The rise in women’s share of nonfarm employment during the 2007–2009 recession: a historical perspective
Whereas CES payroll employment estimates are restricted to nonagricultural wage and salary jobs and exclude private household workers, CPS estimates are based on a broader employment definition. Employment estimates from the CPS provide information about workers in both the agricultural and nonagricultural sectors and about all types of work arrangements; that information captures workers with wage and salary jobs (including employment in a private household), the unincorporated self-employed, unpaid absences from work, multiple job holders, and those doing unpaid work for at least 15 hours a week in a business or farm operated by a family member. As a result of this broader definitional scope, employment estimates derived from the CPS survey are higher than those derived from the CES survey. In addition, while periodic discrepancies in employment trends have occurred, the payroll and household surveys track well over the long term.
Because of the rich demographic detail available in CPS data, these data are generally preferred for the analysis of employment trends among women and men. In addition, the CPS survey estimates employment series separately for women and men, whereas the CES survey estimates employment series for total employment and women, as well as production or nonsupervisory employees. As a result, the employment series for men, as estimated by the payroll survey, must be calculated by the data user.
However, the CES survey offers certain benefits that provide for an enhanced analysis in this paper. While both surveys are subject to sampling error, the payroll survey has a much larger sample size than the household survey.4 The payroll survey’s monthly sample includes approximately 554,000 business establishments of all sizes, representing about one-third of total nonfarm employment. By contrast, the monthly sample size of the household survey is much smaller, at 60,000 households, and covers a smaller fraction of total employed people. Over-the-month changes in household survey employment are therefore subject to a larger sampling error, about four times that of the payroll survey. Because of its larger sample size, smaller sampling error, and more accurate estimation processes, the CES survey allows a more representative analysis of employment trends among men and women across industries.
The household survey first began to estimate civilian employment of men and women in 1948, whereas the establishment survey first produced women’s employment estimates in 1964. Although long-term employment trends have been similar for both surveys, the women’s share of employment, as measured by the household survey, has not risen as steeply as it has risen when measured with the use of payroll data. (See figure 2.) In terms of employment shares, in January 1964, women accounted for 34 percent of household employment and 32 percent of payroll employment; by December 2013, these shares had grown to 47 percent and 49 percent, respectively.
Employment shares, when broken down by gender, reveal an interesting and not entirely explainable shift between the two surveys: in earlier periods, the household survey accounted for a larger share of women’s employment (relative to the payroll survey) than it does today. It is likely that this shift is partly due to the broader definition of employment adopted in the household survey. For example, civilian employment, as measured by the household survey, includes unpaid family workers in both agriculture-related and nonagricultural industries. Historically, this class of workers has been represented predominantly by women and has suffered a long-term contraction, with its share of total employment falling from 1.8 percent in 1964 to less than 0.1 percent in 2013. An additional reason for the decline in the employment share of unpaid family workers is the 1994 CPS redesign, which classified a substantially smaller proportion of women as unpaid family workers.5