Are workers with disabilities more likely to be displaced?

Abstract

The literature on employment and disability has been relatively silent regarding the job loss experience of persons with disabilities. We document the gap in job displacement rates across disability status in the United States over the 2007–2013 period using data from the 2010, 2012 and 2014 Displaced Worker Supplements of the Current Population Survey. We find that men and women with disabilities are, respectively, 75 and 89% more likely to experience an involuntary job loss than men and women without disabilities in the United States over the 2007–2013 period, with gaps in displacement rates of eight and seven percentage points for men and women, respectively. A significant gap is found in most occupation-education subsamples. Using a logit decomposition, we find that differences in observable characteristics do not explain the gap in the job loss rate across disability status. Longitudinal tests following workers over a one-year period point to a causal effect of disability on the likelihood of displacement. While the disability gap may be due to unobservable characteristics, job mismatch and employer discrimination are also possible explanations, highlighting the potential importance of employer and public policies in improving the job security of workers with disabilities.

Keywords:

• disability
• discrimination
• employment
• job displacement
• job mismatch

Acknowledgments:

We have benefited from insightful comments on an earlier version from Stephen Kaye, from participants at the 2010 Annual Meeting of the Labor and Employment Relations Association, and from other participants in the Disability and Demand-side Employment Placement Models project funded by the National Institute on Disability and Rehabilitation Research (NIDRR). All errors or omissions are those of the authors. This work was supported by the US Department of Education, NIDRR [grant number. H133A060033].

Notes

1. A review of this literature can be found in Mitra and Stapleton (2006).

2. There is now a large literature on the health consequences of job loss. For instance, there is evidence (Price et al., 2002) that job loss may via depression lead to impaired functioning. The literature on how job loss might affect persons with preexisting health problems or disability seems more limited. For a review on these literatures see Madrian (2006).

3. The results are very similar when using the disability measure based just on functional limitations (not shown here), which is not surprising given that almost everyone in the full disability sample reported some type of functional limitation.

4. A similar gap was also found for each year of data separately (January 2010, 2012 and 2014).

5. Baldwin and Schumacher use a broader activity limitation measure: they consider that a person has a disability if he or she indicates that a health condition ‘limits their ability to work at a job or around the house’ or ‘limits their mobility or ability to communicate.’ This broader measure used by Baldwin and Schumacher might include persons with less severe disability compared to the one used in this paper, which may explain part of the difference in the finding. In addition, the sample of interest is also constructed differently. Baldwin and Schumacher include all people employed 20 months ago and follow their job changes (voluntary or involuntary) for 20 months. Table 1 focuses on people who, as of January, are employed or are not employed but report that they lost a job in the past three years. People who are not employed but left their jobs for reasons other than displacement (e.g. early retirement) are not included in the sample. This may lead to an upward bias in the estimated disparity in the job loss rate and the probability of losing a job across disability status, if the probability of voluntarily leaving a job is different across disability status. Finally, Baldwin and Schumacher use SIPP data for 1990 and 1993 while Table 1 relies on CPS data for January 2010, 2012 and 2014. If disparities in job loss across disability status changed between the early 1990s and 2010s, this could contribute to explain the difference in results between the two studies. In addition, it should be noted that Baldwin and Schumacher did not break down their analysis by gender.

6. As an alternative, we coded occupation by first using the occupation of the current job and second that of the job lost if the occupation of the current job is not available and the person was displaced. There was a high correlation between current occupation and the occupation of the job lost, so results were overall consistent with these two different ways of coding occupation.

7. For respondents not in the 4th or 8th month of their CPS rotation in January when the Displaced Worker Supplement was conducted, union status was added by merging CPS data from the following three months for those who did not switch jobs.

8. For respondents in months 5 to 8 of the CPS sample when the DWS supplement was conducted, we matched to their observations in months 1 to 4 in the preceding January using household and household member identifiers.