The African-American Labor Supply after Reconstruction: Added Worker Effects in Urban Families

“The added worker effect” refers to the entrance of a family member into the labor force after its head of the household becomes unemployed. Most commonly, a married woman who enters the labor force when her husband loses his job is the eponymous added worker. Although the description has intuitive appeal, present-day empirical evidence of the added worker effect has been thin. Unemployment-insurance benefits mitigate the need for a wife to begin a job search, at least in the short run. In addition, relatively high employment rates of married women may make it statistically difficult to detect movement of a wife into the labor force. Finally, in aggregate data, and perhaps in microdata as well, a discouraged worker effect may work against an added worker effect, if, in response to their husband’s job loss, some wives remain out of the labor force, or if they give up on their own job search.

Historical data have supported the added worker hypothesis, but data availability has limited the chances to conduct tests. Although the 1880 and 1900 censuses asked about “gainful employment,” exactly what the Census Bureau meant by that phrase, in present-day terms, and how respondents interpreted it when answering, is muddled. Nonetheless, from census records, as well as federal cost of living surveys, a small literature has emerged to confirm the existence of an added worker effect in the late nineteenth and early twentieth centuries.

This article uses a unique dataset to make three distinct but intertwined contributions to that literature. First, the sample consists of African Americans, whose family economies at the end of the nineteenth century remain understudied for lack of data. Second, questions about periods of ill health among family members enabled us to address problems with previous measures of gainful work. Third, heads of households who had not worked at some point during the previous year reported the various financial methods that they used to manage their intervals of joblessness. Together, the survey’s characteristics enable definite conclusions about labor-market decision making within African-American families at a time previously understood only through anecdote.

The Added Worker Effect and African-American Unemployment The first empirical study of the added worker effect, by Mincer, differentiated between effects of husbands’ short-term (strong) and long-term (weak) unemployment on their wife’s entry into the labor supply. Long asserted that Mincer’s lack of data about African-American women presented “a severe challenge” to claims of having modeled the entire labor market. He urged that “the case of Negro wives calls for more discussion.” An early and influential study by Heckman and MaCurdy found no added worker relationship with longitudinal data. One reason for this finding is a possible endogeneity problem due to assortative mating, for which longitudinal data might control. In this theory, spouses with similar levels of human capital or degrees of labor-market attachment marry. Lundberg, also using longitudinal data, found a small added worker effect in white families, but strong evidence of endogeneity in African-American families, possibly due to this assortative mating process. ¹

The added worker effect seems to have been stronger in the past than in more recent times. Toward the end of the Great Depression, according to Finegan and Margo, wives of men who were unemployed but not on work relief in 1939 were half again as likely to be in the labor market as were wives with employed husbands, in part due to screening for the assignment of public-works positions. Analogous results have been found in the present day; Cullen and Gruber reported strong evidence that unemployment insurance reduced wives’ tendency to enter the labor force in response to their husbands’ unemployment. The implication is that historical data collected in the absence of unemployment benefits may yield evidence of an added worker effect that is simply latent in the present day. ²

More general added worker effects have appeared in historical data that were collected when unemployment was first recognized as a general social and economic...