We thank Amy Bertin for first-class research assistance. Earlier, Boris Simkovich was instrumental in laying the foundations. We have had helpful comments from a number of seminar and conference audiences and particularly useful suggestions from Charles Calomiris and Helen Manning Hunter at the Economic History Association meetings. This work was supported by the Division of Research of the Harvard Business School, by a special grant from the Dean's discretionary funds, and by NSF Grant SES-9023021. The usual disclaimer applies.Google Scholar

¹ See Bernanke, Ben S., “Employment, Hours, and Earnings in the Depression: An Analysis of Eight Manufacturing Industries,” American Economic Review, 76 (03 1986), pp. 82–109.Google ScholarThe representative firm paradigm is also widely used in the literature on short-run increasing returns to labor that started with Hultgren, Thor, Changes in Labor Cost During Cycles in Production and Business (New York, 1960), and that has again become a focus of interest.Google ScholarSee, for example, Summers, Lawrence H., “Some Skeptical Observations on Real Business Cycle Theory,” Federal Reserve Bank of Minneapolis Quarterly, 10 (Fall 1986), pp. 23–27;Google Scholarand Bernanke, Ben S. and Parkinson, Martin L., “Procyclical Labor Productivity and Competing Theories of the Business Cycle: Some Evidence from Interwar U. S. Manufacturing Industries” (National Bureau of Economic Research Working Paper No. 3503, Oct. 1990).Google Scholar

² See the foreword by Thorp, Willard L. and “Trends in the Scale of Manufacturing Operations” by Thorp, Willard L., Humphrey, Don D., and Porter, Martha H. in Thorp, et al. , Temporary National Economic Committee Monograph No. 27: The Structure of Industry (Washington, DC, 1941).Google Scholar

³ Hounshell, David A., in From the American System to Mass Production: The Development of Manufacturing Technology in the United States (Baltimore, 1984), gives a careful treatment of the development of the methods at Ford. He also provides an authoritative discussion of the general slowness of the diffusion of American System methods.Google Scholar

⁴ See Griliches, Zvi, “Hybrid Corn: An Exploration in the Economics of Technical Change,” Econometrica, 25 (10 1957), pp. 501–22.CrossRefGoogle Scholar

⁵ On central office management, see Crowder, Walter F., “The Integration of Manufacturing Operations,” in Thorp et al., The Structure of Industry.Google ScholarOn the technical advantages of mass production, see Fetter, Frank, “The Fundamental Principle of Efficiency in Mass Production,” Appendix D, in Temporary National Economic Committee Monograph No. 13: Relative Efficiency of Large, Medium-Sized, and Small Businesses (Washington, DC, 1941).Google ScholarThe key modern study is Chandler, Alfred D. Jr, The Visible Hand: The Managerial Revolution in American Business (Cambridge, MA, 1977).Google Scholar

⁶ See Crowder, Walter F., “The Concentration of Production in Manufacturing,” in Thorp et al., The Structure of Industry.Google Scholar

⁷ But see Fetter, “The Fundamental Principle,” for the view that efficiencies are insufficient to explain the increase in concentration.Google Scholar

⁸ See Means, Gardner C., “Industrial Prices and Their Relative Inflexibility,” Document No. 13: Letter from the Secretary of Agriculture (Washington, DC, 1935).Google ScholarFor some statistical calculations, see Crowder, “The Concentration,” chap. 5.Google Scholar

⁹ See Thorp et.al, The Structure of industry.Google Scholar

¹⁰ The U.S. Bureau of the Census conducted a Census of Manufactures covering each of the years 1929, 1931, 1933, and 1935. See Bohme, Frederick, “U.S. Economic Censuses, 1810 to the Present,” Government Information Quarterly (1987), pp. 221–43. The overwhelming majority of the manuscript returns have survived. We are engaged in a long-term project of coding, editing, and analyzing data from them. Then, as now, the sampling frame for these censuses is establishments. Questions varied from year to year, and budgetary restrictions prevented tabulation of all the data. Nonetheless, the enumerators collected a huge and fascinating body of data on the costs of variable inputs, output, revenue, output composition, and some aspects of capital investment and organization.Google Scholar

¹¹ More precisely from Census Industry 1408, “Motor Vehicles.” The upstream industry, “Motor Vehicle Bodies and Parts,” is classified separately as Census Industry 1407.Google Scholar

¹² All these data were taken from the 1939 Census of Manufactures summary volume.Google Scholar

¹³ The greatest departures from modern practice are in the treatment of capital. There were no questions about the value of the capital stock or about investment. For 1929 and 1935, however, there were questions about the physical capital stock. These asked about the horsepower of installed machines. The 1935 figures were neither tabulated nor reported by the Census Bureau.Google Scholar

¹⁴ The actual trough probably occurred somewhat earlier: 1933 was in fact a year of very slight growth. We will refer to 1933 as the trough throughout this article, however. The usage is conventional, and more accurate descriptions are cumbersome.Google Scholar

¹⁵ The census schedule asked for “[the] number [of wage earners] on [the] payroll for [the] week which included [the] 15th day of [the] month, if this was a normal week” for each month in the year. (if that-week was not normal, respondents were to give the number for one that was.) We summed these over the 12 months to calculate WEM.Google Scholar

¹⁶ We are able to report the 1933 Salaried Employees figure because the Census Bureau collected the information, even though it did not publish any statistics.Google Scholar

¹⁷ Unique identifiers were a key element in the Census Bureau's construction of the longitudinal establishment database from the postwar Census of Manufactures and Annual Survey of Manufactures. See McGuckin, Robert H., “Longitudinal Economic Data at the Census Bureau: A New Database Yields Fresh Insights on Some Old Issues” (U.S. Bureau of the Census Center for Economic Studies Discussion Paper CES 90–1, 1990).Google Scholar

¹⁸ Note that the census was actually conducted in the first quarter of the following year. A plant might therefore have been open during the census year but closed and unresponsive by the time of the survey. We do not believe this problem seriously affects our data.Google Scholar

¹⁹ These considerations apply principally to small, single-plant firms. If the firm had only one plant in a city, we called it a match even if the address changed. This reflects the reality of the way the address fields were filled in. It also reflects our analytical concerns. The ongoing assets of the establishments and firms under study here are not so much buildings as firm-specific human capital, both managerial and blue-collar. These assets would most likely be preserved in a short move of the physical plant.Google Scholar

²⁰ We searched through the 1929 body-and-parts industry forms to see whether these apparently new establishments were simply relabelings of establishments previously classified in the other industry. We found that only two (out of well over a thousand) might conceivably qualify. One was an Auburn plant in Connersville, Indiana. the other a McCann (truck) Corporation plant in Cumberland County, Maine. The match was not compelling in either case.Google Scholar

²¹ We searched the pages of Automotive Industries and Cram's Automotive Reports (the ancestor of Ward's) for references to plant openings and closures. This resolved the status of a number of the missing establishments. Unfortunately, it left 30 unresolved. As the problem was overwhelmingly worst at single-plant firms, we had hopes of using Thomas' Register of Manufactures to resolve many more.Google Scholar This source has been used to document shakeouts before—for example, in the influential article by Gort, M. and Klepper, S., “Time Paths in the Diffusion of Product Innovations,” Economic Journal, 92 (1982), pp. 630–53. Unfortunately, a comparison of plant closure news stories in the industry trade sources with Thomas's showed that deletion of a Thomas listing could lag actual closure by four or more years.CrossRefGoogle Scholar

²² See Wiesmyer, M. L., Reminiscences (typescript in the Ford Archives, Dearborn, MI), for example, p. 58.Google Scholar

²³ These data, like most microdata, are very noisy. The noise levels documented in Table 2(B) notwithstanding, the mean levels of these variables are, statistically as well as economically, significantly different. Simple F-tests reject equality of the 1929 means of continuing and closing plants for all the variables mentioned in the text. The F-tests assume normality, perhaps falsely; but the test statistics are very far from the margin of rejection.Google Scholar

²⁴ This statement uses a “jobs” definition rather than the more familiar “employed persons” definition. This is because we can follow plants, but not people, over time.Google Scholar

²⁵ Treating openings as negative closures only reduces the figure to 30.2 percent.Google Scholar

²⁶ The total number of establishments operating in 1937 and 1939 rose slightly again. We do not know whether this small net entry masks a larger gross flow.Google Scholar

²⁷ See, for example, Hall, Bronwyn H., “The Relation between Firm Size and Firm Growth in the U.S. Manufacturing Sector,” Journal of Industrial Economics, 35 (06 1987). pp. 213–36.CrossRefGoogle Scholar

²⁸ See Bresnahan, Timothy F. and Raff, Daniel M. G., “Intra-Industry Heterogeneity and Composition Effects in the Supply Response to the Great Depression: The American Motor Vehicle Industry 1929–1935” (Hoover Institution Domestic Studies Program Working Paper, Oct. 1990), p. 15.Google Scholar

²⁹ For Gibrat's law, see Hall, “The Relation.”Google Scholar

³⁰ Bresnahan and Raff, “Intra-Industry Heterogeneity,” pp. 10–18.Google Scholar