The Reconstruction Finance Corporation, the gold standard, and the banking panic of 1933.

• Author: Butkiewicz, James L.

[Publication of RFC loans] was the most damnable and vicious thing that was ever done. It counteracted all the good we had been able to do.

Atlee Pomerene, Chairman of the RFC

1. Introduction

   By the summer of 1932 the United States banking system had been severely battered by three years of depression, recurring banking panics, and a run on the dollar following Britain's departure from gold. Depression and price deflation depreciated asset values, rendering bank portfolios illiquid. The establishment of the Reconstruction Finance Corporation in February 1932 afforded relief to the beleaguered banking system. The upturn in industrial production in August gave evidence of an incipient recovery.

   However, the recovery was short-lived. Industrial production peaked in October, contracting thereafter. The stability provided by RFC lending was undermined by publication of the names of banks receiving new RFC loans, beginning in August, and culminated with publication of the names of all banks receiving RFC loans in January 1933.

   Concomitant with the economic downturn was an increase in bank suspensions, leading to the final crisis and bank holiday in March 1933. Every discussion of the Banking Panic of 1933 attributes the crisis to one or more specific causes. Einzig (1933), Hodsen (1938), James (1938),(1) Kennedy (1973), Cleveland and Huertas (1985), and Keehn and Smiley (1988, 1993) blame the run on banks on the publication of the names of banks receiving loans from the Reconstruction Finance Corporation. Sullivan (1936), Friedman and Schwartz (1963), and Yeager (1966) cite the publication of RFC loan lists as well as the ultimately well-found fear that Franklin Roosevelt would devalue the dollar upon assuming the presidency as possible causes of the final panic of the Great Depression. Donaldson (1992) and Wigmore (1987) argue that a run on the Fed's gold reserve through the banking system forced the bank holiday. Wicker (1996) blames the crisis on the declaration of a bank holiday in Michigan in February 1933. However, none of these studies presents tests to distinguish among alternative explanations.

   Evidence presented in this paper demonstrates that neither general fear of devaluation nor the publicity about the RFC loans ignited the final crisis. Although publication of the identity of banks receiving RFC loans did increase failures, the increase in failures is small relative to the number of banks identified as recipients of RFC loans.

   Similarly, the daily behavior of the major dollar exchange rates shows no evidence of a speculative attack on the dollar until after the declaration of a banking holiday in Michigan. Several short-lived depreciations of the dollar resulting from political events and local banking difficulties are evident in the exchange rate data, indicating that events affecting confidence in the dollar did affect exchange rates. However, the dollar strengthened relative to the franc following Roosevelt's election and remained strong until the Michigan crisis. Banking difficulties in Illinois, Nevada, and Iowa had minor transitory effects on the value of the dollar, whereas a one-day "business holiday" in Louisiana did not weaken the dollar. However, the Michigan holiday, closing the major banks in a major industrial center, and the ensuing series of bank holidays resulted in a sharp depression of the dollar. The declaration of bank holidays in state after state increased uncertainty about the stability of the banking system. These events aroused fears that the situation in the United States compared with the banking crises that spread through Europe during the summer of 1931, driving country after country off gold.(2) Although the Michigan crisis centered around a loan request made to the RFC, the real cause was a clash of egos between old rivals.

   Section 2 of this paper examines the effect of publication of RFC loans to banks on bank suspensions during the final months of 1932 and the first months of 1933. Section 3 reviews daily exchange rate data for evidence of a speculative attack on the dollar. Section 4 compares the Michigan crisis with earlier crises? Section 5 summarizes the findings of the paper.

2. The Reconstruction Finance Corporation and the Final Banking Panic

   Butkiewicz (1995) has demonstrated that RFC loans to banks reduced bank suspensions prior to the publication of the names of banks receiving loans in late August 1932.(4) The effectiveness of RFC lending decreases at that point, due possibly to loss of depositor confidence resulting from publicity and/or from banks' reluctance to borrow from the RFC due to a fear of adverse effects of publication. The number and dollar value of bank loan applications to the RFC declined substantially in the months following the beginning of publication of the names of banks receiving RFC loans,(5) which indicates banks were reluctant to borrow.

   If publication caused bank runs, then bank suspensions should be higher in states where a high percentage of banks are identified as receiving RFC loans. Keehn and Smiley (1988, 1993) compare RFC lending records with the banks that failed between March 1932 and June 1933. Comparing the publication lists with the names of banks which suspended activity, they find that publication may have been a factor in 264 bank suspensions between March 1, 1932, and March 4, 1933. Although they compare the list of banks identified as receiving RFC loans with lists of suspended banks, it is possible to assess the effect of the publication of the loan lists econometrically, which controls for other possible causes of bank suspensions. Also, publication may have resulted in suspensions of banks not borrowing from the RFC, as bank runs spread from borrowing banks to other local banks.(6,7) Finally, public confidence in banks in general may have been eroded when local banks were identified as having borrowed from the RFC.

   In modeling bank suspensions during the interwar period, several factors have been established as determinants of suspensions. Previous research by Friedman and Schwartz (1963), Temin (1976), McCallum (1990), and Butkiewicz (1995) found bank suspensions to be related to changes in either aggregate or agricultural income and a continuation or persistence of failures, typically proxied by previous suspensions. The basic model relates suspensions to changes in agricultural income and to previous suspensions. A measure of the extent of publicity regarding RFC loans is added to this basic model.

   First, two simple cross-sectional econometric models of bank suspensions by state are used to assess the effect of publication of RFC loan recipients. The models are:

   [Mathematical Expression Omitted] (1)

   and

   [Mathematical Expression Omitted], (2)

   where:

   [PS.sub.it] = Number of banks suspending operations in a state, i, in a given month, t, as a percentage of the number of banks in operation in a state on December 31, 1931.

   [CPS.sub.it] = Cumulative number of banks suspending operations in a state, i, in a given month, t, as a percentage of the number of banks in operation in a state on December 31, 1931. Cumulative totals are from October 1932 (September and October) through February 1933 (September 1932 through February 1933).

   %[Delta][FI.sub.i] = Percent change in nominal farm income in a state from 1929 through 1932.

   [PVS.sub.i] = Bank suspensions in state i from January through August 1932 as a percentage of banks in operation in state i on December 31, 1931.

   [Pub.sub.i, t-1] = Cumulative number of banks listed as receiving RFC loans in state i in all months prior to the current month, t, as a percentage of banks in operation in state i on December 31, 1931.

   The numerators of the dependent variables are the number of banks suspending operations in a given state for each month, September 1932 through February 1933, and the cumulative number of failures in a state in each month, beginning in September 1932. The first monthly cumulative total is September and October, and each subsequent month's failures are added to the previous total.(8) The denominator is the number of banks in operation in each state on December 31, 1931.

   In earlier empirical work (Butkiewicz 1995), previous suspensions and a measure of farm income were found to significantly affect bank failures. The farm income variable used here is the percent change in nominal farm income in each state from 1929 through 1932. Nominal income is used since farmers were adversely affected by both price declines and loss of real income. The previous suspension variable is the total number of bank suspensions in a state from January 1932 through August 1932 as a percentage of the number of banks in the state on December 31, 1931.

   The publicity variable is the cumulated number of banks in a state listed as receiving RFC loans in a given month, as a percentage of banks in the state on December 31, 1931. Because some banks borrowed from the RFC in more than one month, cumulation double-counts these banks. However, repeated borrowing could have been interpreted as indicative of continuing problems for a bank, increasing the likelihood of a run. For September, the variable is the number of banks reported in the August 23 newspaper report. For October, the number of banks reported in the October 8 list is added to the previous total. For November, the October 23 list is added to the previous total. The December variable adds the November 29 list, the January variable includes the number of banks in the December 23 report, and the February variable adds the January 27 lists to the previous totals.

   The list of RFC loans published in August was short; only a 10-day time period was covered, so any effect in September might not be evident in the data.(9) The lists published in January were the most extensive, including all loans made in December and in the period February 2 through July 21, 1932.(10) However, the effect in February is not expected to be...