Analyzing data from extensive surveys of U.S. nonfarm families in 1889-90-predominantly families from eastern states-the authors of this paper show that workers affiliated with labor organizations had earnings that were 22 percent higher than the earnings of other workers, an effect comparable in magnitude to estimates for modern unions. Wage differentials varied considerably across the eight industries studied and across skill levels within each industry. LABOR economists generally agree that modern trade unions have raised unionized workers' earnings relative to the earnings of nonunionized workers; the estimates vary widely, but average about 15 percent (see Lewis 1963Lewis , 1986; Hirsch and Addison 1986, Chap. 5). Empirical findings of a union earnings premium are consistent with the theory of factor markets (the Hicks-Marshall laws of derived demand), which predicts such an effect given the conditions that have obtained since World War I-a strong union presence in high-wage industries that rely on relatively scarce skilled workers.'The same theory has caused economists to doubt the possibility that such an earnings premium could have occurred in earfor criticism and suggestions.1 The Hicks-Marshall laws of derived demand state that demand for a factor is a function of the elasticity of substitution between factors, the elasticity of supply of competitive factors, the elasticity of demand for the final product, and the input share of the factor in the cost of the final product. See Addison and Siebert (1979:240-41) or Ulman (1955:305-7). lier times. Over the years, several important labor historians (in particular, Douglas 1930:562; Slichter 1941, Chap. 12; Ulman 1955, Chap. 17) have stated that union earnings differentials were likely in some industries at particular times in the late nineteenth century; but the profession has tended to look at the proportion of the labor force organized as a measure of market power (as in Wolman 1936:3-14) and, relying on the Hicks-Marshall conditions, to decide that early unions could not have significantly affected earnings.This paper re-examines that conclusion through an analysis of data on 6,375 nonfarm U.S. families in 1889-90. The data source is a survey conducted by the U.S. Commissioner of Labor to gain information on the composition of industrial costs of production (in connection with congressional debate over the McKinley tariff).2 That survey, which gener-2 This tariff, adopted in 1890, was introduced by Representative William McKinley and was much emphasized in his successful 1896 presidential campaign on the Republican ticket. It imposed prohibitively high duties and was meant to protect "infant industries." The tariff provoked bitter opposition abroad, including threats of reprisals, and was a major cause of the Democratic victory in the congressional elections of 1890.