I

Beginning in 1920 through 1921, the US economy suffered a second, more severe post-war economic contraction, lasting 18 months.Footnote ¹ The unemployment rate jumped from 1.4 in 1919 to 11.7 percent in 1921.Footnote ² Consumer prices, which had more than doubled since the beginning of the war in 1914, fell by almost 21 percent.Footnote ³ The agricultural sector that had expanded significantly in the preceding decade suffered a sizable deflation. The resulting agricultural distress was a potential cause of problems for the banking system.

Federal Reserve policy moves contributed to this cyclical fluctuation. Following the war, the Fed kept interest rates low to support Treasury bond prices. The ensuing inflation resulted in a loss of gold reserves, forcing the Fed to increase interest rates. The New York Fed increased its benchmark discount rate three times from November 1919 through June 1920, from 4 to 7 percent, and then cut, from May through November 1921, to 4.5 percent. The other reserve banks adjusted their rates similarly during this time frame, but the timing and magnitude of the rate changes sometimes differed.Footnote ⁴ Friedman and Schwartz (Reference Friedman and Schwartz1963, p. 234) note that the collapse of prices followed the final June rate increase.

Other than changes in Federal Reserve policy, this particular cycle is reputed to have taken its natural course with no governmental involvement. Eichengreen notes that this cyclical episode led some Federal Reserve leaders to conclude that a policy of liquidation was desirable, as a strong recovery quickly followed the contraction (Eichengreen Reference Eichengreen1992, p. 122). Grant describes the episode as a period of ‘instructive inaction’ (Grant Reference Grant2014, p. 8).

However, this perception of the episode is not entirely accurate. Early in the recession, Federal Reserve banks in agricultural regions made loans to banks to support the farm economy. Also, a little-known federal agency, the War Finance Corporation (WFC), actively provided financial assistance, first by providing financing for exports of US products, and shortly after the cyclical trough, lending to banks in troubled agricultural regions. The WFC's lending was channeled to farmers indirectly, by lending to banks, cooperative marketing associations and livestock loan companies to then lend to farmers. If WFC lending provided relief to farmers, it also would mitigate banking difficulties that otherwise would have resulted from agricultural distress.

This article reports the results of an assessment of the WFC's lending program. Using a regression discontinuity in time design and archival data for bank suspensions, estimates are obtained for the impact of the WFC lending program on bank suspensions. The estimates provide support for the view that the lending program reduced bank suspensions.

The article is organized as follows. Section II discusses the economic impacts of the recession, focusing on the agricultural recession and resulting banking problems. The next section recounts the activities of the War Finance Corporation, including the 1921–2 lending program. The fourth section reports the results of the econometric evaluation of WFC lending on bank suspensions. Section V concludes.

II

WFC lending to banks was designed to assist farmers, by providing credit to allow farmers to carry their inventories for longer periods of time. Federal Reserve discount loans for agricultural purposes were limited to six months’ maturity. However, changes in the pattern of sales necessitated loans for longer periods. WFC lending was intended to fulfill farmers’ credit needs, while simultaneously relieving pressure on country banks.

The agricultural sector prospered during the 1910–20 decade, with a significant increase in land under cultivation. As World War I disrupted European agricultural production, US exports soared, adding to the prosperity. In 1920, agricultural output contributed over 12 percent of GDP and by one estimate, farm workers comprised 15.3 percent of the total labor force.Footnote ⁵

While the recession affected both industry and agriculture, the farm economy was hit particularly hard. Industrial production contracted 31.6 percent.Footnote ⁶ An index of prices received by farmers fell 52.5 percent during the downturn. The prices of two major crops, cotton and wheat, fell 72.2 and 64.4 percent respectively, from their cyclical high to low values.Footnote ⁷

Key indicators of the health of the farm economy, indices of prices received, land value per acre, farm exports and farm-mortgage debt, are depicted in Figure 1.Footnote ⁸ All measures rose dramatically during the war years, and then fell, perhaps even more dramatically, except for mortgage debt that was not easily liquidated. Farm exports peaked in 1919, and output prices and land values peaked in 1920. Mortgage debt continued to accumulate, peaking in 1923. The recession of 1920–1 and accompanying deflation caused severe economic problems for farmers.

Figure 1.

Farm economy: overview

Sources: Index of farm prices – NBER Macrohistory database, farm value, exports, and mortgage debt – United States Department of Agriculture 1942.

The drop in exports, price deflation and concomitant decline in farm income were certainly the primary problems. An underlying source of difficulty was likely increased indebtedness incurred during the war-time boom (Rajan and Ramcharan Reference Rajan and Ramcharan2015). Shifts in the demand for and supply of farm credit may have combined to increase farm debt. On the demand side, optimism resulting from the 1917–20 boom led farmers to undertake excessive commitments (Johnson Reference Johnson1973–4).

The increase in the supply of credit to farmers resulted from the creation of the Federal Farm Land Bank system in 1916, providing mortgage loans to farmers (O'Hara Reference O'Hara1983). The system consisted of 12 regional federal farm land banks with member local cooperative associations, and private joint stock land banks. Both could issue tax-exempt bonds to raise funds for mortgage lending to farmers. The dramatic increase in income tax rates and the number of brackets enacted to help finance the war made the tax-exempt bonds issued by the land bank system especially attractive (Kang and Rockoff Reference Kang and Rockoff2015). As a result, funds were available to farmers at subsidized rates and farm debt increased commensurately, especially following the war.

Then, the collapse of prices and foreign trade greatly increased the burden of agricultural debts. Also, the pattern of exports changed. Before the war, exports were sold in large quantities, but following the war, Europeans bought in small amounts, requiring producers to carry stocks longer, increasing their credit needs. Those farmers who had increased their debt loads during the boom years were subsequently hard pressed to service their loans. The resulting agricultural distress was severe.

As the banking system at this juncture in US history was primarily a unit banking system, especially in agricultural states, problems on the farm led to banking problems. The fortunes of unit banks in rural areas were inexorably intertwined with the fortunes of farmers.

The Federal Reserve's preferred measure of banking distress during this period is bank suspensions. The data on state bank failures are fragmentary, and the suspensions data beginning in 1921 are considered to be the most reliable measure of banking problems (Board of Governors of the Federal Reserve System 1943, p. 281). Bank suspensions from 1921 through 1929 averaged 635 per year. Since most suspensions occurred in agricultural states, the high average level of suspensions is often blamed on farming difficulties. Figure 2 depicts the combined annual number of national and state bank suspensions each year, 1921–32, per 100 active banks as of 30 June of each year.Footnote ⁹

Figure 2.

Suspensions per 100 active banks, 1921–32

Source: Board of Governors of the Federal Reserve System (1932), p. 5.

One study finds that agricultural problems were the primary determinant of bank suspensions during the 1920s, and that the existence of state deposit insurance systems that likely increased moral hazard also contributed to higher suspension rates (Alston, Grove and Wheelock Reference Alston, Grove and Wheelock1994). White (Reference White1984), focusing on national bank failures in 1930, found that agricultural problems, along with unit banking, declining bond prices and tight monetary policy explain failures.Footnote ¹⁰

Alternatively, a Federal Reserve internal study (Board of Governors of the Federal Reserve System 1932) attributed banking difficulties during the 1921–31 period to overbanking. A table in this study compares bank suspensions per 100 banks in operation in 1920 (Suspensions) to the growth in the number of banks from 1900 to 1920 (Growth) and population per bank in 1920 (Pop).Footnote ¹¹ An ordinary least squares estimate with Huber-White-Hinkley heteroskedasticity consistent standard errors and covariance obtains the following result, with t-statistics in parentheses:

$$\eqalign{& {\rm Suspensions} = 23.8 + 0.051 {\rm Growth} - 0.0013\ {\rm Pop} \cr & \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,( {4.09} ) \;( {3.28} ) \;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;( {2.02} ) } $$

These results support the Fed's ‘overbanking’ hypothesis.

Initially, Federal Reserve banks in agricultural districts attempted to provide relief to farmers by lending to member banks. Tallman and White (Reference Tallman and White2019) document that early in the 1920–1 downturn, Federal Reserve banks in agricultural regions lent aggressively. However, the lending to assist banks in agricultural regions, particularly by the Atlanta Fed which had been an especially active lender, was criticized within the system, and the Atlanta Fed was pressured to decrease its discount lending in 1921 (White Reference White, Rousseau and Wachtel2017; White and Roberds Reference White and Roberds2020). Total Federal Reserve credit fell steadily beginning in January 1921 until the final months of 1922, as discount lending at this time was the primary source of Federal Reserve credit available to banks to serve as reserves, as seen in Figure 3.

Figure 3.

Discount lending, Federal Reserve credit and bank reserves

Source: Board of Governors of the Federal Reserve System (1943).

Correspondingly, with the decline in Federal Reserve Credit, available data show that member bank lending declined steadily from late 1919 through mid 1922. Following substantial wartime deficits, federal revenue exceeded federal expenditures in every fiscal year from 1920 through 1930, with an accompanying yearly decrease in gross debt. Fiscal policy focused on repaying war-related debt. Nominal interest rates did fall from 1920 through 1922. Lower rates certainly helped, but what farmers needed most was longer-term, and possibly renewable credit (Board of Governors of the Federal Reserve System 1943, pp. 72, 513 and 450).

With the decline in Federal Reserve lending, the ability of banks to fulfill the credit needs of the agricultural economy declined. As both monetary and fiscal policy were contractionary, pressure increased on both farmers and their banks. Attempting to provide relief for farmers, Eugene Meyer, managing director of the WFC, obtained authority to lend to banks, at maturities longer than allowed by the Fed's discount lending. While WFC lending was directed to farmers, this lending would relieve financial distress and likely reduce bank suspensions.

III

The WFC was created to help finance the war effort, not to assist banks. The United States’ entry into World War I placed a huge burden on federal finances. Gross federal debt increased from $2,976 million in June 1917 to $25,834 million by December 1919.Footnote ¹² Treasury Secretary William Gibbs McAdoo worried that debt financing of the war diverted capital from private industry and strained the financial system (McAdoo Reference McAdoo1931, p. 441). To address the war financing issue, McAdoo recommended and in April 1918 Congress legislated the creation of the War Finance Corporation as an off-budget federal agency.Footnote ¹³

To assist private-sector industries deemed essential to the war effort, the WFC was given lending power. The Treasury provided $500 million of capital and the WFC was authorized to issue bonds up to $3,000 million.Footnote ¹⁴ However, as the war ended shortly after it began operations, WFC lending to industry was limited to only $301.5 million. Of this total, $204.8 million consisted of 1919 loans to railroads to assist the United States Railroad Administration, financed in large part by the WFC's sole bond issue of $200 million (War Finance Corporation 1919).

Legislation for the third Liberty bond issue was in process at the same time as the WFC legislation. The first two Liberty bond issues, which had been marketed to the public through appeals to patriotism, were trading at discounts, arousing public discontent. The Third Liberty Bond Act contained a provision creating a sinking fund to repurchase Liberty bonds, to stabilize their prices. The Treasury delegated the war bond purchases to the WFC, and for a two-year period the WFC actively repurchased Liberty bonds and later Victory notes that it resold to the Treasury, until Treasury Secretary David F. Houston suspended the regular repurchase program in 1920.Footnote ¹⁵

The original WFC legislation authorized the corporation to operate for ten years, but its activities were to cease, except for the processing of loans and winding up its activities, six months after the end of the war. Thus, in late 1918 the WFC prepared to end its active operations. However, managing director Eugene Meyer, who believed that temporary government intervention was at times warranted and necessary, proposed WFC lending to support US exports.Footnote ¹⁶ Congress enacted authorizing legislation in March 1919, and the WFC lent $46 million to support exports until Treasury Secretary Houston suspended this lending in May 1920.

Frustrated with Houston's decisions, Meyer resigned from the corporation but pressed for a resumption of WFC lending. A Congressional resolution passed in December 1920 authorized renewed lending. Houston wrote ailing President Wilson's veto message, which Congress overrode in January 1921 (Meyer Reference Meyer1974, box 180). The WFC subsequently lent an additional $38.7 million to finance exports.

As the domestic agricultural economy deteriorated in 1921, Meyer believed that export financing was inadequate and pressed for expanded lending powers. On 24 August 1921, Congress amended the WFC legislation with the Agricultural Credits Act, authorizing lending not only for exports and to banks and financial institutions, but also to cooperative marketing associations and newly created livestock loan companies.

The WFC's expanded powers received strong governmental support. President Harding proposed the legislation to Congress with the support of Treasury Secretary Mellon and Commerce Secretary Hoover.Footnote ¹⁷ The WFC legislation was proposed as an alternative to a proposal for a new farm credit agency proposed by Senator Norris of Nebraska. The senate ultimately supported the WFC proposal without a record vote, and the house approved the bill by a vote of 314–21.Footnote ¹⁸

The WFC began lending in late October 1921 and lent extensively during the next several months, providing relief for farmers and their banks. The WFC's Fourth Annual Report (1922a) provides the agency's rationale for expanded lending powers:

Slow sales abroad were attributed to losses incurred from fluctuating exchange rates, as the gold standard was not yet restored (War Finance Corporation 1922a, pp. 5–6). The need to carry inventories for longer periods highlighted another concern of farmers, the need for intermediate credit.

The Federal Reserve Act authorized discount lending for 90 days but allowed lending for agriculture for six months. The Federal Land Banks made mortgage loans to farmers. However, there was a gap for credit of intermediate term. Meyer consistently argued that loans for livestock should be of a term of two years. Loans of this maturity would not be eligible for discounting or as collateral for advances from the Federal Reserve.

One contemporary author (Benner Reference Benner1926, pp. 50–1) felt the Fed's credit provisions to agriculture had worked well until the 1920–1 contraction. Due to the collapse of prices, farmers were unwilling and unable to repay their loans. Loans to farmers had been made with the expectation that they would be extended, but many country banks were overextended, and their correspondents and the Fed wanted loans to be liquidated (Benner Reference Benner1926, pp. 72–3). The lending powers Meyer obtained for the WFC were designed to fill this gap.

The 1921 Agricultural Credits Act authorized the WFC to make loans to exporters, banks, cooperative marketing associations and livestock loan companies to carry stocks.Footnote ¹⁹ The loans were collateralized and had an initial maturity of one year but could be extended for up to three years from the date of the initial advance. Total outstanding balances were limited to $1 billion. The interest on loans was between 5½ and 6 percent, with the rate on most loans being 5½ percent.Footnote ²⁰ Banks could add a margin of up to 2 percent above the rate charged by the WFC. Footnote ²¹ The initial authority to lend was to end on 1 July 1922, but this authority was extended three times to 31 December 1924. Liquidation of the WFC began on 1 January 1925.Footnote ²²

The WFC opened 33 loan agencies in 31 states that were deemed to be important agricultural and livestock districts.Footnote ²³ Volunteer committees received and reviewed applications. These committees made recommendations to the board of directors in Washington, DC, for final determination.Footnote ²⁴

Using its new powers, the WFC lent aggressively attempting to relieve the agricultural distress. The majority of WFC lending occurred between November 1921 and April 1922. From late October 1921 through December 1924 the WFC lent $269 million, of which 93 percent of financial institution loans and 82 percent of all loans were advanced by June 1922.Footnote ²⁵ Lending through December 1922 totaled 96 percent of loans to financial institutions and 90 percent of total lending. A total of $298.7 million was advanced to banks and other agricultural financial organizations during the life of the lending program, with all loans after 31 December 1924 being expense advances (Acting Secretary of the Treasury 1943, pp. 18–19).

Total lending of $298.7 million was small relative to total agricultural credit and agricultural product. A United States Department of Agriculture study in 1920 estimated that bank loans to farmers totaled $3,879 million, or 18 percent of total bank credit (Benner Reference Benner1926, p. 55). Total WFC lending was 3.1 percent of the NBER estimate of gross farm product and 2.9 percent of the BEA estimate of gross farm product.Footnote ²⁶ While total lending was relatively small, it was not necessarily ineffective. Also, as federal reserve bank credit was declining, WFC lending was the only important source of assistance during this period.

Bank suspensions, which were low in the middle months of 1921, spiked up in the final three months of the year, just as WFC lending began.Footnote ²⁷ Bank suspensions began a decline in February 1922 and the total for the year was the lowest of any year in the decade and much lower than the yearly average for the decade (366 compared to 635). Figure 4 depicts monthly WFC authorizations and advances to financial institutions, co-ops and livestock loan companies, and bank suspensions for the years 1921 through 1929.

Figure 4.

WFC advances, authorizations and bank suspensions

Sources: Acting Secretary of the Treasury, Liquidation of the War Finance Corporation, and Board of Governors of the Federal Reserve System, Federal Reserve Bulletin, September 1937.

WFC officials reported that the new lending authority was successful, providing needed relief to farmers and the financial system.Footnote ²⁸ They maintained that their loans not only helped weak banks, but also strong banks that had not received WFC funding, as with WFC loans helping weak banks, the strong banks were more willing to function normally. They also asserted that just authorizing loans, regardless of whether the funds were advanced, provided relief as banks knew funds were available (War Finance Corporation 1922a, p.14):

The WFC's Fifth Annual Report stated that lending was especially important for nonmember banks.Footnote ²⁹ If WFC loans ended forced liquidation that would force weak banks to suspend operations and possibly fail, then WFC lending could be a reason, at least in part, for the drop in suspensions in 1922.

Loan applications peaked in December 1921, and the Fifth Annual Report also states that the results of its lending began to have an effect in January 1922. WFC loans ‘strengthened the banking situation in the country districts and relieved the necessity of forced liquidation’ (War Finance Corporation 1922b, p. 3). Figure 5 depicts the raw data for several states on WFC advances and bank suspensions in 1921 and 1922. The first bar ranks each state by WFC advances in millions of dollars, for total advances greater than $3 million.Footnote ³⁰ The second and third bars are the number of suspensions in each state for 1921 and 1922 respectively. In most states the number of suspensions decreased, but the positive impact was not universal. In five states, Kansas, Oklahoma, Montana, South Carolina and South Dakota, suspensions increased in 1922 compared to 1921. In three of these states, Kansas, Montana and South Dakota, total bank debits were lower in 1922 than in 1921, indicating that the economies of these states remained weak in 1922. Combined suspensions for the states in Figure 5 decreased from 357 in 1921 to 264 in 1922.

Figure 5.

WFC advances and bank suspensions

Sources: WFC advances and authorizations – Records of the War Finance Corporation. National Archives Record Administration II. Record Group 154. Bank suspensions by state by month – Records of the Federal Reserve System. National Archives Record Administration II. Record Group 82.

IV

The analysis in this section explores whether the introduction of the WFC lending program in November 1921 caused a reduction in bank suspensions. Given the relatively short period of sizable WFC lending, the analysis should be conducted by state by month. Monthly data for bank suspension by state are available in the Federal Reserve archives (Records of the Federal Reserve System, box 1851, National Archives and Records Administration II, Record Group 82). Bank debits are used as a measure of economic activity. Monthly bank debits data by reporting center are summed to obtain monthly data by state (Federal Reserve Board 1924). Only debits from centers having complete data for both years are used. Also, New York City debits were excluded from the New York state data. Bank debits data are not available for the state of Vermont, which was dropped from the sample. The District of Columbia is also excluded as it had no measured agricultural sector. Wheat prices are a national variable, the same for all states (National Bureau of Economic Research, Series 04001a).

Published WFC data do not include monthly lending by state. A search of the WFC archives found monthly data by state for the number and value of advances and authorizations only beginning in June 1922, after most of WFC loans were extended (Records of the War Finance Corporation, National Archives and Record Association II, Record Group 154).Footnote ³¹ The June 1922 data are cumulative totals of authorizations and advances by state. Data by state for the individual months of November 1921 through May 1922 could not be found.

Because of this data limitation, we take advantage of a sharp discontinuity in WFC lending from no lending prior to November 1921 to WFC support during the period from November 1921 to December 1922.Footnote ³² Since there is no available cross-sectional variation in the implementation of the WFC lending, a difference-in-difference framework is not possible.Footnote ³³ Thus, this regression discontinuity in time design compares two groups of bank suspensions that are very similar except for the treatment, which depends discontinuously on the WFC lending cutoff. On the other hand, a different approach would compare two groups that could have some pre-existing differences on top of treatment, but the effect of that treatment is assumed to be constant over time.

Furthermore, there are many potential time-varying confounding factors. The estimates include measures to control for the level of economic activity (bank debits), the farm economy recovery (wheat prices), and seasonal and state fixed effects. Other less important factors are assumed to change smoothly across the date of the implementation of WFC lending. It is also reasonable to assume that the start date of WFC lending is not randomly assigned. Thus, our regression discontinuity in time framework can be thought of as the ‘discontinuity at a threshold’ interpretation of regression discontinuity and not as local randomization (Hausman and Rapson Reference Hausman and Rapson2018). While the regression discontinuity requires different assumptions and less data compared to a difference in difference design, its estimates capture a more local effect around the WFC lending cutoff.

We estimated a regression-discontinuity in time (RDiT) design, in which time is the running variable that determines whether US banks are exposed to WFC lending or not. More specifically, we use the following regression:Footnote ³⁴

$$\eqalign{& Suspensions_{it} = \alpha + \beta _0WFC_t + \beta _1Debits_{it} + \beta _2Wheat\;Prices_t \cr & \,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\,\, + {\rm \;}\mathop \sum \limits_{k = 2}^{47} \mu _k( {\pi_i} ) + \mathop \sum \limits_{m = 2}^{12} \omega _m( Q_m) + {\rm \;}\varepsilon _{it}} $$

where Suspensions _it is the number of bank suspensions in state i and month t. Because the outcome variable is a count process adopting only positive integer values, the regression is estimated using a negative binomial model.Footnote ³⁵

The variable WFC is a dummy variable taking the value 1 for all the months after WFC lending started in November 1921, and zero otherwise. The variable Debits _it contains the annualized change in the log of monthly bank debits by state. These represent the turnover of bank deposits, which vary with the business cycle. This variable is intended to capture state-specific business cycle dynamics. The variable Wheat Prices_t is the annualized change in the log of monthly wheat prices, as the recovery of prices increased farm income. Finally, we control for monthly and state fixed effects, {Q _m} and {π _i}, respectively, to account for seasonality in bank suspensions, as well as socioeconomic differences across the 47 states included in our analysis.

WFC lending was concentrated in those states having experienced significant banking suspensions. Banks in 11 New England and northeastern states and the District of Columbia received no WFC loans.Footnote ³⁶ A single bank in New York received one WFC loan. Thus, an entire section of the country, where agriculture was less important and banking troubles and suspensions were rare, received no WFC loans. WFC lending was highest in those states having a large number of bank suspensions.

Table 1 displays descriptive statistics on US bank suspensions before and after WFC lending. The monthly average rate of suspensions for all states decreased by a statistically significant 13 percent once WFC lending began. Four states, Iowa, Montana, North Dakota and South Dakota, received 25 percent of WFC lending. For this group of states, the rate of suspensions dropped 25 percent, although the decrease is not statistically significant. Similarly, for the 11 states in Figure 5, the decrease in means is 11 percent, but not statistically significant. However, for both groups of states, the decrease in the variance is significant.

Table 1.

Descriptive statistics (monthly average suspensions by state)

Notes: Statistical significance based on the Van der Waerden non-parametric test. *** difference is statistically significant at the 1% level. ^a difference in variances is statistically significant at the 1% level, based on the Bartlett test.

Group 1: Iowa, Montana, North and South Dakota

Group 2: Georgia, Illinois, Iowa, Kansas, Minnesota, Missouri, Montana, Nebraska, North Carolina, North Dakota, Oklahoma, South Carolina, South Dakota, Tennessee, Texas, Utah, Wisconsin

Table 2 presents estimates of the regression model described above for the WFC lending program using different specifications. Column 1 presents estimates from a model that controls only for seasonality, state fixed effects, and the regression discontinuity dummy variable for the WFC lending program. Column 2 adds to the specification in column 1 a control for state business cycles, namely the change in the log of bank debits. Our baseline specification in column 3 adds the change in the log of wheat prices. The coefficient of interest, β ₀, captures the causal effect of WFC lending on the number of bank suspensions. A negative and statistically significant coefficient would provide evidence that the WFC's lending program decreased the number of bank suspensions.Footnote ³⁷

Table 2.

Regression estimates (RDiT) of the WFC lending on bank suspensions

^a Debits and wheat prices are modeled as annual percentage changes to capture the impact of changes in the business cycle.

Notes: The unit of observation is state-month. Standard errors (in parentheses) are clustered at month level. The symbols ***, **,* indicate that coefficients are statistically significant at the 1%, 5%, 10% level, respectively. All the models include state and seasonal fixed effects.

Overall, the monthly estimates provide support that WFC lending reduced bank suspensions. The WFC variable is statistically significant in all specifications. The magnitude of the effect of WFC lending is presented in terms of incidence rate ratios (IRR). According to the estimates in Model 3, WFC lending reduced US bank suspensions by 24 percent (~e ^−0.28 − 1) between November 1921 and December 1922. The incidence rate ratios for the model three estimates are reported in Table 3.

Table 3.

Negative binomial regression results from Model 3 as incidence rate ratios

^a Incidence rate ratios (IRR) are obtained by taking the exponent of the coefficient estimates.

The bank debits variable is a proxy for the business cycle and is a state by month measure. This variable is also statistically significant at the 10 percent level, indicating that as state economies recovered from the recession, bank suspensions fell, although the magnitude of the effect (IRR) is small.Footnote ³⁸

Wheat prices is a month only variable, so it is the same for all states. One quarter of the total funds advanced went to four states – Iowa, Montana, North Dakota and South Dakota – and several other states receiving substantial WFC funding were important wheat producers. Increased wheat prices enabled more profitable sales, providing farmers with the funds needed to repay their loans. The estimates in Model 3 are supportive, as the wheat prices variable has a negative but insignificant impact on suspensions. The estimated IRR is only -0.06 percent, but wheat prices fell substantially, 64.4 percent from their high to low, indicating that the fall in wheat prices may have contributed to bank suspensions.

As noted above, anticipation effects may affect the RDiT estimates. Hausman and Rapson (Reference Hausman and Rapson2018) recommend estimating a ‘donut’ RDiT, removing the observations around the threshold to determine if anticipation effects bias the results. Model 3 is re-estimated deleting the months of September, October and November of 1921, the months around the threshold. The results, reported in Table 4, are essentially unchanged compared to the results for Model 3. Thus, anticipation effects appear not to matter.

Table 4.

Regression estimates of the WFC lending on bank suspensions

The ‘donut’ model is estimated on truncated data, where observations around the date of the start of WFC lending, September–November 1921, were removed to mitigate concerns about short-run anticipation effects.

^aDebits and wheat prices are modeled as annual percentage changes to capture the impact of changes in the business cycle.

Notes: The unit of observation is state-month. Standard errors (in parentheses) are clustered at month level. The symbols ***,**,* indicate that coefficients are statistically significant at the 1%, 5%, 10% level, respectively. All the models include state and seasonal fixed effects.

In sum, the estimates support the hypothesis that WFC lending did contribute to the reduction of bank suspensions. Increased economic activity, as proxied by bank debits, also provided relief. Wheat prices increased slightly from January 1922 through the end of the year but remained low relative to previous years. The estimates support the view that WFC lending did reduce bank suspensions during the period of aggressive lending.

V

The 1920–1 recession was a severe downturn coupled with a significant deflation. The deflation of farm prices was much greater than the consumer price deflation. Farm debt had expanded significantly in the previous decade. The increase in nominal debt, combined with the fall in farm prices and the change in buying patterns for farm output imposed serious financial burdens on the agricultural sector and the banks with which farmers dealt. Farmers were forced to carry stocks of output that previously had sold quickly. Financing of these stocks proved difficult, as Federal Reserve rules limited agricultural loans to six months. The methods of financing that had been effective before the recession now proved inadequate.

Federal Reserve banks in agricultural districts lent liberally in 1920, financed by borrowing gold reserves from Reserve banks located in primarily industrial districts. At the end of 1920, discount lending began a steady decline. Eugene Meyer, managing director of the War Finance Corporation, requested lending authority to assist farmers and their banks. Congress passed the Agricultural Credits Act of 1921, approving the requested authority.

The WFC quickly established a lending program, providing funding to banks, marketing coops, and livestock loan companies that in turn would lend to banks. Although lending was largely concentrated in the months of November 1921 through April 1922, WFC authorities felt their lending provided significant relief to farmers and to the banking system.

A quasi-experimental approach is used to study the causal impact of WFC lending at the state level between November 1921 and December 1922 on US bank suspensions. According to regression discontinuity in time estimates, WFC lending decreased the number of bank suspensions by about 24.2 percent. Improved economic activity and possibly increased wheat prices also contributed to reduced banking difficulties.

Although the WFC only redirected existing funds, it did not create new money, its success in limiting bank suspensions demonstrates the positive impact that lender of last resort operations can have on the financial system.