A History of the Federal Reserve, Volume 1

by Allan H. Meltzer

  139. Early in the morning of March 4, Governor Lehman, acting “on the request of the New York Clearinghouse banks and with the advice and recommendation of the Federal Reserve of New York” declared a state bank holiday for March 4 and 6. Federal Reserve banks closed along with commercial banks. This posed a problem. Ohio had not declared a holiday on March 4, so the Cleveland reserve bank remained open. Dallas also remained open until it received a wire from a bank in Pittsburgh asking for $10 million in cash. Told that a plane was on its way, Dallas closed (CHFRS, interview with Joseph P. Dreibilbis, March 9, 1954). Dreibilbis was counsel to the Dallas bank.

  140. Hoover’s letter also said that the “authorities on which you were relying were inadequate unless supported by the incoming administration.” This point had been made forcefully by Secretary Mills on many occasions, most recently at the midnight meeting of the Board. Mills reported that the attorney general had advised the president not to issue the proclamation. Todd (1995, 21–22) reports a conversation between Glass and Roosevelt in Roosevelt’s hotel room at 11:30 p.m. on March 3. Roosevelt told Glass he had rejected Hoover’s request that they act jointly. Glass then asked Roosevelt what he planned to do. Roosevelt replied: “Planning to close them, of course.” Glass pointed out that Roosevelt lacked the authority to close any banks and especially state banks, but Roosevelt insisted he would have the authority as president.

  The act extended and broadened the president’s powers to close, liquidate, license, and reopen banks under the Trading with the Enemy Act, removing any possible challenge to the legality of his proclamation. The act also strengthened the Reconstruction Finance Corporation, authorized national banks to issue preferred stock, and permitted the RFC to purchase shares in national and state banks. And it prepared for the nationalization of gold holdings by empowering the secretary of the treasury to order all domestic gold owners to sell their holdings to the Treasury.

  The bank holiday was a climax to the depression because it forced the government and the Federal Reserve System to respond to the domestic financial and economic collapse. Actions that had seemed beyond consideration were no longer unthinkable. In the next few months the administration chose domestic expansion over fixed exchange rates and dismissed the opportunity to return to the gold standard. Employment, agricultural prices, and other domestic concerns replaced the gold price and real bills as guides to economic policy.

  EMPIRICAL STUDIES: THE ROLE OF MONEY

  The Great Depression was mainly a monetary event in two senses. Monetary policy could have mitigated or prevented the decline but failed to do so. A different set of Federal Reserve policy actions could have avoided the severe deflation and reduced the depth and severity of the economic decline. In this sense the Great Depression was a response to monetary policy.

  There is another sense in which the depression was a monetary event. The initial decline could have been a response to a negative monetary impulse or sequence of impulses. A few writers have taken this view (Anderson, Shugart, and Tollison 1988). Other students of the period suggest that monetary forces had no role (Temin 1976).

  The extreme positions—that monetary policy was the only cause or that monetary policy played no role—are difficult to sustain.143 A more plausible explanation is that the depth and severity of the Great Depression were the consequence of a series of shocks that the Federal Reserve neglected or failed to offset completely. The shocks include French gold policy, banking panics, increased demand for currency, departure of Britain from the gold standard, the stock market decline, failure of banks in Austria and Germany, collapse of United States export markets in Latin America, the effects of tariffs and retaliation on prices and thus on gold movements, and other events. Some of these events are both the effect of prior changes and the proximate cause of subsequent changes. We are unlikely to develop a complete list of “true” causes that operated independently of other events.

  141. There were few legal challenges. South Carolina’s decision to close its banks was upheld by the Supreme Court in December 1933 (Board of Governors File, box 2165, December 5, 1933).

  142. Walter Wyatt, the Board’s legal counsel, prepared the act. According to Joseph Dreibilbis, one of the Federal Reserve attorneys, there was only one copy of the act when it passed (CHFRS, Dreibilbis, March 9, 1954).

  One alternative is to look for outliers, or large changes, in output and the money stock. A Kalman filter, developed by Bomhoff (1983), uses the past history of a series to predict future values. The difference between predicted and actual values is a measure of changes that could not have been foreseen from the history of the series. Using quarterly data reported in Balke and Gordon (1986) from 1890 or 1915 through 1984, the filter predicts each quarterly observation, then uses the error to revise subsequent predictions. Predictions of GNP, money, and prices are made independently, so it is possible to check on the consistency of the predictions of GNP by summing the errors in predicting real output and prices. For 1928–33, these sums are generally in the same direction and have similar magnitude as the error in GNP. Table 5.26 shows all errors in M1 and errors in GNP and prices greater than or equal to 1 percent for this period.

  The data support five principal implications. First, the depression was caused by a series of unanticipated changes or shocks, not by a single event. There were large shocks to nominal and real GNP, both positive and negative, throughout the period. Despite the reversals in sign, the cumulative sum of the shocks to nominal GNP from the peak in third quarter 1929 to the trough in first quarter 1933 (17.4 percent) represents about one-third of the decline in nominal GNP.144

  Second, most of the large shocks to nominal GNP were also large shocks to real GNP in the same direction. There are fewer large shocks to the price level, suggesting that price changes were mainly the result of system response to current and past shocks acting on output and spending. Large price-level shocks typically have the opposite sign from contemporary output shocks, suggesting that the shock affected supply. Since most of the supply shocks are positive, they cannot explain the long and deep decline in output.

  143. After a thorough examination of several types of data, Hamilton (1987) draws a similar conclusion. He interprets much of the deflation as unanticipated based on commodity prices, interest rates, and newspaper accounts.

  144. This calculation excludes the continuing effects of the shocks in subsequent quarters, hence it understates their impact. Ohanian (2001) attempted a nonmonetary explanation of the decline in productivity. He concludes that nonmonetary factors explain only one-third of the decline.

  Third, shocks to money either are contemporaneous or lead shocks to GNP by a quarter or more. The largest monetary shock comes early in 1930, when M1 fell 11 percent (annual rate), the largest decline in any quarter since 1921. Negative shocks to money dominate 1931, particularly following the Federal Reserve’s response to the British devaluation. The monetary shocks change sign in 1932 following (or accompanying) the relatively large open market purchases in second and third quarter 1932. Positive shocks to real and nominal GNP follow. Although the money stock continued to fall during most of that year, the rate of decline slowed for a time and money stock rose in the fourth quarter. Industrial production and stock prices rose in fall 1932. These data suggest that, contrary to some Federal Reserve interpretations, the 1932 open market purchases did not fail. Continuation of the positive shocks by more expansive actions in 1931 and 1932 or earlier would likely have changed the course of the depression.

  Fourth, some periods show negative shocks to output that are large relative to current or past shocks to money. Fourth quarter 1929 and third quarter 1931 are prominent examples. In both quarters there was some prominent event: for third and fourth quarter 1929 we have the peak in the economy in August 1929, the spread of recession abroad, and the fall in United States stock prices in October 1929; in third quarter 1931 there were banking problems in Germany and the suspension of gold payments by the Bank of England in Sep
tember 1931. These changes may have affected monetary velocity.145 Waves of bank failures and suspensions in 1930.4, 1931.2, 1931.3, 1931.4, and 1932.1 had mixed effects. Shocks to money and nominal GDP were positive in 1930.4, relatively small but positive for nominal GDP in 1931.2, commingled with the effect of the British suspension of gold payments in 1931.3, and accompanied by a large negative shock to money and nominal GDP in 1931.4. Only the bank closings in 1932.1 are accompanied by negative shocks to output and a positive monetary shock that would support a major role for nonmonetary factors associated with bank suspensions. None of this evidence rules out a nonmonetary channel, but it does not suggest a dominant effect of nonmonetary shocks.

  Fifth, there is not much evidence of a decisive monetary surprise, or series of surprises, in the year preceding the start of the depression. The cumulated monetary shocks in the year ending 1929.2 is a small negative value. Price data show a sequence of small deflationary shocks (or errors) for the year ending 1929.2. Nonmonetary factors may have contributed to the deflation and the start of the depression.

  An alternative for investigating nonmonetary shocks uses errors computed from a demand function for money to see if there were large unexplained increases in the demand for money as suggested by Temin (1976). Table 5.27 shows the percentage errors from a demand function in which the logarithm of real money balances depends on the logarithms of interest rates and wealth or expected income. The equation is estimated separately for the logarithms of levels and changes of real money balances, using annual data for 1902 or 1903 to 1958. An appendix shows the equations. Errors are actual values minus estimates from the equation.

  For both equations, most of the errors are comparatively small. These data give no evidence of a sudden large, unexplained desire to accumulate real money balances. Most of the errors in log levels are negative, suggesting that real money balances fell below predicted values. This typically occurs when money growth falls more than anticipated. The years 1928 and 1930 are notable in that regard.

  145. Since monetary velocity is measured by GNP/M1, the numerator reflects the fall in nominal GNP. It is not possible to measure velocity shocks independently using the Kalman filter. An alternative procedure is discussed below.

  The years 1928 and 1932 are the only ones before the middle 1930s with relatively large errors in the change in real balances. Chart 5.2 shows that real balances fell much more in 1928, and rose much more in 1932, than anticipated by the demand equation. The negative error in 1928, for both levels and changes, suggests that before the recession demand for real balances fell more than actual balances.146 Factors other than income, wealth, and interest rates played a role in reducing growth of desired real money balances. In 1932 actual growth of real balances is 6 percent above the growth expected at a time of increased nominal money following several years of falling nominal money. Chart 5.2 shows the prediction errors for 1922–40.

  Gandolfi and Lothian (1977) estimated a demand for money equation using data for a cross section of states for the years 1929–68. Their findings also suggest that the demand function for money remained relatively stable during the Great Depression. They reject the presence of a liquidity trap. Their measure of the interest elasticity declined as interest rates fell, contrary to the liquidity trap. (See also Brunner and Meltzer 1968a.)

  Would a more expansive monetary policy have prevented the Great Depression or reduced it to a typical recession? McCallum (1990) simulated the response of nominal GNP assuming the Federal Reserve followed a monetary base rule from 1923 to 1941.147 McCallum’s rule is activist but not discretionary. The Federal Reserve adjusts the growth of the monetary base each quarter to reflect past changes in base velocity and deviations from a 3 percent growth rate.148

  146. This finding contradicts Field’s (1984) claim that stock market speculation increased the demand for money. A more likely explanation is the unanticipated decline in money growth following the French stabilization. Typically in periods of unanticipated decline in money growth, velocity (the ratio of income to money) rises. Consequently, the demand for money (per unit of income) falls.

  147. Bordo, Chaudri, and Schwartz (1995) take a similar approach using a rule that keeps growth of M2 constant. In some of their simulations, there is no depression. In others, as in McCallum’s, a typical recession would have occurred.

  148. McCallum adjusts his rule to allow for differences between the growth rates of M1 and the monetary base arising from currency drains and bank failures and suspensions.

  The simulations by McCallum (1990) and by Bordo, Chaudri, and Schwartz (1993) support two propositions. First, the Federal Reserve’s inaction converted a modest or possibly moderate recession into the Great Depression. In this limited sense, the depression was caused by monetary policy. Second, nonmonetary events contributed to the decline. All of McCallum’s simulations, and most of the simulations by Bordo, Chaudri, and Schwartz, show a recession in 1929 and 1930.

  Taken together, the estimates of the role of money in 1929–33 point to a relatively large role for money growth both as a factor deepening the recession and, at times, reversing the fall in output. There is no evidence that money was the unique cause of the decline. Systematic effects of other factors, including tax increases or expenditure reductions to balance the budget or tariff increases and retaliation abroad, have not been ruled out.

  Sweden avoided severe deflation. Its central bank, the Riksbank, followed the policy advocated by some members of Congress in the 1920s and at the time; acting under parliamentary guidance, the Riksbank worked to stabilize the domestic price level. Sweden could not offset the real effects of an international decline, but after leaving the gold standard in 1931, the country avoided the deflation and its effects on output, financial institutions, firms, and households (Berg and Jonung 1998). The Swedish recession was comparatively mild.

  Bernanke (1983, 1994), Bernanke and James (1991), and others link monetary and nonmonetary factors in the Great Depression and at other times. These authors accept that bank failures and suspensions during the depression reduced the money stock. They propose, in addition, that bank failures and suspensions reduced bank lending. Since small and medium-sized firms depend disproportionately on bank loans to produce and finance output or sales, reductions in bank lending have a large impact. Further, during the depression, these authors claim, deflation had a nonneutral effect on debtors by forcing contraction, lowering net worth, and reducing access to bank credit. The last of these effects, debt deflation, requires borrowers to be affected more, or more quickly, than creditors.

  Bernanke does not dispute the monetary effect of the Federal Reserve’s failure to stop the bank runs by open market operations. That bank loans declined with bank deposits is an expected consequence of monetary contraction. The extraordinary real rates of interest and high-risk premiums on Baa bonds after late 1930 testify to a general reluctance to extend credit to any borrowers, particularly lower-rated or unrated borrowers.

  More problematic is the particular, nonneutral effect of the decline in bank loans on smaller firms. For this effect to have aggregate consequences, bank loans must decline relative to open market lending by nonbank firms. The data for the second half of the depression show the opposite. Short-term open market lending fell relative to bank lending. Table 5.28 shows the ratio of commercial paper plus banker’s acceptances to loans at 101 weekly reporting member banks at each three-month interval from the peak in August 1929 to February 1933.

  The data show a relative expansion of open market lending during the early months of the decline. The ratio reached a peak in the first six months; thereafter open market lending declined relative to bank lending. The relative decline accelerated when suspensions (measured by deposits of suspended banks) rose beginning in November 1930. During the peak period of bank suspensions in second half 1931, the ratio fell below its value in August 1929.149

  149. Commenting on this section in oral discussion, Bernanke attributed the result to a
change in the commercial paper market. This does not explain banker’s acceptances. Further, as shown in Greef 1938, most of the change occurred before the 1930s.

  Table 5.28 gives little support to the argument that the decline in bank lending had a nonneutral effect that augmented the monetary effect. The common decline in lending by banks and nonbanks suggests a reduction in desired borrowing in response to poor opportunities and widespread beliefs that the recession would continue. These beliefs are documented in the minutes of the Federal Reserve.

  Table 5.29 compares the percentage decline in lending for different groups of banks to the decline in external finance. Weekly reporting banks show the smallest percentage decline. To get a better measure of small banks, subtract weekly reporting banks from all banks. Line 4 of the table shows that this class declined by about the same percentage as banker’s acceptances and less than commercial paper.

  The relative share of credit by large banks rose despite the sharp decline in acceptances and commercial paper. These markets were much smaller than the bank loan market in absolute size. If we assume that large banks lend mainly to large firms, the evidence suggests that credit to large firms declined less than credit to other firms. This conclusion is tempered, however, by the comparison of all member banks and all banks. These groups declined in the same proportion.

  These data do not separate a decline in the demand for loans from restrictions on supply to small firms. The data are entirely consistent with a relative decline in loans demanded by small firms. Data are not available on sales by size of firm, so an examination of the proposition is incomplete.