Industrial Dynamics in a Historical Setting

Cliometric Sessions at 1990 ASSA Meetings, December 30, 10:15 AM

Plant Turnover and Labor Market Dynamics in Late Nineteenth Century Manufacturing: Evidence from Milwaukee

Susan B. Carter*
University of California-Riverside

This paper uses a newly-developed longitudinal data set of firms in the Milwaukee manufacturing sector in the 1890s to quantify the contribution of plant creation, expansion, contraction, and closing to employment dynamics in the late nineteenth century. In the post-World War II era firm entry and exit are now seen as an important determinant of labor market outcomes. Assar Lindbeck and Dennis Snower conclude that, "the entry and exit of firms may play an important long term role in the transmission of product market shocks to the labor market" [1988, p. 43]. Jonathan Leonard notes, "Short durations of employment and high frequencies of disemployment are typically thought of in terms of the characteristics of people. The statistics reveal tremendous turnover of the jobs themselves" [Leonard 1987, p. 147]. His analysis of Wisconsin firms shows that,

"About one in every nine jobs disappears each year. More than one in every eight jobs is created every year. This is not during a great depression, nor a great boom. These are the magnitudes of gross job flows experienced in the average year between 1977 and 1982" [Leonard 1987, pp. 147-9].

Timothy Dunne, Mark J. Roberts, and Larry Samuelson (1989) come to similar conclusions.

Firm turnover might be expected to have been an even more important source of labor market dynamics in the nineteenth century. As Ruth Hutchinson, Arthur Hutchinson, and Mabel Newcomer note, "The high mortality rate of small business enterprises has long been recognized as one of the costs of a system of free competition" [1938, P. 497]. In the past a larger fraction of firms were small [Atack 1986]. Their careful study of business mortality in Poughkeepsie, New York certainly documents high mortality rates. "Thirty percent of the business enterprises started in Poughkeepsie between 1844 and 1936 failed to survive the first year. Less than half lasted more than three years" [Hutchinson, Hutchinson, and Newcomer 1938, p. 497]. Nor was this situation unique to New York. Robert Nesbit describes the Wisconsin economy in the latter part of the nineteenth century in similar terms. "Chance and geography played large roles in...the prospects of entrepreneurial success. The two were inextricable tied together in an emerging economic order that often baffled the shrewd and calculating as well as the majority for whom life was simple a happening...It was an awkward age, caught between the acceptance of scarcity as the prudent rule of life and an abundance of production looking for customers. Steam power and the factory system were creating new abundance, while the new economic order was organizing consumers strictly to the ends of production, marketing, and profits, but without thought to their roles as consumers. Not only was it an awkward age, it was an insecure one" [Nesbit 1985, p. 208]. Many historians would concur with his conclusion, "Business failures were a commonplace" [Nesbit 1985, p. 209]. Very high turnover rates among late nineteenth-century firms might help explain the observed homogeneity of unemployment in that period. In contrast to the modern labor market in which the burden of unemployment falls on relatively small proportion of the labor force who then find themselves out of work for a long time [Clark and Summers 1979; Murphy and Topel 1987}, Alexander Keyssar found that in the late nineteenth-century Massachusetts labor market, "the burden of joblessness was widely shared among the working people" [Keyssar 1986, p. 55; See also Margo 1990, p. 1]. Perhaps this was because most jobs were short lived. The Data Our analysis is based on firm-level data from 434 Milwaukee manufacturing concerns surveyed by the Wisconsin Bureau of Labor and Industrial Statistics in 1891-92 and 637 concerns surveyed in 1897-98. These data were collected during the Bureau's inspection of factories and published in its Fifth and Eighth Biennial Reports [Wisconsin Bureau of Labor and Industrial Statistics 1892, 1898]. Milwaukee was Wisconsin's largest manufacturing city. In 1900 it was home to 20 percent of all manufacturing firms and 34 percent of wage earners in the State [United States Census Office 1902, p. 956]. Its industrialation began with the processing of agricultural products, but quickly expanded to include iron foundries, agricultural implements producers, knitting mills, tanners and boot and shoe manufacturers. Brewing was the leading industry in 1890. The City ranked third nationally in the production of trunks and valises [Nesbit 1985, pp. 159- 183]. The number of manufacturing establishments increased 123 percent in 1870s, 56 percent in the 1880s, and 16 percent in the 1890s while the number of wage earners increased 148, 86, and 24 percent respectively [Still 1965, p. 576]. Thus firm dynamics in Milwaukee are probably representative of those in many urban areas in this period. Our data were collected just before and after the severe depression of 1893 which, according to Bayard Still, "hit Milwaukee with great severity." He reports that "Five banks were forced to close; and though no bad failures occurred among the manufacturing plants, work was suspended and wages were reduced." The "gradual resumption of manufacturing" did not take place until 1895 [Still 1965, p. 324]. Thus these data may show more business failures than comparable data for a more stable period. Since our objective in developing these data is to measure the fraction of jobs created and destroyed as the result of firm births, growth, decline and death, a crucial issue is the completeness of the surveys' coverage. While the Reports contain no explicit discussion of coverage, the Report for 1897-98 notes that the law of April 12th, 1883, which created the Bureau, "made it the duties of its commissioner to inspect all factories and workshops"[Wisconsin Bureau of Labor and Industrial Statistics 1899, p. 199. Emphasis added.]. To assess the success of the Bureau in carrying out its charge we compared the number of firms and number of wage earners enumerated by the Factory Inspection of 1897-98 in Milwaukee with figures from the federal Census of Manufactures of 1899. The results by census industry category are shown in Table 1. In all sectors the Factory Inspection survey reports far fewer firms that the Census. But only in textiles and the hand trades do the Factory Inspection survey employment totals differ from those of the Census by as much as a third. Overall, the Census employment total of 48,328 is only 2 percent higher than the 47,053 employment total from the Factory Inspection. This suggests that firms omitted from the Factory Inspection were small, family enterprises.1 Since our interest is in the effects of firm dynamics for wage labor market dynamics, these omissions do not affect the usefulness of these data for our purposes. We matched the names of firms in each of the two surveys to create a data set which would allow us to measure firm birth, expansion, contraction, and death over the six years between the survey dates. This proved to be a relatively easy task since firms were arranged by city and in alphabetical order in the published reports. In some cases there were annotations regarding name change. Of the 434 firms with complete records in 1891, 158 or 36 percent were no longer in operation in Milwaukee in 1898. Of the 633 firms with complete records in the 1898 report, 357 or 56 percent were not listed in 1891 report. These figures are consistent with descriptive evidence emphasizing high rates of firm creation and mortality in this era.2 Employees per firm averaged 73 in 1891-92 and 57 in 1897-98, or similar to Jeremy Atack's findings for manufacturing firms in the period. Firm Dynamics in Milwaukee Manufacturing, 1891-1898 Something of a surprise, given the rapid growth of the manufacturing sector in Milwaukee and descriptive evidence of the high rates of firm turnover, is the fact that the average firm in each of these surveys had already been in operation for 16 years! Moreover, because older firms also tended to have more employees, the average employee was working for a firm which had already been in operation for 22 years. In a steady state, in which a firm is equally likely to be surveyed at any point in its life, the completed lifespan of firms would be twice the length of the average lifespan at the time of the survey. Thus, these figures imply an average lifespan for ninteenth-century manufacturing firms of 32 years! There is, of course, no inconsistency between our evidence of high turnover rates and evidence that the typical firm remained in operation for a relatively long period of time. The seemingly contradictory nature of these finding can be explained by reference to demographers' distinction between life expectancy at birth and life expectancy of the average person. Although turnover among newly established firms (infant mortality) is high and expected longevity of new firms (life expectancy at birth) is brief, the life expectancy of the average firm (life expectancy of the population) is lengthy [Clark and summers 1979; Akerlof and Main 1981; Hall 1982; Carter and Savoca 1990]. Yet, the figures on the life expectancy of firms indicate considerably more stability of jobs than one might infer from the descriptive evidence. They also provide additional support for our earlier conjecture that non- union, male industrial workers in this era remained with a single firm for about 13 years [Carter and Savoca 1990]. In a later draft we will report the results of a continuous time model of firm longevity. This approach will allow us to calculate the effects of industry, capitalization, work force size, and motive power on the length of time a firm would remain in operation. Composition of Net Change in Milwaukee Manufacturing Employment, 1891-1898 We use these data to disaggregate the net change in manufacturing employment between 1891-92 and 1897-98 into the contributions made by firm birth, expansion, contraction, and death. We compare the results with evidence from the post-World War II period developed by Dunne, Roberts, and Samuelson [1989]. Their figures summarize the dynamics of all U.S. manufacturing plants with more than five employees present in any of the five Censuses of Manufactures conducted between 1963 and 1982. These Census surveys are quite comparable to those conducted by the Wisconsin Bureau except that they are five years rather than six years apart. Preliminary findings are reported in Table 2. The first column reports the net change in manufacturing employment between the two Factory Inspection surveys as a proportion of manufacturing employment in the earlier survey. Between 1891-92 and 1897-98, Milwaukee manufacturing employment grew 12.6 percent. The second through fifth columns show that this net employment change resulted from a 30 percent increase in jobs in newly-established plants, a 17.3 percent increase in jobs in expanding plants, a 16 percent cut in jobs in contracting plants, and an 18.7 percent job loss due to plant closings.3 While the figures indicate an enormous amount of employment instability resulting from turnover of the jobs themselves, what seems most striking to us is their similarity to measures for the post-World War II period. The greater importance of small, poorly capitalized firms in the early period led us to expect much more job turnover. In later drafts we will expand our description of sources of job turnover, reporting, for example, on the relative importance of across-industry employment flows versus flows across plants within an industry. We will also investigate the determinants of the relative job stability we measure. * I would like to thank Ruth McCormick for her careful, imaginative research assistance. The data used in this paper were made available by the University of California Historical Labor Statistics Project. The processing of the original published data into a computerized format was supervised by Susan Carter and Richard Sutch with the financial support of the National Science Foundation, the University of California, and Smith College. The data are archived at the Laboratory for Historical Research at the University of California, Riverside, California 92521. All errors are the responsibility of the present author. 1The employment figures are not strictly comparable. Those from the federal census of manufacturing refer to average number of employees while those from the Factory Inspection survey refer to full capacity employment. Since completing Table 1 we have coded figures for actual employment in 1897-98. Actual employment figures are used in the calculations in Table 2. 2Of the 357 firms recorded in the 1898 but not in the 1892 report, 181 or about half were established prior to 1892. This appears to suggest a serious underenumeration of firms in the 1891-92 survey. However the average employment level of firms omitted from the 1891-92 report was 27, less than a third of the average employment level of 99 recorded for those firms established prior to 1892 which were listed in both the 1892 and in the 1898 reports. This suggests that omissions were concentrated among the smaller employers. 3As Dunne, Roberts, and Samuelson note, these calculations indicate a lower bound on the sum of year-to-year employment changes over the six-year period. Table 1 Firms and Employment by Industrial Category, Federal Manufacturing Census and Factory Inspection Survey, Milwaukee, 1899 and 1897-98 Number of Firms Employment Census Factory Census Factory Inspection Inspection Food and Kindred Products 314 46 2396 3257 Textiles 178 35 4815 3606 Iron and Steel 116 85 10117 7581 Lumber and its Remanufacture 98 50 3251 3229 Leather and its Finished Products 79 34 4461 4998 Paper and Printing 138 61 2581 2263 Liquors and Beverages 23 30 3068 3024 Chemicals and Allied Products 25 6 98 68 Clay, Glass, and Stone Products 149 32 1506 1428 Metals and Metal Products other than Iron and Steel 134 24 1330 3837 Tobacco 236 8 943 760 Vehicles for Land Transportation 49 54 1355 4260 Miscellaneous Industries 148 46 3493 2346 Hand Trades 1557 40 5363 2046 All Other Industries 98 42 3551 3110 Non-manufacturing Industries* 0 43 0 1358 MISSING -- 1 -- -- TOTAL 3342 637 48328 47071 *Includes 32 laundries, nine wood and coal distributors one manufacturer of coal and water gas, and one miner and shipper of coal and iron. Table 2 Manufacturing Employment Change by Source 1891-97 and 1963-1982 Net Change Components of Net Change as in Employment as Proportions of Beginning-Period Employment Years ^L(t)/L(t) Births Expansions Contractions Closings 1891-97 .126 .300 .173 -.160 -.187 1963-67 .145 .141 .194 -.078 -.111 1967-72 z-.027 .196 .106 - .140 -.189 1972-77 .027 .158 .136 -.113 -.153 1977-82 -.038 .176 .117 - .154 -.177 Notes: All twentieth-century data refer to plants with five or more employees. Births are the number of employees in the second period in all plants that first appear in the second period; Expansions refer to the net change in the number of employees in plants in operation in both periods that expanded employment between the two dates; Contractions refer to the net change in the number of employees in plants in operation in both periods that contracted employment between the two dates; Deaths refer to the number of employees in plants in operation in the first period byt not in the second. Sources: 1891-97; Wisconsin Bureau of Labor, Census, and Industrial Statistics, 1892 and 1898, 1963-1982; Timothy Dunne, Mark J. Roberts, and Larry Samuelson, "Plant Turnover and Gross Employment Flows in the U.S. Manufacturing Sector," Journal of Labor Economics 7(1) (June 1989), p. 55.