NOTE: The summary which appeared in the October 1997 The Newsletter of The Cliometric Society contained graphics which cannot be accurately reproduced in this version.

1 The Old South's Last Hour
After World War II, the Old South virtually disappeared as a separate economic region in the United States. This rapid change had profound consequences throughout, as well as beyond, the South. Southern outmigration associated with the demise of the Old South permanently altered the economic and racial structure of northern urban America. The political results were equally far-reaching. Paternalism and the southern plantation declined while the region's resistance to the "welfare state" weakened, thereby enabling the Great Society programs of the Johnson era to pass Congress (Alston and Ferrie, 1993). Also, since African-Americans were disproportionately employed in southern agriculture a decline in agricultural labor demand may have slowed black progress relative to whites in the 1950s (on the black-white gap, see Donahue and Heckman, 1991; Maloney, 1993). The demise of the "Old South" was thus no minor affair, despite its neglect in the economic history literature compared to the region's earlier days.
This paper examines one of the most dramatic changes in the South after 1945, the fall of "King Cotton," symbol of the South's economy for well over a century before 1960. Despite fitful attempts at diversification, cotton remained at the center of southern agriculture in the 1940s even after significant changes in incentives occurred during World War II. By the 1960s, however, agriculture in the region had been completely "remade" (Fite, 1984, pp. 173, 231). Focusing on the states of Alabama, Georgia and South Carolina in the 1950s, the paper summarized here examines three hypothesized causes of the decline of cotton in the Southeast.1 US government programs designed to curb cotton production are an obvious candidate for the South's eroding reliance on cotton. In addition, higher labor costs brought on by World War II outmigration from the South caused cotton to give way to the production of less labor-intensive products. Third, relative price movements in favor of alternative products contributed to the shift away from cotton. Unlike labor cost movements, however, the effects of relative price changes from 1940 to 1950 were at least partly reversed from 1950 to 1960.
The next section of this summary addresses long-term trends in acreage from 1920 to 1970, focusing on the role of government programs and placing the 1950s within a long run context. The third section examines the relative profitability of cotton and an important alternative, beef cattle production. Counterfactual exercises are conducted to examine the importance of labor and output price changes.

2 Trends in Southeastern Acreage and Shares
Figure 1 shows that the South Carolina share of US cotton production decreased markedly from 1920 to 1970, turning decisively downward at the end of World War II.2 Absolute acreage in South Carolina exhibits a downward trend that began much earlier (see Figure 2). Government programs enacted in the 1930s, as well as the 1956 Soil Bank program, could have been instrumental in reducing cotton acreage in the Southeast. A good deal has been written on government programs of the 1930s (see Whatley, 1983). Relatively little attention in the economic history literature has focused on the 1956 Soil Bank. Yet under this program, in 1958, the US harvested the smallest cotton acreage in its history since 1878.
The Soil Bank, like the allotments program, was a response to "surplus production" of major crops in the US.3 It had two major components: the acreage reserve and the conservation reserve. The former had a large effect on cotton production. Under the acreage reserve, farmers were paid an amount close to the net returns per acre to plant amounts of cotton less than allotted acreage. Instituted in 1956, the Soil Bank had little effect that year, but by 1958 the program led to 4.9 million acres of cotton being diverted throughout the US. That amount constituted about 29% of 1956 planted acreage (Cochrane and Ryan, p. 225).
Separating the effects of government programs from the general downward movement in cotton acreage (or its share of the US acreage) with linear regression requires that the series are at least weakly stationary. Controlling for a break in the South Carolina share series in 1946, tests reject the null hypothesis of a unit root for the share series, but not for the absolute acreage series.4 The share series at least should therefore yield reliable results when used in a linear regression.5
To separate the effects of government acreage programs from other changes that would affect the trend, the share and absolute acreage series were regressed on a linear time trend, (including a break at 1946), and dummy variables for the cotton allotments and the Soil Bank (see Table 1 for results). For the share series, the time trend shows the result expected from the visual impression given by Figure 1, turning downward after 1946. In fact, before 1946 the trend is not statistically significant. Strikingly, neither the allotment nor the Soil Bank coefficient is statistically significant at conventional levels. Apparently, government programs did not play a crucial role independent from the 1946 downward trend. The results are similar for the entire Southeast (not reported in this summary). Without government programs cotton production would have shifted away from the Southeast at a steady, if slow, rate.6
The absolute acreage regression result (shown in Table 1) exhibits a downward trend dating from the 1920s, that became less pronounced after 1946. Government allotments during the 1930s and continuing through the 1960s had the effect of decreasing cotton acreage (although the Soil Bank had no independent effect).7 The acreage allotment effect resulted in a reduction of over 200 thousand acres in South Carolina; the trend accounted for about 345,000 acres (about one-third of 1945 acreage) being eliminated from 1946 to 1960.

3 Relative Profitability of Cotton
What was responsible for the downward trend in cotton production after World War II? Labor costs and relative price changes are examined as possible factors.8 Alternative products that required less labor replaced cotton. This was particularly the case when mechanical harvesting methods had already been perfected for such alternatives. Given technological constraints in the 1940s and relatively low yields in the 1950s, cotton harvest mechanization was not profitable in the Southeast until the 1960s. Only in high-yielding areas did producers remain with cotton and mechanize harvesting (Heinicke, 1994).
To evaluate the effects of changes in labor cost and output prices, the paper estimates net returns (per acre, and per hundred dollars invested) in cotton and a major emerging alternative, beef cattle production. 1950 serves as a reference year. Counterfactual wage rates and prices are in turn imposed to examine their effects on net returns. The estimates developed below use data from Southern Agricultural experiment station reports. The Piedmont production region was chosen because: 1) it was an important production sub-region in the Southeast; and 2) relatively intensive studies were made of costs in that sub-region. The agricultural reports contain the information necessary to estimate average relative returns, but many omissions exist in those studies. For example, experiment station reports generally fail to provide the opportunity cost of capital in animals. Often also, they ignore labor used for establishing permanent pasture, a key input for the final product of beef cattle. The estimates here therefore use a combination of experiment station reports from the Piedmont of South Carolina, Georgia, and North Carolina, as well as eastern (non-Delta) Mississippi, and rely on USDA publications for such omitted price series.
Livestock and cotton returns were estimated under the following assumptions regarding technological conditions: tractor preparation (land breaking, fertilizing, seeding) of land for permanent and temporary pasture; tractor-powered operations for cotton in pre-harvesting; and hand harvesting of cotton. Moving from cotton to livestock production thus involved a change from more to less labor intensity. Such changes were typical of the transformation occurring in the Southeast during the 1950s (Fite, 1984). In computing the counterfactual returns, the production technology was "held constant" in the sense that tractor power was assumed except for cotton harvesting.
Table 2 presents the ratio of returns per dollar invested. While the estimates account for most of the relevant costs, they should not be regarded as expected returns even for the "average" farmer. Among other things, a farmer's assessment of the risk inherent in one crop versus another is not included in these calculations. Also, an important cost of cotton production with resident labor typical of the South is not estimated: non-pecuniary transfers are not included. Some of these transfers, (e.g., allowing a sharecropper to hunt or fish on plantation ponds or woods), may have had low opportunity costs, but others (e.g., offering protection to black workers from racial violence), may have been significant. Since livestock production rarely used sharecropping or resident labor beyond the owner's family, such costs would have been much lower for livestock than for cotton. This may explain why livestock gained ground in South Carolina during the 1940s, despite the figure in Table 2 which shows a return for cotton that is twice that of beef cattle in 1950. Finally, learning effects associated with the movement from cotton to livestock are not measured in the estimates.
The estimates should, however, be sufficient for the purpose of assessing the effects of specific labor cost changes on relative returns of cotton and livestock. The first row of Table 2 provides the ratios and differences for rates of return per dollar and returns (in dollars) per acre. As noted above, cotton returns are twice as high per dollar invested, but this does not account for non-pecuniary transfers. Returns were slightly lower (not reported) if sharecropping contracts were used given actual 1950 prices and average cotton yields. The latter typically entailed an arrangement whereby the worker received half of the cotton lint and seed revenue, supplied one-half of the fertilizer and ginning costs, and all of the tractor labor. (An experiment station study shows this was the modal contract arrangement on farms using tractor power for cotton cultivation in the Piedmont.) The second row of Table 2 shows the relative counterfactual returns for 1950 (ratios and differences), if wages had returned to the pre-World War II (1940) level.
According to the calculations reported in Table 2, lower wage rates would have meant that cotton returns per hundred dollars invested were 2.17 times as high as livestock returns (as opposed to 1.99 in 1950). The difference between cotton and livestock returns would have risen from 4.55 to 7.06 percentage points. The net return in dollars per acre shows a similar movement. Although the changes are not large, they clearly favor livestock production over cotton from 1940 to 1950.
In the 1950s, wages were relatively constant in the Southeast. Therefore, if 1960 counterfactual real wages are imposed, the relationship between cotton and livestock returns would be similar to those resulting from actual 1950 prices and wages (compare rows 1 and 3 of Table 2). Thus, the permanently higher wage rates that occurred in the 1940s lowered cotton returns, and provided an important incentive for its disappearance throughout the Southeast in the 1950s.
Several qualifications are in order. First, how could such small changes in relative returns have such large effects on cotton acreage? Second, why did it take until the late 1940s and 1950s for cotton to be reduced decisively as a share of US acreage? How do increased labor costs relate to the widely held view that mechanization "pushed" workers from the South in the 1950s?
First, the small estimated changes in relative profitability assume that both cotton and livestock production used tractors extensively. Yet if wages had returned to 1940 levels after World War II, the incentive to adopt tractors would have been reduced, and many producers may have found that the old mule technology was the least-cost method for cotton production. No allowance has been made here for such potential adjustments.
The delayed move away from cotton in the 1940s requires more study, but is attributable in part to high war-time demand for this product.
Finally, high wages promoting the decrease in cotton seem to be at odds with the hypothesis that workers, many of them black, were "pushed" from the South in the 1950s (see Day, 1967; Wright, 1986). But that argument may apply more immediately to the Mississippi Delta, where complete mechanization came early, than to the Southeast where mechanization of cotton harvesting did not occur on a widespread scale until after 1960 (Heinicke, 1994). To the extent that the argument applies to the Southeast, it would be because indivisibilities inherent in the tractor and related implements that were developed with midwestern US conditions in mind, meant that, once adopted, these machines were capable of displacing labor that otherwise would have stayed on southern plantations.
Turning to output prices, during World War II relative product price movements favored alternatives to cotton, since these products were in high demand, and because substitutes for southern cotton were becoming widespread (Fite, 1984, p. 165, 166, 177). Were such price movements sustained into the 1950s, and did they play a major role in cotton's decline after the World War II? A change in relative prices favored products other than cotton during World War II. If we impose 1940-2 (three year average) relative prices, cotton would have benefited relative to livestock. The gap between cotton and livestock returns would have risen from 4.55 to 9.90 percentage points (row 4 versus row 1 of Table 2). One should not, however, be misled that 1940-2 relative prices were highly advantageous for cotton producers: absolute cotton returns per hundred dollars (not reported in the table) were actually lower in 1940-2, (3.38%), than in 1949-51 prices (9.14%). The key here is that counterfactual livestock returns fared even worse than cotton when 1940-2 output prices are used, showing average losses of -6.52%. As livestock prices increased throughout the 1940s, however, livestock returns turned positive, and rose relative to those of cotton.
Was this movement sustained in the 1950s? Table 2 shows that it was not. If 1958-60 (three year average) prices for cotton and livestock are imposed on 1950 prices and technology, the gap between cotton and livestock returns would have risen from 4.55 to 6.47 percentage points (row 5 versus row 1). Again, livestock returns would have become negative on average with 1958-60 average prices. Cotton returns would have fallen from 9.14% to 4% on average (not reported in the table), but would have remained positive.9
For comparison with the effects of movements in relative prices and labor costs, it is worth briefly considering the economic incentives under the Soil Bank program. Payments under that program were close to those that could be earned by planting. Using the estimates above for returns, and subjecting them to 1958 input and output prices, and machine efficiency adjustments, net returns per acre from the government Soil Bank acreage reserve program would have been about $53.20. The average Piedmont producer could have expected about $63.35 per acre if the average of cotton prices and yields for 1955-57 had prevailed. Given the reduced risk associated with the government contract, the producer was compensated adequately to withdraw land from cotton production, and cotton acreage fell abruptly. When the Soil Bank acreage reserve was suspended in 1959 (the conservation reserve allowed for contracts up to 10 years in duration for planting trees on land previously used for crops, and did not result in land being diverted from cotton), cotton acreage increased dramatically in the Southeast and reached a level roughly consistent with the post-1946 trend.

4 Conclusion
The wage movements that occurred in the 1940-50 period meant that livestock gained an advantage relative to cotton, and one that was not eroded during the 1950s when wages were roughly constant. This provided an important incentive to reduce cotton acreage relative to other products. Relative prices favored non-cotton products during the 1940-60 period, but not as uniformly in the 1950s as did labor price changes. Cotton allotments had effects throughout the period, but the 1956 Soil Bank program had no independent effect on the share or absolute acreage of cotton in South Carolina.

Endnotes
1The Mississippi Delta to the west exhibited developments less typical of the old Cotton Belt.
2 South Carolina time series are used because this is the most homogenous agricultural area in the Southeast for which time series are available for relevant variables. (The trend for the entire Southeast is similar.)
3 From 1933 through the 1970s, government cotton allotments intermittently restricted acreage.
4With a break in a trend, it is difficult to reject the hypothesis of a unit root (Perron, 1989).
5 The share series test statistic, ta, is -4.77, which is greater in absolute value than the critical value of -4.56 (1% significance) for l = 0.5 (see Perron, p. 1377 for critical values). The results for absolute acreage are less clear. The test statistic value is -4.47, which rejects a unit root at the 5% (critical value of -3.96) but not at the 1% level. In addition, the absolute acreage test is sensitive to inclusion of a time trend.
6 The share and absolute acreage series were also tested for a permanent break and/or shift in 1958 with the Soil Bank program; both hypotheses were rejected.
7 Caution should be used here; we cannot reject the hypothesis of a unit root at the 1% level.
8 Fite, 1984, forwards several other factors, including development of southern breeds of cattle and pasture grasses, the emergence of more accessible credit, and transportation advances.
9 Relative prices of other products showed similar, if not identical movements. Oat prices (the most important small grain of the Piedmont) increased relative to cotton in the 1940s. These price increases continued in the 1950s, but the upward movement was very modest.

References
Alston, L. and Ferrie, J. (1993), "Paternalism in Agricultural Labor Contracts in the US South: Implications for the Growth of the Welfare State." American Economic Review 83, pp. 852-876.
Burford, R. L. (1966), "The Federal Cotton Programs and Farm Labor Force Adjustments." Southern Journal of Economics 33, 223-236.
Day, R. H. (1967), "The Economics of Technological Change and the Demise of the Sharecropper." American Economic Review 57, 3, 427-449.
Donahue, J. and Heckman, J. (1991), "Continuous Versus Episodic Change: The Impact of Civil Rights Policy on the Economic Status of Blacks. Journal of Economic Literature 29, 1603-1643.
Fite, G. C. (1984), Cotton Fields No More. Lexington: University of Kentucky Press.
Heinicke, C (1994), "African-American Migration and Mechanized Cotton Harvesting, 1950-60." Explorations in Economic History 31, 501-520.
Maloney, Thomas (1994), "Wage Compression and Wage Inequality Between Black and White Males in the United States, 1940-60." Journal of Economic History 54, 358-381.
Perron, P. (1989), "The Great Crash, the Oil Price Shock, and the Unit Root Hypothesis." Econometrica 57, pp. 1361-1401.
Whatley, W. (1983), "Labor for the Picking: the New Deal in the South." Journal of Economic History 43, pp. 905-929.
Wright, G. (1986) Old South, New South. N.Y.: Basic Books.