¹ Williamson, Jeffrey and Lindert, Peter, American Inequality: A Macroeconomic History (New York, 1980).Google Scholar

² The Weeks report is Weeks, Joseph D., Reports on the Statistics of Wages in Manufacturing Indusries with Supplementary Reports (Washington, D.C., 1886).Google Scholar The Aldrich report is Aldrich, Nelson W., Wholesale Prices, Wages, and Transportation, 52d cong., 2d sess., S.R. 1394 (Washington, D.C., 1893).Google Scholar

³ Coehlo, Philip R. P. and Shephard, James F. in “Regional Differences in Real Wages: The United States, 1851–1880”, Explorations in Economic History, 13 (01 1976), pp. 20330, use the Weeks data to analyze regional differences in real wages after 1850. They point out, however, that outside the Northeast the Weeks data “exhibit a high degree of variability before 1860, probably due to the scarcity of data for … the 1850s” (p. 205). Except for Ohio and Maryland, all of the Aldrich data pertain to New England or the Mid-Atlantic states;Google Scholar see Mitchell, Wesley C., A History of the Greenbacks (Chicago, 1903), p. 285.Google Scholar

⁴ Layer, Robert G., Earnings of Cotton Mill Operatives, 1825–1914 (Cambridge, Mass., 1955);Google ScholarSmith, Walter B., “Wage Rates on the Erie Canal,” this JOURNAL, 23 (09 1963), pp. 298–311;Google ScholarLebergott, Stanley, Manpower in Economic Growth (New York, 1964), pp. 257–333;Google ScholarAdams, Donald R. Jr, “Wage Rates in the Early National Period: Philadelphia, 1785–1830” this JOURNAL, 28 (09 1968), pp. 404–26;Google ScholarZabler, Jeffrey, “Further Evidence on American Wage Differentials,” Explorations in Economic History, 10 (Fall 1972), pp. 109–17;CrossRefGoogle ScholarAdams, Donald R. Jr, “The Standard of Living During American Industrialization: Evidence from the Brandywine Region, 1800–1860”, this JOURNAL, 42 (12 1982), pp. 903–17;Google Scholar“Prices and Wages in Maryland”, this JOURNAL, 46 (09. 1986), pp. 625–45;Google ScholarSokolof, Kenneth, “The Puzzling Record of Real Wage Growth in Early Industrial America: 1820–1860” (unpublished manuscript, University of California Los Angeles, 12. 1986).Google Scholar

⁵ “Reports of Persons and Articles Hired, 1818–1905” Record Group 92, the National Archives.Google Scholar

⁶ The small number of observations before 1820 have been excluded from the analysis.Google Scholar

⁷ See Heppner, Francis and John, Harry, “Wage Data Among Nineteenth Century Military and Naval Records” (unpublished manuscript, National Archives and Record Service, Washington, D.C., 1968).Google Scholar

⁸ The restrictions eliminate observations from the most remote frontier posts and from posts experiencing significant military conflict.Google Scholar

⁹ The Southwest may be an exception; see Frazer, Robert W., Forts and Supplies: The Role of the Army in the Economy of the Southwest (Albuquerque, 1983).Google Scholar

¹⁰ The conclusion is tentative until we can make similar comparisons for other regions.Google Scholar

¹¹ See Rosen, Sherwin, “Hedonic Prices and Implicit Markets,” Journal of Political Economy, 82 (01/02 1972), pp. 34–55 for a discussion of hedonic regressions.CrossRefGoogle Scholar

¹² Twenty-six days per month is the standard figure; see Lebergott, Manpower, p. 245.Google Scholar

¹³ There is precedent for combining common laborers and teamsters; see Abbott, Edith, “The wages of Unskilled Labor in the United States, 1850–1900,” Journal of Political Economy, 13 (06 1905), pp. 350–51. Some teamsters brought their own teams. Such cases are easily spotted because the common practice was to double the regular teamster wage.CrossRefGoogle Scholar

¹⁴ One odd result was a positive coefficient on the number of rations in several regressions. Payment of rations was concentrated at certain forts, and it proved difficult to separate out the effect of rations from the effect of fort location. The content of a ration is known, however, so it is possible to value rations at wholesale prices. Based on available price information, the value of a ration was set at 12 cents, and the regressions with positive coefficients were reestimated, omitting the rations variable. The value may be too low for certain forts and years, but the figure is close to the Army's own valuation; see Frazer, Forts and Supplies, p. 193.Google Scholar

¹⁵ The pay gap between carpenters and other skilled occupations is usually explained by differences in the stability of employment; see Lebergott, Manpower, p. 243. In contrast, there were no systematic differences in pay between teamsters and common laborers. Teamsters were better paid than common laborers in the North, but the opposite was true in the South.Google Scholar

¹⁶ See Lebergott, Manpower, p. 245.Google Scholar

¹⁷ The percentage difference in pay between slave and free labor was larger among artisans. The lower slave wage may indicate a lower skill level. Slaves may have been more costly to supervise, or less likely than free workers to supply their own tools. The lower skill differential among slaves is consistent with evidence that slave and free artisans were closer substitutes than slave and free laborers;Google Scholar see Fogel, Robert and Engerman, Stanley, Time on the Cross (New York, 1974).Google Scholar

¹⁸ Two further adjustments were made. The skilled regression for the South Central states significantly overpredicted wages in Kentucky and Tennessee, and the coefficient for forts in both states was adjusted downward. Additional data pertaining to workers at the Boston Naval Yard (“Naval Hospital Payrolls”, Bureau of Yards and Docks, Record Group 71, the National Archives) were used to estimate wage changes in the Northeast from 1835 to 1837. Details are available from Robert Margo in an appendix (see fn. 19).Google Scholar

¹⁹ For example, the population weight for forts in upstate New York in the 1830s is the average of the 1830 and 1840 shares of the northeastern population living in rural New York state. The occupational weights were derived from the regional distribution of teamsters, laborers, the building trades (carpenters, masons, painters and plasterers), and blacksmiths. The purpose of the weighting is to adjust for the geographic and occupational differences between the sample and the antebellum population. An appendix containing the regressions, the weights, and a detailed discussion of the construction of the regional estimates is available from Robert Margo.Google Scholar

²⁰ In his study of manufacturing wages Sokoloff (“The Puzzling Record”) finds that rural wages rose relative to urban wages from 1820 to 1860. If this were generally true our estimates of the rate of growth of wages in the Northeast would be biased downwards, because urban observations are overrepresented.Google Scholar

²¹ For example, if two of the time period dummies were 1825 and 1826/27, the wage estimate for 1826 would be interpolated (that is, the coefficients of the time period dummies are centered on the midpoint of the period). All of the estimates for 1849 are interpolated, because we have yet to find reports for the year.Google Scholar

²² For example, the fort location dummies could be weighted by the nonagricultural labor force, making use of the new labor force estimates being prepared by Thomas Weiss. Because the distribution of fort locations within regions is less representative of the distribution of population within the South and the Midwest, the regional estimates for these regions are more sensitive to changes in the fort location weights than the estimates for the Northeast. For example, it is likely that high-wage locations are overrepresented in the South and Midwest, even after weighting.Google Scholar

²³ Following convention, we present the estimates to two decimal points (for example, $1.23). A good rule of thumb, however, is to regard changes in the second decimal point (for example, from $0.77 to $0.79) as economically insignificant unless proven otherwise.Google Scholar

²⁴ The 1850 census contains a table showing state averages of daily wages paid to common laborers and carpenters in 1849 (without board)Google Scholar. Weighting the state averages by the regional distribution of carpenters and laborers in 1850 the gives the following regional averages:

Adjusting our wage estimates for 1849 to remove the weights attached to teamsters and skilled occupations other carpenters gives the follwing regional averages:

Except in two cases our estimates exceed the census figures, although the differences are not large. Regional contrasts within the North and South, however, are the same in both sets of estimates. Allowing for the differences in methodology and locations covered (most likely the census figures were unweighted averages of county data) the general consistency of the two sets of estimates is quite good.

²⁵ Williamson and Lindert, American Inequality, appendix G, p. 319. We set the index equal to 100 in 1856. Williamson and Lindert also report a skilled cost-of-living index (p. 101), but the index extends back only to 1844.Google Scholar

²⁶ Ibid., chap. 4.Google Scholar

²⁷ Ibid., pp. 71.Google Scholar

²⁸ On capital-skill complementarity see James, John A. and Skinner, Jonathan, “The Resolution of the Labor Scarcity Paradox,” this JOURNAL, 45 (09 1985), pp. 513–40.Google Scholar

²⁹ Grosse, Scott, “On the Alleged Antebellum Surge in Wage Differentials: A Critique of Williamson and Lindert,” this JOURNAL, 42 (06 1982), pp. 413–18.Google Scholar

³⁰ Lindert, Peter and Williamson, Jeffrey, “Antebellum Wage Widening Once Again,” this JOURNAL, 42 (06 1982), p. 422.Google Scholar

³¹ According to Habbakuk, H. J., American and British Technology in the Nineteenth Century: The Search for Labor-Saving Inventions (New York, 1962), p. 23, the “premium on artisan skills was generally lower in America than in England in the early ninetenth century,” Our estimates of the skill differential, which are higher than contemporary estimates for England, cast doubt on Habakkuk's assertion. For a similar conclusion, see Adams, “Wage Rates in the Early National Period”.Google Scholar

³² Because the skill differential was higher in the Midwest than in the Northeast, the western movement of population could have caused an aggregate surgeGoogle Scholar. Weighting the regional differentials by population shares gives the following estimates for the North:

The differential still does not rise from the 1820s to the 1850s. Thus, our conclusion is unaffected bu population redistribution.

³³ Calculated from Appendix Tables 5 and 6.Google Scholar

³⁴ In the twentieth century the skill differential has risen during downturns; see Reder, Melvin, “The Theory of Occupational Wage Differentials,” American Economic Review, 45 (12 1955), pp. 833–52.Google Scholar

³⁵ Calculated from U.S. Bureau of the Census, Historica Statistics of the United States (Washington, D.C., 1975), series C-132, C-136, p. 111.Google Scholar

³⁶ Williamson and Lindert (American Inequality, appendix D, series 5, p. 307) also report a 22 percent rise in the skill differential in the Northeast from 1844 to 1856. The skill differential in the Midwest may have been affected by the end of the boom in railroad construction. According toGoogle ScholarFogel, Robert (Without Consent or Contract: The Rise and Fall of American Slavery [unpublished manuscript, University of Chicago, 1986], p. 620) “[t]he flood of labor released from the railroads created downward pressure on the wages of established workers and led to heavy unemployment throughout the Midwest.” Because the majority of the workers released were unskilled, the effect was evidently greater on unskilled wages, and the skill differential rose.Google ScholarPubMed

³⁷ Lindert and Williamson, “Antebellum Wage Widening”, p. 421.Google Scholar

³⁸ The weights are the occupational shares in Wright's sample. The calculation excludes machinists (there are no observations on machinists from the 1820s).Google Scholar

³⁹ Massachusetts, , Sixteenth Annual Report of the Bureau of Statistics of Labor (Boston, 1885), pp. 318–36.Google Scholar

⁴⁰ Massachusetts, Sixteenth Annual Report, p. 433. The point settles a matter of confusion in the literature. Grosse (“On the Alleged Antebellum Surge”, p. 413) believed he had found an error in Wright's decadal average for carpenters in the 1850s. Lindert and Williamson (“Antebellum Wage Widening”, p. 420) agreed with him. In fact, nearly all of Wright's averages for the 1850s appear to be incorrect because he included the additional data for 1860. The sensitivity of Wright's averages is troublesome, because other evidence shows no change in the skill differential from 1859 to 1860; see Williamson and Lindert, American Inequality, p. 306–7.Google Scholar

⁴¹ Williamson and Lindert (American Inequality, p. 307) report a series of urban skill differentials for the Northeast showing a 34 percent rise from 1821–1830 to 1851–1856. The denominator of their series, an index of unskilled wages, does not differ significantly in behavior from an index based on our unskilled wage estimates. But the numerator of their series, an index of skilled wages, does differ significantly from an index based on our skilled wage estimates. According to Williamson and Lindert, skilled wages in the Northeast grew by 93 percent from 1821–1830 to 1851–1856, compared with our estimate of 24 percent (calculated from Appendix Table 5). Virtually all of the difference between the two figures is explained by differences in the level of skilled wages in the 1820s, and to a lesser extent, the 1830s. Williamson and Lindert constructed their index of skilled wages by splicing Zabler's (“Further Evidence”) estimates of skilled wages for the Pennsylvania iron industry in the 1820s with data from the Aldrich report after 1840. Figures for the 1830s were interpolated using wages paid for teamwork on the Erie Canal. The most questionable part of their calculation was not the interpolation, but the use of Zabler's estimates. According toGoogle ScholarAdams, Donald R. Jr, in “Wage Rates in the Iron Industry: A Comment,” Explorations in Economic History, 11(Fall, 1973), pp. 89–94, Zabler's estimates are too low. We agree with Adams. An unweighted average of Zabler's estimates of monthly wages of carpenters and blacksmiths gives an implied daily wage of 0.57 cents for 1821–1830, far below any other estimate for the decade.CrossRefGoogle Scholar

⁴² David, Paul, “The Growth of Real Product in the United States Before 1840: New Evidence, Controlled Conjectures,” this JOURNAL, 27 (06 1967), p. 184.Google Scholar

⁴³ See, for example, Hirsh, Susan E., Roots of the American Working Class: The Industrialization of Crafts in Newark, 1800–1860 (Philadelphia, 1978);Google ScholarWilentz, Sean, Chants Democratic: New York City and the Rise of the American Working Class, 1788–1850 (New York, 1984);Google Scholar and Ross, Steven J., Work, Leisure, and Politics in Industrializing Cincinnati, 1788–1890 (New York, 1985).Google Scholar Important early studies are Commons, John R. et al. , History of Labor in the United States (New York, 1916);Google ScholarSullivan, William A., The Industrial Worker in Pennsylvania, 1800–1840 (Harrisburg, 1955);Google Scholar and Ware, Norman, The Industrial Worker, 1840–1860 (New York, 1924).Google Scholar

⁴⁴ Margo, Robert A. and Steckel, Richard H., “Heights of Native-Born Whites during the Antebellum Period,” this JOURNAL, 43 (03 1983), pp. 167–74;Google ScholarPubMedFogel, Robert W., “Nutrition and the Decline in Mortality since 1700: Some Preliminary Findings,” in Engerman, Stanley and Gallman, Robert, eds. Long-Term Factors in American Ecopomic Growth (Chicago, 1986), pp. 439–537;Google ScholarKomlos, John, “The Height and Weight of West Point Cadets: Dietary Change in Antebellum America,” this JOURNAL, 47 (12 1987), pp. 897–927.Google Scholar

⁴⁵ Taylor, George Rogers, The Transportation Revolution (New York, 1951), p. 253; Lebergott, Manpower, pp. 151–53.Google Scholar

⁴⁶ Bridges, Amy, A City in the Republic: Antebellum New York and the Origins of Machine Politics (New York, 1984); Fogel, Without Consent or Contract.CrossRefGoogle Scholar

⁴⁷ Sullivan, The Industrial Worker, p. 31; Ware, The Industrial Worker, p. 32; Wilentz, Chants Democratic, pp. 117, 363.Google Scholar

⁴⁸ The low rate of growth conflicts with the high rate of growth of labor productivity in manufacturing before 1840; see Sokoloff, Kenneth, “Productivity Growth in Manufacturing During Early Industrialization: Evidence From the American Northeast, 1820–1860”, in Engerman, and Gailman, , eds., Long-Term Factors, pp. 679–729.Google Scholar

⁴⁹ David, Paul and Solar, Peter, “A Bicentennary Contribution to the History of the Cost of Living in America,” in Uselding, Paul, ed., Research in Economic History, vol. 2 (New York, 1977), p. 16. If the David-Solar index is used, real wages rose 27 percent for artisans and 22 percent for unskilled labor from 1821–1830 to 1831–1840.Google Scholar

⁵⁰ The David-Solar index splices pre-1850 prices from rural Vermont with the post-1850 Hoover consumer price index. According to Lebergott (Manpower, pp. 334–35), the Vermont data exclude the majority of foods consumed by workers, and exhibit “incredible” price declines for textiles. The Hoover index is a national price index, not a northeastern index. For a critique of the Hoover index, see Lebergott, Manpower, pp. 338–52. The Williamson-Lindert index is a consistent northeastern deflator, because it is based solely on wholesale prices in New York City. The use of wholesale prices, however, may affect the interpretation of the timing and magnitude of the fluctuations in real wages, and trends in urban prices may be misleading for forts in rural areas. The most serious problem with the Williamson-Lindert index is the exclusion of rents; see fn. 52.Google Scholar

⁵¹ If the David-Solar index is used, real wages increase by 13 percent for artisans and 21 percent for unskilled labor from 1837 to 1843.Google Scholar

⁵² The surge in immigration is believed to have caused an increase in the relative price of housing, especially in large cities in the Northeast; see Fogel, Without Consent or Contract, chap. 10. If so, the decrease in real wages would be larger because the Williamson-Lindert deflator excludes the price of housing. If the David-Solar index is used, real wages increase from 1841–1850 to 1851–1856. There is reason to believe, however, that the David-Solar index understates the increase in the cost of living in the 1850s. Wholesale prices in New York and Philadelphia rose by 36 and 29 percent, respectively, from 1844 to 1856, compared with a 13 percent increase in the David-Solar index (calculated from U.S. Bureau of Census, Historical Statistics, series E-52, E-97, pp. 201, 205).Google Scholar

⁵³ Berry, Thomas S., Western Prices before 1861 (Cambridge, Mass., 1943);Google ScholarNorth, Douglass, The Economic Growth of the United States, 1790–1860 (New York, 1961).Google Scholar

⁵⁴ North, The Economic Growth, p. 207.Google Scholar

⁵⁵ Adams, “Prices and wages in Maryland.”Google Scholar

⁵⁶ Lebergott, Manpower, pp. 74–76.Google Scholar

⁵⁷ Coehlo, Philip R. P. and Shephard, James F., “Differences in Regional Prices in the United States, 1851–1880,” this JOURNAL, 34 (09., 1974), pp. 551–91.Google Scholar

⁵⁸ Ross, Workers on the Edge, pp. 9, 43, 143.Google Scholar

⁵⁹ Berry, Western Prices, p. 564. The Berry index is a suitable proxy because 50 to 75 percent of workers' budgets were spent on food.Google Scholar

⁶⁰ Ross, Workers on the Edge, p. 43.Google Scholar

⁶¹ For a similar conclusion based on data on farm wage rates, see Lebergott, Manpower, p. 134–35.Google Scholar