and other preconditions. Thus it will be of interest to comparethe real effects of deflation between countries with and without severe bankingdifficulties.On \"debt deflation,\" that is, the problems of nonfmancial borrowers, muchless has been written than on the banking crises. Only for the United Stateshas the debt problem in the 1930s been fairly well documented (see the sum-mary in Bernanke 1983 and the references therein). In that country, large cor-porations avoided serious difficulties, but most other sectors—small busi-nesses, farmers, mortgage borrowers, state and local governments—wereseverely affected, with usually something close to half of outstanding debtsbeing in default. A substantial portion of New Deal reforms consisted of var-ious forms of debt adjustment and relief.For other countries, there are plenty of anecdotes but not much systematicdata. Aggregate data on bankruptcies and defaults are difficult to interpretbecause increasing financial distress forced changes in bankruptcy practices57 Financial Crisis in the Great Depressionand procedures; when the League of Nations' Monthly Bulletin of Statisticsdropped its table on bankruptcies in its December 1932 issue, for example,the reason given therein was that \"the numerous forms of agreement by whichopen bankruptcies are now avoided have seriously diminished the value of thetable\" (p. 529). Perhaps the most extreme case of a change in rules was Ru-mania's April 1932 Law on Conversion of Debts, which essentially eliminatedthe right of creditors to force bankruptcy. Changes in the treatment of bank-ruptcy no doubt ameliorated the effects of debt default, but the fact that thesechanges occurred indicates that the perceived problem must have been severe.More detailed country-by-country study of the effects of deflation on firm bal-ance sheets and the relation of financial condition to firm investment, produc-tion, and employment decisions—where the data permit—would be ex-tremely valuable. A similar comment applies to external debt problems,although here interesting recent work by Eichengreen and Portes (1986) andothers gives us a much better base of knowledge to build on than is availablefor the case of domestic debts.2.5 Regression ResultsIn this section we present empirical results based on our panel data set. Theprincipal question of interest is the relative importance of various transmissionmechanisms of deflation to output. We also address the question, so far notdiscussed, of whether banking crises could have intensified the deflation pro-cess itself.The basic set of results is contained in table 2.9, which relates the log-differences in industrial production for our set of countries to various combi-nations of explanatory variables. The definitions of the right-hand-side vari-ables are as follows:AlnPW: log-difference of the wholesale price index;AlnEX: log-difference of nominal exports;AlnW: log-difference of nominal wage;DISC: central bank discount rate, measured relative to its 1929 value (agovernment bond rate is used for Canada; since no 1929 interest ratecould be found for New Zealand, that country is excluded in regres-sions including DISC);PANIC: a dummy variable, set equal to the number of months during theyear that the country experienced serious banking problems (see be-low);AlnMO: log-difference of money and notes in circulation.Exports are included to control for trade effects on growth, including thebenefits of competitive devaluation discussed by Eichengreen and Sachs(1986); and the wage is included to test for the real wage channel of transmis-sion from deflation to depression. Of course, theory says that both of these58 Ben Bernanke and Harold Jamesvariables should enter in real rather than in nominal terms; unfortunately, inpractice the theoretically suggested deflator is not always available (as wenoted in our discussion of the real wage above). We resolve this problem bysupposing that the true equation is, for example,(1) Aln/P = pe (AlnEX - AlnP,) + $w (AlnW - AlnPJ + errorwhere Pe and Pw, the optimal deflators, are not available. Let the projectionsof log-changes in the unobserved deflators on the log-change in the wholesaleprice deflator be given by(2) AlnP, = \\\\ftMnPW + ui i = e,wwhere the uj are uncorrelated with Aln/W and presumably the i|i. are positive.Then (1) becomes(3) Aln/P = -(PA + $wtyJA\\nPW + p,Aln£X + (^AlnW + new errorThis suggests allowing AlnPW and the nominal growth rates of exports andwages to enter the equation separately, which is how we proceed.18 PuttingAlnPW in the equation separately has the added advantage of allowing us toaccount for any additional effect of deflation (such as debt deflation) not ex-plicitly captured by the other independent variables.The discount rate DISC is included to allow for the interest rate channel andas an additional proxy for monetary policy. Since AlnPW is included in everyequation, inclusion of the nominal interest rate DISC is equivalent to includ-ing the actual ex post real interest rate, that is, we are effectively assumingthat deflation was fully anticipated; this should give the real interest rate hy-pothesis its best chance.In an attempt to control for fiscal policy, we also included measures of cen-tral government expenditure in our first estimated equations. Since the esti-mated coefficients were always negative (the wrong sign), small, and statisti-cally insignificant, the government expenditure variable is excluded from theresults reported here.Construction of the dummy variable PANIC required us to make a judgmentabout which countries' banking crises were most serious, which we did fromour reading of primary and secondary sources. We dated periods of crisis asstarting from the first severe banking problems; if there was some clear de-marcation point (such as the U.S. bank holiday of 1933), we used that as theending date of the crisis; otherwise we arbitrarily assumed that the effects ofthe crisis would last for one year after its most intense point. The bankingcrises included in the dummy are as follows (see also table 2.7):1. Austria (May 1931-January 1933): from the Creditanstalt crisis to thedate of official settlement of the Creditanstalt's foreign debt.2. Belgium (May 1931-April 1932; March 1934-February 1935): for oneyear after the initial Belgian crisis, following Creditanstalt, and for one59 Financial Crisis in the Great Depressionyear after the failure of the Banque Beige de Travail led to a generalcrisis.3. Estonia (September 1931-August 1932): for one year after the generalbanking crisis.4. France (November 1930-October 1932): for one year following each ofthe two peaks of the French banking crises, in November 1930 and Oc-tober 1931 (see Bouvier 1984).5. Germany (May 1931-December 1932): from the beginning of the majorGerman banking crisis until the creation of state institutes for the liqui-dation of bad bank debts.6. Hungary (July 1931-June 1932): for one year following the runs in Bu-dapest and the bank holiday.7. Italy (April 1931-December 1932): from the onset of the banking panicuntil the takeover of bank assets by a massive new state holding com-pany, the Istituto por le Riconstruzione Industriale (IRI).8. Latvia (July 1931-June 1932): for one year following the onset of thebanking crisis.9. Poland (June 1931-May 1932): for one year following the onset of thebanking crisis.10. Rumania (July 1931-September 1932): from the onset of the crisis untilone year after its peak in October 1931.11. United States (December 1930-March 1933): from the failure of theBank of the United States until the bank holiday.The inclusion of Austria, Belgium, Estonia, Germany, Hungary, Latvia,Poland, Rumania, or the United States in the above list cannot be controver-sial; each of these countries suffered serious panics. (One might quibble onthe margin about the exact dating given—for example, Temin [1989] and oth-ers have argued that the U.S. banking crisis did not really begin until mid1931—but we doubt very much that changes of a few months on these dateswould affect the results.) The inclusion of France and Italy is more controver-sial. For example, Bouvier (1984) argues that the French banking crisis wasnot as serious as some others, since although there were runs and many banksfailed, the very biggest banks survived; also, according to Bouvier, Frenchbanks were not as closely tied to industry as other banking systems on theContinent. For Italy, as we have noted, early and massive government inter-vention reduced the incidence of panic (see Ciocca and Toniolo 1984); how-ever, the banks were in very poor condition and (as noted above) eventuallysigned over most of their industrial assets to the IRI.To check the sensitivity of our results, we reestimated the key equationsomitting first the French crisis from the PANIC variable, then the French andItalian crises. Leaving out France had a minor effect (lowering the coefficient60Ben Bernanke and Harold Jameson PANIC and its f-statistic about 5% in a typical equation); the additionalexclusion of the Italian crisis has essentially no effect.19As a further check, we also reestimated our key equations omitting, inseparate runs, (i) the United States; (ii) Germany and Austria; and (iii) alleastern European countries. In none of these equations were our basic resultssubstantially weakened, which indicates that no single country or small groupof countries is driving our findings.The first seven equations in table 2.9 are not derived from any single model,but instead attempt to nest various suggested explanations of the link betweendeflation and depression. Estimation was by OLS, which opens up the possi-bility of simultaneity bias; however, given our maintained view that the defla-tion was imposed by exogenous monetary forces, a case can be made for treat-ing the right-hand-side variables as exogenous or predetermined.The principal inferences to be drawn from the first seven rows of table 2.9are as follows:201. Export growth consistently enters the equation for output growthstrongly, with a plausible coefficient and a high level of statistical significance.2. When wage growth is included in the output equation along with onlywholesale price and export growth (row 5), it enters with the wrong sign.Table 2.9Determinants of the Log-difference of Industrial Productiori(dependent variable: Aln/P)Independent VariablesEquation(1)(2)(3)(4)(5)(6)(7)(8)Aln/W.855(.098).531(.095).406(.121).300(.111).364(-141).351(.128).296(.123)Aln£XAlnW DISCPANICAlnMO-.0191(.0026).231(.043).148(.041).231(.046).150(.044).103(.044).217*(.048)-.0157(.0027).272(.206)-.072(.197)-.119 -.0358(.189) (.0102)-.015(.189)-.0156(.0029)-.0138(.0028)-.0126(.0031).405(.098)Note: For variable definitions, see text. The sample period is 1930-36. The panel consists oftwenty-four countries except that, due to missing wage data, Finland, Greece, and Spain areexcluded from equations (5)-(8). Estimates of country-specific dummies are not reported. Stan-dard errors are in parentheses.*Export growth is measured in real terms in equation (8).61 Financial Crisis in the Great DepressionOnly when the PANIC variable is included does nominal wage growth havethe correct (negative) sign (rows 6 and 7). In the equation encompassing allthe various channels (row 7), the estimated coefficient on wage growth is ofthe right sign and a reasonable magnitude, but it is not statistically significant.3. The discount rate enters the encompassing equation (row 7) with theright sign and a high significance level. A 100-basis-point increase in the dis-count rate is estimated to reduce the growth rate of industrial production by3.6 percentage points.4. The effect of banking panics on output is large (a year of panic is esti-mated in equation (7) to reduce output growth by 12 X .0138, or more than16 percentage points) and highly statistically significant (f-statistics of 4.0 orbetter). The measured effect of the PANIC variable does not seem to dependmuch on what other variables are included in the equation.5. There may be some residual effect of deflation on output not accountedfor by any of these effects. To see this, note that in principle the coefficient onAlnPW in equation (7) of table 2.9 should be equal to and opposite theweighted sum of the coefficients on AlnEX, AlnW, and DISC (where theweights are the projection coefficients of the respective \"true\" deflators onAlnPW). Suppose for the sake of illustration that each of the projection coef-ficients equals one (that is, the wholesale price index is the correct deflator).Then the expected value of the coefficient on AlnPW should be approximately.052; the actual value is .296, with a standard error of .123. Thus there maybe channels relating deflation to depression other than the ones explicitly ac-counted for here. One possibility is that we are simply picking up the effectsof a simultaneity bias (a reverse causation from output to prices). Alterna-tively, it is possible that an additional factor, such as debt deflation, should beconsidered.As an alternative to the procedure of nesting alternative channels in a singleequation, in equation (8) of table 2.9 we report the results of estimating thereduced form of a simple aggregate demand-aggregate supply (AD-AS) sys-tem. Under conventional assumptions, in an AD-AS model output growthshould depend on money growth and autonomous spending growth (repre-sented here by growth in real exports21), which shift the AD curve; and onnominal wage growth, which shifts the AS curve. In addition, we allowPANIC to enter the system, since banking panics could in principle affect bothaggregate demand and aggregate supply. The results indicate large and statis-tically significant effects on output growth for real export growth, moneygrowth, and banking panics. Nominal wage growth enters with the correctsign, but the coefficient is very small and statistically insignificant.We have so far focused on the effects of banking panics (and other vari-ables) on output. There is an additional issue that warrants some discussionhere; namely, the possibility that banking panics might have themselves wors-ened the deflationary process.Some care must be taken with this argument. Banking panics undoubtedly62 Ben Bernanke and Harold Jameshad large effects on the composition of national money supplies, money mul-tipliers, and money demand. Nevertheless, as has been stressed by Temin(1989), under a gold standard, small country price levels are determined byinternational monetary conditions, to which domestic money supplies and de-mands must ultimately adjust. Thus banking panics cannot intensify deflationin a small country.22 Indeed, a regression (not reported) of changes in whole-sale prices against the PANIC variable and time dummies (in order to isolatepurely cross-sectional effects) confirms that there is very little relationshipbetween the two variables.The proposition that bank panics should not affect the price level does notnecessarily hold for a large country, however. In econometric language, undera gold standard the price level of a large country must be cointegrated withworld prices; but while this means that domestic prices must eventually adjustto shocks emanating from abroad, it also allows for the possibility that domes-tic shocks will influence the world price level. Notice that if banking panicsled to deflationary shocks in a large country and these shocks were transmittedaround the world by the gold standard, a cross-sectional comparison wouldfind no link between panics and the price level.The discussion of the gold standard and deflation in section 2.2 cited Ham-ilton's (1987) view that the initial deflationary impulses in 1928-29 camefrom France and the United States—both \"big\" countries, in terms of eco-nomic importance and because of their large gold reserves. This early defla-tion obviously cannot be blamed on banking panics, since these did not beginuntil at least the end of 1930. But it would not be in any way inconsistent withthe theory of the gold standard to hypothesize that banking panics in Franceand the United States contributed to world deflation during 1931-32.23Empirical evidence bearing on this question is presented in table 2.10. Weestimated equations for wholesale price inflation in the United States andFrance, using monthly data for the five-year period 1928-32. We included anerror-correction term in both equations to allow for cointegration between theU.S. and French price levels, as would be implied by the gold standard. Thiserror-correction term is the difference between the \\og-levels of U.S. andFrench wholesale prices in period t — 1; if U.S. and French prices are in factcointegrated, then the growth rate of U.S. prices should respond negatively tothe difference between the U.S. price and the French price, and the Frenchgrowth rate of prices should respond positively. Also included in the equationsare lagged inflation rates (to capture transitory price dynamics), current andlagged base money growth, and current and lagged values of the deposits offailing banks (for the United States only, due to data availability).The results are interesting. First, there is evidence for cointegration: Theerror-correction terms have the right signs and reasonable magnitudes, al-though only the U.S. term is statistically significant. Thus we may infer thatshocks hitting either French or U.S. prices ultimately affected both price lev-els. Second, both U.S. base money growth and bank failures are important63 Financial Crisis in the Great DepressionTable 2.10 Error-correction Equations for U.S. and French Wholesale PricesDependent VariableMnUSAWPIConstantLog USAWPI - log FRAWPI(lagged once).044(t = 3.81)-.166(t = 2.77)-.530(F = 1.57;/> == .202)1.412(F = 5.62; p =.0005)MnFRAWPI-.0061.57)(t =.071(t =1.10).320(F == 2.48; p = .057).519(F == 0.78; p = .569)Four lags of own WPI growthCurrent and four lags of basemoney growthCurrent and four lags ofdeposits of failing U.S.banks, in logsR2D-W-.020(F = 5.6l;p =.0005).5311.62.3071.87Note: Deposits of failing banks are from the Federal Reserve Bulletin. USAWPI and FRAWPI arewholesale price indexes for the United States and France, respectively. Monthly data from 1928to 1932 are used.determinants of the U.S. (and by extension, the French) deflation rates; thesetwo variables enter the U.S. price equation with the right sign and marginalsignificance levels of .0005.With respect to the effect of banking panics on the price level, then, theappropriate conclusion appears to be that countries with banking panics didnot suffer worse deflation than those without panics;24 however, it is possiblethat U.S. banking panics in particular were an important source of world de-flation during 1931-32, and thus, by extension, of world depression.2.6 ConclusionMonetary and financial arrangements in the interwar period were badlyflawed and were a major source of the fall in real output. Banking panics wereone mechanism through which deflation had its effects on real output, andpanics in the United States may have contributed to the severity of the worlddeflation.In this empirical study, we have focused on the effects of severe bankingpanics. We believe it likely, however, that the effects of deflation on the finan-cial system were not confined to these more extreme episodes. Even in coun-tries without panics, banks were financially weakened and contracted their64 Ben Bernanke and Harold Jamesoperations. Domestic debt deflation was probably a factor, to a greater orlesser degree, in every country. And we have not addressed at all the effect ofdeflation on the burden of external debt, which was important for a number ofcountries. As we have already suggested, more careful study of these issues isclearly desirable.Notes1. The original diagnosis of the Depression as a monetary phenomenon was ofcourse made in Friedman and Schwartz (1963). We find the more recent work, thoughfocusing to a greater degree on international aspects of the problem, to be essentiallycomplementary to the Friedman-Schwartz analysis.2. Eichengreen and Sachs (1985) discuss several mechanisms and provide somecross-country evidence, but their approach is somewhat informal and they do not con-sider the relative importance of the different effects.3. To be clear, gold inflows to France did increase the French monetary base di-rectly, one for one; however, in the absence of supplementary open market purchases,this implied a rising ratio of French gold reserves to monetary base. Together with thevery low value of the French money multiplier, this rising cover ratio meant that themonetary expansion induced by gold flowing into France was far less significant thanthe monetary contractions that this inflow induced elsewhere.4. Temin (1989) suggests that German monetary policy provided yet another con-tractionary impetus.5. There remains the issue of whether the differences in timing of nations' departurefrom the gold standard can be treated as exogenous. Eichengreen and Sachs (1985)argue that exogeneity is a reasonable assumption, given the importance of individualnational experiences, institutions, and fortuitous events in the timing of each country'sdecision to go off gold. Strong national differences in attitudes toward the gold stan-dard (e.g., between the Gold Bloc and the Sterling Bloc) were remarkably persistentin their influence on policy.6. The countries in our sample are listed in table 2.1. We included countries forwhich the League of Nations collected reasonably complete data on industrial produc-tion, price levels, and money supplies (League of Nations' Monthly Bulletin of Statis-tics and Yearbooks, various issues; see also League of Nations, Industrialization andForeign Trade, 1945). Latin America, however, was excluded because of concernsabout the data and our expectation that factors such as commodity prices would play amore important role for these countries. However, see Campa (forthcoming) for evi-dence that the gold standard transmitted deflation and depression to Latin America in amanner very similar to that observed elsewhere.7. We define abandonment of the gold standard broadly as occurring at the first datein which a country imposes exchange controls, devalues, or suspends gold payments;see table 2.1 for a list of dates. An objection to this definition is that some countriescontinued to try to target their exchange rates at levels prescribed by the gold standardeven after \"leaving\" the gold standard by our criteria; Canada and Germany are twoexamples. We made no attempt to account for this, on the grounds that defining adher-ence to the gold standard by looking at variables such as exchange rates, moneygrowth, or prices risks assuming the propositions to be shown.8. In constructing the grand averages taken over gold and non-gold countries, if a65 Financial Crisis in the Great Depressioncountry abandoned the gold standard in the middle of a year, it is included in both thegold and non-gold categories with weights equal to the fraction of the year spent ineach category. We use simple rather than weighted averages in the tables, and similarlygive all countries equal weight in regression results presented below. This was donebecause, for the purpose of testing hypotheses (e.g., about the relationship betweendeflation and depression) it seems most reasonable to treat each country (with its owncurrency, legal system, financial system, etc.) as the basic unit of observation and toafford each observation equal weight. If we were instead trying to measure the overalleconomic significance of, for example, an individual country's policy decisions,weighted averages would be more appropriate.9. The use of the terms Ml and M2 should not be taken too literally here, as thetransactions characteristics of the assets included in each category vary considerablyamong countries. The key distinction between the two aggregates is that commercialbanks, which were heavily involved in commercial lending, were much more vulner-able to banking panics. Savings banks, in contrast, held mostly government securities,and thus often gained deposits during panic periods.10. However, it must be mentioned that recent exponents of the real wage explana-tion of German unemployment invoke it to account for high levels of unemploymentthroughout the mid and late 1920s, and not just for the period after 1929 (Borchardt1979).11. In the French case, however, there may have been some fear of governmentdefault, given the large deficits that were being run; conceivably, this could explain thehigher rate on French bills.12. A possible response to this point is that fear of devaluation added a risk pre-mium to assets in gold standard countries. This point can be checked by looking atforward rates for foreign exchange, available in Einzig (1937). The forward premia ongold standard currencies are generally small, except immediately before devaluations.In particular, the three-month premium on dollars versus the pound in 1932 had amaximum value of about 4.5% (at an annual rate) during the first week of June, but formost of the year was considerably less than that.13. The effect of deflation on banks, and the relationship between deflation andbank runs, has been analyzed in a theoretical model by Flood and Garber (1981).14. An important issue, which we cannot resolve here, is whether government take-overs of banks resulted in some restoration of intermediary services, or if, instead, thegovernment functioned primarily as a liquidation agent.15. In the next section we divide our sample into two groups: eleven countries withserious banking problems and thirteen countries without these problems. In 1930, theyear before the peak of the banking crises, the countries that were to avoid bankingproblems suffered on average a 12% deflation and a 6% fall in industrial production;the comparable numbers for the group that was to experience panics were 13% and8%. Thus, there was no large difference between the two groups early in the Depres-sion. In contrast, in 1932 (the year following the most intense banking crises), indus-trial production growth in countries without banking crises averaged -2%; in thegroup that experienced crises the comparable number was — 16%.16. Although this correlation seems to hold during the Depression, we do not wantto conclude unconditionally that branch banking is more stable; branching facilitatesdiversification but also increases the risk that problems in a few large banks may bringdown the entire network.17. Causality could run in both directions. For example, Wigmore (1987) arguesthat the U.S. banking panic in 1933 was in part created by a run on the dollar.18. It has been pointed out to us that if nominal wages were literally rigid, then thisapproach would find no effect for wages even though changes in the real wage might66 Ben Bernanke and Harold Jamesbe an important channel for the effects of deflation. The reply to this is that, if nominalwages are completely rigid, the hypothesis that real wages are important can never bedistinguished from an alternative which proposes that deflation has its effects in someother way.19. In another sensitivity check, we also tried multiplying PANIC times the changein the deposit-currency ratio, to allow for differential severity of panics. The resultsexhibited an outlier problem. When Rumania (which had a change in the deposit-currency ratio of — .76 in 1931) was excluded, the results were similar to those ob-tained using the PANIC variable alone. However, inclusion of Rumania weakened boththe magnitude and statistical significance of the effect of panics on output. The \"rea-son\" for this is that, despite its massive deposit contraction, Rumania experienced a5% growth of industrial production in 1931. Whether this is a strong contradiction ofthe view that panics affect real output is not clear, however, since according to theLeague of Nations the peak of the Rumania crisis did not occur until September orOctober, and industrial production in the subsequent year fell by 14%. Another reasonto downplay these results is that the change in the deposit-currency ratio may not be agood indicator of the severity of the banking crisis, as the Italian case indicates.20. Results were unchanged when lagged industrial production growth was addedto the equations. The coefficient on lagged production was typically small and statisti-cally insignificant.21. Deflation is by the wholesale price index.22. A possible exception to this proposition for a small country might be a situationin which there are fears that the country will devalue or abandon gold; in this case thecountry's price level might drop below the world level without causing inflows of re-serves. An example may be Poland in 1932. A member of the Gold Bloc, Poland'swholesale price level closely tracked that of France until mid 1931, when Poland ex-perienced severe banking problems and withdrawals of foreign deposits, which threat-ened convertibility. From that point on, even though both countries remained on thegold standard, money supplies and prices in Poland and France began to diverge. Fromthe time of the Polish crisis in June 1931 until the end of 1932, money and notes andcirculation dropped by 9.1% in Poland (compared to a gain of 10.5% in France); Polishcommercial bank deposits fell 24.5% (compared to a 4.1% decline in France); andPolish wholesale prices declined 35.2% (compared to a decline of 18.3% in France).Despite its greater deflation, Poland lost about a sixth of its gold reserves in 1932,while France gained gold.23. This hypothesis does not bear on Temin's claim that there was little that centralbanks could do about banking crises under the gold standard; rather, the argument isthat if, fortuitously, French and U.S. banking panics had not occurred, world deflationin 1931-32 would have been less severe.24. Indeed, if banking panics induced countries to abandon gold, they may haveindirectly contributed to an eventual rise in price levels.ReferencesBernanke, Ben. 1983. Non-monetary effects of the financial crisis in the propagationof the Great Depression. American Economic Review 73: 257-76.. 1986. Employment, hours, and earnings in the Depression: An analysis ofeight manufacturing industries. American Economic Review 76: 82-109.Bernanke, Ben, and Mark Gertler. 1990. Financial fragility and economic perform-ance. Quarterly Journal of Economics 105: 87-114.67 Financial Crisis in the Great DepressionBoard of Governors of the Federal Reserve System. 1943. Banking and monetary sta-tistics, 1919-41. Washington, DC: Government Printing Office.Borchardt, Knut. 1979. Zwangslagen und Handlungsspielraume in der grossen Wirt-schaftskrise der fruhen dreissiger Jahren: Zur Revision des uberlieferten Geschi-chtesbildes. Jahrbuch der Bayerische Akademie der Wissenschaften, 87-132. Mu-nich.Bordo, Michael, and Finn Kydland. 1990. The gold standard as a rule. Typescript,Rutgers University and Carnegie-Mellon University.Bouvier, Jean. 1984. The French banks, inflation and the economic crisis, 1919-1939.Journal of European Economic History 13: 29-80.Campa, Jose Manuel. Forthcoming. Exchange rates and economic recovery in the1930s: An extension to Latin America. Journal of Economic History.Choudhri, Ehsan U., and Levis A. Kochin. 1980. The exchange rate and the interna-tional transmission of business cycle disturbances: Some evidence from the GreatDepression. Journal of Money, Credit, and Banking 12: 565-74.Ciocca, Pierluigi, and Gianni Toniolo. 1984. Industry and finance in Italy, 1918-40.Journal of European Economic History 13: 113-36.Eichengreen, Barry. 1986. The Bank of France and the sterilization of gold, 1926-1932. Explorations in Economic History 23: 56-84.. 1987. The gold-exchange standard and the Great Depression. Working Paperno. 2198 (March). Cambridge, Mass.: National Bureau of Economic Research.Eichengreen, Barry, and T. J. Hatton. 1987. Interwar unemployment in internationalperspective: An overview. In Interwar unemployment in international perspective,ed. B. Eichengreen andT. J. Hatton, 1-59. Boston: Kluwer Academic Publishers.Eichengreen, Barry, and Richard Portes. 1986. Debt and default in the 1930s: Causesand consequences. European Economic Review 30: 599-640.Eichengreen, Barry, and Jeffrey Sachs. 1985. Exchange rates and economic recoveryin the 1930s. Journal of Economic History 45: 925-46.. 1986. Competitive devaluation in the Great Depression: A theoretical reas-sessment. Economic Letters 21: 67-71.Einzig, Paul. 1937. The theory of forward exchange. London: Macmillan.Fisher, Irving. 1933. The debt-deflation theory of great depressions. Econometrica 1:337-57.Flood, Robert P., Jr., and Peter M. Garber. 1981. A systematic banking collapse in aperfect foresight world. NBER Working Paper no. 691 (June). Cambridge, Mass.:National Bureau of Economic Research.Friedman, Milton, and Anna J. Schwartz. 1963. A monetary history of the UnitedStates, 1867-1960. Princeton: Princeton University Press.Haberler, Gottfried. 1976. The world economy, money, and the Great Depression.Washington, DC: American Enterprise Institute.Hamilton, James. 1987. Monetary factors in the Great Depression. Journal of Mone-tary Economics 19: 145-69.. 1988. The role of the international gold standard in propagating the GreatDepression. Contemporary Policy Issues 6: 67-89.Kindleberger, Charles P. 1984. Banking and industry between the two wars: An inter-national comparison. Journal of European Economic History 13: 7-28.League of Nations. 1926. Memorandum on Currency and Central Banks, 1913-1925.Geneva.. 1935. Commercial banks, 1929-1934. Geneva.. 1944. International currency experience: Lessons of the inter-war period.Geneva.Lindert, Peter. 1969. Key currencies and gold, 1900-1913. Princeton Studies in Inter-national Finance, no. 24.Newell, Andrew, and J. S. V. Symons. 1988. The macroeconomics of the interwar68 Ben Bernanke and Harold Jamesyears: International comparisons. In Interwar unemployment in international per-spective, ed. B. Eichengreen and T. J. Hatton, 61-96. Boston: Kluwer AcademicPublishers.O'Brien, Anthony. 1989. A behavioral explanation for nominal wage rigidity duringthe Great Depression. Quarterly Journal of Economics 104: 719-35.Temin, Peter. 1976. Did monetary forces cause the Great Depression? New York:W. W. Norton..1989. Lessons from the Great Depression. Cambridge, Mass.: MIT Press.Wigmore, Barrie. 1987. Was the Bank Holiday of 1933 a run on the dollar rather thanthe banks? Journal of Economic History 47: 739-56.