Structural Causes of Rural Bank Failures in the Twenties: Parallels with the Eighties

Lee J. Alston (University of Illinois)
Wayne Grove (University of Illinois)
David C. Wheelock (University of Texas)

August 1990

Until recently, bank failures seemed to be a problem of the past. From 1934 to 1980 there were few failures as deposit insurance, heavy regulation, and relatively steady economic growth made banking a particularly stable industry. But, since 1980 bank failures have surged, as illustrated in Figure 1.

Researchers attribute the rise in bank failures during the 1980s to two principal factors: i) heavy loan defaults following sharp downturns in the agriculture, petroleum and real estate sectors, and ii) deregulation of the banking and savings and loan industries without concomitant deposit insurance reform. Because deposit insurance removes the incentive for depositors to monitor the health of their banks, financial institutions are freed from having to pay risk premia on their customers' deposits. As a result, banks and S&L's have an incentive to make more risky investments than they would otherwise. Deregulation gave banks and S&L's more freedom to attract deposits and allowed them greater latitude in their choice of investments. In other words, it permitted them to exploit the incentives extant with deposit insurance. They did so, lending heavily to the agricultural, energy and real estate sectors during their booms, only to face large defaults as each went bust.

Calomiris (1989) and Thies and Gerlowski (1989) have recently examined the relationship between deposit insurance and banking instability in the early part of the twentieth century. While there was not federal insurance of bank deposits until 1933, there were insurance funds in eight states, and both studies conclude that the presence of deposit insurance increased the number of bank failures during the 1920s.

There are a number of other similarities between the bank failures of the 1980s and those of the 1920s. In both decades, failures seem largely attributable to conditions in certain sectors or regions of the country. For example, the failure rate of agricultural lenders exceeded that of other banks in both periods. From 1982-87, the closure rate of agricultural banks was 1.95 per thousand, which was statistically greater than the 1.35 per thousand closure rate for non-agricultural banks. During the twenties bank failure was also largely a rural phenomena. From 1921-1929, 79% of all bank suspensions occurred in towns of less than 2500 population. Comparisons for each year are plotted in Figure 2. The failure rate of rural banks was nearly twice that of banks located in cities larger than 2500 population. The relatively high frequency of failure among agricultural banks during the 1980s has been attributed to the boom in agricultural land prices in the late 1970s which was not followed by a proportional increase in agricultural income. Johnson (1973) speculates that a similar boom-bust cycle in agriculture following World War I led to the relatively high frequency of rural bank failures during the 1920s. Government lending to agriculture also seems to have accentuated the problem of rural bank failures in both decades. For the eighties, Smith (1987) argues that "[b]ank lending to agriculture was sorely pressed by intense competition from the cooperative Farm Credit System, particularly during the 1970s and early 1980s." We believe a similar phenomena occurred during the 1920s when increased competition from new federal agricultural lenders forced many rural banks out of business.

This paper takes a new look at the causes of rural bank failures during the 1920s. Previous research has highlighted the fact that a large proportion of failing banks during this decade were small, rural banks, and hence identified changes in agricultural income as a leading cause of bank failure. But none has focused on the particular channels through which agricultural distress caused rural bank failures, nor captured the changing impact of agricultural distress over time. We have constructed a new data set, consisting of annual failure rates of banks located in towns of less than 2500 population for 1921 to 1929, which enable us to examine closely the importance of agricultural distress and government policy on the largest class of bank failures during this decade.

II. Bank Failures in the Twenties

Small banks in small towns covered the U.S. in the twenties. Of the nearly 29,000 banks in 1920 two-thirds were in towns of less than 2,500 population. By the end of the decade many of these banks had failed. The number of banks in towns less than 2500 fell by 27% over the twenties, despite the fact that the population in those communities increased marginally over the period. The corresponding decline for banks in towns greater than 50,000 was less than 0.4%. By 1929 there were about 23,700 banks in the U.S., compared to some 29,000 banks in 1921.

Banks began to fail with the general economic downturn of 1920. For the U.S. as a whole about 500 banks failed in 1921. Failures continued to mount in the early twenties, averaging over 680 from 1923 to 1929, and peaking in 1926 at more than 950 failures.

The national averages obscure the considerable regional and urban/rural variation. Figure 2 highlights the disproportionate share of rural bank failures: of the 5712 bank suspensions during the 1920s, 4515 (79%) occurred in towns of less than 2500 population. In Figures 3 and 4 we illustrate the regional variation in rural bank failures. In the first part of the decade the Mountain, Plains and Southeastern states suffered the most banking distress. Banks on the West Coast, those north of the Mason/Dixon Line, and those east of the Mississippi River fared well. Failure rates ranged from 0 in New Hampshire, Vermont, Massachusetts and Connecticut to 12 per thousand in Montana. In the latter part of the twenties the West North Central and Southeastern states experienced the highest failure rates, and again the Northeastern states suffered little. Failure rates ranged from 0 in seven Northeastern states to 15 per thousand in Florida.

III. The Causes of Rural Bank Failures

Contemporary observers and historians have suggested at least three general causes of rural bank failures during the twenties. Most scholars contend that adverse conditions in agriculture in the twenties, or ex-post overly optimistic expectations of future agricultural prices in the late teens, was the leading cause of the increase in bank failures during the 1920s. O. M. W. Sprague, professor of banking and finance at Harvard, testified before the House Committee on Banking and Currency in 1930 that the great majority of banks failed because they were unable to withstand the stress exerted by the persistence of unprofitable prices for the products of agriculture and animal husbandry -- stress that was particularly severe because it was experienced after years of a bounding prosperity and extreme appreciation in value of farm property, and a large increase in the number of farms mortgaged and the amount of mortgage indebtedness.

Whether agriculture was depressed in the twenties is still debated; it depends on one's standards. If the late teens are used as the base, then agricultural income in the twenties was depressed, but if one compares agricultural income or its growth with the pre-war period, then conditions in the twenties appear normal. Our view is that there was no general agricultural depression, and that the majority of farmers fared well in the twenties. Nonetheless, a significant minority, who had entered into farmland mortgages in the late teens, experienced systematic distress in the twenties and hence put pressure on rural banks who held the mortgage debt. In addition to indebtedness, all farmers were squeezed by increased real estate taxes and deteriorating terms of trade. An index of real estate taxes of 100 in 1920 rose to 114 by 1929. Wholesale prices weakened during the 1920s reversing a longstanding favorable relationship between farmer's output prices and prices paid for inputs and consumer goods.

In a sense, the reason some farmers had trouble in the twenties is because times were too good from World War I to 1920, particularly livestock prices and the prices for the cash crops of cotton, wheat and corn. Land values soared in states such as Iowa where most of the land was already under the till: land value in 1920 as a percentage of 1912 was over 220%. Land values increased over 100% in Minnesota, North Carolina, South Carolina, Georgia, Kentucky, Tennessee, Mississippi, Arkansas and Louisiana. In more marginal farming areas improved acreage increased dramatically: improved acreage increased over 150% between 1910 and 1925 in all the Mountain states plus Texas and Oregon. In Wyoming, New Mexico and Arizona improved acreage increased over 1000%. Crops were planted on much of the land west of the 100th meridian for the first time, prompted by more liberal homesteading provisions, high wartime prices, and relatively abundant rainfall. Further west cattle took over and herds increased spectacularly during the boom years of the teens. The number of cattle in Wyoming, for example, nearly doubled between 1914 and 1918.

The agricultural expansion brought with it an increase in mortgage debt, much of which was held locally by banks and private individuals. Owner- operated farms having mortgage debt increased from 33% in 1910, to 37% in 1920, and to 42% in 1930. The increase in the proportion of farms mortgaged was particularly great in western states. In North Dakota 71% of owner-operated farms carried mortgages in 1920, and 50 to 60% of farms in neighboring states were similarly encumbered. From 1910 to 1920 mortgage debt per acre of farm land increased 135%, about the same as the rise in the per acre value of ten leading crops. When the agricultural boom ended in 1920, indebted farmers had difficulty meeting their mortgage payments and farm failures reached historic highs. As farms failed the value of farm debt in many banks' portfolios fell, causing their insolvency. Small rural unit banks with few resources, limited facilities, and restricted activities found it particularly difficult to cope. Agricultural distress was not the only factor causing bank failures. Various state and federal policies may have augmented or lessened the likelihood of bank failure. At the federal level, in 1916 Congress passed the Federal Farm Loan Act which created joint stock land banks, federal land banks and the Federal Farm Loan Board which oversaw the activities of the "federal" banks. The legislation was intended to foster competition among lenders and to make low cost credit available to farmers in all regions. Yet, there may have been good reasons why some farmers had been unable to obtain credit from private sources: the loans would have been too risky. The timing of the legislation was also unfortunate. By encouraging the entry of banks into the farm lending market at the height of the agricultural boom the legislation may have caused "excessive" competition. Local rural banks, who already were in the business of supplying credit, may have been forced to lower interest rates and relax standards below long-run profit maximizing levels. Consider the following excerpt from a study in Indiana: "Receivers, liquidating agents, and other persons familiar with the affairs of failed banks suggested, in 41 instances...that many of the practices leading to bank failures, were directly caused by 'cut-throat' competition which sprang up in various communities as a result of too many banks." When the agricultural prosperity bubble burst in 1920 the excess number of agricultural lenders had to be eliminated.

Joint stock land banks were private institutions but had the privilege of issuing tax-exempt bonds. In return they were subject to supervision by the Federal Farm Loan Board and a cap was put on the interest rates they could charge. Many joint stock land banks arrived on the scene at, or just following, the peak of agricultural prosperity: of the eighty-three banks chartered in the first ten years, twenty-one were created in 1919 and forty in 1922. The lending presence of joint stock land banks varied considerably across regions: as an average yearly percentage of total farm mortgage debt, joint stock bank lending was most concentrated in South Carolina (11%), North Carolina (11%), Illinois (9%), Kentucky (9%), West Virginia (8%), Ohio (7%), Indiana (7%), Wyoming (7%) and Oregon (7%). Joint stock bank lending was completely absent in seventeen states. Federal land banks held an even greater share of the outstanding farm mortgage debt, and therefore posed more of a competitive threat in more states [include 1922 figures here from the source I gave a copy of??]. In 1920 federal land banks held 3.5% of total farm mortgage debt whereas joint stock banks held less than 1%. The federal land banks were non- profit farm cooperatives and like joint stock banks their presence varied regionally. Federal land banks made relatively more loans in the twenties in the Southern, Mountain and far Northwest Coast states. From 1921 to 1925 the average yearly percentage of farm mortgage debt held by federal land banks varied from 2% in New Jersey and California to 18% in Utah. The link between the new federal lending institutions and commercial bank failures was fairly direct, while the impact of state policies on bank failures was less so. Branch banking and deposit insurance were state policies that may have affected the likelihood of bank failures. Branch banking may reduce a bank's susceptibility to failure for any given downfall in deposits or assets prompted by distress in a particularly location. To the extent that farm failures were somewhat localized, branch banking would allow a troubled bank to draw on funds from other branches within its system which were not affected by farm distress. This would most likely be the case where branches covered both rural and urban areas.

Branch banking was quite limited in the United States during the 1920s. Over half the branches in the country were located in California, New York City and Detroit. Only ten states permitted state-wide branch banking, most in New England, but also a few southern states and California. Others limited branches chiefly to certain large cities, and the majority of banks with branches had only one branch, most of which were located in the same town as the head office.

Deposit insurance was even less prevalent than branch banking: eight states had deposit insurance schemes in the early twenties and only six after 1925. The Great Plains states along the 100th meridian favored deposit insurance. North Dakota, South Dakota, Nebraska, Kansas, Oklahoma and Texas, along with the nonplains states of Mississippi and Washington, had deposit insurance in the early twenties. By 1923 the Oklahoma system had failed, as did Texas' in 1925.

Deposit insurance eliminates the incentive for individuals to monitor the health of their banks or to demand risk premia on deposit interest rates (at least to the extent of insurance coverage). Because precise information on the quality of a bank's portfolio is costly to acquire, prior to the introduction of federal deposit insurance in 1933 the failure of individual banks sometimes triggered general deposit runs which brought down numerous otherwise healthy banks. Deposit insurance deters runs because depositors do not fear losing their funds in the event of bank failure. Hence even if they observe an increase in the probability of their bank failing, depositors have little incentive to withdraw their funds. Although deposit insurance tends to limit the likelihood of widespread banking panics, the U.S. experience with deposit insurance has not been altogether successful. In essence, most deposit insurance schemes have failed precisely because they remove the incentive for depositors to monitor the performance of their banks. Because bankers have little of their own capital exposed, the absence of monitoring by depositors encourages banks to undertake more risky investments than they would otherwise. Indeed, fraud may become more prevalent as bankers find it easier to escape immediate detection. Fraud or incompetence are frequently cited as causes of bank failures in the twenties. Ex-post fraud and incompetence are difficult to untangle as is incompetence and overly optimistic expectations. Incompetence as judged with hindsight is more likely to occur in areas typified by booms and busts. For example, in Wyoming it was noted that there was a "lack of bankers to run the rapidly growing number of banks"[Cite ??] Incompetence ex-ante should not vary across regions unless one assumes a constraint on the human capital necessary for banking.

Deposit insurance alone may not be sufficient to lead to fraud or mismanagement. First, depositors could and did create an incentive for bank owners to act properly by making bankers residual claimants and by forcing them to expose their capital to risk. Second, in the twenties there was not a significant principal/agent problem between shareholders and bank managers because the majority of rural banks were closely held and managed by shareholders. For example, a majority of the stock of 82% of South Dakota banks in 1930 was held by five or fewer individuals. Nevertheless, deposit insurance does limit the incentive to monitor and if the capital of the residual claimants has been eroded by an exogenous shock such as farm distress, then bankers have little to lose by engaging in fraud or excessive risk-taking. For this reason if deposit insurance did have an impact on bank failures we expect it to interact with farm distress, i.e., in states with deposit insurance funds we expect a larger impact of farm distress on bank failures.

A further influence leading to the high rate of farm failures in the twenties was a dramatic reduction in the cost of transportation. Beginning in the middle teens, many rural residents purchased automobiles and country roads were improved dramatically. Both factors made it easier to bank further from home at banks in larger towns which could offer better terms. In the language of central place theory, the automobile changed the network of interdependent cities which were based on a hierarchy of central services. Transportation costs determine the minimum and maximum sizes of city's market area and growth potential. In addition to economies of scale in banking there are most likely economies of scope in shopping. That is, there is a given fixed cost associated with going to town. Once there, the marginal cost of additional shopping declines. It is not that banking was the sole or indeed the primary reason for changing a trade center but it was a part of the transactions that rural residents had to get done while in town. As one contemporary statement makes clear, "People who formerly did their business in the local small town can now get to the county seat or a larger town in a few minutes. They go there because there is a movie there and better stores, etc. Naturally, their banking has followed."

For the U.S. as a whole the estimated average percentage of farmers purchasing automobiles for the first time was 37% for 1915-1919; 20% for 1920-1924; and 15% for 1925-1929. Though there was considerable variation across states in the percentage of automobile ownership, the absolute percentage increases were remarkably similar. The variation in automobile ownership in 1920 ranged from 5.5% in Mississippi to over 75% in Nebraska, and by 1930 from 26% in Louisiana to 92% in Nebraska. In contrast, the absolute percentage increase from 1920 to 1930 ranged only from 12% in Arizona to 40% in Delaware, and in all but seven states the increase was between 20 and 40%.

IV. Rural Bank Failures during the Twenties: Testing the Hypotheses

Hypotheses abound to explain why so many banks failed in the twenties but there is a lack of empirical testing, in part because the data on rural banks is not easily obtainable. We have computed annual failure rates for commercial banks located in towns of less than 2500 population for each state from 1921 to 1929. Bank failure was primarily a rural phenomena during the twenties, and by focusing on bank suspensions in small towns only we hope to identify more clearly the channels through which agricultural distress and government policy affected the likelihood of failure.

To test the hypotheses developed in the previous section we use state level data to estimate the parameters of the following regression equation: Rural Bank Failure Rates = b0 + b1 Agricultural Distress + b2 Federally Sponsored Banks + b3 Branch Banking + b4 Deposit Insurance + b5 Improved Transportation + b6 Farm Population as % of Rural Population. We divide the twenties into two periods, 1921-25 and 1926-29, to allow us to test whether different factors influenced failures in the early twenties as compared to the later twenties.

If agricultural distress caused banks to fail we expect b1>0. For the earlier period we have two proxies for agricultural distress: the expansion of improved acreage from 1910 to 1920 and the increase in land values from 1910 to 1920. We use these variables because we do not have state level data on farm failures. However, the extant evidence suggests that farm failure rates were greatest in the early twenties in those areas that experienced the most expansion in improved acreage and the greatest increases in land values. For the late twenties our proxy for agricultural distress is the yearly average farm failure rate.

If federally sponsored banks crowded out private banks then we expect b2>0. Our proxy for federally sponsored banks is the percentage of outstanding farm mortgage debt held by joint stock land banks and by federal land banks. If branch banking spread risks sufficiently to prevent rural banks from failing then we expect b3<0. Our proxy for branch banking is a dummy variable which is set equal to 1 in branching states and equal to 0 in unit banking states. If deposit insurance increased the likelihood of fraudulent banking practices then expect b4>0. Our proxy for deposit insurance is also a zero/one dummy variable.

If improvements in transportation made some small rural banks no longer economically viable, then we expect b5>0. Our proxy for improvements in transportation is the annual percentage increase in automobiles on farms. We include the percentage farm population in the regression analysis to control for the varying importance of the farm sector across rural areas. We expect that the more concentrated in farming is the rural population the greater the likelihood of bank failures. If true, then we expect b6>0. We present the regression results in Table 3. The dependent variable is the number of rural bank suspensions per 1000 banks. Equations 1 and 2 are the results for 1921-25. For these years our results suggest that no single factor alone can account for bank failures. Overall the variables have considerable explanatory power. The agricultural distress variables perform as predicted with more explanatory power for improved acreage [confirm this by looking at coefficients and standard deviations]. The coefficients on the state banking variables, branch banking and deposit insurance, are not statistically different from zero, suggesting that these policies had little impact on bank failure rates.

It could be that branch banking and deposit insurance affected the likelihood of bank failures less directly. Branch banking may be a safeguard when agricultural stress occurs and conversely once agricultural stress sets in deposit insurance may exacerbate its impact. We test for this by interacting our agricultural distress variables with the branch banking and deposit insurance dummy variables (see Equation 2). The results are very similar to those in Equation 1 except that there is now a positive association between deposit insurance and bank failures. The coefficients on the interactive terms are not statistically significant, however. Nor are the coefficients on branch banking and its interactive term.

Equations 3 and 4 are the results for 1926-29. Like the earlier period, the overall explanatory power is good and agricultural distress provides most of the explanation. The state banking laws do not offer much explanation and the role of the federal banks seems to have played itself out by the late twenties. Though no doubt causing bank failures in some instances, the increased use of the automobile is not well correlated with bank failures in the aggregate.

In testing for the interactive effects of branch banking and deposit insurance (Equation 4), we get no support for either variable reducing or augmenting bank failures in the presence of agricultural distress. The regression analysis suggests agricultural distress was the dominant explanation of bank failures in the twenties. Most likely this is due to the lack of diversification by rural banks: they held too many of their assets in farm mortgages. As indicated by the qualitative evidence presented in the previous section, improvements in transportation drove some rural banks out of business but our regression analysis indicates that the total impact from the adoption of the automobile was small. If federally sponsored banks crowded out private banks, our results indicate that they did so in the early twenties, shortly after their introduction. It appears that the increase in the federally sponsored banks' share of the farm credit market came at the expense of private individuals. Branch banking appears not to have influenced bank failures but this is no surprise given the limited amount of branch banking in rural areas. Our results for the impact of deposit insurance are the least stable and therefore we are leery of embracing an interpretation as for its role.

V. Conclusion

The parallels between the banking and savings and loan failures of today and the bank failures of the 1920s are striking. Aggregate economic growth was strong in both the twenties and the eighties, but agriculture and some natural resource industries, such as petroleum, suffered declining fortunes following heady booms. In the 1980s falling agricultural and oil prices depressed real estate values, increased loan defaults, and pushed numerous financial institutions into failure. The situation was exacerbated by rapid increases in real estate prices and loans in the preceding boom. The same sequence of events characterized the twenties. Expectations of a continuation of rapidly increasing demand for farm output proved too optimistic ex-post, and forced heavily indebted farmers and their bankers out of business.

Government policy played a role in the bank failures of both the twenties and the eighties as well. Many of the government subsidized federal land banks and joint stock land banks opened just as farm prices were beginning to fall and rural lenders were experiencing loan losses. Banks in those states where the government sponsored lenders captured a relatively large share of the market suffered a higher rate of failure than those in states where the presence of the government banks was not large. Similarly, in the eighties, competition from government lenders has been cited as a contributory factor in rural bank failures.

Deposit insurance and the incentives it creates for risk taking are often cited as the chief culprits in recent bank failures. And other researchers have concluded that deposit insurance also contributed to bank failures during the 1920s. Our study has found the relationship between deposit insurance and bank failure to be sensitive to specification, however, and we are unable to offer any conclusive evidence on the issue. Similarly, we do not detect any clear evidence of a relationship between branch banking and bank failure during the twenties. Both deposit insurance and branching were quite limited and there was variation across states in the nature of branching laws and insurance plans, making it difficult to identify precisely the impact each had on the failures of the twenties. Moreover, of the few states with deposit insurance, a majority were among those states suffering the greatest agricultural distress, again making it difficult to disentangle their separate roles on bank failures.

Friedman and Schwartz (1963, pp. 434-442) emphasize the contribution of Federal deposit insurance to the reduction in bank failures since 1933. Gambs (1977, p. 20) writes that "much of the decline in the rate of bank failures since the 1930's can probably be attributed to the improved stability of the economy in the post-World War II period."

For comparison, from 1930-33 there was an average of 2274 bank suspensions per year (Federal Reserve Bulletin, September 1937, p. 868). The average for 1934-79 was 15 per year, and from 1943-74 there never was more than 9 failures per year (Federal Deposit Insurance Corporation Annual Report 1987, p. 49).

See, for example, Kaufman (1987, pp. 208-210) and Garrison, Roger W., Short, Eugenie D. and O'Driscoll, Gerald P, "Financial Stability and FDIC Insurance," in The Financial Services Revolution, Policy Directions for the Future, edited by Catherine England and Thomas Huertas (Boston, 1988), pp. 187-207. For one of the first articles to note the potential pitfalls of bank deregulation without deposit insurance reform see Kareken, John H., "Deposit Insurance Reform: or, Deregulation is the Cart, Not the Horse," Quarterly Review, Federal Reserve Bank of Minneapolis, Spring 1983, pp. 1-9.

Numerous studies analyze the economics of deposit insurance, including Garrison, et. al., "Financial Stability...", and Calomiris and Kahn (forthcoming).

Thies, Clifford F. and Gerlowski, Daniel A., "Deposit Insurance: A History of Failure," Cato Journal Vol. 8 (no. 3), Winter 1988, pp. 677-693. Calomiris, Charles, "Do "Vulnerable" Economies Need Deposit Insurance?: Lessons from the U. S. Agricultural Boom and Bust of the 1920's," working paper, Northwestern University, October 1989.

This is in contrast to the 1930s when, except for the initial failures, bank failures occurred in all parts of the country and were associated with panics and the general economic depression.

Smith, Hilary H., "Agricultural Lending: Bank Closures and Branch Banking," Economic Review, Federal Reserve Bank of Dallas, September 1987, p. 32. Here, an "agricultural bank" is defined as one whose ratio of agricultural loans to total loans exceeds the average of such ratios for all banks. Because of data limitations, this is not the definition we adopt in this paper. Bank "closure" is a more restrictive measure than bank "failure," as the former includes only those insolvent banks which are closed or sold to other banks, while the latter also includes those insolvent banks which regulators permit to remain open.

Federal Reserve Bulletin (September 1937, p. 901).

From 1921-1929, the annual failure rate for banks located in towns under 2500 population was 2.5%. For banks in larger towns the failure rate was 1.3%. These rates are calculated by taking the ratio of the total number of bank suspensions from 1921-29 to the total number of banks operating on June 30, 1920 and dividing by 9. Hence, the figures are not adjusted to account for new entrants. The figures also exclude private banks. Source: Federal Reserve Bulletin, September 1937, pp. 901-906).

See Belognia, Michael T. and Gilbert, R. Alton, "Agricultural Banks: Causes of Failures and the Condition of Survivors," Review, Federal Reserve Bank of St. Louis, May 1987, pp. 30-37.

Johnson, H. Thomas, "Postwar Optimism and the Rural Financial Crisis of the 1920's," Explorations in Economic History Vol. 2, no. 2, Winter 1973-74, pp. 171-192.

Smith, "Agricultural Lending," p. 32.

Gambs, for example, found that variation in agricultural income was a primary determinant of state differences in bank failure rates during the 1920s. Gambs, Carl M., "Bank Failures--An Historical Perspective," Monthly Review, Federal Reserve Bank of Kansas City, June 1977, pp. 10-20.

Federal Reserve Bulletin, Sept. 1937, p. 906.

Ely, Bert ..., 1986, p. 33.

The net decrease of 5,506 banks masks the entries that occurred: 9,928 total banks exited the market, while 4,422 entered (B, G, and C Banking Commission, Vol. 3: 53).

Hearing before the House Committee on Banking and Currency 1930, p. 18- 21.

For the view that depressed agricultural prices are in part to blame for rural bank failures, see Hearings of the House Committee on Banking and Commerce (1930) and Bremer (1935). Johnson (1973) argues that the post World War I speculative bubble is to blame for farm failures and rural bank failures. On the other hand, Charles Holt observes that "the evidence indicates that the farmer was one of the few beneficiaries of the prosperity of the twenties" (Holt 1977, p. 284).

U.S. Department of Agriculture, Bureau of Agricultural Economics 1942, p. . U.S. Department of Commerce, Bureau of the Census 1927, State Table 1 and U.S. Department of Commerce, Bureau of the Census 1913, Volume 5 Table 24.

Huntoon 1982, p. 36.

Branch, Chain, and Group Banking Study by the Banking Commission, Vol. 10: 30-31.

Alston, Lee J., "Farm Foreclosures in the United States During the Interwar Period," Journal of Economic History Vol. 43, no. 4, Dec. 1983, pp. 885-903. The best discussion of the federal government's role in farm lending is found in Horton, Larsen and Wall (1942).

Report of Study Commission for Indiana Financial Institutions 1932, p. 87. Horton, Larsen and Wall 1942, p. .

These figures are for 1925 and calculated from Horton, Larsen and Wall (1942: ?).

Horton, Larsen and Wall 1942, p. 222.

Horton, Larsen and Wall 1942, p. .

Other state specific regulations, such as minimum capital requirements and lending restrictions, as well as differences in the extent of supervision, may have contributed to variation in failure rates among states. We limit our consideration to branch banking and deposit insurance since they are most typically cited as causes of state by state differences. Future research should examine the impacts of other regulatory differences. In contrast to the United States, Canada allowed nation-wide branch banking. While many branch offices closed during the 1920s and early 1930s, as business declined in depressed areas, there were virtually no bank failures (Haubrich 1990--JME).

Branch banking, which was fairly common before the Civil War in the south, disappeared almost entirely soon after the passage of the National Bank Act of 1863, and the majority of branch banks existing in the 1920s had been established since the turn of the century.

Branch, Chain, and Group Banking Commission 1931 Vol. 2, Branch Banking in the U.S., p. 28.

Calomiris surveys the histories of state deposit insurance schemes in both the pre-Civil War period and the 1920s. Calomiris, Charles, "Deposit Insurance: Lessons from the Record," Economic Perspectives, Federal Reserve Bank of Chicago, May/June 1989, pp. 10-30.

Calomiris and Kahn, forthcoming p. 3; Gambs 1977, p. 14. It has been suggested that the twenties was a period of too many banks and not enough bankers [ cite the study of Montana banks]. A Federal Reserve study of bank failures in the twenties indicates that failed banks had a higher proportion of questionable assets and loans to officers, directors, and their interests than did banks that did not fail [B,C, and G Banking Comm.] It is important to note that the study included urban and rural banks and did not indicate where fraud was more prevalent.

Bremer 1935, p. 39. Bremer argues that the distribution of stockholdings for South Dakota was representative of country banks in general. One bank examiner in 1925 remarked: "the x banks (in the larger town) have made a practise [sic] of handling farm paper at 8 per cent, it being customary for small country banks to charge 10 per cent" (Federal Reserve Board 19??: 209).

Numerous studies conducted by sociologists, and researchers at state agricultural experiment stations highlight the impact of the automobile on rural trade centers. A bibliography of such studies is Manny (1956). Quoted in Ely 1986, p. ??. The statement by George Harrison, Governor of the Federal Reserve Bank of New York, in 1931 on the causes of the bank failures of the twenties is similar: "with the automobile and improved roads, the smaller banks...with nominal capital, out in the small rural communities, no longer had any real reason to exist. Their depositors welcomed the opportunities to get into their automobiles and go to the large centers where they could put their money" (Hearings of the Subcommittee on Banking and Currency, United States House of Representatives, 1931, p. 44).

McKibben and Griffen 1938, p. 109.

U.S. Census 1932, p. 535.

The location and year of each bank suspension from 1921 to 1929 are from U.S. House of Representatives, Hearings before the Committee on Banking and Currency, Vol. 1, Part 1 1930, pp. 314-418. The 1920 Census of Population was used to determine which incorporated towns had fewer than 2500 people. March issues of Polk's Bankers Encyclopedia (Bankers Encyclopedia Co.) were used to determine the total number of banks located in such towns in each state in each year. A particularly difficult task was determining the population of unincorporated towns. Although Polk's lists a population figure for each town, it seems in many instances to be inaccurate (a conclusion reached by comparing the population figures in Polk's for several incorporated towns with those given in the census). Thus, except for obvious cases, we considered most unincorporated towns to have a population less than 2500.

Also contributing to our selection of this break was the absence of data on farm foreclosures, one of our measures of agricultural distress, before 1926. Descriptive statistics on the data used in the regression analysis along with definitions of the variables and the data sources are presented in Table 1.

Alston, "Farm Foreclosure...."

The following states were designated branch banking states: Arizona, California, Delaware, Maine, Maryland, North Carolina, Rhode Island, South Carolina, Tennessee, and Virginia. Each allowed state-wide branching, or had large numbers of branches outside the home-office county, as of 1930 (Federal Reserve Bulletin, December 1930, pp. 811-12). A few states had no laws specifically prohibiting branching, but had few if any branches. These states were considered unit banking states.

For 1921-25 the following states had deposit insurance: Kansas, Mississippi, Nebraska, North Dakota, Oklahoma, South Dakota, Texas, and Washington. Except for Oklahoma and Texas, whose funds were suspended in 1923 and 1925, each had insurance in the 1926-29 period as well (Calomiris, "Deposit Insurance: Lessons from the Record").

The functional form used in all regression is log/log. For the most part the substantive results are the same regardless of functional form. We will note when functional form matters.

When the linear functional form is used the coefficient on land value is no longer significant and the t statistic on the coefficient for deposit insurance increases to 1.5 and the coefficient is 1.45.

The coefficient on deposit insurance is sensitive to functional form. When we ran equation 2 linearly, the coefficient on deposit insurance was not significantly different from zero but the interactive term was 12.8 with a t statistic of 1.31.

When the linear functional form is used the coefficient on deposit insurance is 1.79 (t=1.37) and the coefficient on federal banks is -.15 (t=1.4).

When we ran the interactive terms with the linear functional form, the results differ considerably. R2 drops to .39; the coefficient on agricultural distress falls to .05 (t statistic =.62); the coefficient on branch is -2.9 (t=- 1.28); the coefficient on deposit insurance is -4.39 (t=-1.4); the coefficient on federally sponsored banks is -.15 (t=-1.7); the coefficient for the deposit insurance interactive term is .29 (t=2); and the coefficient for the branch interactive term is .19 (t=1.2). The results are puzzling. It is the only run when agricultural distress does not offer the most explanatory power and the remaining variables pick up the slack. We are at a loss in explaining this outlier.

There were some important differences among the state insurance schemes of the 1920s that are not captured by our zero/one dummy variable which could account for instability in its coefficient. In three states, Kansas, Texas and Washington, state-sponsored insurance was voluntary. In Texas, banks chosing not to enter the state plan were required to obtain a private insurance bond. In Kansas and Washington banks could simply chose to carry no insurance, or to leave the system at any time. Calomiris (1989: 17) notes that this provision reduces the risk subsidization among banks and hence the likelihood that insured banks will engage in excessively risky practices. Calomiris further argues (1989: 18-19) that deposit interest rate ceilings and extended stockholder liability limited the ability and incentive for bankers to exploit insurance. And, limitations on the funding of state insurance systems reduced the credibility of insurance, and hence depositors had more incentive to monitor the performance of banks than under the current FDIC system. Thus, we would expect that the state plans of the 1920s had less to do with the bank failures of that decade than the current insurance and regulatory environment have to do with the failures of the 1980s.

1 A longer version of this paper, complete with figures, footnotes, references and data appendix will be available at the Cliometric Session at ASSA or from any author. The Causes of Rural Bank Failures in the Twenties Dependent Variable: Suspensions per 100 banks

		(1)			(2)			(3)			(4)	
Intercept		-30.08			-47.9			-5.11
		(1.5)			(1.9)			(1.1)			(0.7)

Land Value		5.26			8.65
		(1.5)*		   	(2.0)**
Improved Acreage		3.74			5.12
		   (2.8)***	   	(2.8)***
Farm Foreclosures								3.1	
								   (2.3)**		   (1.9)**

Branch Banking		0.62			35.98			-0.36
		(0.6)			(0.8)			(0.3)			(0.5)

Deposit Insurance		0.66			43.9			-0.27
		(0.8)			 (1.6)*			(0.2)			(0.6)

Government Bonds		1.92			1.69			0.17	
		   (1.8)**		   (1.8)**		(0.2)			(0.6)

Automobile		0.99			0.93			0.28		
		(1.3)			(1.2)			(0.2)			(0.1)

Rural Population		2.78			1.4			6.42		
		 (1.4)*			(0.6)			   (2.6)**		   

D.I. x Land Value					-8.3
D.I. x Improved Acre					-6.5
D.I. x Farm Fore										
B.B. x Land Value					-6.8
B.B. x Improved Acre				-2.1
B.B. x Farm Fore										
R2		.55			.57			.56			.57
F		5.27			3.4			8.8			6.57	
Notes: T-statistics are in parentheses, ***, **, * indicate statistically 
significant at the .01, .05, and .10 levels.