Job promotion in midcareer: gender, recession, and “crowding”
The NLSY79 has a number of advantages over other surveys. One is that it allows obtaining an individual’s actual labor market experience from the number of weeks worked since the last interview. Because women may work more discontinuously than men, capturing that experience corrects for the potential measurement error in the standard indicator based on age and education. Another advantage of the survey is that it contains detailed information on promotions. The survey questions related to promotions always concern in-house promotions, namely, those with the current employer. Specifically, the survey asks respondents (who are not self-employed) for information on up to five jobs, as follows: “Since [date from which information about employer will be collected (start date or date of last interview if last interview employer) (jobs 1–5)], have you experienced a promotion, a demotion, or any other type of position change?” If multiple promotions are recorded, subsequent questions regarding the nature of the promotion are asked for the most recent promotion in any given job. In this study, only the workers who have experienced a promotion with their current employers (that is, employers for job 1 in NLSY79, in any round) are counted as promoted. The appendix illustrates that promotions of this type constitute the vast majority of cases.6
Although labor market activity has been surveyed in great detail in the NLSY79 from the outset, occupation codes have not been recorded consistently across different waves of the survey. Between 1979 and 2000, occupations were coded on the basis of 1970 census occupation codes. Since 2002, however, occupations have been identified with the use of an updated classification system that captures new and emerging occupations.7 In the present analysis, occupation codes were mapped in order to enable comparisons of 1996 occupations with occupations in the 2006 and 2010 rounds of the survey (as well as those in all rounds for some of the analyses). Specifically, crosswalks provided in the literature were used to match all NLSY79 occupation codes to the 1990 census occupation codes.8
The characteristics of promotions
In 1996, when NLSY79 respondents were between 31 and 39 years of age, their careers were most likely to be taking off. In 2006, these same respondents, now ages 41 to 49, were at the peak of their careers. Table 1 captures differences in the promotion characteristics of workers at these two points in career development. Do the returns to promotion increase as one moves up the career ladder? Do later promotions come with more responsibility, if not necessarily more pay? And do the answers to these questions vary by gender? Besides addressing these questions, the present investigation attempts to gain some insight into the effects of adverse macroeconomic conditions on the promotion experiences of men and women by comparing NLSY79 round 2010, which encompasses the Great Recession, with round 2006. Moreover, in order to strengthen the discussion of the effects of the business cycle on promotions, the analysis subsequently utilizes all rounds of the survey and compares the age-specific experience of a younger cohort of respondents (those 31 to 35 years of age in 1996) with that of an older cohort of respondents (those 36 to 39 years of age in 1996) over the entire data period, but with a focus on the economic downturn of 2001 and the Great Recession of 2007–2009.
6 Because the NLSY79 became biennial in 1994, calculated promotion rates should be adjusted appropriately when compared with rates from earlier rounds of the survey.
7 According to NLSY79 attachment 3 (Industrial and Occupational Classification Codes), three-digit 2000 census codes are used in the 2002 survey, four-digit 2002 census codes are used in the 2004 survey, and four-digit 2003 census codes are used in the 2006, 2008, and 2010 surveys. Based on the codes in the attachment, dividing the four-digit codes used between 2004 and 2010 by 10 gives the same three-digit codes as those used in the 2000 census, except for the unemployed, the military, those not in the labor force, and uncodable items not included in the sample for this article. Instead of using one-digit occupation codes, the present study uses three-digit codes for the 2002–2010 period and then employs crosswalks to make these codes comparable to 1970 census codes. Attachment 3 of the NLSY79 can be found at http://www.nlsinfo.org/content/cohorts/nlsy79/other-documentation/codebook-supplement/nlsy79-attachment-3-industrial-and.
8 In particular, the analysis followed the crosswalks provided in David Dorn, "Essays on inequality, spatial interaction, and the demand for skills" (Ph.D. dissertation, University of St. Gallen, September 2009) and David H. Autor and David Dorn, "The growth of low-skill service jobs and the polarization of the U.S. labor market," American Economic Review, vol. 103, no. 5, 2013, pp. 1553–1597. These two papers provide three-digit occupation codes, or occ1990dd codes, that can be used as a link between the occupation codes of the 1970, 1990, and 2000 censuses. The analysis first used the crosswalk between the 1970 census occupation codes and occ1990dd codes and then the crosswalk between the 2000 census occupation codes and occ1990dd codes to code all occupations in the sample on a consistent occ1990dd basis. The crosswalk between the 1990 census occupation codes and occ1990dd codes also was used to integrate feminization measures from the Integrated Public Use Microdata Series (IPUMS) into the study’s dataset (see below). Further, Autor and Dorn’s aggregation was used to group all occupations to the one-digit level. These occupation codes were downloaded from Dorn’s website at http://www.cemfi.es/~dorn/data.htm. In the mapping of the 1970 census codes to occ1990dd codes, there were 66 occupations not observed in NLSY79 and 13 occupations that could not be directly mapped. For one of these occupations, namely occ1990dd occupation “274,” the code 280 from the 1970 census codes was assigned with the use of occupational definitions contained in Peter B. Meyer and Anastasia Osborne, “Proposed category system for 1960–1970 census occupations,” working paper 238 (U.S. Bureau of Labor Statistics, September 2005) and the appendix in Dorn, “Essays on inequality.” Altogether, the analysis lost only 10 observations from unsuccessful mapping. Similar problems were encountered in mapping the 2000 census codes to occ1990dd codes. Specifically, 20 occupations were not observed in the NLSY79, while 18 occupations could not be mapped. To minimize observation loss after mapping (to 10 or fewer observations), the sources cited above were used to assign the approximate occ1990dd codes for 16 occupations in the 2000 census.