Hostname: page-component-78c5997874-8bhkd Total loading time: 0 Render date: 2024-11-15T02:20:43.577Z Has data issue: false hasContentIssue false
  • English
  • Français

Real Estate Mortgages, Foreclosures, and Midwestern Agrarian Unrest, 1865–1920

Published online by Cambridge University Press:  03 March 2009

James H. Stock
James H. Stock
Affiliation:
Assistant Professor of Public Policy at the Kennedy School of Government, Harvard University, Cambridge, Massachusetts 02138.
Get access Rights & Permissions [Opens in a new window]

Abstract

The hypothesis that the fear of foreclosure of farm mortgages provided an important impetus to American agrarian reform movements of the late nineteenth and early twentieth centuries is reconsidered. This hypothesis is consistent with the observation that farm income, although volatile, on average improved over this period.Indeed, despite low average foreclosure rates, the temporary effects of foreclosures on specific regions in which there was unrest appears to have been dramatic. An examination of indebtedness data and measures of unrest both for the period of the Alliance movement and for the era of the Nonpartisan League in North Dakota appears to support the hypothesis linking the fear of foreclosure to agrarian unrest.

Type
Articles
Information
The Journal of Economic History , Volume 44 , Issue 1 , March 1984 , pp. 89 - 105
Copyright
Copyright © The Economic History Association 1984

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 North, Douglass C., Growth and Welfare in the American Past (Englewood Cliffs, 1974), pp. 137–48Google Scholar, and Higgs, Robert, The Transformation of the American Economy, 1865–1914 (New York, 1971), pp. 86–102.Google Scholar For reviews of this research and further references, see Lee, Susan P. and Passell, Peter, A New View of American Economic History (New York, 1979), pp. 292–303;Google ScholarMayhew, Anne, “A Reappraisal of the Causes of Farm Protest in the United States, 1870–1900,” this JOURNAL, 32 (06 1972), 464–75;Google Scholar and McGuire, Robert A., “Economic Causes of Late-Nineteenth Century Agrarian Unrest: New Evidence,” this JOURNAL, 41 (12. 1981), 835–52.Google Scholar Several new explanations for this unrest in the absence of continual distress have been proposed. Some of these can be classified as sociological or psychological: Higgs, American Economy, p. 101, suggests that the loneliness of farm life could exacerbate difficult circumstances and perhaps lead to activism, while North, Growth and Welfare, p. 145, suggests that the disenchantment of the farmers stemmed in part from the drop in farm income relative to nonfarm income. Mayhew, “A Reappraisal,” p. 469, proposes that the protest can be seen as a reaction to increasing farm commercialization; that is, that farmers were objecting “to the increasing importance of prices” rather than to the prices themselves. It would be difficult to refute these explanations by statistical analyses of historical data, nor do they shed any light onto the important question of the geographic and temporal distribution of protest during this period. Thus it seems desirable to search for other plausible solutions to this apparent puzzle.Google Scholar

2 Lee and Passell, A New View, p. 301, note that the growing importance of markets and crop specialization, coupled with price and yield variability, intensified farm net income fluctuations. The increasing importance of international markets could also have exacerbated income fluctuations, for as prices grew to reflect global rather than local conditions the correlation between price and yield fluctuations would have become less negative.Google Scholar

3 McGuire, “Agrarian Unrest,” 837–52.Google Scholar

4 This risk averse farmer would still prefer a certain to an uncertain income stream. The worst effects of “feast or famine” harvests could be overcome, however, by using financial markets to smooth consumption.Google Scholar

5 Foreclosure did not necessarily entail dispossession, since many states had a statutory period after a foreclosure sale during which the mortgagor could redeem his or her debt. Indeed, these redemption periods can be seen in part as an outcome of agrarian unrest; this point is discussed by Skilton, Robert H., “Developments in Mortgage Law and Practice,” Temple University Law Quarterly, 27 (08. 1943), 315–84. Presumably, it was the fear of dispossession, not of foreclosure, that drove farmers to participate in protest movements. Since a foreclosure sale was the key step towards dispossession, however, the fear of foreclosure and the fear of dispossession generally will be treated as interchangeable in this paper.Google Scholar

6 Evidence of sporadic distress on the plains is provided by Bogue, Allan G., From Prairie to Corn Belt (Chicago, 1963)Google Scholar, and by Bowman, John D., “An Economic Analysis of Midwestern Farm Land Values and Farm Land Income,” Yale Economic Essays, 5 (Fall 1965), 316–52.Google Scholar

7 For example, see Buck, Solon J., The Agrarian Crusade (New Haven, 1920), pp. 20–22 and 105–107;Google ScholarBriggs, H. E., “Grasshopper Plagues and Early Dakota Agriculture, 1864–1876,” Agricultural History, 8 (04. 1934), 51–63;Google Scholar and Schell, Herbert S., “The Grange and the Credit Problem in Dakota Territory,” Agricultural History, 10 (04. 1936), 59–83.Google Scholar

8 Shannon, Fred A., American Farmers' Movements (Princeton, 1957), p. 53.Google Scholar

9 These proxies for the fear of foreclosure have limitations and would be complemented by actual foreclosure data. Unfortunately, there are two drawbacks to using such data. First, the level of foreclosures is itself an endogenous variable, since some reform movements succeeded in stopping threatened foreclosures (as did, for example, the Farm Holiday Movement in North Dakota in the early 1930s). The second drawback is practical: Data on foreclosures are not readily available and are often untrustworthy. On this latter point, see Bogue, Allan, Prairie to Corn Belt, p. 180, and footnote 27 below.Google Scholar

10 The Nonpartisan League is discussed by Taylor, Carl C., The Farmers' Movement, 1620–1920 (New York, 1953)Google Scholar, by Saloutos, Theodore and Hicks, John D., Agricultural Discontent in the Middle West, 1900–1939 (Madison, 1951)CrossRefGoogle Scholar, and by Shannon, Farmer's Movements.Google Scholar

11 U.S. Bureau of the Census, Eleventh Census of the United States, 1890, v. 12, “Real Estate Mortgages,” p. 107.Google Scholar

12 Bogue, Prairie to Corn Belt, p. 179.Google Scholar

13 Buck, Agrarian Crusade, p. 106.Google Scholar

14 Bogue, Allan G., Money at Interest (Lincoln, 1955), p. 256.Google Scholar Bogue notes that, among the mortgagors, “a significant percentage of individuals were not primarily farmers” (p. 257), and he warns that “it is possible to exaggerate both the amount of land mortgage[d]… and …foreclosed” (pp. 259–61). Taking these points into account, however, the conclusion still remains that the incidence of farm foreclosures would have been great. Alston, Lee, “Farm Foreclosures in the United States During the Interwar Period,” this JOURNAL, 43 (12. 1983), 885–904, presents similar evidence on the geographical concentration of foreclosures during the 1920s and 1930s.Google Scholar

15 The assumption of 15 indebted “neighbors” is perhaps conservative. For example, Walsh County was typical of the major wheat-producing counties in eastern North Dakota in 1920. Using county averages for farm acreage, the fraction of land in farms, and average reported rates of indebtedness, if a “neighboring farm” is (rather arbitrarily) defined as one within three miles, then a typical farmer would have 22 indebted “neighbors.”Google Scholar

16 McGuire, “Agrarian Unrest,” p. 842. New York and Pennsylvania were excluded because of the belief that industrial and urban mortgages in these states represented such a large fraction of overall indebtedness that the effect of farm mortgages could not be gleaned from aggregate real estate mortgage data.Google Scholar

17 This follows from supposing that mortgages taken to relieve distress would be subject to the variability of the weather, pests, and global prices, while mortgages taken to support plans of expansion would be based on long-term expectations and thus might be somewhat stable (or following a trend) from year to year.Google Scholar

18 These two groups can be viewed as reflecting the rapidly changing, or “high frequency,” components and slowly changing, or “low frequency,” components of new mortgages. The variance of the three-period moving average is an intermediate statistic between these two extremes, since very rapid changes in mortgages will be smoothed by the moving average.Google Scholar

19 The Mann-Whitney statistic tests the hypothesis that the means of the cells are the same against the alternative that they increase in a specific order. This test is described by Lehmann, E. L., Nonparametrics: Statistical Metkods Based on Ranks (San Francisco, 1975), pp. 232–38. The asymptotic distribution of this statistic was used to compute the marginal significance levels reported in Table 3. This test has two advantages over that based on the Spearman rank correlation coefficient. First, it is hoped that its asymptotic distribution more closely approximates its exact distribution than would be the case for the Spearman test applied to this data set, since the approximation to the distribution of the rank correlation assumes the number of cells to grow without bound. Second, the null and alternative hypotheses of the Mann-Whitney test correspond exactly to those of this theory, while the null and alternative hypotheses of the Spearman test are complicated relationships between the distribution of the two random variables and are difficult to interpret precisely.Google Scholar

20 It can be argued that these results understate the effect of the fear of dispossession (as opposed to the fear of foreclosure) in the regions of greatest protest during this period, since the older states generally had longer debt redemption periods than the younger states. According to Robert H. Skilton, “Mortgage Law and Practice,” during this period Illinois, Indiana, Michigan, Iowa, and Missouri had statutory redemption periods of one year; Wisconsin and Minnesota had redemption periods of two and three years, respectively. In contrast, Kansas and South Dakota had no statutory redemption periods, while in Nebraska, the law only set an upper limit of nine months. Of the four states in which the Alliance movement was most active, only North Dakota had a redemption period of a full year after sale.Google Scholar

21 This hypothesis is weaker than that usually used in relation to asset pricing, specifically, that the price reflects rational expectations. The assumption here is simply that the prospective buyer acts on a consideration of future net income from the farm. Thus, the ratio of current net income to land value measures the degree to which current farm profitability exceeds, or falls short of, the market assessment of future profitability. Whether this assessment is either ex ante or ex post rational is inconsequential for this conclusion.Google Scholar

22 This proxy has two obvious shortcomings. First, it examines farm income only in one year, 1889. Second, it is based on gross rather than net farm income. Although income measures could be constructed for earlier years using an approach such as McGuire's (“Agrarian Unrest”), it would be inappropriate to use the farm value data from the 1890 Census. Furthermore, the level of farm income per acre in 1889 is negatively correlated with the centers of unrest, as expected under the simple (unconditional) hypothesis that regions with low income can be expected to have greater discontent. This latter point suggests that this proxy is appropriate for purposes of this section.Google Scholar

23 The logit model assumes there to be a “true” continuous measure of the degree of unrest in the i-th state, Zi, and that this measure can be written as a linear function of the proxy variables P1i and P2i and an independent error term; that is, Zi = b0 + b1p1i + b2p2i + ui. Only is observed, however, where if Zi ≦ a0, if a0 < Zi ≦ a1, and if a1 <Zi, where a0 and a1 are constants. If ui has a logistic distribution with scale parameter c, then this becomes the logit model with three possible outcomes. The parameters (b0, b1, b2, a0, a1, c) are not all identifiable; the estimated values of the functions of the parameters which are identifiable are not particularly meaningful in the analysis at hand and therefore are not reported. For an introduction to the logit model, see Judge, George G., Hill, R. Carter, Griffiths, William E., Lutkepohl, Helmut, and Lee, TsoungChao, Introduction to the Theory and Practice of Econometrics (New York, 1979)Google Scholar, Ch. 18. For a broader treatment, see Amemiya, Takeshi, “Qualitative Response Models: A Survey,” Journal of Economic Literature, 19 (12. 1981), 1483–1536.Google Scholar

24 Dakota, North, 1919 Legislative Manual (Bismark, 1919), p. 372.Google Scholar

25 U.S. Department of Agriculture, Agricultural Marketing Service. “Wheat…,” p. 10.

26 North Dakota Agricultural Experiment Station, “Report of the Dickenson Substation for the Years 1914 to 1918 Inclusive,” Bulletin #131 (October. 1919), p. 10.Google Scholar

27 Sheriff's deeds were the instrument used by the sheriff to pass title on land used to secure a delinquent mortgage. These percentages are taken from Porter, John W., Land Transfers in Cass County, North Dakota, 1865–1935, a period of Sixty-One Years,” mimeograph, North Dakota Agricultural College Experiment Station, Department of Agricultural Economics (August. 1936), p. 11, which summarizes Porter's examination of 33,430 deeds filed over these years in Cass County. The absolute number of sheriff's deeds filed in each of these years was 25, 14, 7 and 105, respectively. Unfortunately, Porter presents only decennial data on sheriff's deeds. Furthermore, these statistics should be viewed cautiously since the data on foreclosures during this period appear to be inconsistent. For example, Porter reports a total of 744 transfers of farm deeds during 1920 in Cass County, North Dakota, for a total of 215,074 acres. However, the U.S. Department of Agriculture, Bureau of Agricultural Economics (BAE), “Transfers of Farm Real Estate” (August 1939), reports only 474 transfers for a total of 112,461 acres in Cass County that year. The total number of voluntary transfers reported by the BAE is 448, while Porter reports 536 Warranty Deeds, 48 Quit Claim Deeds, and 14 Patents (transfers of land from the government) for a total of 598 voluntary transfers. Interestingly, the two sources agree on the number of administrative and executive sales (13). The greatest discrepancy occurs in the reporting of foreclosures. Porter reports 105 transfers by sheriff's deed, while the BAE reports only six involuntary transfers (including foreclosures, sales for taxes, assignment to creditors, and bankruptcies and other distress transfers). Similar discrepancies occur for other years as well.Google Scholar

28 These and other unreported statistical results are available from the author upon request. It should be mentioned that the correlations are weaker between the indebtedness data and 1918 voting patterns, and are weaker still (or even negative) with the 1916 voting data. Since mortgages generally had short durations, however, there is no strong reason to believe that 1920 indebtedness data would be a good proxy for the indebtedness data of 1916. As a result, these correlations are difficult to interpret and are not reported here.Google Scholar

29 Suppose that the fraction voting for Frazier in the i-th county, Vi, can be expressed as vi = F(Xib + ui) where F is a cumulative distribution function, b is a parameter vector, Xi is the vector composed of a constant, the yield variable, and the proxy for the fear of foreclosure, and ui is an independent and identically distributed error term. This can be rewritten as F-1(Vi) = Xib + ui. If F is assumed to be logistic, then F-1(Vi) = log(Vi/(l - Vi)), that is, F-1 (Vi) is the natural logarithm of the odds ratio Vi/(1-Vi). Thus this model can be estimated by regressing Iog(Vi/(1 - Vi)) on Xi using ordinary least squares.

30 The particular influence curve (IC) used was the “Tukey bi-square”: IC(z) = z(l - (z/k)2)2 if |z| < k, and IC(z) = 0 otherwise, where k was set at six times the median absolute value of the estimated residuals. For a discussion of this and some other robust regression estimators, see Hill, Richard, “Robust Regression When There are Outliers in the Carriers,” Communications in Statistics–Theory and Methods, 11 (1982), 849–67.Google Scholar For a more theoretical discussion, see Huber, Peter, Robust Statistics (New York, 1981), pp. 153–198.CrossRefGoogle Scholar A critical introduction to the subject robust estimation with historical references is given by Stigler, Stephen M., “Do Robust Estimators Work with Real Data?,” Annals of Statistics, 5 (11. 1977), 1055–98.CrossRefGoogle Scholar

31 The tests were also performed for the subsample of the fifteen counties planting over 200,000 acres of wheat in 1920. Not surprisingly the marginal significance levels were lower in the larger sample than the smaller one. The estimated coefficients were essentially unchanged, however, by dropping the counties with smaller wheat planting, adding confidence to the conclusions stated in the text. The statistics of Table 6 were also computed for another debt proxy, obtained by subtracting the number of wholly owned farms reporting debt in the 1920 census from the total number of farms and dividing by the total number of farms. A typical value of this variable is 85 percent. Of course, many mortgaged farms were partially rented and thus not wholly owned. Also, the fraction of wholly owned farms not reporting mortgage data to the Census Bureau was substantial—up to 10 percent of all wholly owned owner-operated farms in some counties. Not surprisingly, this proxy performed poorly relative to the other measures of indebtedness. Although the signs of the relevant statistics based on this variable always agree with the theory, the estimators have considerably larger standard errors than do their counterparts for the other proxies.Google Scholar