MONETARY INTERVENTION REALLY DID MITIGATE BANKING PANICS

DURING THE GREAT DEPRESSION: EVIDENCE ALONG THE ATLANTA

FEDERAL RESERVE DISTRICT BORDER

By

Andrew J. Jalil+

Acknowledgments. I am indebted to J. Bradford DeLong, Barry Eichengreen, Christina D.

Romer, David H. Romer and James A. Wilcox for their continuous guidance, advice and support. I am grateful to Lauren Bloomquist, Thomas Verghese and Joseph Warren for exceptional research assistance and to numerous seminar participants and the referees for helpful comments and suggestions.

+ Department of Economics, Occidental College, 1600 Campus Road, Los Angeles, CA 90041. Email: [email protected]. MONETARY INTERVENTION REALLY DID MITIGATE BANKING

PANICS DURING THE GREAT DEPRESSION: EVIDENCE ALONG THE

ATLANTA FEDERAL RESERVE DISTRICT BORDER

Abstract

Throughout the Great Depression, the Federal Reserve Bank of Atlanta aggressively

intervened to stabilize its banking system. To assess these policies, I analyze the

performance of banks along counties straddling the border of the Atlanta Fed. My results

indicate that expansionary initiatives designed to inject liquidity into the banking system

reduced the incidence of bank suspensions by 34 to 72 percent in some regions.

Moreover, an analysis of the balance sheets of individual Federal Reserve Districts

suggests that liquidity intervention did not expend large resources.

“Why was monetary policy so inept?...The monetary system collapsed, but it clearly need not have done so….pursuit of the policies outlined by the System itself in the 1920’s, or for that matter by Bagehot in 1873, would have prevented the catastrophe.”

[Friedman and Schwartz, A Monetary History of the United States, 1867-1960, p. 407]

Introduction

The causes of bank failures during the Great Depression remain one of the most important issues scholars face in attempting to explain the severity of the contraction of 1929-

1933. Two schools of thought dominate the debate. Milton Friedman and Anna J. Schwartz

(1963), argue that the massive waves of banking panics reflected liquidity crises. As fear spread throughout the country, deposit withdrawals accelerated and bank runs became self-fulfilling panics, forcing many solvent but illiquid institutions to fail or suspend operations. By contrast,

1 advocates of the second school of thought, such as Peter Temin (1976), contend that the banking panics were crises of fundamental solvency. Mounting default rates, deteriorating bank loan portfolios, and a downward trend in the value of bank assets contributed to a decline in the fundamental health of banks, undermining the solvency of financial institutions throughout the country. According to this view, banks failed because they were insolvent, rather than simply illiquid.1

Two recent studies have made substantial headway in empirically testing these competing claims. Charles W. Calomiris and Joseph R. Mason (2003) use a disaggregated panel data set of individual bank characteristics to model the determinants of bank failure risk. They conclude that fundamentals describe bank failure risk well and that, as a consequence, Friedman and Schwartz exaggerated the extent to which the banking panics of 1929-1933 were liquidity crises.

According to their analysis, efforts by the Federal Reserve to act as a lender of last resort to provide liquidity assistance to struggling banks would have proven ineffective.

Gary Richardson and William Troost (2009) analyze the performance of banks in

Mississippi during the depression. The Federal Reserve Act of 1913 separated the southern half of Mississippi from its northern half. The southern half fell under the jurisdiction of the Atlanta

Federal Reserve District, whereas the northern half fell under the St. Louis Federal Reserve

District. These two districts adopted divergent policies in response to banking panics. The

Atlanta Fed aggressively intervened to stabilize its banking system, adhering to a principle known as Bagehot’s Rule. This rule dictated that during times of panic, the central bank should act as a lender of last resort, extending credit to banks and rushing cash to ailing financial institutions. By contrast, the St. Louis Fed favored nonintervention in the midst of panics. It was a staunch advocate of the Real Bills doctrine, a principle that called for a contraction of the supply of credit during recessions and limited intervention in the face of banking panics. Richardson and Troost track bank survival rates in Mississippi during the major banking panic of 1930 and conclude that bank performance was far superior in the portion of Mississippi that fell under the jurisdiction of

2 the Atlanta Federal Reserve District. As a consequence, Richardson and Troost conclude that liquidity intervention mitigated banking panics during the depression.2

Both studies provide first-rate analysis. Yet, because these two studies offer contradictory policy implications, it is difficult to determine the extent to which liquidity intervention could have mitigated the banking panics of the Great Depression. Moreover, both studies have limitations. As they acknowledge, Calomiris and Mason confine their analysis to

Fed member banks although most failing banks were nonmembers. Due to data deficiencies, they rely on biennial data (available on two dates: Dec 31, 1929 and Dec 31, 1931) in assembling their bank-specific characteristics—a constraint that may render their data unrepresentative of the health of institutions at the timing of failures. Regarding the second study, Richardson and

Troost limit their analysis to one state—Mississippi—raising the possibility that some force other than Fed policies could be driving their results and generating concerns about the external validity of their findings.

A New Test

The debate over the nature of the banking panics of the Great Depression continues.

Economists remain divided as to whether liquidity intervention could have alleviated the nation’s banking panics. A new test—one that could provide additional evidence to evaluate the merits of

Friedman and Schwartz’s classic claim that the Federal Reserve could have stabilized the country’s banking system had it acted as a lender of last resort—would be beneficial for this debate. As a consequence, I propose a new test in this paper—one that is similar in nature to that proposed by Richardson and Troost, but more expansive in geographic scope.

Throughout the course of the depression, the Federal Reserve Bank of Atlanta championed monetary activism and aggressively intervened to support its banking system. In the midst of panics, the Atlanta Fed acted as a lender of last resort, extending credit and rushing money to ailing financial institutions. Richardson and Troost (2009) draw on myriad sources to

3 document the policy regime of the Atlanta Fed. They cite memos from Eugene Black, the

Governor of the Atlanta Fed, showing that agents rushed large sums of cash to banks during the panic of 1930—a panic that broke out following the collapse of Caldwell and Company, an investment banking firm that controlled the largest financial conglomerate in the South.

Richardson and Troost also cite Richard Gamble’s History of the Federal Reserve Bank of

Atlanta, 1914-1989 to document that such interventionist actions were not isolated instances and that the Atlanta Fed continually responded to banking panics with expansionary measures.

Moreover, they document that these policies predated the Great Depression and that the Atlanta

Fed adhered to Bagehot’s Rule as far back as the early 1920s.

Richardson and Troost (2006) and Friedman and Schwartz (1963) document that, by contrast, the Federal Reserve System as a whole did not intervene to support its banking system.

Fed officials from other districts espoused liquidationist ideologies, believing that occasional recessions helped purge economies by removing inefficient firms and institutions from the market. According to this view, expansionary monetary initiatives would impede these healthy market corrections and hinder economic recovery. As a consequence, the Federal Reserve

System as a whole did not intervene to support the country’s banking system.

Richardson and Troost describe in detail how the St Louis Fed adhered to the Real Bills doctrine and actively opposed intervention during the banking panic in the fall of 1930. Lyster V.

Chandler (1971) provides a similar account of the St. Louis Fed. Chandler (p. 142) reports a speech by William McC. Martin, Sr., Governor of the St. Louis Federal Reserve Bank, indicating his opposition to monetary intervention:

I cannot see how the situation can be benefited by putting fifty millions of dollars, or, in

fact, any other amount, into the general market at this time…The reason that more money

is not being used is because it is not needed, and when there is already sufficient money

to meet the expressed needs, it seems to me unwise artificially to add to the amount

already sufficient in order to encourage a use which because based on a redundancy of

money rather than on actual needs may be hazardous.

4 Chandler extends this characterization to most of the other Federal Reserve Districts, including those that shared a border with Atlanta. For example, Chandler (p. 139) describes Lynn P.

Talley, Governor of the Dallas Fed, as “yet another opponent of actively expansionary policies” and George J. Seay, Governor of the Richmond Fed, as espousing views similar to those of

Martin.3 David Wheelock (1991, p. 100) corroborates these accounts, noting that “Within the

Reserve System there remained considerable acceptance of the Real Bills view that the supply of credit should decline during recessions, because lacking ‘productive’ outlets an excess supply would generate speculation or inflation.” Wheelock cites correspondence from the Richmond

Fed, indicating its adherence to the Real Bills view, as well as correspondence from other Federal

Reserve Districts.4

Given this diversity in policy regimes, I propose a new test of the effects of monetary intervention on bank performance during the Depression. Specifically, I analyze the performance of banks in counties located within 50 miles of the Atlanta Federal Reserve District. The Atlanta

Fed (District 6) shared a border with four other Federal Reserve Districts: Richmond (District 5),

St. Louis (District 8), Cleveland (District 4) and Dallas (District 11).

The logic of my approach is as follows. Due to geographic proximity, neighboring counties should share a similar propensity to suffer from a depositor panic. However, by restricting my analysis to counties straddling the Atlanta Federal Reserve District border, I isolate counties that are geographically close to one another, but that are under the jurisdiction of different Fed policy regimes. The Atlanta Fed aggressively intervened to stabilize its banking system by rushing cash to banks undergoing runs and acting as a lender of last resort. The other districts did not. Consequently, an analysis of bank performance in counties along the Atlanta

Federal Reserve District border should shed insights into the effects of monetary intervention on banking panics. For example, if the Atlanta Fed’s interventionist policies had no effect on outcomes, then bank suspension rates in counties under the jurisdiction of the Atlanta Federal

Reserve District should be similar to bank suspension rates in neighboring counties across the Fed

5 border. On the other hand, if the Atlanta Fed’s interventionist policies had substantial effects, then bank suspension rates should be lower in counties under the jurisdiction of the Atlanta

Federal Reserve District than bank suspension rates in neighboring counties across the border.

To conduct this test, I construct a panel regression model. The sample includes all counties located within 50 miles of the Atlanta Federal Reserve District border. I index observations by county and year and I divide observations into distinct geographic regions and monetary regimes. Since the Atlanta Federal Reserve District shared its border with four other

Federal Reserve Districts, I define four geographical regions: Atlanta-Richmond, Atlanta-

St.Louis, Atlanta-Cleveland and Atlanta-Dallas. Each county is allocated to one geographic region, based on closest proximity. Figure 1 presents these regions. A “monetary regime” includes all of the counties under the jurisdiction of a particular policy regime. One monetary regime—the Atlanta Federal Reserve District—aggressively intervened to stabilize its banking system, whereas the other districts did not. Therefore, I formulate the following linear model:

4 1933 1933

Sit = ∑ ∑ β jt (Rij Dt )+ ∑ αt (AiDt )+ Ψi +εit j=1 t=1927 t=1927

where Sit represents bank suspension rate in county i in year t, Rj represents geographical region dummies (j=1 Atlanta-Richmond, j=2 Atlanta-St. Louis, j=3 Atlanta-Cleveland, j=4 Atlanta-

Dallas), A represents a dummy variable for the Atlanta Federal Reserve District, Dt represents yearly dummies (t=1927, 1928, 1929, 1930, 1931, 1932, 1933), and Ψ represents time-invariant county-level controls.5

I interact region and year dummies to capture region-year fixed effects. These interaction terms measure the average level of bank suspension rates among counties in a region in a given year. They reveal the extent to which banking panics gripped a particular region in a given year.

Moreover, I interact the Atlanta Fed dummy with year dummies to capture monetary regime-year fixed effects. These interaction terms measure the average effect on bank suspension rates of being in the Atlanta Federal Reserve district in each year. Hence, they provide yearly estimates

6 of the effectiveness of monetary intervention in mitigating banking panics. Lastly, to control for economic, demographic and regulatory differences across counties, I include a set of county-level control variables.6

Table 1 presents descriptive data on the sample size. It reports the number of counties, the number of banks in operation on January 1, 1927, and the number of bank suspensions in each year of the contraction, along with the two previous years to identify any potential trends in bank suspensions in the run-up to the depression. The sample includes 356 counties (166 in Atlanta and 190 in non-Atlanta) and 1614 banks (682 in Atlanta and 932 in non-Atlanta).

Table 2 reports economic, demographic and regulatory characteristics of counties for the four geographical regions. The table separates each region into its Atlanta and non-Atlanta component and reports averages across counties. The table reveals a large degree of homogeneity across each region along a variety of dimensions.

Table 3 presents the regression results. It reports two specifications. Regression 1 presents the full model with county-level economic, demographic, and regulatory control variables. Regression 2 excludes the county controls. The results are similar across the two specifications.

First, consider the results before the start of the depression—1927 and 1928. In both regressions, the region-year fixed effects are small and in most cases, statistically insignificant, indicating that these regions were not seriously affected by banking panics. The Atlanta Fed fixed effects are also small and statistically insignificant, indicating that bank suspension rates did not differ substantially across opposite sides of the Atlanta Fed border in the run-up to the depression. These findings suggest that bank performance was relatively homogenous along opposite sides of the Atlanta Fed border before the start of the depression.

Next, consider the results during the Great Contraction. The region-year fixed effects increase substantially during the contraction. In both regressions, the region-year fixed effects show that during the early stages of the contraction—1929 to 1930, banking panics were most

7 severe in Atlanta-Richmond and Atlanta-St. Louis in 1929 and 1930 and in Atlanta-Cleveland in

1929. In 1931 and 1932, region-year fixed effects are large and significant in most regions, but in

1933, drop to levels that are, in general, insignificantly different from zero.

The main result from the empirical model, however, is the Atlanta Fed fixed effects. In both regressions, in 1929 and 1930, the early stages of the depression, the Atlanta Fed fixed effects are large, negative and strongly significant, indicating that being under the jurisdiction of the Atlanta Federal Reserve District reduced the incidence of banking panics. These results support the hypothesis that monetary intervention mitigated banking panics. According to these results, by the end of 1930, the percentage of bank suspensions in counties located in the Atlanta district were, on average, roughly 10 to 12 percentage points lower than the percentage of bank suspensions in neighboring counties across the border in the St. Louis, Richmond, Cleveland, or

Dallas district. Moreover, the magnitude of these results is comparable to those of Richardson and Troost (2009). Richardson and Troost show that approximately 75 percent of the banks in the St. Louis district of Mississippi were still in operation by mid 1931, whereas approximately

85 percent of the banks in the Atlanta district of Mississippi were still in operation by mid 1931— a difference of roughly 10 percent.7

During the later stages of the contraction, in 1931, 1932 and 1933, the Atlanta Fed fixed effects are statistically indistinguishable from zero, indicating that bank suspension rates did not differ substantially across opposite sides of the Atlanta Fed border. Thus, while the results from these regressions suggest that monetary intervention alleviated banking panics during the early stages of the depression, they also indicate that bank suspension rates began to converge in 1931 and 1932, in spite of high levels of bank suspension rates. What explains this convergence?

There are several potential explanations. One explanation is that liquidity intervention became less effective over time. As the severity of the depression intensified between 1929 and 1933, the nature of the banking panics might have changed. Indeed, as economic conditions and fundamentals deteriorated over the course of the depression, the banking panics, which initially

8 began as crises of illiquidity, might have transformed into crises of fundamental solvency during the later stages of the depression. This might—at least in part—reconcile the findings of

Richardson and Troost (2009) with those of Calomiris and Mason (2003).8 However, another explanation is that other organizations—such as the National Credit Corporation and the

Reconstruction Finance Corporation, which began lending money to banks in 1931 and 1932, respectively—smoothed out differences in credit lending to banks across Federal Reserve

Districts during the later stages of the depression. Furthermore, a third explanation is that other

Federal Reserve Districts, which initially opposed intervention, moved closer to the policies of the Federal Reserve Bank of Atlanta over time. Richardson and Troost (2009) document that in response to the successes of the Atlanta Fed, the St. Louis Fed, beginning in the summer of 1931, started to adopt policies similar to those of the Atlanta Fed. Future research is needed to sort out these competing potential explanations.9

These results strongly support the hypothesis that monetary intervention mitigated banking panics during the early stages of the depression. To gauge the magnitude of these effects, consider the following calculation. Dividing the Atlanta Fed fixed effects, which measure the impact on bank suspension rates of being in the Atlanta Federal Reserve District in each year, by the region-year fixed effects, which measure the average level of bank suspension rates among counties in a region in a given year, provides estimates of the extent to which monetary intervention reduced the incidence of bank suspensions. For example, the results from regression

2 indicate that the Atlanta Fed reduced the extent of bank suspensions by approximately 43 percent in 1929 and 40 percent in 1930 in Atlanta-Richmond and by approximately 48 percent in

1929 and 34 percent in 1930 in Atlanta-St. Louis. Using the results from regression 1 yields even stronger statistics: these estimates indicate that after controlling for economic, demographic and regulatory differences across counties, the Atlanta Fed reduced the extent of bank suspensions by approximately 51 percent in 1929 and 57 percent in 1930 in Atlanta-Richmond and by approximately 72 percent in 1929 and 66 percent in 1930 in Atlanta-St. Louis.

9 The Impact of Monetary Intervention on the Fed Balance Sheet

The results indicate that monetary intervention—on the part of the Atlanta Fed— mitigated banking panics during the early stages of the contraction. This suggests that interventionist policies from the Federal Reserve System as a whole would have alleviated the banking panics of the initial stages of the depression.

However, this finding poses a related question: would such a concerted system-wide response have been feasible? The United States was on a fixed exchange rate regime, the gold standard, during the Great Contraction. As a consequence, a potential constraint on Fed policy might have been its reserve position. If liquidity intervention expended large resources, then it might be the case that all twelve Federal Reserve Districts would not have been able to extend support to their banking systems without jeopardizing the reserve position of the Federal Reserve

System as a whole.10

To investigate this issue further, I analyzed the balance sheet of the Atlanta Fed to determine whether its actions impaired its reserve position. Figure 2 displays the reserve position of the Atlanta Fed from July 1, 1929 to December 31, 1930. It shows the ratio of total reserves against note and deposit liabilities—the reserve ratio. For sake of comparison, I include the reserve position of two other Federal Reserve Districts that experienced serious banking panics during the first two years of the depression, but that did not intervene to support their banking systems—St. Louis and Richmond.11

The Federal Reserve Act required each Reserve bank to hold reserves in the form of gold and lawful money equal to at least 35 percent of its deposit liabilities and reserves in the form of gold equal to at least 40 percent of its outstanding Federal Reserve notes; as a consequence, the average reserve requirement against Federal Reserve notes and deposits was between 35 and 40 percent. Figure 2 reveals that the reserve position of the Atlanta Fed remained substantially above this requirement throughout 1929 and 1930. This suggests that monetary intervention did not impair the balance sheet of the Atlanta Fed during the early stages of the depression.12

10 What explains the strong reserve position of the Atlanta Fed in the midst of such large- scale intervention? To answer this question, I read contemporary newspaper reports that described the banking panics that occurred in the Atlanta Federal Reserve District in 1929 and

1930. In July 1929, during the outbreak of a panic in Florida, aggressive efforts to rush cash to banks undergoing runs and strong declarations from Fed officials restored confidence and quelled the panic. Creed Taylor, Deputy Governor of the Atlanta Federal Reserve Bank, proclaimed

“local bankers could ‘have all the money with which they need to meet the situation’” and The

New York Times reported, “confidence had been restored…the arrival of $5,000,000 here today and yesterday from the Atlanta Federal Reserve Bank and the sight of the money in huge stacks in the cages of the bank tellers had a reassuring effect.”13 In November 1930, during the panic that coincided with the collapse of Caldwell and Company, the newspaper reports indicate that the actions taken by Atlanta Fed officials—of rushing cash and pledging support to banks— reassured depositors, inducing them to redeposit their money. The New York Times reported,

“Many depositors who had withdrawn their money from the East Tennessee National and the City

National Banks redeposited it today and business was again almost normal…huge stacks of cash and currency [from the Federal Reserve Bank of Atlanta] reassured the most timid.”14 The newspaper accounts indicate that the Atlanta Fed was able to rush cash to banks without impairing its reserve position because its promise to extend aid to banks, coupled with visible displays of rushing cash to banks, had a powerful effect in restoring confidence.

Conclusions

This paper extends the analysis of Richardson and Troost (2009) beyond Mississippi to a larger region—the entire border of the Atlanta Federal Reserve District. It directly tests a chief concern of the work of Richardson and Troost—the possibility that the results may be specific to

Mississippi and consequently, may be driven by some force other than differences in monetary policy. By analyzing bank performance in counties straddling the border of the Atlanta Federal

11 Reserve District during the Great Depression, this paper conducts a new test of the effects of liquidity intervention on bank performance in a more expansive range of regions. According to my empirical estimates, liquidity intervention reduced the incidence of bank suspensions by 34 to

72 percent, depending on the model specification, across a wide array of areas. These findings corroborate those of Richardson and Troost and support the assertion that the Federal Reserve

System could have mitigated the banking panics of the early stages of the depression had it acted as a lender of last resort. They also indicate that the conclusions of Richardson and Troost are not specific to one state, but rather are generalizable to a much larger, more diverse area.15

Would monetary intervention during the early stages of the depression have been sufficient to prevent subsequent panics? According to Friedman and Schwartz (1963), the banking panics of the Great Depression were liquidity crises. In their view, the Federal Reserve could have contained the banking panics had it rushed cash to banks undergoing runs and injected liquidity into the country’s banking system. Unfortunately, because the counterfactual—bank performance between 1929 and 1933 with the Federal Reserve System acting as a lender of last resort—is unobservable, it may be impossible to know with certainty whether monetary intervention would have been sufficient to stop a nationwide banking crisis and avert the worst of the depression. Nonetheless, the results of this paper suggest that monetary intervention was strongly effective during the early stages of the depression. As a consequence, the failure of the

Federal Reserve System to act as a lender of last resort during the initial stages of the depression was a squandered opportunity.

12 Endnotes

1 The second school has two variants: bank insolvency as a result of bad conditions versus insolvency due to bad lending. Temin (1976) is an advocate of the former variant.

2 In related work, Carlson, Mitchener, and Richardson (2011) describe how liquidity intervention—on the part of the Atlanta Fed—helped arrest a banking panic in Florida in 1929.

3 While Chandler makes it clear that there were important differences among the individual Federal

Reserve Districts, Chandler describes the Federal Reserve Banks, except for those of Atlanta and New

York, as in general being opposed to aggressive expansionary initiatives. Chandler summarizes, “With ideas such as those described above so widely prevalent within the System, it is no wonder that the Reserve banks other than New York and Atlanta were unwilling to support positive and aggressive expansionary policies, including prompt and sharp reductions in discount and bill-buying rates and large purchases of government securities” (p. 142).

4 For a detailed discussion of the views at the various Federal Reserve Banks, see Chandler (1972), Meltzer

(2003), and Wheelock (1991).

5 The data on bank suspensions come from a study conducted by the Federal Deposit Insurance Corporation

(FDIC). This data has been transferred to computer format by the University of Michigan’s Inter-

University Consortium for Political Science Research (ICPSR). The FDIC data set contains county level, annual data that reveal the number of banks in operation, the number of banks suspending, and the total number of deposits from 1920 to 1936. Two caveats accompany the use of the FDIC data. First, the FDIC data reports bank suspension rates—not failure rates. It includes both permanent closures (failures) and temporary closures (suspensions). The distinction could potentially be important for the empirical findings that follow since not all banks that suspended operations ultimately failed or were liquidated. For example, it could be possible that actions taken to contain banking panics reduced suspension rates, but not failure rates. Second, not all bank suspensions (i.e. the closing down of an unsound bank by a bank examiner or a voluntary liquidation) necessarily accompany banking panics.

6 The economic and demographic county-level control variables come from the census of population, manufacturing, and agriculture for 1930 and include total population, persons per square mile, fraction of population in labor force, urban population share, unemployment rate, size of farms, number of

13 manufacturing establishments, percentage of cropland with failures, and ratio of debt to farm value. The regulatory control variables come from Polk’s Bankers’ Encyclopedia (1929) and the Federal Reserve

Bulletin (November 1928, September 1930) and include state minimum capital and reserve requirements.

7 These results suggest that monetary intervention mitigated banking panics during the early stages of the depression. However, one potentially complicating factor is the collapse of Caldwell and Company, which occurred in the Atlanta Federal Reserve District portion of Tennessee in 1930 and could bias the results against the Atlanta Fed. Richardson and Troost (2006), for example, mention that they focus on Mississippi rather than Tennessee—both of which are divided by the Atlanta Fed District border—in part because the collapse of Caldwell and Company occurred in the 6th district portion of Tennessee and could bias the results against the Atlanta Fed in that area. Thus, the strength of these results—in spite of the collapse of

Caldwell and Company—is noteworthy. Nonetheless, to check whether this source of bias might be present, as a robustness check, I re-estimate the preceding regressions, but exclude the area surrounding the collapse of Caldwell and Company—specifically, all counties in a Tennessee Atlanta versus non-Atlanta pairing in 1930—that is, all counties in Tennessee and all counties whose closest neighbor in the Atlanta

Federal Reserve District is Tennessee. The monetary regime-year and region-year estimates for 1927-1929 and 1931-1933 are virtually unchanged from the previous regressions. However, the Atlanta Fed fixed effect for 1930 increases dramatically to -0.099 (standard error = 0.029) for regression 1, which includes county-level control variables, and to -0.085 (standard error = 0.028) for regression 2, which omits the control variables. These results indicate that this source of bias is likely real: excluding the area surrounding the collapse of Caldwell and Company roughly doubles the Atlanta Fed fixed effect in 1930.

8 This would also indicate that the scope for action from the Federal Reserve—via liquidity support–was strongest during the initial stages of the depression.

9 An additional finding that merits attention involves the regulatory control variables from regression 1.

The coefficients on the capital and reserve requirement variables are significant and indicate that an increase in capital requirements reduces bank suspension rates, but that an increase in reserve requirements increases bank suspension rates. The direction of these effects is the same as those found in Mitchener

(2005). Thus, these results along the Atlanta Fed border corroborate his findings. Most importantly, these results indicate that even after controlling for differences in capital and reserve requirements, the Atlanta

14 Fed fixed effects are strongly significant during the early stages of the depression—evidence that monetary intervention mattered. For a detailed discussion of the role of capital and reserve requirements in bank stability during the depression, see Mitchener (2005).

10 See, for example, Hsieh and Romer (2006) for a discussion of the role of the gold standard in Federal

Reserve policymaking during the Great Contraction.

11 The reserve ratio data come from the monthly Federal Reserve Bulletin from 1929-1930.

12 In addition, the Federal Reserve Act required that the outstanding notes of each Federal Reserve Bank be backed 100 percent by collateral in the form of gold and eligible paper. An analysis of balance sheet data from the monthly Federal Reserve Bulletin reveals that this condition was also met in each month in 1929 and 1930.

13 July 20, 1929. Commercial and Financial Chronicle. “The Florida Bank Failures.” p. 422.

14 November 14, 1930. New York Times. “Money Redeposited in Knoxville.” p. 19.

15 Other scholars—see, for example, Ziebarth (2013)—have used the Mississippi natural experiment identified by Richardson and Troost to study the effects of monetary policy on outcomes other than bank failures. A key contribution of this paper is that it makes the case for extending the analysis beyond

Mississippi to the entire Atlanta Fed border.

REFERENCES

Annual Reports. Federal Reserve Banks. 1929-1930.

Calomiris, Charles W., and Joseph R. Mason. “Fundamentals, Panics, and Bank Distress During

the Depression.” American Economic Review 93, no. 5 (2003): 1615-47.

Carlson, Mark, Kris J. Mitchener, and Gary Richardson. “Arresting Banking Panics: Federal

Reserve Liquidity Provision and the Forgotten Panic of 1929.” Journal of Political

Economy 119, no. 5 (2011): 889-924.

Chandler, Lester V. American Monetary Policy, 1928-1941. New York: Harper & Row, 1971.

Commercial and Financial Chronicle. 1929.

15 Federal Reserve Board. Bulletin, 1928-1930.

Friedman, Milton and Anna J. Schwartz. A Monetary History of the United States, 1867-1960.

Princeton: Princeton University Press, 1963.

Hsieh, Chang-Tai, and Christina D. Romer. “Was the Federal Reserve Constrained by the Gold

Standard During the Great Depression?” Journal of Economic History 66, no. 1 (2006):

140-76.

Meltzer, Allen H. A History of the Federal Reserve. Vol 1, 1913-1951. Chicago: Univ. Chicago

Press, 2003.

Mitchener, Kris. “Bank Supervision, Regulation, and Financial Instability during the Great

Depression.” Journal of Economic History 65, no. 1 (2005): 153-85.

Polk’s Bankers Encyclopedia Co. Polk’s Bankers Encyclopedia, March 1929. Detroit, MI.:

Polk’s Bankers Encyclopedia Co., 1929.

Richardson, Gary and William Troost. “Monetary Intervention Mitigated Banking Panics During

the Great Depression: Quasi-Experimental Evidence from the Federal Reserve District

Border in Mississippi, 1929 to 1933.” NBER Working Paper No. 12591, Cambridge,

MA, October 2006.

______. “Monetary Intervention Mitigated Banking Panics during the Great

Depression: Quasi-Experimental Evidence from a Federal Reserve District Border, 1929-

1933.” Journal of Political Economy 117, no. 6 (2009): 1031-73.

Temin, Peter. Did Monetary Forces Cause the Great Depression? New York: W.W. Norton,

1976.

Wheelock, David. The Strategy and Consistency of Federal Reserve Monetary Policy, 1924-

1933. Cambridge: Cambridge University Press, 1991.

Wicker, Elmus. The Banking Panics of the Great Depression. Cambridge: Cambridge

University Press, 1996.

16 Ziebarth, Nicholas L. “Identifying the Effects of Bank Failures from a Natural Experiment in

Mississippi during the Great Depression.” American Economic Journal:

Macroeconomics 5, no. 1 (2013): 81-101.

17 TABLE 1 SAMPLE SIZE

Counties Banks Bank Suspensions 1927 1928 1929 1930 1931 1932 1933 Atlanta-Richmond Atlanta 73 275 3 20 19 16 12 19 1 Richmond 62 338 9 14 52 26 14 36 0 Atlanta-St. Louis Atlanta 60 274 4 8 19 21 11 22 2 St. Louis 78 403 4 4 56 38 20 32 0 Atlanta-Cleveland Atlanta 6 29 0 0 4 0 0 4 0 Cleveland 18 67 3 0 3 4 7 9 0 Atlanta-Dallas Atlanta 27 104 1 0 2 8 3 25 0 Dallas 32 124 1 0 6 1 1 19 0

Total 356 1614 25 46 161 114 68 166 3 Note: The data come from the Federal Deposit Insurance Corporation (FDIC), available through the Inter- University Consortium for Political Science Research. The table reports the number of counties, the number of banks in operation in 1927 and the number of bank suspensions in each year from 1927 to 1933.

18 TABLE 2 CHARACTERISTICS OF COUNTIES Atlanta - Richmond Atlanta - St. Louis Atlanta - Cleveland Atlanta - Dallas Atlanta Richmond Atlanta St. Louis Atlanta Cleveland Atlanta Dallas Economic and Demographic Variables Population (1000s) 21.5 27.6 26.4 22.2 21.9 28.3 28.7 21.8 Persons per Square Mile 56.7 57.0 44.9 46.4 87.5 67.1 46.5 33.7 Fraction of Population in Labor Force (%) 35.9 34.2 36.1 36.5 32.1 28.6 37.6 37.9 Urban Population Share (%) 10.5 6.2 5.5 7.6 10.2 4.5 14.6 8.0 Unemployment Rate (%) 1.9 2.4 1.6 1.8 1.5 2.6 2.8 2.0 Average Size of Farms (in acres) 84.0 84.6 75.3 74.3 82.2 83.2 61.1 153.2 Number of Manufacturing Establishments 25.5 31.5 27.6 16.8 17.8 48.2 16.4 20.4 Percentage of Cropland with Failures 2.1 1.5 2.4 2.1 1.4 1.7 2.7 4.7 Ratio of Debt to Farm Value 34.4 30.0 36.5 44.4 29.0 25.7 35.5 31.8 Regulatory Variables Minimum Capital Requirement (in $1000s) 20.1 20.6 10.0 12.1 10.0 15.0 10.0 10.0 Reserve Requirement (%) 13.4 11.2 12.4 12.1 10.0 10.0 18.1 18.0

Sample Size (# of Counties) 73 62 60 78 6 18 27 32 SOURCE: Economic and demographic statistics come from the censuses of population, manufacturing, and agriculture for 1930. The data can be electronically downloaded at NHGIS (http://www.nhgis.org) or ICPSR (http://www.icpsr.umich.edu/icpsrweb/ICPSR). Minimum capital and reserve requirements come from Polk's Bankers' Encyclopedia, 1929, and the Federal Reserve Bulletin, Nov 1928 and Sep 1930. NOTE: The table reports averages across counties.

19 TABLE 3 PANEL REGRESSION MODEL DEPENDENT VARIABLE: COUNTY BANK SUSPENSION RATE

Regression 1 Regression 2 Coefficient Standard Error Coefficient Standard Error Monetary Regime Fixed Effects Atlanta1927 -0.011 0.006 -0.003 0.006 Atlanta1928 0.005 0.010 0.013 0.010 Atlanta1929 -0.078** 0.024 -0.071** 0.023 Atlanta1930 -0.046** 0.017 -0.037* 0.017 Atlanta1931 -0.033 0.019 -0.024 0.019 Atlanta1932 -0.018 0.030 -0.010 0.029 Atlanta1933 0.000 0.007 0.008 0.006

Region Fixed Effects 1927 Atlanta-Richmond 0.003 0.026 0.015* 0.006 Atlanta-StLouis -0.028 0.024 0.013** 0.004 Atlanta-Cleveland 0.007 0.027 0.026 0.018 Atlanta-Dallas -0.059* 0.028 0.006 0.004 1928 Atlanta-Richmond 0.031 0.028 0.043** 0.011 Atlanta-StLouis -0.029 0.026 0.011 0.007 Atlanta-Cleveland -0.023 0.022 -0.003 0.003 Atlanta-Dallas -0.071* 0.029 -0.006 0.004 1929 Atlanta-Richmond 0.152** 0.036 0.165** 0.025 Atlanta-StLouis 0.107** 0.032 0.149** 0.022 Atlanta-Cleveland 0.128* 0.064 0.147* 0.059 Atlanta-Dallas 0.008 0.038 0.074** 0.025 1930 Atlanta-Richmond 0.081* 0.033 0.093** 0.020 Atlanta-StLouis 0.069* 0.031 0.110** 0.020 Atlanta-Cleveland 0.033 0.031 0.052* 0.023 Atlanta-Dallas -0.022 0.032 0.043** 0.013 1931 Atlanta-Richmond 0.080* 0.035 0.092** 0.023 Atlanta-StLouis 0.020 0.030 0.060** 0.017 Atlanta-Cleveland 0.071 0.043 0.090* 0.037 Atlanta-Dallas -0.038 0.030 0.026* 0.011 1932 Atlanta-Richmond 0.147** 0.043 0.160** 0.032 Atlanta-StLouis 0.076* 0.032 0.120** 0.023 Atlanta-Cleveland 0.136* 0.055 0.156** 0.051 Atlanta-Dallas 0.183** 0.049 0.252** 0.042 1933 Atlanta-Richmond -0.009 0.025 0.003 0.005 Atlanta-StLouis -0.041 0.024 -0.001 0.003 Atlanta-Cleveland -0.021 0.021 -0.002 0.002 Atlanta-Dallas -0.069* 0.027 -0.004 0.003

County-Level Controls Yes No

Capital Requirement -0.003** 0.001 ------(in $1000s) Reserve Requirement 0.004* 0.002 ------R2 0.125 0.119 * = Significant at the 5 percent level. ** = Significant at the 1 percent level. Note: Robust standard errors, clustered at the county level are in parentheses.

20 FIGURE 1

COUNTIES WITHIN 50 MILES OF ATLANTA FED DISTRICT BORDER

Note: The map shows counties located within 50 miles of the Atlanta Federal Reserve District border. The map comes from the U.S. Census Bureau. The 50-mile buffer was generated using a Geographic Information System (GIS) program.

21 FIGURE 2

RESERVE POSITION OF ATLANTA, ST. LOUIS AND RICHMOND FEDERAL

RESERVE BANKS

Note: The reserve ratio data come from the monthly Federal Reserve Bulletin. The reserve ratio measures the ratio of total reserves against note and deposit liabilities.

22