Disability prevalence and social insurance program participation among workers in selected nonstandard work arrangements

Lawrence F. Katz and Alan B. Krueger have identified both supply- and demand-side explanations for this labor market development.³ Supply-side explanations suggest that the observed rise in nonstandard work arrangements has been due to increases in the age and education of the U.S. workforce, as well as to shifts in individual preferences toward flexible work hours and work–life balance. Demand-side explanations focus on technological innovation, suggesting that new technologies have eased the monitoring and collection of information about current and potential workers and thus reduced the potential advantages of internal labor markets (those in which higher level job vacancies are filled from within the firm). In addition, rising income inequality and declines in unionization have limited workers’ collective bargaining power and ownership opportunities. Relying on external employees allows firms to reduce the amount of rent sharing with their workforce.

In this article, we explore the relationship between subsets of nonstandard forms of employment and worker disability status. Since the 1990s, several countries have introduced new types of labor contracts to increase labor market flexibility, and some industry-specific studies have found that, compared with permanent workers, temporary workers have significantly higher rates of mental and physical health problems and suffer more severe injuries in the workplace.⁴ Over the last 20 years, similar evidence has been collected and presented for the newly emerging categories of nonstandard workers, indicating that these workers face greater mortality and morbidity risks, as well as higher rates of mental problems and physical injuries.⁵ According to Michael Quinlan and Philip Bohle, three main factors explain such adverse outcomes: “pressures,” which arise from the precariousness of jobs and the consequent income insecurity;⁶ “disorganization,” which stems from the lack of clarity about who, among employers, is responsible for investment in human capital and for health and safety on the job;⁷ and “regulatory failure,” which arises from the lack of worker representation, enforcement of safety norms and regulations, and social insurance benefits.⁸

Studying workers in vulnerable and nonstandard employment status has been challenging. As noted by Dana Madigan, Linda Forst, and Lee S. Friedman, “One of the difficulties in researching temporary employment is the wide variety of definitions used to capture this group of workers. … [T]erms used to describe vulnerable employment status include temporary, contingent, precarious, non-standard, and sub-contractor…. Particularly in the United States, the inability to better define the effects of temporary employment may stem from a variety of limitations in available data. Additionally, occupational health and safety regulations and procedures can vary by state, giving wide fluctuations in data available.”⁹ Because of these challenges, much of the existing research on worker disabilities and nonstandard work arrangements has been based either on state workers’ compensation data or on datasets from outside the United States. Our research enhances the relevant literature by using a large, nationally representative U.S. dataset. We examine the frequency and typology of disabilities among two subsets of nonstandard workers (temporary agency workers and a proxy category for other nonstandard workers) whom we compare with other employees. We also explore differences in those workers’ use of selected social insurance programs (Medicaid, Medicare, and workers’ compensation). To the extent that the nonstandard labor contracts selected for study are associated with higher rates of long-term health problems and disabilities, our results reinforce the need for further research on the effectiveness of potential public policies or new contractual mechanisms designed to help disabled workers in newly emerging nonstandard jobs.

Data

Our analysis uses data from the Current Population Survey Annual Social and Economic Supplement (CPS ASEC), the largest U.S. labor force dataset, which is cosponsored by the U.S. Census Bureau and the U.S. Bureau of Labor Statistics (BLS). The CPS ASEC is an annual household survey that supplements CPS-collected monthly labor force information with data obtained from additional questions on work experience, income, noncash benefits, and migration of people ages 15 years and older. We limit our data analysis to the 2010–14 period for three reasons: (1) the CPS six-question sequence on disability (about difficulties related to hearing, seeing, concentrating, physical movements, activities of daily living, and instrumental activities of daily living) was not added to the survey until June 2008, with annual averages becoming available in 2009;¹⁰ (2) the economy was still in a recession in 2009 (the 2007–09 Great Recession), which can skew our results; and (3) many expansions of the Affordable Care Act went into effect in 2014, including voluntary Medicaid expansions and state health insurance exchanges. These three factors can complicate the comparability of findings across years.

It is important to note that previous research has found substantial consistency in the lower socioeconomic and health status reported by people with and without disabilities across the CPS ASEC and other U.S. nationally representative surveys (e.g., the American Community Survey, the National Health Interview Survey, and the Survey of Income and Program Participation).¹¹ In this article, we focus on the potential precarious nature of selected jobs. The CPS ASEC includes data organized by North American Industry Classification System (NAICS) or other industry codes, as well as by occupation codes. To capture the characteristics of the workforce experiencing greater employment insecurity, we break down the broad category of nonstandard workers into two distinct subcategories: temporary agency workers and a proxy category for other nonstandard workers. Temporary agency workers are defined as individuals who reported being employed in jobs covered by the NAICS code for temporary help services. The proxy category for other nonstandard workers corresponds to a variable we created by using selected occupation codes for occupations that typically do not require higher education, are not managerial, are not in agriculture, and result in a high percentage of Internal Revenue Service 1099 forms (used to report income paid by an individual or a business other than one’s employer).¹² These occupations are among those with the largest number of self-employed individuals as reported by BLS.¹³ In 2012, our selected occupations—which include maids and housecleaners, grounds maintenance workers, barbers, hairdressers, childcare workers, carpenters, construction laborers, taxi drivers, and delivery drivers¹⁴—employed at least 170,000 individuals in jobs that did not require more than a high school diploma.¹⁵

Results

This section presents descriptive statistics for the 2010–14 period, comparing the demographic and socioeconomic attributes, disability status, and social insurance program participation of individuals in the overall sample and our selected groups of nonstandard workers. In addition, the section presents results from analyses based on propensity-score matching, facilitating group comparisons in terms of both observed and unobserved worker characteristics.

Descriptive statistics

To shed light on our data and the extent of disabilities throughout the workforce, tables 1a and 1b present selected attributes of the entire population surveyed from 2010 to 2014 (a total of 731,092 individuals), showing data on a range of demographic and socioeconomic characteristics, individuals’ disabilities, and the prevalence of our two categories of nonstandard work.¹⁶ Among individuals in the overall sample, 51.6 percent were female and 66.2 percent were White. The average annual income was $34,332, showing substantial variance; its median value was $21,352, and 0.14 percent of the population reported income losses. Only 0.5 percent of individuals in the sample worked at a temporary agency, but 5.1 percent fell in our proxy category for other nonstandard workers. In terms of disability prevalence, 11.4 percent of respondents answered affirmatively to at least one of the six questions on disability, a figure that rose to 15.7 percent with the inclusion of a work limitation (when individuals reported a health problem or disability preventing them from either working or performing the kind and amount of work they could otherwise do).

Table 2 reports the percentage of temporary agency workers and other nonstandard workers (proxy category) in the overall samples for each year in the 2010–14 period. As shown in the table, the prevalence of workers in both categories of nonstandard work was quite stable over the period. This finding suggests that our data capture the beginning of platform work (for example, TaskRabbit and Uber were founded in 2008 and 2010, respectively) and precede the major expansion of the gig economy that occurred after 2014.¹⁷ By 2017, jobs with work arrangements defined as “contingent” and “alternative” by BLS were estimated to make up 3.8 and 10.0 percent, respectively, of all U.S. employment.¹⁸

Given that many people with disabilities are not in the labor force, tables 3a and 3b compare the demographic and socioeconomic characteristics, disability status, and social insurance program participation of individuals in and out of the labor force. For these comparisons, we conduct a statistical t-test to assess whether the proportion and mean values for the two groups are different at the 5-percent level of significance in the weighted samples.¹⁹ (The results of this test are reported in a separate column, and the same test applies to tables presented later in the article.)

As expected, the tables show statistically significant group differences for all observed characteristics for the combined years 2010–14. People not in the labor force were more likely to be female, Black, native born, less educated, older, and reliant on Medicare and Medicaid. Most importantly, they were much more likely to have a disability compared with people in the labor force. For instance, 24.3 percent of those not in the labor force responded affirmatively to at least one of the six questions on disability, compared with 3.8 percent of those in the labor force who did. People not in the labor force also were much more likely to be under the poverty line (20.7 percent, compared with 8.4 percent of those in the labor force) and, hence, much more likely to receive Medicaid benefits (19.2 percent, compared with 6.4 percent of those in the labor force). Among workers who received workers’ compensation benefits, those who were not in the labor force received benefits that were, on average, 50 percent greater than the benefits received by individuals still in the labor force.²⁰ This finding may reflect the effect of more severe occupational injuries among individuals not in the labor force, with those injuries resulting in permanent disability and prompting workers to leave employment. The greater social insurance program participation (in Medicare, Medicaid, and workers’ compensation) of people not in the labor force is also reflected in their lower average out-of-pocket medical expenses.

Because our focus is on selected nonstandard workers, tables 4a and 4b limit our calculations to individuals in the labor force, comparing those who work at a temporary agency with those who do not. This comparison shows that almost all proportion and mean values for these two groups are statistically different at the 5-percent level of significance (the only exceptions are the variables for hearing, vision, physical disability, activities of daily living, instrumental activities of daily living, and participation in workers’ compensation). Compared with the rest of the labor force, temporary agency workers in 2010–14 were more likely to be women, minorities, less educated, immigrants, and single. They also showed higher frequency of affirmative response to the complete six-question sequence on disability, work limitations, and the combination of the six-question sequence and work limitations. The difference for difficulties in concentrating, remembering, or making decisions is particularly striking, with 2.1 percent of temporary agency workers reporting such difficulties, compared with 1.0 percent of all other workers. Respondents at temporary agencies also were twice as likely to participate in Medicaid and to be in poverty.

Tables 5a and 5b compare the subset of workers in our proxy category for other nonstandard workers with all other workers in the labor force. As shown in the tables, most of the proportion and mean values for these two groups are statistically different at the 5-percent level of significance, except for age, the value of workers’ compensation benefits, participation in Medicare, and the disability variables for hearing, vision, physical disability, activities of daily living, and instrumental activities of daily living. Although the proxy category for other nonstandard workers captures relatively fewer women, it does include more minorities, immigrants, and less educated workers. Like temporary agency workers, workers in the proxy category faced more difficulties concentrating, remembering, or making decisions and were more likely to experience work limitations, to be on Medicaid, to have lower income, to be under the poverty line (twice as likely), and to receive workers’ compensation benefits.

Propensity-score matching

The descriptive analysis suggests that jobs in our selected categories of nonstandard work are not only characterized by lower earnings, but also employ more individuals who are disabled. However, comparisons across groups are complicated by the possibility that workers in nonstandard jobs are not randomly selected and that these jobs potentially attract more vulnerable individuals. To overcome this limitation, this section presents results obtained by implementing propensity-score matching that makes our selected nonstandard work groups and traditional work samples more similar. Propensity-score matching is a method used to create “similar” groups that can be compared more accurately in terms of both observed and unobserved characteristics.²¹ Propensity scores are predicted probabilities, estimated from a probit or a logit model, that show the probability of being in a treatment group (in our case, our two groups of selected nonstandard workers). Propensity-score matching takes selected characteristics of the treated (selected) group of nonstandard workers and “draws a sample from the rest of the population so that the proportions of certain demographic characteristics of the control group align with the [selected nonstandard work] group.”²²

In this analysis, we again focus only on workers in the labor force. The propensity of a respondent to be in one of our subsets of nonstandard workers is estimated by using age, sex, race, ethnicity, highest schooling attained, marital status, and immigrant status as independent regressors. With the help of the Stata PSMATCH2 command, we match nonstandard-worker respondents to control-group respondents by using their respective propensity scores and nearest-neighbor matching with replacement and common support.²³ We use the Stata PSTEST command to conduct t-tests for differences in means and to assess the balancing between the two samples after matching.

Table 6 compares propensity-score-matched samples of temporary agency workers and the rest of the working population for specific outcomes. The two samples are well matched on the independent variables (see bottom panel). The results reported in the table confirm our previous findings in terms of outcomes: workers at temporary agencies were significantly less likely to be employed at the time of the survey; were more likely to have difficulties concentrating, remembering, or making decisions, as well as a disability and/or work limitation; earned lower income; and were twice as likely to be under the poverty line and on Medicaid.

Table 7 compares propensity-score-matched samples of workers in our proxy category for other nonstandard workers and the rest of the working population. Again, the two samples are well matched on the independent variables. Like temporary agency workers, matched-sample respondents in our proxy category were significantly less likely to be employed at the time of the survey, earned substantially lower income, and were more likely to be in poverty and on Medicaid. In terms of disability, they were once again having more difficulties concentrating, remembering, or making decisions, and they received more in workers’ compensation benefits.

Discussion

Nonstandard work arrangements imply a variety of outcomes for different stakeholders. Firms using these arrangements can gain in terms of increased flexibility, potentially reducing their hiring and transaction costs, as well as their cost of contributing to private or social insurance programs (e.g., health insurance, disability, unemployment, pension, and workers’ compensation benefits). For consumers, nonstandard work arrangements may bring gains in terms of lower prices and greater access to markets. Workers’ outcomes, however, are more difficult to assess because of the wide variety of nonstandard contractual arrangements. On the one hand, some workers may prefer these new arrangements if they permit them to control and manage their work, accommodate their individual needs, or foster their own entrepreneurship. This is likely to be the case among freelancers, independent contractors, and some gig workers.²⁴ However, such advantages also carry costs: lower earnings (nonstandard workers are less likely to reap the benefits of any increases in the profits of their employers and are much less likely to be protected by collective bargaining agreements); limited guarantee of full-time or stable employment; lack or potential loss of pension, health, unemployment, and workers’ compensation benefits; and greater odds of incurring negative consequences from limited job-specific training and potential violations of labor standards. In addition, previous studies have shown that workers in nontraditional jobs are at higher risk of injuries and adverse health outcomes.²⁵

In this article, we have explored a large, nationally representative U.S. dataset to shed more light on certain aspects of precarious work. Given the cross-sectional nature of our data, the present analysis does not aim to identify potential causal mechanisms between nonstandard work and workers’ disability status, but rather to assess the extent to which disabled individuals are employed in selected nonstandard contracts. The existing literature has already established that while disability status does not necessarily predict landing “quality jobs” (i.e., jobs offering higher pay and better fringe benefits),²⁶ disability is related to shorter tenure and predicts labor force participation.²⁷

Our data confirm that U.S. individuals who are not in the labor force are more likely to be disabled. For individuals in the labor force, our descriptive analysis suggests that workers in our selected subsets of nonstandard jobs have lower socioeconomic status and less stable employment. We also detect that, compared with other full-time employees, these nonstandard workers report statistically significantly higher frequencies of any disability and of disabilities that limit work. One of our most interesting findings is the higher prevalence of difficulties concentrating, remembering, or making decisions among our selected nonstandard workers. Again, determining whether this higher prevalence occurs because job insecurity affects mental health, a hypothesis suggested by Torkel Rönnblad et al.,²⁸ or because individuals with poor mental health self-select into temporary jobs, a hypothesis suggested by Chris Dawson et al.,²⁹ is beyond the scope of this article. However, given that labor markets increasingly rely on nonstandard work arrangements, our finding points to the need to increase awareness of mental health disabilities on worksites.

Our analysis based on propensity-score matching, which improves the comparability between our two subsets of nonstandard workers (the treatment groups) and other workers in the labor force (the control group), largely corroborates our descriptive findings. In this analysis, we still find significant group differences, with our treatment groups exhibiting higher prevalence of work-limiting disabilities and difficulties concentrating, remembering, or making decisions.

Besides highlighting the frequency of disabilities among U.S. workers in selected nonstandard jobs, our results show differences in the use of some social insurance programs between workers in those jobs and the rest of the labor force. Consistent with other studies,³⁰ we find that our nonstandard workers are much more likely to live in poverty. In addition, these workers are more frequently covered by Medicaid, a finding possibly explaining why they report smaller amounts of out-of-pocket medical expenses (despite the fact that temporary workers are less likely to have health insurance coverage³¹). This finding suggests that changes to Medicaid eligibility could affect both the economic and health outcomes of the precarious workforce.³²

Our simple descriptive analysis also indicates that recipients of workers’ compensation benefits more often report being not in the labor force. Among individuals in the labor force, those captured by our proxy category for other nonstandard workers are much more likely to have received workers’ compensation benefits. Although this difference in frequency does not hold in our analysis based on propensity-score matching, we do find that workers in our proxy category received larger compensation benefits. Again, the present analysis does not permit establishing whether these workers were injured before or during their nonstandard employment. However, our finding indicates that our proxy category for other nonstandard workers more often captures injured employees and/or employees with more severe occupational injuries. This finding, together with the evidence about the prevalence of work-limiting disabilities in our proxy category for nonstandard jobs, highlights the importance of retraining and accommodation in those jobs. It also brings urgency to the need to clarify and inform both employers and employees about who is in charge of protecting nonstandard employees against work incidents through workers’ compensation insurance.³³

Conclusion

Our descriptive analysis has shown that our selected nonstandard jobs are indeed employing more vulnerable workers: minorities, immigrants, less educated and poorer people, and disabled or injured individuals. Future research should investigate the safety nets, working conditions, and physical and mental health accommodations needed by this ever-growing part of the workforce. Given the continuous growth of nonstandard jobs, our results also hint at the risk of rising income inequality among workers and of potential cost shifting from employers to social insurance programs. Future studies should explore new data sources based on more specific definitions of nonstandard jobs and assess the magnitude of the aforementioned issues. These studies should also exploit longitudinal data to cast light on the causal mechanisms linking worker disabilities, socioeconomic status, and job quality.