He wishes to thank Thomas Weiss, Price Fishback, De-Min Wu, Don Lien, Bill Sundstrom, and Michael Haines for their many helpful suggestions. Lee-Pyng Tyan provided able research assistance. This investigation was supported by the University of Kansas General Research Allocation No. 3899-20-0038.Google Scholar

¹ Recent examples include Rothenberg, Winifred B., “The Emergence of Farm Labor Markets and the Transformation of the Rural Economy: Massachusetts, 1750–1855,” this Journal, 48 (09 1988), pp. 371–87;Google Scholarand Rosenbloom, Joshua L., “One Market or Many? Labor Market Integration in the Late Nineteenth-Century United States,” this Journal, 50 (03 1990), pp. 85–107.Google Scholar Geographic integration is, of course, only one aspect of labor market efficiency. An overall analysis of labor market efficiency would also need to consider the allocation of labor between different sectors and between firms within sectors. However, Holzer, Harry J., in “Employment, Unemployment and Demand Shifts in Local Labor Markets” (NBER Working Paper No. 2858, 1989), found that for contemporary data, frictions involved in the geographic reallocation of labor are a more important source of inefficiency than are those related to supply adjustments at any particular location.Google Scholar

² As far as I know, the only attempts to date to analyze nineteenth-century wage differentials and migration as simultaneously determined endogenous variables are those by Bjork, G. C., “Regional Adjustment to Economic Growth: the United States, 1880–1950,” Oxford Economic Papers, 20 (1968), pp. 81–97:CrossRefGoogle Scholarand Orsagh, Thomas J. and Mooney, Peter J., “A Model for the Dispersion of the Migrant Labor Force and Some Results for the United States, 1880–1920,” Review of Economics and Statistics, 12 (08 1970), pp. 306–12.CrossRefGoogle Scholar

³ The use of cross-sectional data reflects an implicit assumption that the different markets not completely integrated, because in a truly unified market there would be only one wage, and observed variation would reflect compensating variations, differences in labor quality, or measurement errors. At the same time, it is necessary that the locations not be completely unintegrated. this case, different cities will obtain labor from different sources, and the resulting estimates supply elasticities will confound the different elasticities faced by employers at different places. The hypothesis of complete integration cannot be tested within a model of local wage determination, but it seems reasonable to characterize the sample of U.S. cities examined here as less perfectly integrated, it is possible to test for differences in labor supply elasticities across subsets of the sample, and this possibility is considered in subsequent pages.Google Scholar

⁴ This result is consistent with work by Friedman, Gerald. “Skill Differentials. Unions, American Labor Markets” (mimeo, Department of Economics, University of Massachusetts, 1989), which found that building trades unions were more effective than others in controlling entry into local labor markets.Google Scholar

⁵ U.S. Department of Commerce and Labor, Nineteenth Annual Report of the Commissioner Labor, 1904, “Wages and Hours of Labor” (Washington, DC, 1905);Google ScholarU.S. Department Commerce and Labor, Eighteenth Annual Report of the Commissioner of Labor, 1903, “Cost Living and Retail Prices of Food” (Washington, DC, 1904):Google ScholarU.S. Department of the Interior, Census Office, Eleventh Census, 1890. “Report on Population of the United States.” part 2 (Washington, DC, 1897):Google ScholarU.S. Department of the interior, Census Office, Eleventh Census, 1890, “Report on Manufacturing Industries in the United States,” Part 2: Statistics of Cities (Washington, DC, 1895);Google Scholar and U.S. Department of the Interior. Census Office, Twelfth Census, 1900, “Abstract” (Washington, DC, 1901).Google Scholar

⁶ The industrial and occupational coverage of the study was dictated largely by the need to consider comparable occupations throughout the country.Google Scholar

⁷ Inspection of price indices in Haines, Michael R., “A State and Local Consumer Price Index for the United States in 1890,” Historical Methods. 22 (Summer 1989), pp. 97–105. suggests that variations in the cost of food were highly correlated with variations in a broader basket of goods.CrossRefGoogle Scholar

⁸ U.S. Congress, Senate, Retail Prices and Wages. Senate Report No. 986, 52nd Cong., 1st sess. (Washington, DC, 1892).Google Scholar

⁹ Coelho, Philip R. P. and Shepherd, James F., “Regional Differences in Real Wages: the United States, 1851–1880,” Explorations in Economic History, 13 (04 1976), pp. 203–30; and Rosenbloom, “One Market or Many?”CrossRefGoogle Scholar

¹⁰ The census industries used were (with the occupations to which they were matched in parentheses) blacksmithing and wheelwrighting (blacksmiths); foundry and machine shop products (pattern makers, iron molders, machinists, and boilermakers); carpentering (carpenters): masonry, brick, and stone (bricklayers): painting and paperhanging (painters); plumbing and gas fitting (plumbers). The sum of employment in all the building trades was used as a proxy for employment of building trades laborers, and employment of foundry and machine shop laborers was measured as the sum of employment in foundry and machine shop products and architectural ironwork. The manufacturing census data are preferred to occupation statistics from the population census because workers reporting a particular occupation may well have been employed in industries other than those included in the wage data, and because at least in more recent times the labor market appears to be segmented along industrial as well as occupational lines.Google Scholar

¹¹ North, Douglass C., “Locational Theory and Regional Economic Growth,” Journal of Political Economy, 43 (06 1955), pp. 243–58.CrossRefGoogle ScholarWilliamson, Jeffrey G., in Late Nineteenth-Century American Development: A General Equilibrium History (Cambridge, 1974), pp. 21–51, offers an alternative but related characterization of midwestern growth.Google Scholar

¹² On the rate and variability of urban growth, see Madden, Carl H., “On Some Indications of Stability in the Growth of Cities in the United States,” Economic Development and Cultural Change, 4 (04. 1956), pp. 236–52,CrossRefGoogle Scholar and Madden, Carl H.. “Some Spatial Aspects of Urban Growth in the United States,” Economic Development and Cultural Change, 4 (07 1956), pp. 371–87.CrossRefGoogle Scholar The determinants of differential urban growth rates are discussed in Pred, Allan, The Spatial Dynamics of U.S. Urban-Industrial Growth, 1880–1914: Interpretive and Theoretical Essays (Cambridge, MA, 1966);Google Scholar and Meyer, David R., “Midwestern Industrialization and the American Manufacturing Belt in the Nineteenth Century,” this JOURNAL, 49 (12. 1989), pp. 921–38.Google Scholar

¹³ The census also reports the value of the capital stock and total revenues. The capital stock data appear to be unreliable, as there were no common standards for valuing equipment or adjusting for depreciation. Total revenue offers a more plausible alternative to the raw materials variable used here. Presumably, the lower elasticity of substitution characterizing the relationship of materials to fixed capital would make materials a better proxy than total revenue for all nonlabor inputs into the production function. In practice, expenditures on raw materials and total revenue are highly correlated and yield nearly identical results in the estimates reported hereafter.Google Scholar

¹⁴ Theoretical determinants of migration are discussed in Nelson, Philip, “Migration, Real Income and Information,” Journal of Regional Sciences. I (Spring 1959), pp. 43–73;CrossRefGoogle Scholar and Gill, Flora, “Economics and the Black Exodus: An Analysis of Negro Emigration from the Southern United States” (Ph.D. diss., Stanford University, 1974).Google ScholarOn the historical importance of friends and relatives, see Erickson, Charlotte, American Industry and the European Immigrant, 1860–1885 (Cambridge, MA, 1957);CrossRefGoogle Scholar and Dunlevy, James A. and Gemery, Henry A., “Economic Opportunity and the Responses of ‘Old’ and ‘New’ Migrants to the United States,” this Journal, 38 (12 1978), pp. 901–18.Google ScholarThese scholars found that migrant stock was an important determinant of immigrant destinations in the 1890s. For additional references, see Rosenbloom, Joshua L., “Padrones, Employment Agencies. and Labor Exchanges: An Examination of the Methods of Labor Market Adjustment in the Late Nineteenth Century United States” (Research Paper in Theoretical and Applied Economics No. 90–1. University of Kansas, 1990).Google ScholarEvidence on railroad rates is from Rosenbloom, “One Market or Many?” p. 121.Google Scholar

¹⁵ Skilled metaiworkers were employed in railroad repair shops and agricultural implements manufacturing as well as in foundry and machine shops. Building trades workers—especially carpenters, painters, and plumbers—were often employed in manufacturing establishments.Google Scholar

¹⁶ It could be argued that LFORCE should be used as the measure of labor supply in the model rather than as an exogenous variable. The rationale for choosing manufacturing employment as the quantity variable is discussed in fn. 10. Preliminary exploration of alternative specifications suggests that LFORCE measures something very different from industry employment.Google Scholar

¹⁷ It is worth noting that the assumption that wages adjust to achieve local labor market equilibrium is also necessary to interpret the magnitudes of wage differentials as evidence for the extent of geographic integration. For anecdotal evidence on labor market adjustment, see Ozanne, Robert, A Century of Labor Management Relations at McCormick and International Harvester (Madison, 1967), pp. 14–33.Google Scholar

¹⁸ An alternative is to assume that the prices received by producers varied in the same way as retail prices did. Estimates of the model in this form are qualitatively quite similar to those reported below. As the truth probably lies somewhere between these two extremes, it may be concluded that the results are robust to different specifications of the impact of product prices on labor demand.Google Scholar

¹⁹ For simultaneous equation systems, R ² may not be an appropriate measure of goodnesss of fit. A more reliable indication of the model's accuracy is the fit of the reduced form equations that it implies. For the wage equation, R ² values range from about 0.2 to 0.6, suggesting that the model does indeed account for a considerable fraction of intercity wage variation.Google Scholar

²⁰ The strong positive correlation between raw materials expenditures and labor demand indicates that the potential endogeneity bias noted earlier is not, in fact, particularly severe. The availability of alternative local sources of labor appears to have been a particularly important influence on labor supply. Although LFORCE is highly correlated with city size (as measured by the number of gainfully employed workers in all occupations), substituting that variable for LFORCE would alter the estimated coefficient values considerably, eliminating all the positive wage elasticities in Table 2.Google Scholar

²¹ Because of the small number of southern and western cities in the sample, it was not possible to estimate separate labor supply elasticities for each of these regions.Google Scholar

²² The labor supply elasticities within the Northeast and Midwest were (with standard errors in parentheses) skilled building trades, 2.32 (3.22): skilled foundry and machine shop trades, 0.02 (1.21); and unskilled labor, 0.34 (1.13). Interaction effects between real wages and a dummy for cities outside the Northeast and Midwest were skilled building trades, 4.36 (1.59): skilled foundry and machine shop trades, –1.11 (0.39); and unskilled labor, 0.95 (0.24).Google Scholar