Geographic Aspects of Labor Market Integration Before the Civil War

Robert A. Margo
Department of Economics
Calhoun Hall
Vanderbilt University
Nashville TN 37235
MARGORA@ctrvax.vanderbilt.edu

A set of spatially distinct, but "integrated" labor markets has two characteristics: first, a tendency for wages to equalize across locations, net of cost of living differences or locational amenities or disamenities, for otherwise identical labor (the "law of one price"); and second, adjustment to economic disparities that appears sensible. If the size of the set of integrated labor markets within a country expands geographically over time, a "national" labor market is said to be "emerging".

For the most part, the conventional wisdom is that a national labor market in the United States began to truly emerge after the Civil War. Regional labor markets in the North allegedly became integrated as early as the 1870s or 1880s. The process of regional integration was quite different in the South. "The defining economic feature of the South prior to World War Two was not poor performance or failure," according to Gavin Wright but the "isolation ... of the southern labor market from national and international flows". The "isolation" of Southern labor markets left its imprint in the form of persistently low real wages which did not begin to increase appreciably relative to other regions until after World War Two.

With few exceptions, relatively little work has been done on regional labor market integration before the Civil War. Using state-level data on farm wages, Lebergott made comparisons of coefficients of variation between paired census dates: all such pairwise comparisons showed a decline, suggesting the beginnings of market integration. Lebergott also showed (graphically) that population growth at the state level between census dates (for example, 1850 to 1860) was positively correlated with initial level of wages in the state, which he interpreted as the response of in-migrants to cross-state wage differentials.

1. The Westward Movement of Population, 1800-1860

The United States underwent a massive east-west redistribution of population before the Civil War. At the beginning of the nineteenth century, virtually the entire labor force -- 93 percent of it -- lived in the Northeast or South Atlantic regions, both long settled. But, following 1800, a process of "westward movement" began. In the case of the Midwest, the growth rate of its labor force share was very rapid between 1800 and 1810, but then decelerated for the next two decades. During the 1830s, however, the Midwest experienced a 60 percent increase in the growth rate of its labor force share. Growth in the share declined in the 1840s, but then stabilized in the 1850s. By 1860, the Midwest claimed 41 percent of the Northern labor force.

The South followed a broadly similar "east-west" pattern early in the nineteenth century. Growth in the South Central's share of the Southern labor force was rapid between 1800 and 1810, but declined during the 1810s and 1820s, but then accelerated in the 1830s. The growth rate declined sharply in the 1840s, and continued to remain very low in the 1850s. By 1860, 52 percent of the Southern labor force resided in the South Central region.

Why should labor have moved west before the Civil War? The simplest answer is that locations in the Midwest and South Central regions were perceived to have economic value net of the costs of migration. The simple answer, however, runs into an empirical puzzle. Estimates of per capita income show substantially lower values in the Midwest relative to the Northeast in 1840 and 1860, while in the South, per capita incomes in the East South Central region were virtually identical to those in the South Atlantic. Economic theory suggests that individuals generally move from low income to high income locations, not the other way around.

One way around the puzzle is to adjust in some manner -- or dispute -- the per capita income figures. Aside from questioning the data, the puzzle can be resolved in various ways. Perhaps migration west selected individuals from the lower half of the eastern income distribution -- the so- called "safety-valve" hypothesis. Another explanation is that migration to the frontier was prompted by the possibility of capital gains, not current income.

A third explanation, originally suggested by Coehlo and Shepherd, is that the value of the marginal product of labor per unit of time -- the real wage -- was initially higher on the frontier than in settled areas of the East Coast. The existence of a real wage gap provided a potential economic gain to migration -- if not immediately, then perhaps over time (assuming migrants stayed on the frontier).

To resolve the paradox in the manner suggested by Coehlo and Shepherd, it is necessary to develop a model in which real wages are higher on the frontier, per capita income lower, and migration to the frontier optimal. Further, it is apparent that the settlement was not completed overnight, so migration in the model must be spread out over time.

Margo (1998, appendix 5.6) develops a theoretical model of regional settlement along these lines. In the primary version of the model, there are two regions -- A, a settled region, and B, a frontier region. Initially, most of the labor force is concentrated in region A. The economic problem is to allocate labor between A and B so as to maximize the present discounted value of aggregate production.

Because of the initial regional imbalance in factor proportions (and possibly other reasons), labor is more productive at the margin on the frontier. Thus, there is an incentive for labor to be reallocated from A to B. However, migration to the frontier incurs two costs -- a direct cost, such as a transportation cost, plus an assimilation cost. The assimilation cost is modeled as a delay in productivity -- that is, labor must wait for some time before it can be economically productive on the frontier. Thus migration is costly in part because output that would have been produced during the assimilation period had migrants remained in region A is foregone. Nevertheless, as long as the direct costs of migration are not too great, migration from A to B will generally be optimal.

The assumption that migration includes an assimilation cost implies that, at the margin, aggregate costs of migration were rising, or convex. With convex "adjustment" costs, migration will be spread out over time. The pace of migration is quickest initially, but then slows down until no longer optimal, and "settlement" is complete. Wages are initially higher in region B (the frontier) than in region A (the settled region), but as the share of the labor force in region B increases, the gap in wages diminishes. As is true of migration, the pace at which the wage gap diminishes is greatest initially, but declines over time.

A key implication of the model is that per capita output in region B may be less than in region A, at least initially and possibly for some period of time. The assimilation cost is the reason. Labor is willing to migrate because the present value of migration, net of all costs, is positive. However, the ratio of migrants to currently productive labor in region B is highest during the initial phase of the settlement process. Because (by assumption) migrant labor does not contribute to current production in region B, per capita output in B can be less than in region A, all the while migration from A to B is occurring. But, over time, the ratio of migrants to currently productive labor in B falls, and per capita income in region B converges on (and may surpass) per capita income in region A.

The movement from A to B in response to an initial regional imbalance in factor proportions can be thought of as a "supply-side" response -- a benchmark case, in other words, for analyzing the evolution of regional wage gaps and the migration process. Migration in response to "demand-side" shocks can also be modeled, by allowing the productivity of labor in one region relative to the other to change. An unexpected rise in relative productivity on the frontier -- prompted, for example, by rising demand for frontier-produced goods -- can accelerate both the pace of migration and cause the regional wage gap to temporarily increase.

2. Regional Wage Differentials Before the Civil War

Margo (1998, ch. 3) presents annual series of real wages (1860 = 100) for three occupations -- common labor, artisans, and white collar workers -- in each of four census regions -- Northeast, Midwest, South Atlantic, and South Central -- over the period 1821 to 1860. In Margo (1998, ch.3) a regional cost of living deflator is constructed for 1850, using data on the weekly cost of board. The real wage series are then benchmarked to the regional cost of living deflator, so that comparisons can be made across regions in levels and trends. I also construct regional estimates of the (free) labor force in each occupation, and use these to compute aggregate real wage series by occupation. Further details may be found in Margo (1998, ch. 5).

Comparisons between the Midwest and Northeast, and the South Central and South Atlantic, are examined first. In the 1820s, real wages in the Midwest exceeded real wages in the Northeast by 0.28 in log terms for common labor, and by considerably more for skilled artisans and clerks. By the 1850s these wage gaps had undergone a pronounced decline; for example, in the case of artisans, the wage gap fell by -0.304 in logs. The declines were not monotonic, however. The regional wage gap for common labor rose slightly in the 1830s and 1850s, as did the gap for clerks in the 1840s and skilled labor in the 1850s.

The initial real wage advantage of the South Central region over the South Atlantic was smaller than the gap between the Midwest and the Northeast; and, unlike in the North, there was no clear trend towards wage equalization. For common labor, the wage gap between the South Central and South Atlantic states was about 13 percent in the 1820s. The gap rose in the 1830s, but then fell back in the 1840s. The gap declined slightly in the 1850s but was only marginally lower in the 1850s than in the 1820s.

In the case of artisans, the wage gap between the South Central and South Atlantic remained constant at about 8 percent in the 1820s and 1830s, rose slightly in the 1840s, and then increased substantially in the 1850s. The initial regional wage gap was considerably larger for clerks -- about 0.19 in log terms -- and the gap trended slightly upward over the antebellum period, reaching 0.22 in log terms by the 1850s.

Next, I examine the evolution of each region's real wage relative to the national average. Also calculated is the weighted "mean absolute deviation", a a summary statistic. Wages of common labor in the Midwest in the 1820s exceeded the national average, but converged towards the national average over time. Real wages in the Northeast in the 1820s were below average but also converged by the 1850s. Real wages of common labor in the South Atlantic were below the national average in the 1820s, but slightly above levels in the Northeast. However, by the 1830s, the pattern had reversed itself -- real wages of common labor were higher in the Northeast than in the South Atlantic. Real wages in the South Atlantic fell further behind in the 1840s before recovering somewhat in the 1850s, but were still below their relative level (compared with the national average) in the 1850s than in the 1820s. Real wages of common labor in the South Central states were above the national average in the 1820s, but by the 1840s had fallen below the national average, as in the South Atlantic region. Like in the South Atlantic, however, some recovery occurred in the 1850s. Overall, the mean absolute deviation declined but, because of the divergent trends in the 1830s and between the South and the North, the decline was relatively modest in magnitude, and not monotonic. Results for artisans and clerks may be found in Margo (1998, ch. 5).

Next, I report the slope coefficients from time-series regressions of each region's relative real wage on its share of the labor force.. If the slope coefficient is negative, then increases in the region's relative share of the (occupation-specific) labor force are associated with declines in the region's relative real wage, which is consistent with shifts in labor supply as the dominant factor behind shifts in relative wages across regions. However, if the coefficient is positive, then increases (or decreases) in the region's labor force share were associated with increases (decreases) in the region's relative wage, a signal that demand shifts may have occurred.

The clearest evidence that shifts in supply were important appears in the regressions for the Midwest. For all three occupations the coefficient was negative and statistically significant, with elasticities ranging from -0.15 (common labor) to -0.28 (clerks). The coefficients were also negative for the Northeast, but only significant in the case of artisans. For all three occupations in the South Atlantic states, and for white collar workers in the South Central, the coefficients were positive, indicating that, in these cases, real wages and labor force shares moved together -- an indication that demand shifts were present.

The findings just discussed bear on several important aspects of antebellum economic development. First, and most importantly, they are broadly supportive of a "labor markets" explanation of the settlement process, along the lines suggested by Coehlo and Shepherd. Real wages were initially higher on the frontier -- the Midwest and South Central states -- than in the settled East, the Northeast and South Atlantic. The existence of these regional wage gaps provided, at least in principle, an economic incentive to migrate to the frontier, and thus help resolve the paradox noted earlier in the chapter -- the movement of population from higher per capita income regions, such as the Northeast, to lower per capita income regions, such as the Midwest.

In the North, the shift of labor towards the Midwest coincided with a secular decline in the regional wage gap, especially pronounced in the case of artisans and white collar labor. The opposing movement in relative wages and labor force shares in the Midwest is consistent with the view that "supply-side" factors -- migration -- can explain the secular trend in convergence in wages between the Midwest and Northeast.

In the South, the dominant migration pattern was also east-west, but the shift of labor towards the South Central states did not coincide with a narrowing of regional wage gaps -- if anything, the gaps were wider on the eve of the Civil War than in the 1820s. The failure of the regional wage gaps to close in the South is inconsistent with the views of various historians who have argued that the existence of inter-regional slave markets enhanced the general efficiency of east-west factor mobility in the South.

In the case of common labor, the absence of regional wage convergence in the South was not the only adverse pattern. Using the regional labor force shares, it is possible to produce an overall estimate of the North-South wage gap for common labor. This gap was slightly negative in the 1820s, indicating that real wages were initially higher in the South. But, in the 1830s, the wage gap turned markedly in favor of the North. The gap continued to widen in the 1840s before narrowing somewhat in the 1830s, but still remaining positive on the eve of the Civil War.

As noted earlier, economic historians have long been aware of the existence of a North-South wage gap after the Civil War, but the origins of the gap -- in particular, whether it predated the War -- have remained somewhat mysterious. My results suggest that the origins can be dated to the 1830s, a timing that can hardly be considered to be random. Beginning in the late 1820s and continuing through the 1830s, improvements in internal transportation, rising demand for cotton, and various federal land policies that subsidized frontier development helped fuel a land boom in Midwest and South Central regions. This demand shock evidently left its imprint in the labor market in the form of rising real wages for common labor in both regions relative to their respective regions of prior settlement (the Northeast and South Atlantic). However, real wages of common labor in the South Atlantic fell relative to wages in the Northeast, a pattern that is difficult to explain except by a shift in relative labor demand in favor of the Northeast. The most plausible candidate for such a demand shift is early industrialization. Manufacturing first took hold in the 1820s and began to grow rapidly in the 1830s. The early growth of manufacturing was not distributed uniformly across the antebellum landscape; instead, it was concentrated in the Northeast. The combination of demand shocks favoring the Midwest and the Northeast can explain why real wages of common labor in the South fell below those in the North before the Civil War.

4 Conclusion

This paper is a (much) abridged version of Chapter 5 of Margo (1998). A more detailed discussion of results can be found in the complete version. The chapter also presents an analysis of wage convergence at the local level using county level data on wages from the 1850 and 1860 census. The issue is whether high (or low) wage labor markets in 1850 remained high (or low) wage in 1860, or whether there was "regression-to-the-mean", as would be implied by a set of integrated local labor markets. I find that there was considerable regression towards to the mean.

5. Reference

Margo, Robert A. 1998. Wages and Labor Markets Before the Civil War. Chicago: University of Chicago Press, forthcoming.