The author is Assistant Professor of Economics at Yale University, New Haven, Connecticut 06520. A longer version of the paper is available from the author. A future publication on employment estimates is under preparation. Helpful comments from Stanley Lebergott and members of the Macroeconomics Workshop at Yale are gratefully acknowledged. Stanley Lebergott and Christina Romer provided unpublished data from their research. The research assistance of Laura Owen and financial support from a grant to the Department of Economics at Yale are greatly appreciated.

¹ The years correspond to distinct estimation procedures. GNP estimates are from Kuznets, Simon, National Product since 1869 (New York, 1946),Google Scholar and Capital in the American Economy: Its Formation and Financing (Princeton, 1961).Google Scholar Unemployment was estimated by Lebergott, Stanley, Manpower in Economic Growth: The American Record since 1800 (New York, 1964).Google Scholar

² Kuznets, , National Product.Google Scholar

³ Shaw, William H., Value of Commodity Output since 1869 (New York, 1947).Google Scholar

⁴ The average ratio of commodity output to GNP for 1889–1929 (both in 1929 prices) was 47, using figures in Kuznets, , Capital, pp. 553–54 for commodity output and pp. 555–56 for GNP (Variant III).Google Scholar

⁵ Notably Gallman, Robert E., “Gross National Product in the United States, 1834–1909,” in Output, Employment, and Productivity in the United States after 1800, NBER Studies in Income and Wealth, no. 30 (Princeton, 1966),Google Scholar who made some revisions to the underlying commodity output data and made use of estimates of distributive margins by Barger, Harold (Distribution's Place in the American Economy since 1869 [Princeton, 1955]).Google Scholar

⁶ Kuznets, Capital, Appendix C discusses his regression estimates. See Romer, Christina, “New Estimates of Gross National Product, 1872–1918” (unpublished manuscript, Princeton, 1985) for details of her revisions.Google Scholar

⁷ The 1909–1918 estimates are described only in notes to Table R-20 (Kuznets, , Capital, pp. 551–52), except to note that they are “probably subject to a wider margin of error than those for the years beginning with 1919” (p. 535).Google Scholar

⁸ The precise coefficient depends on the specification of trend. Seven-year moving averages will be used here, for comparability with Romer's estimates. On that basis, Kuznets's regression series for GNP have a standard deviation of percentage deviations from trend 85 percent as great as the standard deviation of commodity output for 1872–1914. In log differences, the ratio is 9.

⁹ The estimates reported here are based on percentage deviations from seven-year moving averages—the specification used by Romer. None of the important findings are altered under alternative specifications with shorter moving averages or first differences of logarithmic values.

¹⁰ That by no means contradicts Romer's well-founded criticism of recent users of Kuznets's unadjusted components series (which assumes, in effect, a coefficient of one) such as Friedman, Milton and Schwartz, Anna, Monetary Trends in the United States and the United Kingdom: Their Relation to Income, Prices, and Interest Rates, 1867–1975 (Chicago, 1982);CrossRefGoogle Scholar and Gordon, Robert J., “Price Inertia and Policy Ineffectiveness in the United States, 1890–1980,” Journal of Political Economy, 90 (12 1982), pp. 1087–17.CrossRefGoogle Scholar

¹¹ This was suggested by Robert Gallman during the discussion following the session in which this paper was presented. Kuznets was skeptical of such an approach, citing the only moderate success of the aggregate regressions (Capital, p. 545). But aggregating weak or unstable relationships merely hides the problem.Google Scholar

¹² An alternative approach would be to compare time periods on the basis of some other concept for which consistently comparable data have been collected. Romer's postwar replication of Lebergott's estimates does not constitute such a comparison for reasons discussed below. I have explored the cyclical properties of several alternatives, all of which show substantial stabilization after World War II.

¹³ Lebergott, , Manpower.Google Scholar

¹⁴ Lebergott's estimates for the 1890s and 1930s were on a somewhat different basis than those for 1900–1930 that are the target of Romer's criticisms. The 1890–1899 estimates were never given much credence by Lebergott, though they have been included in standard compendia (compare U.S. Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970, Bicentennial Edition, Part I [Washington, D.C., 1976], Series D-85–86, p. 135).Google Scholar He interpolated all of employment on a manufacturing employment index. Alternative estimates are discussed in Weir, David R., “Stabilization Regained: A Reappraisal of the U.S. Macroeconomic Record, 1890–1980” (unpublished manuscript, Yale University, 09 1985), pp. 32–36.Google Scholar On the 1930s see Darby, Michael R., “Three-and-a-Half Million U.S. Employees Have Been Mislaid: Or, an Explanation of Unemployment, 1934–1941,” Journal of Political Economy, 84 (02 1976), pp. 1–26.CrossRefGoogle Scholar

¹⁵ See Woytinsky, W. S., Three Aspects of Labor Dynamics (Washington, 1942) for a discussion of the concept of the usual labor force.Google Scholar

¹⁶ Most of the nonagricultural wage and salary worker series fall in this category. Trade employment was calculated by inflating census totals (as constructed by Lebergott) by the 1930 ratio of BLS employment to the census total (Lebergott, , Manpower, p. 453). Agricultural employment and some of the service sector and self-employed were benchmarked directly to population censuses.Google Scholar

¹⁷ Lebergott, , Manpower, pp. 356, 20.Google Scholar

¹⁸ I have been unable to find a description of the procedures in Manpower. The methods described here were deduced from an analysis of the appendix tables and confirmed in private communication with the author.

¹⁹ Romer, Christina, “Spurious Volatility in Historical Unemployment Data,” Journal of Political Economy, 94 (02 1986), pp. 1–37.CrossRefGoogle Scholar

²⁰ Romer avoided the problem of the duplicating item by working with sectoral employment estimates from the CPS household surveys and replacing them with alternatives consistent with CPS levels at benchmarks. Since it was never Lebergott's intention to estimate CPS totals by sector, this is not a fair test of the sectoral series. Whether or not it is a fair test of Lebergott's attempt to estimate aggregate employment consistent with CPS measures depends on a great many things not addressed by Romer.

²¹ See regression analyses in Table 3 of Weir, “Stabilization Regained.”

²² Lebergott, , Manpower, p. 436.Google Scholar

²³ Shaw, , Commodity Output, p. 96.Google Scholar

²⁴ See, for example, Woytinsky, , Labor Dynamics, pp. 105–93;Google Scholar and Carson, Daniel, “Changes in the Industrial Composition of Manpower since the Civil War,” in NBER, Studies in Income and Wealth, vol. 11 (1949), p. 145.Google ScholarDurand, John, The Labor Force in the United States, 1890–1960 (New York, 1948) concludes that the net effect of cycles on participation before 1940 is uncertain.Google Scholar

²⁵ Butz, William and Ward, Michael, “The Emergence of Counter-cyclical U.S. Fertility,” American Economic Review, 69 (06 1979), pp. 318–38.Google Scholar