Characteristics of Low-Wage Workers and Their Labor Market Experiences: Evidence from the Mid- to Late 1990s. Subgroup Findings


We have found that the average earnings of low-wage workers improve somewhat over time. At the same time, however, many low-wage workers do not experience positive labor market outcomes. This section addresses the important question: Which groups of low-wage workers experience improvements in their labor market outcomes and which groups do not? Examining differences in overall employment outcomes across subgroups of the low-wage population has important policy implications for targeting appropriate services to those who are at most risk of poor outcomes.

We conducted our subgroup analysis in two interrelated ways. First, we examined key labor market outcomes for selected subgroups one at a time. These subgroups were defined by worker, area, and job characteristics at the time the workers started their low-wage jobs. (31) Second, we conducted a multivariate analysis to examine the association between particular explanatory (subgroup) variables and key labor market outcomes, holding constant the effects of other explanatory variables. The multivariate analysis accounts for correlations among the subgroup variables and also allows us to efficiently examine labor market outcomes for a large number of subgroups.

We examined four key labor market outcomes for the subgroup analysis:

  1. The percentage of months low-wage workers spent in low-wage jobs during the 42-month follow-up period
  2. The percentage of months workers spent in higher-wage jobs (that is, in medium- and high-wage jobs)
  3. The percentage of months workers spent in all jobs
  4. Whether the worker spent less than 25 percent of months in higher-wage jobs

We used the total time employed measure to assess the overall labor force attachment of subgroups of low-wage workers. We examined the average percentage of time that workers held higher-wage jobs to assess the extent to which subgroups of workers were able to escape the low-wage labor market over time. Finally, because focusing on averages can mask important subgroup differences in the distributions of the amount of time workers spent in various labor market activities, we also examined the share of workers who spent little time (less than one-quarter time) in the medium- and high-wage labor market sectors. Together, these summary outcome measures were used to identify subgroups who had the most and least successful labor market experiences.

The subgroup analysis was conducted separately by gender. Furthermore, all figures were calculated using the longitudinal panel weight. We estimated the multivariate models using ordinary least squares methods for the continuous outcome measures (the first three listed above) and logit maximum likelihood methods for the binary outcome measure (the fourth measure listed above). In the multivariate analysis, we conducted statistical tests to gauge the statistical significance of differences in labor market outcomes across subgroups. For some subgroups with small sample sizes (see Table III.1), the standard errors of the estimates are large. Consequently, some relatively large parameter estimates are not statistically significant. (32)

We included the following categories of explanatory variables in the regression models:

  • Individual and household characteristics measured at job start (from the longitudinal panel file)
  • Prepanel employment information (from the wave 1 topical module)
  • Job characteristics measured at the start of the low-wage employment spell (from the longitudinal panel file)
  • Area characteristics and state economic indicators measured at the start of the job, as well as changes in unemployment rate indicators between the start and end of the follow-up period (from published data sources; see the Methodological Appendix A)(33)

View full report


"report.pdf" (pdf, 1.02Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®