Contagion and Bank Failures During the Great Depression ...

Contagion and Bank Failures During the Great Depression:The June 1932 Chicago Banking Panic

By CHARLES W. CALOMIRIS AND JOSEPH R. MASON*

We examine the social costs of asymmetric-information-induced bank panics inan environment without government deposit insurance. Our case study is theChicago bank panic of June 1932. We compare the ex ante characteristics ofpanic failures and panic survivors. Despite temporary confusion about bank assetquality on the part of depositors during the panic, which was associated withwidespread depositor runs and bank stock price declines, the panic did not pro-duce significant social costs in terms of failures among solvent banks. (JEL G28,G21, N22, E58, E32)

Recent work in banking theory and historyhas helped to define the potential causes andcosts of bank panics, which various authorshave argued can be traced to speculative at-tacks on the numeraire (Barry A. Wigmore,1987; R. Glenn Donaldson, 1992), illiquidityshocks (Douglas W. Diamond and Philip H.Dybvig, 1983; Donaldson, 1993), or shocksto bank asset values when there is informationasymmetry between bankers and depositorsabout the incidence of those shocks (Calomirisand Gary B. Gorton, 1991; Calomiris andCharles M. Kahn, 1991; Calomiris and LarrySchweikart, 1991; Sudipto Bhattacharya andAnjan V. Thakor, 1993; George G. Kaufman,1994). In the latter case, when depositors can-not observe whether individual banks are sol-

* Calomiris: Graduate School of Business, ColumbiaUniversity, New York, NY 10027 and National Bureau ofEconomic Research; Mason: Bank Research Division, Of-fice of the Comptroller of the Currency, Washington, DC20219. The authors thank Richard Grossman, Ed Kane,George Pennacchi, Berry Wilson, two anonymous refer-ees, seminar participants at Columbia University, George-town University, and Universidad Torcuato di Telia, andconference attendees at the NBER Development of theAmerican Economy meetings, the American EconomicAssociation/Cliometric Society meetings, and the Finan-cial Management Association meetings for helpfulcomments. We are also grateful to Ven Vitale for researchassistance, and to the National Science Foundation (SBR-9409768) for funding. Opinions expressed are the au-thors', not those of the Office of the Comptroller of theCurrency.

vent, but can observe a shock that affectsbanks' portfolios, they may initiate runs on allbanks, both solvent and insolvent.

Bank panics are short-lived phenomena,historically measured in days or weeks. How-ever, a panic still can have important long-lived costs if it results in the disappearance ofsolvent banking institutions. That concern isoften invoked to justify the significant expan-sion of the government safety net for U.S.banks during the 1930's (Anthony J. Saundersand Berry Wilson, 1994). In this paper, wetake a close look at one of the clearest exam-ples of an asymmetric-induced bank panic dur-ing the Great Depression, and ask whethersolvent banks failed during that panic.

The answer to this question has importantpublic policy implications. Studies of bankpanics argue that panics induced by assetshocks and asymmetric information can behard to resolve with monetary policy alone. Incontrast, speculative attacks on the numeraireinduced by uncertainty about its future valuecan be stopped by policies that resolve uncer-tainty about monetary policy (for example, bya devaluation, as in the United States in March1933). Similarly, bank panics that result fromshocks to liquidity preference and a limitedsupply of aggregate liquid bank assets can beresolved by traditional monetary policy in theform of open-market purchases of securitiesby the central bank (Bruce Champ et al.,1991). But bank panics caused by asymmetricinformation about the condition of banks

863

Digitized for FRASER http://fraser.stlouisfed.org/ Federal Reserve Bank of St. Louis

864 THE AMERICAN ECONOMIC REVIEW DECEMBER 1997

cannot be resolved by monetary policy. Valuesof individual bank loan portfolios (aboutwhich uninformed depositors are concerned)are not controllable by monetary policy(Frederic S. Mishkin, 1991; Calomiris, 1994).

To resolve the problem of asymmetric-information-induced runs one must limit therisk depositors face as the result of asymmetricinformation, and thus remove the incentive fordepositors to demand immediate withdrawal.This can either be done privately or publicly.Privately, banks can either act individually toreassure depositors or agree temporarily tostand behind each other's liabilities. So longas depositors are confident that the coalition ofmutually insuring banks is solvent collec-tively, that collective action can bring thepanic to an end without resort to suspension ofconvertibility. Alternatively, the governmentcan provide insurance of deposits, either in theform of a commitment to pay depositors, or bylending cash to banks against their illiquid as-sets at a subsidized rate.

This paper addresses the empirical questionof whether private bank actions to stemasymmetric-information runs are adequate, orwhether government deposit insurance isneeded. Specifically, we ask whether privateinstitutions can prevent the failure of solventbanks during a bank panic. We examine thatquestion in the context of the banking crisesof the Great Depression. The example we fo-cus on is the Chicago panic of June 1932. Wechoose this example for three reasons. First,we argue it is a quintessential example of anasymmetric-information-induced panic. Sec-ond, this was one of the most publicizedexamples of a run on banks during the bankingcrises of the early 1930's, which coincidedwith the federal government's decisions to es-tablish the public safety net for banks. Third,by focusing on a particular location and epi-sode, we are able to clearly identify theorigins of the panic and to control for the ef-fects of location, time, and macroeconomicenvironment—factors that might otherwisecomplicate our analysis.

Our strategy in the paper is as follows. Weuse a variety of measures to judge whetherbanks that failed during the Chicago panicwere likely to have been solvent. We investi-gate whether they were predictably weaker ex

ante (and thus more vulnerable to asset pricedecline) relative to banks that survived thepanic. We employ data from individual bankfailure experience, balance sheets, income andexpense statements, and stock prices for. fail-ing and surviving Chicago banks before andduring the panic. We analyze characteristics offailing and surviving banks to determinewhether the banks that failed during the panicwere similar ex ante to those that survived thepanic. We find that panic failures were weakerthan panic survivors, and argue that panic fail-ures can be attributed to asset value decline offailed banks rather than to depositor confusionabout the value of bank assets.

While depositors did confuse panic survi-vors with panic failures, the failure of solventbanks did not result from that confusion. Onereason such failures were avoided may be thatsolvent banks knew each other's condition bet-ter than depositors, and had the incentive andthe ability to help each other avoid failure dur-ing the crisis. Private cooperation by the Chi-cago clearing house banks appears to havebeen instrumental in preventing the failure ofat least one solvent bank during the panic.

Section I provides historical background forthe Chicago panic to support our use of theJune crisis as an example of an asymmetric-information-induced bank panic, and our iden-tification of the panic event window. SectionII presents our empirical analysis of the char-acteristics of panic failures, panic survivors,and banks that failed outside the panic win-dow. Section III concludes with a summary ofour findings and a discussion of their impor-tance and limitations.

I. The June 1932 Banking Crisis in Chicago

Our discussion of the Chicago banking cri-sis establishes five "facts" that support its usefor testing the value and limitations of privatecooperation during asymmetric-information-induced panics. Together, these five facts es-tablish that the Chicago panic resulted fromlocation-specific asset shocks that were rele-vant for bank portfolios; that it was a trueasymmetric-information-induced panic in thatall banks (ex post solvent and ex post insol-vent) experienced heavy withdrawals andstock price declines during the panic; that (at


VOL. 87 NO. 5 CALOMIRIS AND MASON: BANK FAILURES DURING THE DEPRESSION

Number ofChicago BankFailures, Daily

FIGURE 1. NUMBER OF FAILED OR SUSPENDED BANKS (DAILY AND MONTHLY), JUNE 1931-DECEMBER 1932

Notes: Bank suspensions are from the Federal Reserve Bulletin, various issues. Bank failures consist of receivershipsand voluntary liquidations, and come from the Annual Report of the Comptroller of the Currency, and the StatementShowing Total Resources and Liabilities of Illinois State Banks at the Close of Business, Superintendent of Banking ofthe State of Illinois, various issues. In order to accommodate the logarithmic scale on the right-hand side, 0.1 is substitutedfor observed values of zero in the monthly data.

least in one case) banks were willing to sup-port each other against the uninformed runs ofdepositors; and that Chicago bank failures thatoccurred outside the panic window did not co-incide with similar (panic) events.

A. Was the June Crisis in Chicagoa Unique Event Nationally?

As Figure 1 shows, mid-to-late June of 1932witnessed concentration of bank failures inChicago, whether measured by the number ortotal assets of failed banks.1 The number ofbank failures in June 1932 was not particularly

1 Asset plots are available from the authors uponrequest.

high at the state, Federal Reserve District, ornational level in comparison to previousmonths. In contrast, Chicago experienced a se-vere concentration of failures during the weekof the panic. Of the 49 bank failures in the stateof Illinois during that month, 40 took place inChicago, and 26 of these failed in the week ofJune 20-27 {Commercial and FinancialChronicle, July 2, 1932 p. 71 ).2

Deposit outflows indicate a similar pattern.As shown in Figure 2, Chicago banks saw anunusually large decline in their deposits during

2 The reported * 'failure dates'' in the Commercial andFinancial Chronicle of June 20—June 27 correspond tofailure dates of June 21-June 28 reported by state andnational bank regulators.

865



Chi TotalDeposits(3MA)

NY TotalDeposits(3MA)

FIGURE 2. PERCENT CHANGE IN THE THREE-PERIOD MOVING AVERAGE OF TOTAL DEPOSITS, WEEKLY-REPORTINGBANKS IN CHICAGO AND NEW YORK CITY, JUNE 1931-DECEMBER 1932

late June, and banks in other areas of the coun-try (notably New York City, the financial cen-ter) did not share in that precipitous decline.

B. Solvent and Insolvent Banks BothSuffered During the Panic

Contemporary chroniclers and economichistorians have pointed to the June bankingcrisis in Chicago as an important example ofhow a systemwide attack by depositors onbanks can produce pressure on solvent and in-solvent banks alike. In 1932, the crisis re-ceived national, as well as local, attention inthe press. Contemporary reports clearly indi-cate that depositors ran ex post solvent as wellas ex post insolvent banks en masse.

The Commercial and Financial Chronicle(July 2, 1932 pp. 70-71) provided a detailedaccount of the runs on Chicago banks, and spe-cifically noted that even healthy banks (includ-ing, for example, First Chicago) were affected.

These reports emphasized that long lines of in-dividual depositors formed at banks. Somebanks that were reported to have experiencedlarge withdrawals (including First Chicago andContinental) were able to withstand their runsand remain open, while other banks (includingone Loop bank—the Chicago Bank of Com-merce), were forced to close.

Initially (before June 22), bank distress waslimited to a few banks, but this soon spreadand was associated with a dramatic decline inaggregate deposits in Chicago banks. The dra-matic withdrawals from downtown banks be-gan on June 22 and reached their peak onFriday, June 24. F. Cyril James (1938 p. 1034)distinguishes the panic in late June from pre-vious periods of banking distress in Chicago:

[previous] runs ... were directed againstparticular banks that were known to beenfeebled; this one was directed againstthe whole Chicago money market and

Note: There were 40 failures inChicago during June, 24 of which

occurred in the panic weekpreceding June 29


VOL. 87 NO. 5 CALOMIRIS AND MASON: BANK FAILURES DURING THE DEPRESSION 867

the First National group, in the center ofthe battle, still had more than a hundredand twenty-five million dollars of cashresources available, even though it hadpaid out fifty millions since Tuesdaynight. In the case of earlier runs, thecrowds had been drawn from a particularlocality or a special group: this time peo-ple from all parts of the city seemed toconverge on the Loop in hysterical fearand anxiety.

Bank stock prices fell during the panic, al-though not all Loop banks experienced largedeclines. Central Republic, First Chicago, andContinental were among the Loop bank survi-vors whose stock prices declined the most. FirstChicago stock fell from 150 (bid) at the closeon June 18 to a low of 131 (bid) at the closeon June 24. Continental's stock fell from 70(bid) on June 18 to a low of 60 (bid) on June27. Central Republic saw its stock price fallfrom 52 (bid, June 18 close) to 47 (bid) onJune 25, and then its stock plummeted to a priceof 4 (bid) on Monday, June 27. Afterwards, itsstock price, and those of the other survivingLoop banks, rebounded rapidly. The stock ofthe only Loop bank to fail during the panic, theBank of Commerce, was trading at 9 (bid) onJune 18-June 24 for a $20 par value. On June25 its price ceased to be reported.

C. Local Adverse Economic NewsPrecipitated the Panic

Contemporary discussion of the crisis em-phasized the adverse local economic news thathad precipitated it. The panic did not reflect ex-ogenous liquidity-demand shocks. Rather thanwithdrawing deposits to spend them, depositorsoften redeposited those withdrawals in then-postal savings accounts. James (1938) arguesthat the panic was triggered by several factors,all of which combined in rapid succession toundermine depositors' faith in the value of Chi-cago banks' assets. The bad news included fall-ing prices for local utility stocks and othercorporate assets, a well-publicized local case ofbank fraud and mismanagement, and a munic-ipal revenue crisis for the city of Chicago (Chi-cago Tribune, June 26, 1932 p. A17).

The city government's revenue problemweakened the banks in three ways. First, it

meant that the banks were forced to bear in-creased risk on their bond portfolios as theflow of coupons was interrupted. Second, Chi-cago banks were called upon to purchase illiq-uid tax warrants to help keep the municipalgovernment afloat. Third, city workers wereforced to draw down their bank deposits to paynormal living expenses, thus reducing bank re-serves and increasing the proportion of (risky)loans and securities in bank portfolios. Notsurprisingly, Chicago bankers saw the viabil-ity of the banking system and the financialproblems of the city as closely related. A del-egation of Chicago city officials and citizensvisiting Congress to request federal govern-ment assistance for the city in June 1932included many prominent bankers. The dele-gation's request for $80 million in aid was re-buffed by Congress on June 22 (ChicagoTribune, June 23, 1932 p. 1).

The municipal revenue crisis was sympto-matic of the deep contraction in local assetprices and economic activity. Among themany victims was the Insull utility empire inChicago, whose stock and debt were widelyheld by institutions and individual investors.Chicago utility companies grew dramaticallyduring the 1920's in anticipation of growingdemand and were caught short by the suddendecline in the local economy. The prices ofCommonwealth Edison (ComEd) and Insullstocks and bonds show precipitous declines inthe first six months of 1932. From Januarythrough March, Insull stock declined by 76percent, Insull bonds fell by 62 percent, andComEd stock lost 25 percent of its value. FromMarch through early June the declines accel-erated. Insull debt lost 98 percent of its value,Insull common stock lost 89 percent, and In-sull preferred lost 47 percent.

Chicago's economic problems were re-flected in several local financial disasters andcases of bank fraud (an activity traditionallymore pronounced in bad times than in good)that made front-page news in Chicago day af-ter day just prior to the crisis.3 John Bain, adefendant in the most important bank fraud

3 Milton Esbitt (1986) emphasizes the importance ofmanagement practices for explaining bank failures in Chi-cago in 1931.



case, was a local real estate developer whoowned a chain of banks. On June 9, the 12banks in the chain failed to open for business(James, 1938 p. 1033; Elmus R. Wicker, 1993p. 15). Not until June 23, however, did it be-come clear just how large the losses from fraudhad been in the Bain chain. On that date thecourt released its estimate that the value of thebanks' assets was roughly $3.5 million, com-pared to total deposits of $ 13 million (ChicagoTribune, June 24, 1932 p. 9) .

An important additional piece of evidencethat the panic was induced by fears of bankinsolvency, rather than by exogenous liquiditydemand by depositors, is that much of thefunds withdrawn from banks were redepositedimmediately in the form of riskless postal sav-ings. The Commercial and Financial Chroni-cle (July 2, 1932 p. 71) reports that, duringthe week of June 20-June 27, "The PostalSavings Department, which normally has 25or 30 windows in the Chicago post office ...increased the number to around 100 ... to ac-commodate deposits." That report noted that"... about $1,000,000 had been received [inpostal savings on June 27 ] , compared to$2,000,000 and $3,000,000 a day at [the] peak[of the crisis], and [compared to] a normaldaily average of $200,000."

In summary, by June 23, Chicago bank de-positors had witnessed, in a matter of only twoweeks, the collapse of some of the largest busi-nesses in their city, several enormously costlycases of bank fraud, and a deepening of themunicipal financial crisis as the result of thedenial of relief to their city government by fed-eral authorities. All of these stories were front-page news day after day in the two weeksleading up to the banking crisis, and the newsgrew progressively worse. In this light, it is notsurprising that depositors became increasinglyconcerned about the ability of banks to pay outtheir deposits, and transferred bank deposits toriskless postal savings accounts.

D. Interbank Cooperation Helped toPreserve Solvent Banks Under Pressure

As already noted, the Loop banks experi-enced severe stock price decline and depositwithdrawals during the crisis. Although twobanks, Central Republic and the Bank of Com-

merce, experienced unusually severe declinesin their stock prices, the Bank of Commercefailed while Central Republic survived. Its sur-vival, however, depended on cooperation bylarge Loop banks to resolve its distress.

During the crisis, Central Republic wasnearly taken down by its depositors. As doubtsabout Central Republic's solvency grew anddeposit withdrawal pressure mounted, thebank's management prepared to close the bankvoluntarily to avoid the risk of its being closedby bank depositors. Solvent banks that had lostdepositors' confidence had an incentive toclose preemptively to preserve stock value byavoiding the costs of liquidating bank assetsduring a run, and the transaction costs asso-ciated with having the bank taken over by aregulator. This incentive was particularlystrong in 1932, when bank stockholders facedthe threat of double liability on deposits, whichmeant that liquidation costs borne by stock-holders could conceivably exceed the com-plete loss of the bank's capital.

Other Chicago banks saw the prospect of Cen-tral Republic's voluntary liquidation as a poten-tial disaster for depositor confidence in theirbanks. Bankers clearly believed that depositorswere reacting to fears of bank insolvency andwere unable to distinguish between solvent andinsolvent banks. Prominent bankers from Chi-cago and New York met as a group to devise aplan to defend the Central Republic Bank andTrust Co. from the continuing withdrawals.Fearing the spillover effects of a decision to liq-uidate Central Republic, these bankers managedto persuade General Dawes (its Chair) to con-tinue operating by offering an arrangement toinfuse Central Republic with new liquidity.

The initial plan for the loan to Central Re-public provided for $10 million in back-up li-quidity from New York and Chicago banksand $80 million from the Reconstruction Fi-nance Corporation (RFC), but the final dealinvolved assistance only from Chicago banksand the RFC. The deal that emerged combinedliquidity assistance from the RFC with, in es-sence, a back-up credit enhancement by theChicago banks.4 RFC liquidity support for the

4 It is important to keep in mind that the RFC was theonly entity charged with helping avoid the insolvency of



Chicago banks—like all RFC lending duringthis period—was fully collateralized by veryhigh-quality, liquid assets; credit risk on theRFC loan to Central Republic was borne inlargest part by the Chicago banks that formedthe lending syndicate.5 Importantly, the RFCagreed to allow municipal tax warrants—$30million of which had been sold to Loop banks(Chicago Tribune, June 25, 1932 p. 6)—toqualify as collateral for its loan.6 Once the cri-sis passed, Central Republic saw its depositoutflows cease and its stock price increase.Central Republic was a solvent bank savedfrom failure by the collective intervention ofother Loop banks.

E. The June 20-27 CrisisWas a Unique Event

Having argued that the June banking crisisis an example of a panic induced by asset valuedecline and asymmetric information, we turnto the question of whether there were othersuch episodes in Chicago during the first sixmonths of 1932. That question is relevant forour discussion in Section II, where we will usethe absence of panics during that period (out-side the window of the June crisis) as an iden-tifying restriction to investigate whetherfailures during the panic were similar to fail-ures outside of the panic.

Bank failures during 1932, and more gen-erally during the Great Depression, for themost part have not been identified by histori-ans as resulting from panics or asymmetric in-

individual banks. At this time, Federal Reserve Banks didnot view the prevention of bank insolvency as their man-date. This is in sharp contrast to more recent experienceduring the 1980' s. For a review of the history of Fed lend-ing policy, see Anna J. Schwartz (1992).

5 Joseph R. Mason (1995) argues that prior to its useof preferred stock purchases to assist banks, the RFC wasnot effective in stemming bank failures. James (1938 p.1044) cites the view, common at the time, that because ofthe strict collateral requirements on RFC lending, RFCassistance often increased the credit risk faced by bankdepositors.

6 Abbreviated bank balance sheets were routinely re-ported in newspapers following call dates. Thus the June30 Reports of Condition published in the Chicago Tribuneon July 2 provided further evidence of the soundness ofChicago banks (Chicago Tribune, pp. 18-24).

formation. With the exception of the Junepanic, no contemporary chronicler or scholarof which we are aware has identified any othertime interval during 1932 as a panic or bankingcrisis in Chicago. Neither has anyone identi-fied any nationwide banking panics as havingoccurred during 1932 (Milton Friedman andSchwartz, 1963; Wicker, 1980, 1993; Ben S.Bernanke, 1983; Calomiris, 1993; CliffordThies and Daniel Gerlowski, 1993; Eugene N.White, 1984).

We have reviewed a variety of facts thatidentify the Chicago bank panic of June 1932as a quintessential example of an asymmetric-information-induced panic, resulting from lo-cal economic problems that affected bankasset values. Just as important for our pur-poses is the uniqueness of the panic window.The crisis, as reflected in sudden depositwithdrawals and stock price declines, andwidespread coverage in the local and nationalpress, was brief and was surrounded by timesin which bank failures did not coincide witha panic.

II. Failures and Survivors During the Panic

We now return to our central question ofwhether the absence of government deposit in-surance promotes the failure of solvent banksduring asymmetric-information-induced pan-ics. Specifically, we investigate whether sol-vent banks were able to survive withdrawalpressures (partly via private coordination)during the June 1932 Chicago bank panic. Toanswer that question we make use of the factthat the panic was a unique event during 1932.We assume that outside the panic window (inearly 1932) banks that failed were actually in-solvent. We then use the characteristics ofthose nonpanic failures to evaluate the likelysolvency or insolvency of banks that failedduring the panic.

If banks that failed during the panic werejust as strong (according to some set of crite-ria) as those that survived during the panic,that would provide evidence in favor of thenull hypothesis that confusion on the part ofdepositors about bank quality produced ran-dom bank failure. If, on the other hand, banksthat failed during the panic were weaker thanbank survivors, then panic failures cannot be



viewed as purely random. That implies a re-jection of the null hypothesis.

It is not difficult to reject this "strong" ver-sion of the null hypothesis, but that does notprove that only insolvent banks failed duringthe panic. Depositor confusion might have pro-duced the failure of some solvent banks, withpotentially large social costs, even if on averagebanks that failed during the panic were weakerthan those that survived it. Thus in our empir-ical work we also address a "weak" form ofthe null hypothesis—that panic-induced fail-ures were not purely random, but were impor-tantly random. This version of the nullhypothesis is inherently difficult to reject for-mally on the basis of ex ante statistical tests ofmeans, medians, and regression coefficients.Combining these with additional evidence,however, we argue that the social costs of theunwarranted closure of solvent institutions (ifany) must have been very small.

Our empirical discussion divides into twoparts. First, in subsection A we present evi-dence that leads to the rejection of the strongform of the null hypothesis. Then we confrontand reject the weak form of the null hypothe-sis, using statistical evidence as well as infor-mation from bank examiner reports.

A. Comparisons of Bank AttributesLeading Up to the Panic

We divide the Chicago banks in our sampleinto three groups: panic failures (banks thatfailed during the panic, June 20-27) , non-panic failures (banks that failed at otherdates), and survivors (banks that did not failin the first seven months of 1932). We thencompare the ex ante attributes of these threegroups.7

In our analysis of survivors, panic failures,and nonpanic failures, we focus on four exante measures of bank condition: (1) the ratio

7 The dates we choose for the panic window are con-sistent with James's (1938) discussion, newspaper ac-counts of the beginning and end of the panic, and the dailymovements of the stock prices of the ten Loop banks re-ported in the Chicago Tribune, which reached their nadiron June 27. The results are robust to reasonable alternativedefinitions of the panic window.

of the market value of equity to the book valueof equity; (2) the estimated probability of fail-ure of banks; (3) the rate of decline in bankdeposits; and (4) the interest paid on bankdebts. These various measures of bank risk areavailable for different subsets of Chicagobanks, depending on the availability of data onstock prices and interest paid on deposits.Stock prices are not available for all Chicagobanks, and interest paid is only available forFed member banks. The data set for the studyconsists of several components: balance sheetdata, income and expense data, and stock pricedata. Balance sheet data from December 31,1931 call reports were collected for all stateand national banks in Chicago, a total of 123banks. Total assets and total deposits were alsocollected for December 31, 1930, to permitcalculation of the changes in those variablesduring 1931. Balance sheet data for the 22 na-tional banks and 11 state banks that weremembers of the Federal Reserve System comefrom microfilm of the original Reports of Con-dition filed with the Office of the Comptrollerof the Currency (OCC) and the Board of Gov-ernors of the Federal Reserve System. Statenonmember bank balance sheet data are fromthe compilation of Statements of State Banksof Illinois issued by the Superintendent ofBanking of the State of Illinois. The disaggre-gated Reports of Condition of member banksfacilitated aggregation of balance sheet cate-gories to reporting standards comparable withthe Statements of State Banks of Illinois. Thestock prices for Chicago banks are end-of-month observations published in the Bank andQuotation Record (of the Commercial and Fi-nancial Chronicle). Interest payments areavailable only for Fed member banks (fromthe Reports of Condition).

1. Market-to-Book Value Ratios. Figure 3plots the means and 50-percent inclusionranges for market-to-book value ratios for thethree separate groups of Chicago banks (sur-vivors, panic failures, and nonpanic failures).To adjust for survivorship bias in plottingthese trends, we retain failed banks in our sam-ple after their date of failure, and assume thattheir stock value after failure is zero.

The striking fact illustrated by Figure 3 isthat as early as January 1931 the banks that



FIGURE 3. RATIOS OF MARKET-TO-BOOK VALUES OF EQUITY, END-OF-MONTH QUOTES, JANUARY 1931-JULY 1932

Notes: Survivorship bias can arise if banks that fail are excluded from subsequent means and standard deviations. Wecorrect for this bias by including failed banks and assuming they had zero market value of equity after the date of failure.

survived the June panic appeared to be a dis-tinct group with higher average market-to-book ratios. The banks that failed during thepanic generally had slightly higher average ra-tios than those that failed at other times, butthroughout the prepanic period (January 1,1931-June 20, 1932) the market-to-bookvalue ratios of panic failures were very closeon average to those of prepanic failures andvery different from those of panic survivors.By January 1932, most of the panic failureshad market-to-book ratios less than unity. Fig-ure 3 shows that all Chicago banks sufferedfrom capital decline during 1931 and 1932,and that the banks that failed during the Junepanic reached and maintained unusually lowmarket-to-book value ratios long before thepanic.

2. Failure Predictions. Next, we use ex anteobservable characteristics of Chicago banks(based on bank data reported in December

1931) to compute scores predicting failureduring the first seven months of 1932. Wecompare the probabilities of failure for thethree groups of banks (panic failures, non-panic failures, and survivors). Our ex antescores indicate that panic failures and non-panic failures on average were weaker banksthan survivors at least as early as the end of1931.

We estimate the probability of failure usinga logit model of the links between bank char-acteristics (e.g., balance sheet ratios) and bankfailure. The danger of using ex post failures toestimate failure risk, of course, is that specialevents with low probabilities may have influ-enced actual failure experience during the panicin ways that were unpredictable ex ante andpossibly unrelated to underlying insolvency.For example, if a panic were a common shockto all banks, then the level of reserves might doan excellent job of forecasting panic failureseven if the banks that failed during the panic

•Panic Failures

•Nonpanic Failures

Survivors



did not have higher ex ante probabilities of fail-ing. Thus using ex post panic failures to con-struct a model of ex ante bank weakness maybias toward identifying panic failures as inher-ently weak when in fact they were not. To avoid(or at least minimize) this problem, we reportlogit failure forecasts constructed from both in-sample and out-of-sample estimation.8 In theout-of-sample forecasts, we exclude banks thatfailed during the panic from the sample whenwe estimate the coefficients relating bank char-acteristics to the probability of failure. Thisconstrains the panic failures to be predicted us-ing model parameters that were constructed toexplain nonpanic failures, and thus preventsspecial unpredictable events during the panicfrom influencing predictions of failure.

Our logit results are reported in Table 1. Weinclude the following variables (all measuredat year-end 1931) in our specification: size(log of total assets), the reserve-to-demanddeposit ratio (where reserve assets are definedas cash and government securities), the realestate loan share (defined as the ratio of loanson real estate to total loans), the ratio of realestate owned to illiquid assets (which mainlyincludes repossessed real estate collateral, andexcludes bank premises), the ratio of lastyear's retained earnings to net worth, and thelong-term debt ratio (bills payable plus redis-counts plus time deposits, divided by totalassets).

This combination of variables also forms thebasis of the logit models of White (1984),David C. Wheelock (1992), Calomiris andWheelock (1995), and Mason (1995), all ofwhich are used to forecast bank failures for the1920's and 1930's. This specification typicallyis interpreted as capturing measures of each ofthe following fundamental bank characteris-tics: bank size, asset liquidity, exposure to realestate market risk, nonperforming loans (realestate owned), recent bank performance (re-tained earnings/net worth), and bank liability

8 We also estimated logit models for prepanic failuresonly (excluding the failures that occurred during or afterthe panic). The results were essentially identical to thosewe report below for nonpanic failures (which include fail-ures that occurred after the panic), and so we do not reportthem here.

composition. Bank liability composition is auseful signal of weakness because—as White(1984), Wheelock (1992) , Calomiris andWheelock (1995) , and Mason (1995)argue—reliance on high-interest, borrowedfunds was an undesirable necessity only suf-fered by higher-risk banks (see also our dis-cussion of debt composition below). Whilenot all variables prove significant in our logits,we retain measures of the basic concepts thatprevious studies have found to be importanteven if they do not prove statistically signifi-cant in our sample. Excluding them would notaffect our results. We also experimented withincluding two other variables (not reported inTable 1): the ratio of book net worth to assetsand the percent changes in deposits or assetsof banks from December 1930 to December1931. In neither case did the regressors addpredictive power to our models.

The results in Table 1 are quite similar forthe in-sample and out-of-sample specifications,which is consistent with the view that failuresduring panics were similar events to nonpanicfailures. The variable coefficients that are sig-nificant are of the expected signs. Banks withhigher reserve ratios, higher ratios of retainedearnings to net worth, and lower proportions oflong-term debt were less likely to fail.

Table 2 reports the mean and median pre-dicted failure probabilities for the logit modelsby category of bank (panic failure, nonpanicfailure, and survivor), and the significance lev-els for tests of differences across categories inmeans and medians. These results indicate thatthe banks that failed during the panic were lessrisky ex ante than banks that failed outside thepanic and more risky than survivors. Compar-isons using predicted values from in-sampleand out-of-sample regressions are similar. Byconstruction, the in-sample results show less ofa difference between panic and nonpanic fail-ures. Also by construction, out-of-sample fore-casts tend to have lower probabilities of failure.Our results are consistent with the notions thatpanic failures were much weaker banks thanpanic survivors, and that failures during thepanic were a continuation of the same processthat underlay other failures.

3. Deposit Withdrawals and Debt Composi-tion. If panic failures had been relatively weak



TABLE 1—LOGIT MODEL RESULTS

Out-of-sample In-sample

Dependent variable of all models: Bank failure N[0, 1] used for significance levels

Number of observations

Number of panic failures

Number of nonpanic failures

Log-likelihood

Restricted (slopes = 0) log-likelihood

Chi-squared statistic (k — 1 df)

Significance level

Variable name

Constant

Bank size (log of total assets)

Ratio of reserves to demand deposits

Real estate loan share

Ratio of other real estate owned to illiquid assets

Ratio of net earnings to net worth

Long-term debt

86

0

18

-18.95

-44.12

50.34

3.38E-09

CoefficientStandard

error

3.317.40

-0.75*0.54

-4.47***1.81

2.102.78

13.8912.16

-25.33***70.27

23.31***6.48

114

28

18

-51.37

-76.88

51.03

2.28E-09

CoefficientStandard

error

1.493.73

-0.200.25

-2.97***0.83

-1.391.71

2.098.25

-15.20***5.30

12.05***2.64

* Statistical significance at = 0.10.** Statistical significance at = 0.05.

*** Statistical significance at = 0.01.

institutions for months prior to the panic, thenthey should have experienced larger rates ofdepositor withdrawal before the panic. It is notpossible to obtain comparable records of de-posit accounts for failed banks after the De-cember 1931 call, but one can ask whetherpanic failures experienced relatively large de-posit withdrawals from December 1930 to De-cember 1931. Table 3 reports data on the rateof decline of deposits over that year. Clearly,panic failures and nonpanic failures experi-enced much larger withdrawals than survivorsin 1931. Panic survivors experienced an av-erage decline in deposits of 41 percent, com-

pared to 55 percent for nonpanic failures, and33 percent for survivors.

The higher rate of decline in deposits forpanic failures and nonpanic failures during1931 is reflected in the debt compositions ofthose banks. Detailed data on the compositionof bank liabilities are available for all banks inour sample, either from Federal Reserve orstate call reports. Table 3 presents data on theliability composition of banks as of December1931.

Interestingly, the shares of the various debtcategories vary systematically across the threegroups of banks. In particular, panic failures



TABLE 2—MEANS AND MEDIANS OF FAILURE PROBABILITIES, BY CLASS OF BANK

Nonpanic failures

Score

Standard error

Number

Panic failures

Score

Standard error

Number

Survivors

Score

Standard error

Number

In-sample logit

Mean

0.753

0.042

18

0.556

0.043

28

0.248

0.030

68

t-statistics for tests of differences

Panic, survivors

Panic, nonpanic

Nonpanic, survivors

5.66***

3.11***

8.13***

Median

0.796

0.033

18

0.625

0.059

28

0.187

0.055

68

4.67***

2.19**

5.61***

Out-of-sample logit

Mean

0.669

0.068

18

0.269

0.055

28

0.088

0.021

68

3.75***

4.57***

10.78***

Median

0.780

0.053

18

0.166

0.106

28

0.005

0.015

68

2.25***

4.42***

19.32***

* Significant at =0.10.** Significant at = 0.05.

*** Significant at = 0.01.

(like nonpanic failures) tend to rely muchmore on borrowed money (defined as billspayable and rediscounts). As we have notedabove, the standard interpretation of thisfinding—which is consistent with observeddifferences in deposit withdrawal rates acrosscategories during 1931 reported in thesetables—is that when demandable debt is with-drawn from risky banks, those banks areforced to rely on high-cost borrowed money,typically collateralized by liquid bank assets.As noted above, other studies have found thatthe share of borrowed money is a reliable pre-dictor of bank failure during the 1920'sand 1930's (White, 1984; Wheelock, 1992;Calomiris and Wheelock, 1995; Mason,1995). Moreover, examiners from the Officeof the Comptroller of the Currency used a re-liance on borrowed money as a clear indica-tion that a bank was having trouble. For

example, in referring to the trouble at the HydePark-Kenwood National Bank, the examinerwrote that: "... practically all of the bank'sbonds and securities are pledged [as collateralfor borrowed money] and the bank is now aheavy borrower, the Chief Examiner advisingthat the total borrowings on January 28amounted to $684,000 due to the heavy de-cline in deposits."

In summary, panic failures and nonpanic fail-ures experienced significantly larger withdraw-als of deposits long before the panic. Consistentwith this deposit withdrawal pressure, panicand nonpanic failures alike experienced fun-damental debt financing reallocations charac-teristic of troubled institutions.

4. Interest Rates on Debt. Interest rates ondebt should indicate debtholders' perceptionsof the risk of bank failure. If panic failures and



TABLE 3—DEPOSIT AND INTEREST RATE COMPOSITION, BY CLASS OF BANK

Changein

totaldeposits

Survivors

Mean -0.3251

Standard error 0.0363

Number of obs. 62

Panic failures

Mean -0.4115


Number of obs. 28

Nonpanic failures

Mean -0.5514


Number of obs. 18

Tests of differences between means (t-

Nonpanic, panic 1.777**

Panic, survivor 1.287*

Nonpanic,survivor 3.245***

Changein

totalassets

-0.2145

0.0491

63

-0.3397

0.0299

28

-0.3979

0.0242

18

statistics)

1.381*

1.637*

1.971**

Demanddeposits

0.5098

0.0226

68

0.4911

0.0314

28

0.3835

0.0336

18

2.263***

0.461

2.672***

Due tobanks

0.0301

0.0078

68

0.0216

0.0058

28

0.0053

0.0029

18

2.131**

0.667

1.618*

Timedeposits

0.4600

0.0245

68

0.4873

0.0315

28

0.6113

0.0325

18

2.629***

0.630

2.99***

Borrowedmoney

0.0197

0.0070

68

0.0831

0.0159

28

0.1872

0.0301

18

3.338***

4.249***

8.192***

Intereston total

debt

0.0062

0.0005

18

0.0093

0.0009

11

0.0116

0.0014

3

1.161

3.386***

4.201***

Notes: Deposits are presented as a proportion of total deposits, equal to demand deposits, interbank deposits, time deposits,and bills payable and rediscounts. Interest is reported as interest expense as a proportion of the relevant deposit category,i.e., demand deposit interest expense/demand deposits. Interest is calculated as the amount of interest paid over the lastsix months as a proportion of the total in each category of debt as of December 31, 1931. Changes in total assets anddeposits are from December 31, 1930 to December 31, 1931.


*** Significant at =0.01.

nonpanic failures were more likely to fail exante they should have been forced to payhigher interest on their debt prior to the Junepanic. For a small sample of Chicago banks(31) that were Fed members, we have data onthe interest paid during the last six months of1931 on each of the categories of debt dis-cussed above (individual demand deposits,bank deposits, time deposits, and borrowedmoney). We report the aggregate amounts ofthese in Table 3 as a fraction of the respectiveoutstanding debts shown on the December 31,1931 balance sheets. The banks are grouped,as before, according to their failure experi-

ence. It is important to keep in mind that ourreported interest rate differences capture theexperience of only a small sample of banks,and are measured with error because we divideinterest flows over a six-month period by end-of-year balances. Therefore, these data maynot provide an entirely accurate picture of in-terest rates paid as of December 1931.

In the column entitled interest paid on totaldebt we compute the means for each of thethree categories of banks of the ratio of totalinterest paid relative to total debt. We find thatpanic failures and nonpanic failures paid sig-nificantly higher interest rates on debt than



survivors. Panic failures paid an overall inter-est cost that was 50 percent higher than that ofsurvivors, and nonpanic failures paid nearlydouble the interest rate paid by survivors.9

In summary, we find that panic failures andnonpanic failures paid higher interest rates ontheir debt than survivors. The higher interestpaid by those two classes of banks reflectedtheir relative reliance on high-cost funds (bor-rowed money and time deposits) rather thanhigher interest costs on demand debt.

B. The Role of Declining Asset Valuesin Chicago Bank Failures

The fact that panic failures appear to havebeen stronger institutions on average than non-panic failures, as measured by their publiclyavailable financial data as of the end of 1931,has several possible explanations. One possibil-ity is the weak form of the null hypothesis (thata significant number of solvent banks failed dur-ing the panic). If one believed that bank char-acteristics (as measured in December 1931)accurately reflect unchanging cross-sectionaldifferences in bank condition throughout the pe-riod January-July 1932, then panic failures con-sequently appear excessive. That is, under theassumption of unchanging bank condition, thefact that panic failures' characteristics lie be-tween those of nonpanic failures and those of

9 We also examined the breakdown of interest paid ac-cording to each category of debt. Differences in total de-posit risk show up in withdrawals of demand depositsfrom banks, in changes in relative weights attached to var-ious types of debt, and in overall debt costs, but not indemand deposit interest rate differences. Other researchexamining links between bank-failure risk and demand de-posit interest rates during the Great Depression has alsofailed to find a positive relationship between demand de-posit interest rates and bank-failure risk. George J.Benston (1964) analyzed banks during the period 1929-1935 and found no significant positive relationship be-tween demand deposit interest rates and failure risk. As inour sample, he sometimes found a negative (and insignif-icant) relationship between the two. One explanation forthese findings is provided by Gary B. Gorton and GeorgeG. Pennacchi (1990), who argue that some bank deposi-tors may be very unwilling to accept increased risk on theiraccounts. Risk-intolerant depositors may prefer to adjustto changes in bank riskiness via changes in the quantityof balances they hold with a bank rather than changes inthe interest paid on those balances.

survivors implies that the failure process wasless discriminating during the panic—that is,that the traits of panic failures reflect a mix ofsolvent and insolvent institutions.

But such an assumption is surely incorrect.In Section I we presented evidence that localasset values (and the value of bank portfolios)declined dramatically in the first half of 1932.This implies that the failure threshold forbanks was shifting over that period. In an en-vironment of persistently declining assetprices, the first banks to fail (nonpanic fail-ures) will appear measurably weaker thanbanks that fail only after asset values havefallen much more (panic failures). To controlfor changes in the probability of failure withinour period, we estimate a survival durationmodel of bank failure. This model supports theargument that declining fundamentals can ex-plain the quality difference between (early)nonpanic failures and (late) panic failures.

Our survival duration model is similar to ourlogit model except that it forecasts the lengthof time the bank will survive (measured indays after December 31,1931), and allows forchanges in the underlying transition probabil-ities during our period, i.e., the conditionalprobabilities of failure on any given day, viaa logistic hazard function. This hazard func-tion effectively separates the effects ofchanges in the probability of failure across in-dividuals from shifts in the baseline probabil-ity of failure associated with diminishingfundamental bank asset prices (Guido W.Imbens, 1994 p. 703). The implied baselineprobability of failure estimated in our modelincreases at a decreasing rate from January toJune, and declines during July. A technicalAppendix, available from the authors upon re-quest, provides a formal discussion of our sur-vival duration model.

The results of our survival duration modelare reported in Table 4. The results are quali-tatively similar to those for the logit model inTable 1, although, of course, coefficients in thetwo models are of opposite sign. As shown inTable 4, our survival duration model is capableof approximating the gaps in time betweennonpanic failures and panic failures. Table 5illustrates that the model overpredicts survivalduration on average for both panic failures andnonpanic failures. That is, using the same scor-



TABLE 4—SURVIVAL MODEL RESULTS

Out-of-sample

Dependent variable: Log of time elapsed (in days) from December 31,

Number of observations

Number of panic failures

Number of nonpanic failures

Log-likelihood

Restricted (slopes = 0) log-likelihood

Chi-squared statistic (k - 1 df)

Significance level

Variable name

Constant

Bank size (log of total assets)

Ratio of reserves to demand deposits

Real estate loan share

86

0

18

-39.32

-67.08

55.51

3.65E-10

CoefficientStandard error

1.775.21

0.530.42

2.69***0.98

-0.611.62

Ratio of other real estate owned to illiquid assets —4.957.29

Ratio of net earnings to net worth

Long-term debt

10.52**6.07

-12.15***3.63

In-sample

1931

114

28

18

-84.36

-117.30

65.88

2.85E-12

CoefficientStandard error

4.53***1.76

0.130.12

1.44***0.39

0.580.72

0.043.33

7.41***2.40

-5.80***1.10

* Statistical significance at = 0.10.** Statistical significance at = 0.05.

*** Statistical significance at =0.01.

ing model (based on December 1931 charac-teristics), and allowing for time variation inthe hazard function, we estimate mean sur-vival duration for nonpanic failures of 192days, compared to 349 days for panic failures,while the actual respective survival meanswere 107 and 177 days. But the model accu-rately estimates the relative health of panic andnonpanic failures. Our model predicts thatnonpanic failures survive (on average) 60 per-cent as long as panic failures, and in fact theyaveraged 55 percent of the survival time ofpanic failures. Thus our model does not un-derpredict panic failures relative to nonpanicfailures, as one would expect if panic failures

were unwarranted and nonpanic failures werewarranted. In other words, when one accountsfor the time-varying probability of failure forall banks, the model does as well estimatingcross-sectional hazards during the panic asprior to the panic.

C. Further Evidence the Panic EntailedLow Social Costs

The duration survival model results areconsistent with the view that only observa-bly insolvent banks failed during the panic.But these findings do not constitute a formalrejection of the weak form of the null



TABLE 5—MEANS AND MEDIANS OF DURATION PREDICTIONS(IN DAYS FROM DECEMBER 31, 1931), BY CLASS OF BANK

Nonpanic failures

Score

Standard error

Number

Panic failures

Score

Standard error

Number

Survivors

Score

Standard error

Number

t-statistics for tests of differences

Panic, survivors

Panic, nonpanic

Nonpanic, survivors

In-sample duration

Mean

192

24

18

349

49

28

1,482

308

68

2.35***

2.46***

2.15**

Median

168

25

18

253

38

28

688

99

68

2.78***

1.66**

2.69***

Actual

Mean

107

NA

18

177

NA

28

NA

NA

68

NA

NA

NA

duration

Median

131

NA

18

177

NA

28

NA

NA

68

NA

NA

NA


*** Significant at = 0.01.

hypothesis—that some solvent banks failedduring the panic, and that the social costsfrom these failures were important. To in-vestigate that possibility, we take a closerlook at panic failures, particularly at "out-liers' ' that appear (on the basis of observabletraits in December 1931) to have beenhealthy institutions. Examination reports onthe condition of these outliers reveal deepproblems in these institutions prior to thepanic, which were publicly known but notcaptured by 1931 balance sheet ratios. Inmany cases, bank fraud and accounting ir-regularities explain why banks that failedduring the panic appear stronger statistically(on the basis of 1931 accounting data) thanthey did to contemporaries, for whom theirproblems were common knowledge by mid-1932.

Table 6 presents data on the distributions oflogit scores for the three groups of banks usingin-sample and out-of-sample estimation. Notethat none of the panic failures has an out-of-sample score that is as low (that is, as good) asthe top quartile of survivors. The minimum(best) out-of-sample score of the panic failuresis 0.00059, and the cutoff for the lowest (best)quartile of survivors is 0.00025. Only six panicfailures had out-of-sample logit scores that werelower than or equal to the median of survivors.

Were these six panic failures examples ofbanks that were solvent but allowed to fail bytheir fellow bankers? If so, were the socialcosts of those failures high? It is easier to an-swer the second question. These six banks col-lectively represented a trivial proportion of thebank assets of Chicago (1.2 percent of totalassets as of December 1931), and while it is



TABLE 6—DISTRIBUTIONAL ANALYSIS OF LOGIT SCORES FOR EX POST PANIC FAILURES AND SURVIVORS

Minimum

25th percentile

Median

75th percentile

Maximum

Minimum

25th percentile

Median

75th percentile

Maximum

In-sample p(fail),panic failures

0.048

0.448

0.626

0.701

0.905

In-sample p(fail),survivors

0.000

0.040

0.162

0.403

0.943

Percent of survivorsestimated belowthat probability

(in-sample logit)

0.311

0.782

0.910

0.932

0.988

Percent of panic failuresestimated belowthat probability

(in-sample logit)

0.000

0.000

0.038

0.222

1.000

Out-of-sample p(fail),panic failures

0.001

0.038

0.175

0.469

0.906

Out-of-sample p(fail),survivors

0.000

0.000

0.006

0.074

0.821

Percent of survivorsestimated belowthat probability

(out-of-sample logit)

0.326

0.697

0.832

0.929

1.000

Percent of panic failuresestimated belowthat probability

(out-of-sample logit)

0.000

0.000

0.139

0.300

0.918

conceivable that some of them were solventbanks, the social costs of their demise cannothave been very large.

To answer the first question we explored thespecific circumstances of these six panic fail-ures and their financial condition prior to fail-ure. We searched the records of the OCC forany relevant examinations and correspondencewith respect to the two national banks includedin the group of six (South Ashland NationalBank and Standard National Bank). We wereable to locate information about both of thesebanks. Prior to the panic, both of these bankshad experienced large loan losses and were un-der investigation by the U.S. Attorney Generaland the OCC for fraud.

The records we discovered for South Ash- land clearly indicate that this bank was on theverge of failure at least two months prior tothe panic, and possibly earlier. While the bankwas closed on June 24 and placed in the handsof a receiver on June 27, South Ashland's VicePresident, Guy Brown, had written to the OCCas early as June 2 to announce that the bankhad decided on May 25 to liquidate itself inresponse to an April 27 examination by the

Comptroller's office. That examination re-vealed that the bank had experienced large un-accounted losses that left its capital impaired,and placed it in violation of its charter. TheDeputy Comptroller wrote that "the officersand directors have been operating the bankalong unsound lines." In particular, the Dep-uty Comptroller criticized the bank's manage-ment for allowing a large loan to the bank'sDirector/Manager that produced an enormousloss for the bank. The examiner, in his May 5letter to the OCC, wrote: "This bank is nowunder the incompetent management of Direc-tor James G. Hodgkinson, who dominates thepolicies; he is absolutely broke and the mannerin which he has furthered his own interests ismost reprehensible. A report on his operationshas been made to the United States DistrictAttorney." The details of the examination re-veal fraudulent activities, including check kit-ing by Hodgkinson.

Standard National Bank was also involvedin a case of fraud. Its Vice President, who wasalso the Vice President of another bank thatfailed during the panic (People's NationalBank and Trust Co.) confessed to appropriating



bank funds to finance his speculation in thestock market. The individual and the twobanks were all under investigation by theOCC and the U.S. Attorney General as earlyas October 1931.

South Ashland and Standard are interestingexamples of banks whose accounts as of De-cember 1931 do not provide pictures of theirtrue position prior to the panic. While theirscores in our models are very strong, theircondition according to the examiners was ex-tremely weak. It is possible that some or allof the four state bank outliers may be expli-cable in similar terms. After all, if the strengthof a six-month-old balance sheet were enoughto conclusively indicate a bank's strength,asymmetric-information panics could neveroccur.

The Comptroller's examination reports indi-cate that information about fraud and risk takingby banks that failed during the panic surfacedbetween December 1931 and April 1932.Clearly, investors in bank stock detected specialproblems in the banks that failed during thepanic in those same months. As Figure 1 shows,the market-to-book value ratios for panic failuresfell sharply from January through April 1932.The market-to-book ratio of surviving banks didnot decline nearly as precipitously.

The wealth of OCC file material we discov-ered led us to search for the examination rec-ords of the other national banks in our samplethat failed during the June panic. It is difficultto quantify the statements of examiners (toconvert them to scores), but it is easy to sum-marize their content. We were able to locateinformation for all but one of the other nationalbanks that failed during the panic. In everycase for which we have records, the bank ex-aminers had indicated extreme problems at thebank at least as early as the end of April 1932.Following are excerpts from (or synopses of)the examiner's remarks about each of thesebanks:

(1) Jackson Park National Bank (Unpub-lished examination report, April 27,1932): "... the condition of this institu-tion remains highly unsatisfactory fromevery angle. It will be noted criticized as-sets have increased materially since lastexamination."

(2) National Bank of Woodlawn (Unpub-lished examination report, April 25,1932): "The report of an examination ...completed April 25 ... shows a bond de-preciation ... which greatly exceeds thebank's entire capital, surplus, and undi-vided profits ...."

(3) Bowmanville National Bank (Unpub-lished examination report, Letter of April6,1932): "The report of the examination... completed February 25 shows such anunsatisfactory condition it is requestedthat you hold a meeting with the directorsor a committee thereof and ascertainwhether or not something further can bedone to strengthen the bank, after whichthis office would be advised fully of theconclusions reached ... the solvency of[the bank] is questioned in view of thedoubtful and loss items and the bonddepreciation."

(4) Midland National Bank (Unpublished ex-amination report, April 27, 1932): "Thislittle institution is struggling along prob-ably as best as could be expected underthe circumstances ... Loss of $1,000,000in deposits within a year has just abouttaken away earning capacity." "The re-port ... shows a bond depreciation of$162,004 and losses of $16,041 in loans,which consume the surplus fund, undi-vided profits and reserve for contingen-cies and impair the bank's capital to theextent of $67,414."

(5) Hyde Park-Kenwood National Bank (Un-published examination report, February10,1932): "The report of an examinationof your bank completed December 28 ...shows an exceedingly unsatisfactory con-dition and that you are confronted with aserious situation. This conclusion is basedupon the slow and doubtful assets ...shown in the report; the fact that practi-cally all of the bank's bonds and securitiesare pledged and the bank is now a heavyborrower, the Chief Examiner advisingthat the total borrowings on January 28amounted to $684,000 due to the heavydecline in deposits."

(6) Ravenswood National Bank (Unpub-lished examination report, April 18,1932): "The report of an examination of



your bank completed April 18 ... dis-closes several unsatisfactory conditionsas set out by the examiner throughout thereport and itemized on page 11, the cor-rection of which should be effected asrapidly as possible. The report shows ahigh percentage of the bank's loans to bein very unsatisfactory condition . . . ." ByMay 9, accounting fraud was also dis-covered to have taken place by a formeremployee.

These qualitative statements about nationalbanks that failed during the panic reinforce theevidence from Tables 2 and 3 that panic fail-ures were among the weakest banks in the sys-tem at the time of the panic.10

III. Conclusion

We have compared the attributes of banksthat failed during the Chicago panic of June1932 to those of banks that failed at othertimes in early 1932, and those of banks thatsurvived the period using a variety of stan-dards of comparison. Comparisons of themarket-to-book value of equity, the estimatedprobability of failure or duration of survival,the rates of withdrawal of debt during 1931,and the interest rates paid on borrowed moneylead to the same conclusion: failures of banksduring the panic reflected the continuation ofthe same process that produced failures beforethe panic. The special attributes of failingbanks are distinguishable months before thepanic and were reflected in stock prices, failure

10 The Chicago Bank of Commerce, the largest bank tofail during the panic and the only Chicago Loop bank tofail, had an estimated probability of failure of 0.00572 inthe out-of-sample logit, and an estimated probability offailure of 0.448 in the in-sample logit. Four panic failureshad lower out-of-sample estimated probabilities of failurethan the Chicago Bank of Commerce, and seven panicfailures had lower in-sample estimated probabilities offailure. The panic failures with lower estimated failureprobabilities (for which we have examination reports)were viewed as severely troubled banks by examiners.While it is not possible to determine whether the Bank ofCommerce was solvent or insolvent during the panic usingits logit or survival scores alone, we are able to say thatits scores were not unusual relative to panic failures thatwere perceived by examiners as troubled institutions.

probabilities, the opinions of bank examiners,debt composition, and interest rates.

We conclude that failures during the panicreflected the relative weakness of failing banksin the face of a common asset value shockrather than contagion. The panic was precipi-tated by exogenous asset-price decline, and thebanks that failed during the panic were amongthe weakest banks in the city. While asym-metric information between depositors andbanks precipitated a general run on banks, ourevidence suggests that this asymmetric-information problem did not produce failuresof solvent banks.

If the risk of solvent banks failing during anasymmetric-information panic is not high (asthe evidence from the Chicago panic sug-gests), that could have important implicationsfor bank regulatory policy. Deposit insuranceand government assistance to banks since theDepression have been motivated in part by theperception that bank failures during the De-pression were a consequence of contagion,rather than the insolvency of individual banks.If private interbank cooperation, buttressed byliquidity assistance from the monetary author-ity (like the assistance provided by the RFCto the Chicago clearinghouse), are adequate topreserve systemic stability, then a far less am-bitious federal safety net might be desirable(Calomiris, 1990).

Our findings lend support to James's (1938)account of the role of interbank cooperation inmitigating the costs of the banking crisis. Thelimited duration and costs of contagion mayhave reflected the cooperative intervention bythe Chicago clearinghouse, which used its liq-uid assets to protect at least one solvent bankfrom unwarranted attack until the runs by un-informed depositors subsided. Absent such co-operation, the failure experience during thepanic of June 1932 could have been very dif-ferent. Our evidence suggests, somewhat con-trary to the portrayals in Friedman andSchwartz (1963) and Bernanke (1983), thatclearinghouses continued to serve an impor-tant function during the Great Depression, anddid not see the Fed or the RFC as ''reliev [ing]them of the responsibility of fighting runs"(Bernanke, 1983 p. 260). How far can onegeneralize from these conclusions? Was theChicago panic of June 1932 representative of



other banking panics during the Great Depres-sion? Because panics and waves of bank fail-ure were scattered across time and locationduring the Great Depression, we believe an-swering that question will require analysis ofother local panics, using detailed bank-leveldata similar to those we have analyzed for theChicago panic. Defining and analyzing thoseevents is an important area for future researchon the causes of bank failures during theDepression.

REFERENCES

Bank and Quotation Record (of the Commercialand Financial Chronicle). New York: Dana,various issues.

Benston, George J. "Interest Payments on De-mand Deposits and Bank Investment Be-havior." Journal of Political Economy,October 1964, 72(5), pp. 431-49.

Bernanke, Ben S. ' 'Nonmonetary Effects of theFinancial Crisis in the Propagation of theGreat Depression." American EconomicReview, June 1983, 73(3), pp. 257-76.

Bhattacharya, Sudipto and Thakor, Anjan V."Contemporary Banking Theory." Journalof Financial Intermediation, October 1993,3(1), pp. 2-50.

Calomiris, Charles W. "Is Deposit InsuranceNecessary?" Journal of Economic History,June 1990, 50(2), pp. 283-96.

"Regulation, Industrial Structure,and Instability in U.S. Banking: AnHistorical Perspective," in MichaelKlausner and Lawrence J. White, eds.,Structural change in banking. Home-wood, IL: Business One Irwin, 1993, pp.19-116.

.. "Is the Discount Window Necessary?A Penn Central Perspective." Review, Fed-eral Reserve Bank of St. Louis, May/June1994, pp. 31-55.

Calomiris, Charles W. and Gorton, Gary B."The Origins of Banking Panics: Models,Facts, and Bank Regulation," in R. GlennHubbard, ed., Financial markets and finan-cial crises. Chicago: University of ChicagoPress, 1991, pp. 109-73.

Calomiris, Charles W. and Kahn, Charles M.' 'The Role of Demandable Debt in Struc-turing Optimal Banking Arrangements."

American Economic Review, June 1991,81(3), pp. 497-513.

Calomiris, Charles W. and Schweikart, Larry."The Panic of 1857: Origins, Transmission,and Containment." Journal of EconomicHistory, December 1991, 57(4), pp. 807-34.

Calomiris, Charles W. and Wheelock, David C."The Failures of Large Southern BanksDuring the Great Depression." Working pa-per, Columbia University, November 1995.

Champ, Bruce; Smith, Bruce D. and Williamson,Stephen D. "Currency Elasticity and Bank-ing Panics: Theory and Evidence." Work-ing paper, University of Western Ontario,1991.

Chicago Tribune. Various issues, 1932.Commercial and Financial Chronicle. New

York: Dana, July 2, 1932, 135(1), p. 71.Diamond, Douglas W. and Dybvig, Philip H.

"Bank Runs, Deposit Insurance, and Li-quidity." Journal of Political Economy,June 1983, 97(3), pp. 401-19.

Donaldson, R. Glenn. "Sources of Panics: Evi-dence from the Weekly Data." Journal ofMonetary Economics, November 1992,30(2), pp. 277-305.

"Financing Banking Crises: Lessonsfrom the Panic of 1907." Journal of Mon-etary Economics, February 1993, 31(1),pp. 69-95.

Esbitt, Milton.' 'Bank Portfolios and Bank Fail-ures During the Great Depression: Chi-cago." Journal of Economic History, June1986, 46(2), pp. 455-62.

Federal Reserve Bulletin. Washington, DC:Board of Governors of the Federal ReserveSystem, various issues.

Friedman, Milton and Schwartz, Anna J. A Mon-etary History of the United States, 1867-1960. Princeton, NJ: Princeton UniversityPress, 1963.

Gorton, Gary B. and Pennacchi, George G. ' 'Fi-nancial Intermediation and Liquidity Crea-tion." Journal of Finance, December 1990,45(1), pp. 49-72.

Imbens, Guido W. "Transition Models in aNon-Stationary Environment." Review ofEconomics and Statistics, November 1994,96(4), pp. 703-20.

James, F. Cyril. The growth of Chicago banks,volume 11. New York: Harper, 1938.



Kaufman, George G. "Bank Contagion: A Re-view of the Theory and Evidence." Journalof Financial Services Research, April 1994,8(2), pp. 123-50.

Mason, Joseph R. "Determinants and Effects ofRFC Lending to Banks During the Great De-pression." Working paper, Office of theComptroller of the Currency, November 1995.

Mishkin, Frederic S. "Anatomy of a FinancialCrisis." National Bureau of Economic Re-search (Cambridge, MA) Working PaperNo. 3934, December 1991.

Office of the Comptroller of the Currency. Un-published examination reports and letterspertaining to national banks in Chicago,1931 and 1932.

Annual report. Washington, DC:U.S. Government Printing Office, variousissues.

Reports of Condition of (Federal Reserve Member)Banks. Unpublished records, various calls.

Saunders, Anthony J. and Wilson, Berry. "Con-tagious Bank Runs: Evidence from the1929-1933 Period." Working paper, NewYork University, 1994.

Schwartz, Anna J. "The Misuse of the Fed'sDiscount Window." Review, Federal Re-serve Bank of St. Louis, September/Octo-ber, 1992, 74(5), pp. 58-69.

Superintendent of Banking of the State of Illinois.Statement showing total resources and lia-bilities of Illinois state banks at the close ofbusiness. Springfield, IL: Journal PrintingCompany, various issues.

Thies, Clifford and Gerlowski, Daniel. "BankCapital and Bank Failure, 1921 -1932: Test-ing the White Hypothesis." Journal of Eco-nomic History, December 1993, 53(4), pp.908-14.

Wheelock, David C. "Deposit Insurance andBank Failures: New Evidence from the1920s." Economic Inquiry, July 1992,30(3), pp. 530-43.

White, Eugene N. "A Reinterpretation of theBanking Crisis of 1930." Journal of Eco-nomic History, March 1984, 44(1), pp .119-38.

Wicker, Elmus R. "A Reconsideration of theCauses of the Banking Panic of 1930."Journal of Economic History, September1980, 40(3), pp. 571-83.

"A Reconstruction of the Gold andBanking Crises in 1931." Working paper,Indiana University, October 1993.

Wigmore, Barry A. "Was the Bank Holiday of1933 Caused by a Run on the Dollar?"Journal of Economic History, September1987, 47(3), pp. 739-56.


Contagion and Bank Failures During the Great Depression ...

Documents

Transcript of Contagion and Bank Failures During the Great Depression ...