Correspondence should be addressed to the authors at the History of Saving Project, Institute of Business and Economic Research, University of California, Berkeley, CA 94720. The data referred to in this article are archived at the Laboratory for Historical Research, University of California, Riverside, CA 92521

Charles Wetherell assisted with the computations and Eric Bails with the bibliographic citations. We are grateful to Susan Carter, Daniel Scott Smith, Thomas Weiss, and Samuel Williamson for helpful comments. We are also most grateful to Daniel Scott Smith and Janice Reiff for providing information on the sample of data they collected from the 1880 and 1900 manuscript censuses. Data tapes on several different samples were provided by the Inter-university Consortium for Political and Social Science Research. Financial support was provided by the Foundation de la Maison des Sciences de l' Homme, Paris; the John Simon Guggenheim Foundation; the Institute of Business and Economic Research; and the Academic Senate of the University of California.

¹ Ransom, Roger L. and Sutch, Richard, “The Labor of Older Americans: Retirement of Men On and Off the Job, 1870–1937,” this JOURNAL, 46 (03 1986), pp. 1–30.Google Scholar

² The explanation for this is twofold. First, gainful occupation included anyone with a source of income, including property income, while the labor force excluded retired persons who receive only property income. Second, the occupation designated for each individual was their “customary occupation” whether or not the individual was employed at that occupation at the time of the census. Thus it appears that many retired individuals were enumerated with their former occupations. See the discussion of this in U.S. Bureau of the Census, 1900, Occupations, pp. ccxxv, ccxxxiii;Google Scholar and Durand, John D., “Development of the Labor Force Concept, 1930–1940,” in Ducoff, J. Louis and Hagood, Margaret, Labor Force Definition and Measurement (Social Science Research Council Bulletin 55, 1947), pp. 80–90, esp. p. 81.Google Scholar

³ This “public-use sample” of the 1900 census was collected by the Center for Studies in Demography and Ecology at the University of Washington under the direction of Samuel Preston. The sample is available on computer tape from the Inter-university Consortium for Political and Social Research. The data on labor force participation are reported in Ransom, Roger L. and Sutch, Richard, “The Decline of Retirement in the Years Before Social Security: U.S. Retirement Patterns, 1870–1940”, in Ricardo-Campbell, Rita and Lazear, Edward P., eds., Issues in Contemporary Retirement (Stanford, 1988), pp. 3–37, appendix table.Google Scholar

⁴ Mushkin, S. J. and Berman, Alan, “Factors Influencing Trends in Employment of the Aged,” Social Security Bulletin, 10 (08 1947).Google ScholarDurand, John D., Labor Force in the United States, 1890–1960 (Social Science Research Council, 1948).Google ScholarGagliardo, Domenico, American Social Insurance (revised edn., New York, 1955).Google ScholarLong, Clarence, The Labor Force Under Changing Income and Employment (New York, 1958).Google Scholar As an alternative to adjusting gainful occupation to estimate labor force, Jon Moen has adjusted the postwar census data on labor force to extend the gainful occupation data beyond the 1930 census; see Moen, Jon, “Essays on the Labor Force and Labor Force Participation Rates: The United States from 1860 through 1950” (Ph.D. Diss., University of Chicago, 1987). Since many retired individuals were gainfully occupied, this measure of work rates is inappropriate for the subject at hand.Google Scholar

⁵ The first year for which there exists an official estimate of the labor force by age is 1930. The 1930 estimates of the labor force were made in the 1940s by Durand, John D. and Goldfield, Edwin, U.S. Bureau of the Census, Sixteenth Census, 1940, Population: Estimates of the Labor Force, Employment, and Unemployment in the United States, 1940 and 1930 (Washington, D.C., 1944).Google Scholar More detail is given in footnote 12 of Ransom, and Sutch, , “Labor of Older Americans,” p. 6.Google Scholar

⁶ The conventional view can be attributed to Durand, who estimated that the retirement rate of men 65 and older (not 60 and older) was 36.8 percent in 1900 and increased 25 percent between 1890 and 1930. See Durand, , Labor Force in the U.S., table A-6, p. 208.Google Scholar

⁷ See Ransom, and Sutch, , “Labor of Older Americans,” for a discussion of these earlier census returns, especially table 2 (p. 14), for the citations to the census volumes.Google Scholar

⁸ Moen, Jon, “The Labor of Older Men: A Comment,” this JOURNAL, 47 (09 1987), pp. 761–67.Google Scholar

⁹ There is no dispute about the trends in labor force participation after 1940. The labor force concept was introduced in 1937 and has been used as the conceptual framework for the routine data collection procedures of the Bureau of the Census and the Bureau of Labor Statistics since the 1940 census. For a discussion of the trends since 1940, see Tuma, Nancy Brando. and Sandefur, Gary D., “Trends in the Labor Force Activity of the Elderly in the United States, 1940–1980,” in Ricardo-Campbell, Rita and Lazear, Edward P., eds., Issues in Contemporary Retirement, pp. 38–83.Google Scholar

¹⁰ U.S. Bureau of the Census, Twenty Censuses: Population and Housing Questions, 1790–1980 (Washington, D.C., 1978), pp. 35–37.Google Scholar

¹¹ Moen, , in “Comment,” p. 761, disputes our statement that the concept of gainful occupation was introduced in 1890, replacing the concept of “productive employment” used in 1870 and 1880. He suggests that the 1870 and 1880 censuses also used a gainful employment concept, although he provides no citations to support this claim. We find the discussion of the definitions of occupation in the published census volumes and the instructions to the enumerators, which we cited in “Labor of Older Americans,” clear on this point and we are at a loss to understand why Moen finds the distinction “difficult to discern”. In any case, this point is not critical to the issue at hand (the meaning of “gainful employment”), although it does bear on the proper interpretation of the 1870 and 1880 data so we shall return to it in the next section.Google Scholar

¹² It is possible to identify individuals in the third group from the manuscript census returns of 1900. They were recorded with occupational designations such as “capitalist,” “landlord,” or “retired”. Moen is incorrect when he suggests that these individuals were excluded from the gainfully occupied in 1900. He correctly notes that the published tabulations for 1900 did not list capitalists or landlord separately; see Moen, , “Comment,” p. 763. However, it can be established from the sample drawn from the manuscript returns that when the occupations were tabulated in 1900, capitalists were combined with and reported as bankers, while landlords were combined with real estate agents.Google Scholar

¹³ Stanley Lebergott argues that a proportion of men recorded with months of nonemployment by the 1900 census should be considered as out of the labor force; Lebergott, Stanley, Manpower in Economic Growth: The American Record Since 1800 (New York, 1964), pp. 403–405.Google Scholar

¹⁴ U.S. Bureau of the Census, Twelfth Census, Occupations at the Twelfth Census (Washington, D.C., 1904), p. ccxxxiii.Google Scholar

¹⁵ The word “competence” in this context means the ownership of assets sufficient to finance the necessaries and conveniences of life without the need to work.

¹⁶ The census inquiry about the number of “months not employed” was accompanied by the following instructions issued enumerators: The object of this question is to get the number of months (or parts of months) in the census year (June 1, 1899, to May 31, 1900) during which each person having a gainful occupation was not employed. For those who have no gainful occupation, leave the column blank. The law does not contemplate that this question shall apply solely to the principal occupation in which the person may have been engaged during the year, but it is the intent to find out the number of months (or parts of months) during which a person ordinarily engaged in gainful labor was not employed at all. A return is required in columns 19 [occupation] and 20 [months unemployed] for each and every person 10 years of age and over who was engaged in gainful labor during any part of the census year (June 1, 1899, to May 31, 1900, inclusive), or who is ordinarily occupied in remunerative work but during the census year was unable to secure work of any kind. In the latter case enter his customary occupation, as carpenter, bricklayer, etc., in column 19 and the figure “12” in column 20 to show that, although he had an occupation or trade, he was not employed at all during the year at that or any other kind of work. U.S. Census, Twenty Censuses, p. 37. The insertions in parentheses are in the original while those in square brackets have been interpolated for clarity.Google Scholar

¹⁷ The data graphed are smoothed by presenting a moving average for five-year cohorts to eliminate some of the jaggedness produced by age heaping. Notice that the rates presented are not unemployment rates as conventionally defined (unemployment as a percentage of the labor force); they are instead population rates. Also note that the months of nonemployment collected by the 1900 census measures the total amount of unemployment experienced any time during the census year and will be higher than the contemporaneous unemployment rate.

¹⁸ A four-month cut-off was used by the Census Bureau to make the 1930 estimate of the labor force; see Durand, and Goldfield, , “Estimates of Labor Force,” p. 11.Google Scholar

¹⁹ For men 60 and older, 2.77 percent reported exactly six months of nonemployment, while for men under 60 only 1.98 percent reported six months.

²⁰ Not all those reporting months of nonemployment were unemployed during the census week. The total number of months of unemployment over the preceding year was recorded whether or not the individual was unemployed at the time. To calculate an estimate of the number of individuals who were temporarily unemployed during a typical week we divided the total number of months of unemployment during the previous year reported by men in the labor force by 12.

²¹ The profile of unemployment rates by age is reproduced as figure 1.9 in Ransom, and Sutch, , “U.S. Retirement Patterns,” p. 21.Google Scholar

²² Moen, , “Comment,” p. 764.Google Scholar

²³ Durand, and Goldfield, , “Estimates of Labor Force,” pp. 2, 11.Google Scholar

²⁴ Moen, , “Comment,” p. 764.Google Scholar

²⁵ Moen, , “Comment,” p. 762.Google Scholar

²⁶ U.S. Bureau of the Census, Twenty Censuses, pp. 18, 20.Google Scholar

²⁷ The published tabulations excluded “persons retired from active pursuits by reason of an acquired competence, of support secured from grown children, or of advanced age”; U.S. Census Office, Ninth Census [1870], Statistics of Wealth and Industry of the United States, III (Washington, D.C., 1872), p. 798.Google Scholar

²⁸ The instructions were identical in 1870 and 1880; U.S. Bureau of the Census, Twenty Censuses, pp. 19, 22.Google Scholar

²⁹ Ibid., p. 29.

³⁰ Ibid.., p. 35.

³¹ Moen, , “Comment,” p. 766.Google Scholar Moen misstates our position on this issue. He suggests that we claim the permanently unemployed were recorded with occupations by the enumerators, but removed from the 1870 and 1880 lists of occupations when the manuscripts were edited in Washington. He asserts that we justified removing the permanently unemployed from our 1900 estimate of the labor force in an attempt to achieve comparability with the earlier edited tabulations. He says such a procedure would be incorrect. “Because information on months of unemployment was not available in 1870 and 1880, these permanently unemployed individuals could not have been excluded from the occupation tables of those censuses,” Moen, , “Comment,” p. 764.Google Scholar However, he misunderstands us. Our method of estimating the permanently unemployed in 1900 was our own device and was not patterned after a method employed in 1880. For the reasons stated both here and earlier, we believe that individuals who can be classified in group 2b, the so-called “permanently unemployed,” were simply not recorded with an occupation in 1870 and 1880 but that many were in 1900. We should also note that the 1880 census did collect the months of unemployment, U.S. Census, Twenty Censuses, p. 20. All experts agree, however, that group 2a was probably underreported in that year.Google Scholar

³² W. Andrew Achenbaum rejected the published tabulation because he suspected an underenumeration of occupations of the elderly by the enumerators; see Achenbaum, W. Andrew, Old Age in the New Land: The American Experience Since 1790 (Baltimore, 1978), p. 179.Google ScholarMoen, , in “Comment,” p. 766, specifically asserts that Achenbaum is wrong, and we agree.Google ScholarRansom, and Sutch, , “Labor of Older Americans,” p. 7.Google Scholar Daniel Scott Smith has also judged the occupational enumerations for the older population in 1880 to be reliable; Smith, Daniel Scot., “A Community-Based Sample of the Older Population From the 1880 and 1900 United States Manuscript Census,” Historical Methods, 11 (Spring 1978).CrossRefGoogle ScholarPubMed

³³ Moen confines himself to the comment “how the Census Office arrived at the published estimates is still not clear,” Moen, , “Comment,” p. 766. Here again he seems confused, for the Census Office was not reporting “estimates” but complete tabulations of the enumerators' returns.Google Scholar

³⁴ The sample is mentioned, but described only in passing, in Smith, , “A Community-Based Sample.” A codebook provides more information;Google ScholarJensen, Richard, et al., “Old Age in the United States, 1880,” Inter-university Consortium for Political and Social Research 8427 (First edition, Winter 1985). The data are available in machine-readable format from the Inter-university Consortium in Ann Arbor, MI.Google Scholar

³⁵ Moen, , “Comment,” table 3, p. 765.Google Scholar

³⁶ If both percentages were correct, they would imply a labor force participation rate for men 60 to 64 of only 44 percent. This would be dramatically lower than the 80 percent rate for this group of men reported for 1900.

³⁷ Moen, , “Essays,” table 3, p. 29. We believe Moen's estimate, based on gainful occupation, is substantially in error; our estimate of labor force participation for men 65 and older in 1900 is 56.8 percent.Google Scholar

³⁸ Smith, , “Community-Based Sample,” p. 67.Google Scholar

³⁹ We have found cases where nonoccupational titles were ruled over with a different pen than the enumerator's. Elsewhere small checkmarks appear next to proper occupations, but the nonoccupational titles were not marked off.

⁴⁰ The “peculiarity of occupation” variable in the Newberry tape included codes for: “no peculiarity of occupation,” “no occupation reported” (which meant the space was left blank), “retired” or “pensioner,” “institutionalized,” or occupation reported as “sick,” “old,” “lame,” etc. See Jensen, , et al., “Old Age,” p. 4.Google Scholar

⁴¹ The fact that we found relatively few people with nonoccupational titles in the 1880 data (and no one reported as a capitalist or landlord) reinforces our contention that the census enumerators followed a definition of productive occupation consistent with modern labor force definitions. Presumably, such individuals—many of whom were listed with gainful occupations in 1900—were explicitly entered as “retired” or were not recorded with occupations in 1880 and 1870. The original occupation data were made available to us through the assistance of Daniel Scott Smith and Janice Reiff.

⁴² Most of the difference in the two estimates is due to the exclusion of the “permanently unemployed.” We suspect that our adjustment remains incomplete, since it seems likely that the unemployment variable was not completely reported in 1880; see U.S. Census Office, 1900, Occupations, p. ccxxvii. The Newberry sample suggests an unemployment rate for men 65 and older of less than 1 percent. Forty of the 65 sampling units contained no individuals that reported unemployment.Google Scholar

⁴³ Smith, , “Community-Based Sample,” pp. 67, 73.Google Scholar

⁴⁴ At the earlier date institutionalization was a common “solution” for the problem of the aged, particularly the infirm and dependent. The progressive reform movement that gained headway in the late nineteenth century sought to prevent the reliance on these institutions as dumping grounds for the elderly. Also, the expansion of savings banks, tontine life insurance (an early form of self-financed pensions), and government pensions for Civil War veterans during this period probably meant that more older men had sufficient resources to avoid the need to resort to public charity in their old age than 20 years earlier. See Ransom, Roger L. and Sutch, Richard, “Tontine Insurance and the Armstrong Investigation: A Case of Stifled Innovation, 1868–1905,” this JOURNAL, 47 (06 1987), pp. 379–90.Google Scholar

⁴⁵ The procedure was to choose every old person whose name appeared on every nth page (50 names to a page). The value of n was chosen to obtain approximately 30 people. Smith, , “Community-Based Sample,” describes the procedures used.Google Scholar

⁴⁶ When he reported results for the larger sample he drew from the 1900 census, Smith employed such a weighting scheme. We adjusted the urban and rural components of each state represented in the sample using population weights taken from the published census; U.S. Bureau of the Census, Historical Statistics of the United States, Colonial Times to 1970 (Washington, D.C., 1975), Series A202–203, pp. 25–37.Google Scholar

⁴⁷ The sampling error is taken as twice the standard error. This gives a 95 percent confidence interval. The standard error for a random sample is calculated as the square root of p times q divided by n; where p is the estimate of labor force participation, q is equal to 1 - p. and n is the sample size (740).

⁴⁸ The labor force participation rates of older men for each state in 1870 and 1880 are discussed in the text and displayed in Figure 2. The proportion of older men working at selected occupations was reported in Ransom, and Sutch, , “Labor of Older Americans,” table 6, p. 22. When referring to that table, note the similarity between the age structures of those with various occupations exhibited in 1870 and 1890 with that in 1880. This suggests that it is very unlikely that an unsystematic failure to fully tabulate the results for older men can explain the divergence between the Newberry sample and the published results.Google Scholar

⁴⁹ Ransom, and Sutch, , “Labor of Older Americans,” pp. 7–8.Google Scholar

⁵⁰ We are grateful to Daniel Scott Smith for suggesting this interesting conjecture.