The association of commuting time and wages for American workers with disabilities

2.1 Data

We used publicly available data from the 2016 American Community Survey (ACS). The ACS is the annual household survey conducted by the U.S. Census Bureau. The survey captures a range of information about housing, population, and the workforce. Information about commuting time to work is included as well. We restricted our sample to employed adults aged 18 to 64 (Unweighted N = 1,439,070) and weighted our data to adjust for the complex sampling design used in the ACS.

2.2 Measures

2.3 Analytical plan

We first ran descriptive statistics of our sample. We tested for differences in sociodemographic and travel characteristics by disability status using Chi square. Next, we ran a series of regressions. We ran two logistic regressions of ‘long travel time’, controlling for the covariates listed above with one key difference. The first model used the ‘any disability variable’ and the second model used the ‘disability type’ variables. Results are reported as odds ratios. These models generally followed the specification below:

F indicates the outcome of interest (long commute time) of individual i who lives in location j.

F _ij is a function of his or her underlying disability (H _ij ), individual characteristics (X _ij ), mode of transportation (Z _ij ), and unobservable factors (e _ij ) as follows: F ij = f ( H ij ; X ij ; Z ij ; e ij )

For the analysis conducted here, X contained age, center city residence, educational attainment, full or part-time employment, race, and sex and Z contained separate dummy variables for modes of transportation, with private vehicle as the reference group.

Third, we ran an ordinary least squares (OLS) regression of the log of average hourly wage. We used individual weights and robust clustering by Public Use Microdata Areas (PUMAs) to account for wage differences that might vary within different areas of the country. PUMAs are statistical geographic areas that nest within states or other entities. The regression followed the specification noted below:

Y indicates the outcome of interest (log of hourly wage) of individual i who lives in location j.

Y _ij is a function of his or her underlying disability (H _ij ), individual characteristics (X _ij ), long commute time (T _ij ), and unobservable factors (e _ij ) as follows: Y j = a + B 1 H j + ; B 2 X ij + B 3 T ij + e ij

X contained age, center city residence, educational attainment, race, and sex.

Again, this regression was conducted using two different models, one that included ‘any disability’ and one that included ‘disability types’, in addition to the covariates listed above. Coefficients and standard errors are reported. In addition, the final equations to predict hourly wages are provided and used to demonstrate wage differences by disability status, holding all else constant.

3.1 Descriptive results

Table 1 shows the sociodemographic characteristic of the sample. Nearly six percent of working-age employed adults had a disability. Some differences in characteristics were noted by disability status. Persons with disabilities were older and had lower levels of educational attainment than people without disabilities. In terms of employment, 68 percent of workers with disabilities worked full-time (35 hours or more per week). In contrast, 79 percent of workers without disabilities worked full-time.

Table 2 shows basic travel characteristics of our sample. While most employees took a private vehicle (85.6%) to work, five percent used public transportation. Five percent worked at home. Significant differences were noted by disability status. Persons with disabilities were less likely to use private vehicles and were slightly more likely to use public transportation to travel to and from work. Persons with disabilities were also slightly more likely to use other forms of transportation (six percent) than persons without disabilities (four percent). Of particular note is that very few American workers, regardless of disability status, reported using public transportation (five percent).

The mean travel time to work was 27 minutes. Twenty-three percent of workers traveled 35 minutes or more to work with little variation by disability status.

Table 3 shows results from two of our travel time logistic regressions, addressing our first two research questions. The first model in Table 3 uses ‘any disability’ as the disability indicator. The second model uses separate dummy variables to indicate each type of disability as well as persons with more than one type of disability. Persons without a disability were the reference group. When controlling for individual characteristics, mode of transportation, and full- or part-time status, the presence of any disability was not associated with a longer commute time. Similarly, individual types of limitations were not associated with longer commute times.

Both models show that men were significantly more likely than women to have a long commute time to work (OR: 1.30; p < 0.001) and that full-time workers were more likely to have long commutes than part-time workers (OR: 1.64, p < 0.001). Compared to those with less than a high school education, persons with higher levels of education had significantly higher odds of having long commutes. Respondents who lived in center cities were significantly less likely to have long commute times (OR:0.092; p < 0.001). Mode of transportation was significantly associated with commute time as well, as people who took public transportation to work were significantly more likely to have long commute times than those who used a private vehicle (OR: 7.31; p < 0.001).

Table 4 addresses our final research question, showing results of the log of hourly wage OLS regressions. Having a disability was significantly associated with lower levels of wages, even when controlling for travel time and other factors. For the general population, having a long commute was associated with higher wages.

In both models, being female, nonwhite, and having less than a high school education were associated with lower earnings.

Table 5 presents estimated hourly wages by disability status and commute time, holding all else constant and controlling for geographic differences. Estimates are based on the average person with a disability in the sample of employed adults. A hypothetical average case of a 49 year old white high school educated male who had a disability and a short commute from a non-center city location with a disability would earn $12.28per hour. A similar person without a disability would earn $16.09 per hour. A wage gap between workers with and without disabilities exists, even for workers who have longer commute times ($14.56 for workers with disabilities, $19.07 for workers without disabilities). Differences in wages are presented by activity, functional and sensory limitation as well.

4.1 Mode of transportation to work

Our descriptive analysis highlighted some important points. First, private vehicles are by far the most important mode of transportation to work in the U.S., for both workers with and without disabilities. In future years, advances to self-driving vehicle technology may increase the ability of persons with disabilities to utilize private vehicles for all purposes (Claypool, Bin-Nun, & Gerlach, 2017). Policymakers should ensure that such advances are inclusive of the heterogeneous population with disabilities, while also acknowledging that assistance might be needed across the entire travel chain, from door to door.

Second, public transportation is only used by a small portion of the population to get to work. A combination of reduced access to public transportation as well as concerns about increases in travel time and loss of flexibility may contribute to this low level of use. In addition, for workers with disabilities, issues of accessibility may come into play. Although federal legislation requires that public transit be accessible, issues remain. Disability advocacy groups in New York City, for example, recently filed a class action suit against the Metropolitan Transportation Authority alleging discrimination against persons with ambulatory disabilities (Disability Rights Advocates, 2017).

4.2 Commute time

As we began our research, we questioned whether workers with disabilities would have shorter commute times than others. We found, however, that workers with disabilities were not significantly more likely than those with no disability to have shorter commute times, even when controlling for mode of transportation, and sociodemographic characteristics. These results held even when examining specific types of activity, functional or sensory limitations. This can be construed as a positive result, suggesting that commuters with disabilities are behaving similarly to commuters without disabilities.

As our data did not include information about the actual miles traveled to work, we are limited in our ability to state with confidence that commute time is reflective of the geographic radius within which persons with disabilities are obtaining work, however. Persons with disabilities could be working closer to home while simultaneously facing longer commute times, possibly reflecting restricted choices in terms of employment. Other surveys which include more detail about travel to work, such as the National Household Travel Survey, are improving their identification of persons with disabilities and may provide a reliable source for further research in this area.

4.3 Wages and commute time

Our results also confirmed prior research, which had suggested that females had shorter commute times than men (Haley-Lock, Berman, & Timberlake, 2013). Females were also earning approximately 26 percent less per hour than men, holding all else constant. These findings highlight that women with disabilities may be at particular disadvantage in terms of employment, compared to men. Prior research has suggested that such joint disadvantage exists for women with disabilities (O’Hara, 2004). Adjusting the data in Table 5 to represent a female worker with a disability and similar characteristics as those described above results in a female worker with a disability earning $9.48 per hour. A male worker with a disability would earn 30 percent more per hour ($12.28). The need for additional gender-specific research and policymaking in this area is clear.

At the outset of our research, we also questioned whether workers who traveled similar lengths of time earned similar hourly wages, controlling for disability status and other factors. Our findings point to substantial monetary differences in wages between workers with and without disabilities. These results are striking considering that we controlled for individual characteristics (including educational attainment) as well as commute time. While workers with disabilities were as likely to have long commute times as other workers, the returns in terms of wages were lower. Indeed, workers with disabilities who commuted 35 minutes or longer earned less ($14.56 per hour) than workers without disabilities who had shorter travel times to work ($16.09 per hour). Wage differences were evident by limitation type as well. Persons with cognitive limitations earned the lowest amounts per hour.

Prior research among the general population (Blumenberg & Fields, 2013; Madden & Chiu, 1990; Rapino & Fields, 2013) has suggested that those who have longer commutes earn higher wages. Within limitation type, those who traveled longer to work did earn more per hour, yet did not match the higher wages of workers without disabilities. Other factors such as discrimination, health issues, or occupation choices may be negatively influencing the earnings of workers with disabilities. Future work which considers the relevance of these factors along with commuting characteristics can provide information that can be used by policymakers to ensure that persons with disabilities are earning to their full potential.

4.4 Limitations

In-depth qualitative research could build off of the findings presented here to better understand the nuances of commuting with a disability, while also addressing some of the limitations inherent in using cross-sectional Census data which relies on self-report data. Greater detail on job location relative to home, as well as disability types, associated health conditions, and personal decisions about employment options would provide more detail in important areas. In addition, it is important to note that this analysis focused on workers with disabilities and thus the results in terms of modes of transportation should not be assumed to apply to persons with disabilities who are not employed. Workers with disabilities are likely to have less severe limitations than other persons with disabilities, which may also influence residential choice and access to private transportation. The work presented here does not provide any detail about transportation barriers, which may be precluding other individuals with disabilities from joining the work force.