¹ Slichter, Sumner, The Turnover of Factory Labor (New York, 1919).Google Scholar

² Hirschman, Albert O., Exit, Voice, and Loyalty (Cambridge, Mass., 1970)Google Scholar; Freeman, Richard and Medoff, James, What Do Unions Do? (New York, 1984).Google Scholar

³ More nuanced treatments of company unions can be found in Brody, David, Workers in Industrial America: Essays on the Twentieth Century Struggle (New York, 1980)Google Scholar and Nelson, Daniel, “The Company Union Movement, 1900–1937: A Reexamination,” Business History Review 61 (Autumn 1982): 335–57CrossRefGoogle Scholar. Paul Douglas's 1921 article is a model of clarity in distinguishing between the efficiency and distributional consequences of company unions. In Douglas's view, company unions fostered “heightened production” through “much improved plant morale” and “greater individual effort,” while independent unions insured that labor received its fair share of the gains. See Douglas, Paul, “Shop Committees: Substitutes For, Or Supplement To, Trades Unions?,” Journal of Political Economy 29 (April 1921): 89–107.CrossRefGoogle Scholar

⁴ Nelson, Daniel, Managers and Workers (Madison, Wise, 1975).Google Scholar

⁵ Chandler, Alfred D. Jr, The Visible Hand: The Managerial Revolution in American Business (Cambridge, Mass., 1977)Google Scholar; Slichter, Turnover of Factory Labor; Clark, Gregory, “Why Isn't the Whole World Developed? Lessons from the Cotton Mills,” Journal of Economic History 47 (March 1987): 141–73.CrossRefGoogle Scholar

⁶ U.S. Department of Labor, Bureau of Labor Statistics, “The Safety Movement in the Iron and Steel Industry, 1907–1917,” Bulletin No. 234 (Washington, D.C., 1918): 13.Google Scholar

⁷ Quoted in Brody, David, Steelworkers in America: The Nonunion Era (New York, 1960), 78.Google Scholar

⁸ Brissenden, Paul F. and Frankel, Emil, “Mobility of Labor in American Industry,” Monthly Labor Review 10 (June 1920): 36–56.Google Scholar

⁹ Slichter, Turnover of Factory Labor, 85–89, 57–74.

¹⁰ Doeringer, Peter B. and Piore, Michael J., Internal Labor Markets and Manpower Analysis (Lexington, Mass., 1971).Google Scholar

¹¹ Lazonick, William, “Technological Change and the Control of Work: The Development of Capital-Labour Relations in U.S. Manufacturing Industry,” in Managerial Strategies and Industrial Relations, Gospel, Howard F. and Littler, Craig R., eds. (London, 1983).Google Scholar

¹² Montgomery, David, Workers' Control in America (Cambridge, Mass., 1979), 96.Google Scholar

¹³ Fishback, Price V. and Kantor, Shawn Everett, “Square Deal or Raw Deal? Market Compensation for Workplace Disamenities, 1884–1903,” Journal of Economic History 52 (December 1992): 826–48.CrossRefGoogle Scholar

¹⁴ Brandes, Stuart, American Welfare Capitalism: 1880–1940 (Chicago, 1970).Google Scholar

¹⁵ Brody, Steelworkers in America.

¹⁶ Brissenden, Paul F. and Frankel, Emil, “Mobility of Labor in American Industry,” Monthly Labor Review 10 (June 1920)Google Scholar; U.S. Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970, Series D85–85 (Washington, D.C., 1975): 135.Google Scholar

¹⁷ Brissenden, Paul F. and Frankel, Emil, “Mobility of Labor in American Industry,” Monthly Labor Review 10 (June 1920): 44.Google Scholar

¹⁸ See the Data Appendix for a complete discussion of the data used in this and other empirical analyses in the paper. A more detailed reporting of the empirical results presented here and elsewhere in the paper is available from the author upon request. Note that some, but by no means all, of the correlations reported in this study are referred to as being “statistically significant” even if they fall within an 80% confidence interval, which constitutes a lower standard than the minimum of 90%, and sometimes 95%, commonly invoked in statistical tests. One reason for adopting a lower standard in this case is that small sample sizes tend to raise the standard errors of estimates, thereby making it more difficult to find in favor of statistical significance than otherwise would be the case with a larger sample size.

¹⁹ Hirschman, Exit, Voice and Loyalty.

²⁰ Olson, Mancur, The Logic of Collective Action (Cambridge, Mass., 1971).Google Scholar

²¹ Joseph Schaffner of Hart, Schaffner and Marx—the site of the 1910 strike that sparked the successful organizing drive of clothing workers—commented: “Careful study of the situation has led me to the belief that the fundamental cause of the strike was that the workers had no satisfactory channel through which minor grievances, exactions, and petty tyrannies of underbosses could be taken up and amicably adjusted.” In Fraser, Steven, Labor Will Rule: Sidney Hillman and the Rise of American Labor (New York, 1991), 51.Google Scholar

²² Lazonick, William H., Competitive Advantage on the Shopfloor (Cambridge, Mass., 1990), 244Google Scholar; Brody, David, Labor in Crisis: The Steel Strike of 1919 (New York, 1965), 129.Google Scholar

²³ Montgomery, Workers' Control in America, 98.

²⁴ Nelson, Managers and Workers, 141.

²⁵ Montgomery, David, The Fall of the House of Labor (Cambridge, Mass., 1987), 415CrossRefGoogle Scholar. Workers in meat packing, for example, were granted the eight-hour day with ten hours' pay and a Federal Administrator to arbitrate issues left unsettled by negotiations. In steel, workers were granted time-and-a-half for work over eight hours and for Sundays and holidays, as well as the elimination of a system of bonus pay widely held by workers to be responsible for a general speed-up in the pace of work. See Brody, Labor in Crisis.

²⁶ See North, Douglass C., Institutions, Institutional Change and Economic Performance (Cambridge, Mass., 1990)CrossRefGoogle Scholar for an interesting analysis of the role ideology plays in mitigating the free-rider problem associated with collective action towards institutional change.

²⁷ The estimated coefficient on the interactive term, weighted by the average level of joint administration, is less in absolute value than the estimated coefficient on the reduced turnover variable. These results are available upon request from the author.

²⁸ Union membership reached a peak of 5,047,800 in 1920, but fell rather dramatically over the next three years to 3,622,000 in 1923. By 1930 union membership accounted for only 10.2 percent of the nonagricultural labor force; it had stood at 19.4 percent in 1920. See Bernstein, Irving, The Lean Years: A History of the American Worker 1920–1933 (Boston, Mass., 1960), 84.Google Scholar

²⁹ Myers, James, Representative Government in Industry (New York, 1924).Google Scholar

³⁰ The average monthly separation rate in manufacturing fell from 10.1 over the period 1910–18 to 4.9 in the years 1920–29, while the quit rate alone declined by half—from 7.4 to 3.7; Jacoby, Sanford, Employing Bureaucracy: Managers, Unions, and the Transformation of Work in American Industry, 1900–1945 (New York, 1985), 268Google Scholar. Turnover rates for the two periods are not strictly comparable, however. The measurements for the latter period are median rates, whereas those for the former period are based on average rates. A comparison of quit rates in the early 1920s with the later 1920s is therefore less fraught with possible error. Monthly quits averaged 5.4 per 100 employees for the period 1919–1923 and 2.7 for the period 1924–29; Lazonick, Competitive Advantage on the Shop Floor, 251.

³¹ Burton, Ernest R., Employee Representation (Baltimore, Md., 1926), 29Google Scholar; Nelson, “The Company Union Movement,” 338, 344. According toxs the U.S. Bureau of the Census, Historical Statistics of the United States, Series D 127–141, p. 137, the labor force of the U.S. stood at roughly 30 million in 1926, and the manufacturing labor force was roughly 10 million. The percent of the nonmanagerial work force in manufacturing involved in company unions could not have exceeded 20 percent.

³² National Industrial Conference Board, The Growth of Works Councils in the United States (New York, 1925), 10Google Scholar; National Industrial Conference Board, Collective Bargaining Through Employee Representation (New York, 1933), 16.Google Scholar

³³ Burton, Ernest R., Employee Representation (Baltimore, Md., 1926), 117, 169, 172–73, 174Google Scholar, reports that roughly 30 percent of a sample of surveyed representation plans contained provisions for arbitration of disputes. However, in only a little over half of these cases was arbitration automatic; the others required the mutual consent of workers and management. What little evidence exists suggests, moreover, that arbitration was a rarelyused device for settling disputes.

³⁴ French, Carroll E., The Shop Committee in the United States (Baltimore, Md., 1923), 54Google Scholar; Commons, John R. et al. , History of Labor in the United States, 1896–1932, vol. 3 (New York, 1935), 348Google Scholar; Leiserson, William M., “The Accomplishments and Significance of Employee Representation,” Personnel 4 (July/Aug. 1928): 127Google Scholar; Bernstein, The Lean Years, 173.

³⁵ Slichter, Sumner, “The Current Labor Policies of American Industries,” Quarterly Journal of Economics 43 (May, 1929): 413.CrossRefGoogle Scholar

³⁶ Brandes reports that through the machinery of the Rockefeller plan at Colorado Fuel and Iron, workers expressed a desire to reduce the workday from twelve to eight hours instead of following U.S. Steel's practice of paying time-and-a-half for the last four hours of the dreaded twelve hour day. Management acceded to the workers' wishes. Although the change meant a cut in pay, it also brought workers home four hours earlier and set the future standard for the industry. See Brandes, American Welfare Capitalism, 130.

³⁷ Cohen, Lizbeth, Making a New Deal (Cambridge, Mass., 1990), 173.Google Scholar

³⁸ Brandes, American Welfare Capitalism.

³⁹ Lazonick, Competitive Advantage on the Shop Floor, 241; Gordon, David, Edwards, Richard, and Reich, Michael, Segmented Work, Divided Workers: The Historical Transformation of Labor in the United States (New Rodaelle, 1982,).Google Scholar

⁴⁰ U.S. Department of Labor, Statistics of Industrial Accidents in the United States to the End of 1927, Bulletin No. 490 (Washington, D.C., 1929)Google Scholar; Bernstein, The Lean Years, 172.

⁴¹ Jerome, Harry, Mechanization in Industry (New York, 1934).Google Scholar

⁴² Interestingly, the introduction of a variable capturing changes in horsepower per worker over the early 1920s into the productivity growth equations (rows 1 and 2) left the results largely unchanged.

⁴³ Nonparametric tests may have an advantage over parametric tests, such as regression techniques, when certain features of the analysis—for example, small sample size—make it likely that standard parametric assumptions are violated. We have utilized a nonparametric test in this case, however, because the simple summation of injury rate and productivity growth performance to arrive at a measure of joint shopfloor benefits would not allow us to distinguish cooperative outcomes from those that are noncooperative (i.e., ones which contain dramatic benefits for one party but losses for the other). Under our ranking system, industries with modest improvements in productivity and safety receive higher rankings than those with, for example, dramatic productivity growth and slight reductions in safety.

⁴⁴ National Industrial Conference Board, Collective Bargaining Through Employee Representation (New York, 1933), 13, 22.Google Scholar

⁴⁵ Spates, T. G., “An Analysis of Industrial Relations Trends,” American Management Association Personnel Series, No. 5 (1937)Google Scholar; Dietz, J. W., “Status of Personnel Men in the Organization,” American Management Association Annual Convention Series, No. 58 (1927)Google Scholar; Niesz, Homer E. and Knapp, A. A., “The Scope of Activities of a Personnel Department,” American Management Association Annual Convention Series, No. 60 (1927).Google Scholar

⁴⁶ Alternative measures of labor productivity were considered, but they resulted in a much reduced sample size and generally left our reported findings unchanged. Using Fabricant's measure of output by industry (as opposed to value of output which is used here) to generate the productivity data yielded a drastically reduced sample size. Replacing the number of wage earners in the denominator of our productivity measure with an index of worker hours from Beney reduced the sample size by one quarter and yet yielded similar results to those presented here. Fabricant, Solomon, The Output of Manufacturing Industries 1899–1937 (New York, 1940)Google Scholar; Beney, Ada M., Wages, Hours, and Employment in the United States 1914–1936 (New York, 1936).Google Scholar

⁴⁷ It might be claimed that this comparison is not legitimate because the two samples are composed of different industries. However, the results are not substantively altered by restricting the analysis of productivity growth to the six industries that are common to both samples. The coefficient on the company union index for this analysis is positive and significant in the early 1920s and positive but insignificant and smaller in size in the late 1920s.

⁴⁸ Once again, the introduction of the mechanization variable into the productivity growth equation left the results largely unchanged.

⁴⁹ The average rate of change in horsepower per worker over the two periods was 19 percent for our sample of industries. If the rate of change in mechanization had been 19 percent during the period 1921–25, a 1 percent increase in company union concentration in an industry would have been associated with a 10 percent decrease in the injury rate. For a similar rate of change in mechanization in the 1925–29 period, however, a 1 percent increase in the company union coverage would have been associated with a 3 percent reduction in the injury rate. Again, it might be claimed that this comparison of injury rate trends in the early versus late 1920s is not legitimate because the two samples are composed of different industries, thereby possibly giving rise to different results even absent any structural change in the ability of company unions to affect shopfloor safety. However, the results are not substantively changed when the analysis is restricted to the six industries that are common to the two samples. The coefficient on the company union index in the simple regression of injury rate changes is negative and significant in the early 1920s and negative but insignificant and smaller in absolute value in the late 1920s.

⁵⁰ Changes in total factor productivity measure real enhancements in the productive efficiency of labor and capital combined. Such measures are not influenced by substitution between inputs in production as are the partial measures of productivity growth—the change in output divided by labor input—used in our earlier analyses. Thus, the change in total factor productivity is a superior measure of productivity growth. Unfortunately, measures of total factor productivity are not available for subperiods of the 1920s, only for the decade as a whole.