6

Workforce and Workplace Trends in Disability and Disability Policy

OVERVIEW

This session of the workshop was guided by a set of questions about disability in the workplace, including these:

• How have functional needs required for work changed over time? How have these changes affected the occupational distribution of workers with disabilities and disability-related unemployment?
• What are the key facilitators to inclusion of individuals with disabilities in the workforce, including workplace accommodations and employment/benefit policies?
• To what extent have COVID-19-era and post-COVID-19 era changes to the nature of the workplace affected disability prevalence overall and by occupation?
• What policies and practices—both positive and exploitive (e.g., subminimum wages)—affect employment of individuals with intellectual or other severe disabilities?
• What have been the effects of programs and disability policies on workers’ participation and well-being (e.g., the effectiveness of the ADA in ensuring employer accommodations and assessing effectiveness)? How do the outcomes of U.S. policies compare with those enacted in other countries?

With workshop chair Emily Agree moderating, Nicole Maestas (Harvard Medical School and committee member), presented findings about the implications of Social Security Disability reform for employment. Ari Ne’eman (Harvard University) discussed the impact of the pandemic on disability employment trends, including compositional changes versus shifts in employment prospects. Douglas Kruse (Rutgers University) discussed how employer disability practices affect disability employment outcomes. Finally, Luigi Pistaferri (Stanford University) presented new evidence on the employment effects of the Americans with Disabilities Act.

SOCIAL SECURITY DISABILITY REFORM AND IMPLICATIONS FOR EMPLOYMENT

By providing benefits to workers who are unable to work due to a medical condition, the Social Security Disability Insurance Program (SSDI) is a key source of support for people with disabilities. Program eligibility is broadly based on a medical definition of disability, although information about functioning may be considered as well if information is present about conditions in a person’s medical records.

In presenting her research on disability policy impacts, Nicole Maestas (Harvard Medical School) pointed out that, for years, SSDI has been known for so-called runaway growth. However, as shown in Figure 6-1, the reality is that the program’s caseload has actually declined over the past decade. This trend, she finds, is partly due to substantial changes in allowance rates—that is, the percentage of applicants allowed—set at the appellate

level. The total number of beneficiaries began to decline in 2010 as the number of program exits started exceeding the number of new awards.

For new awards, Maestas and colleagues found little change in the allowance rate at initial application over the period 1992 through 2022, staying at about 32 percent. What has changed over this period is the allowance rate at the appellate review, which dropped from about 74 percent in 2005 to 50 percent in 2015. If someone’s earnings capacity is above the substantial gainful activity threshold (SGA)—an established level of work activity and earnings—SSDI program guidelines indicate that they should be denied, while it if is below SGA they should be approved. During appellate review for the year 2008, an estimated 16 percent had very high earnings and were always denied (in statistical terms based on the instrumental variables model), and about 33 percent had very low earnings and were always accepted, but there was a range in the middle with about 52 percent of applicants in which judges disagreed. By the end of 2014, after a period of little-noticed administrative policy reforms, the allowance rates declined sharply—the model used by Maestas indicates

denials increased from 16 percent to 30 percent—and the eligibility margin narrowed. There was a stable population in terms of their characteristics, and yet a dramatic change appeared in terms of acceptance rates.

Maestas and colleagues explored the reasons for this change in acceptance rates over the past decade. One factor was the occurrence of a federal employee hiring freeze, which led to mounting backlogs, one result of which was that the Social Security Administration (SSA) began a new series of training tools to try and make decisions more consistent across judges. Another possible factor was a high-profile fraud case that involved a disability lawyer having an arrangement with a disability judge, resulting in the judge having a 100 percent allowance rate for cases involving that lawyer. Maestas’ research found that 30 percent of applicants had their decision affected by SSA policy reforms. Implementation of policy-compliance training contributed to the significant decline in the appellate allowance rate.

Maestas concluded that the end result of changes over the past decade is that the current eligibility margin is at an earnings capacity far below the SGA target, meaning that there are applicants who are being denied who ostensibly have earnings capacity well below the SGA target. Even though the administrative policy reforms did manage to target those applicants with relatively greater earnings capacity, the eligibility threshold was already below the SGA target and the reforms moved it further away.

During open discussion, Montez noted that the number of SSDI beneficiaries peaked in 2010 and then declined, and the trends in U.S. life expectancy also peaked in 2010 followed by a decline. She asked Maestas if the trend among beneficiaries could be affecting the trend in life expectancy. Maestas responded that it is possible and that other research has shown that SSDI has protective mortality effects. However, she added that this would need to be studied further to make a conclusive statement about this relationship.

IMPACT OF THE PANDEMIC ON DISABILITY EMPLOYMENT TRENDS

Ari Ne’eman (Harvard University) continued the discussion of employment trends among people with disabilities; much of the content of his presentation was based on joint research with Maestas. People with disabilities, like people without disabilities, saw a sudden decline in the employment-to-population ratio when COVID-19 began. However, Ne’eman explained that as the public health emergency unfolded, people with disabilities showed much faster employment growth than the non-disabled population. This increase was specifically concentrated in occupations that are most amenable to telework (Liu & Quinby, 2024; Ne’eman & Maestas, 2023).

However, Ne’eman noted that, at the exact same time that the growth in employment was taking place, there was a dramatic increase in reported disability rates, especially among those categorized as having cognitive disability. A key question that has arisen is whether cognitive effects from long COVID-19 may have contributed to these observed trends. In reviewing the research evidence to date, Ne’eman cited the work of Sheiner and Salwati (2022) who, using Current Population Survey (CPS) data, found that the increase in measured employment and labor force participation was primarily due to an increase in reported disability rates. In other words, disability employment improvements were primarily the result of compositional change as employment growth was largest among the cognitively disabled (who also saw the largest disability increase). Guo and Krolikowski (2024) used the longitudinal component of the CPS to show that about half of the increase in disabled labor force participation between 2021 and 2022 came from people already in the labor force switching into self-reported disability status rather than disabled people switching into employment in a way that actually increased the labor supply.

For their own work, Ne’eman and Maestas used decomposition analysis to calculate how much of the difference in disabled employment occurring from 2019 to 2023 was the result of changes in the relationship between demographic characteristics in the disabled population and employment, and how much of it was due to changes in the observed characteristics of the disabled population. They performed the analysis both for the disabled population as a whole and specific to people with and people without cognitive disabilities. They also subdivided the disabled population into those with ongoing disabilities and those with new disabilities. The ongoing disability rate pre- and post-COVID-19 was found to remain fairly constant, while the new disability rate changed dramatically over the course of the public health emergency. The average employment rate in the newly disabled population over the past decade or so is about twice as high as in the ongoing disabled population. The researchers found that, of the 7.8 percentage point increase in disability employment between 2019 and 2023, about 42 percent of that increase can be explained by observed compositional changes. Those compositional changes started before COVID-19; their two most influential variables, educational attainment and physical disability, had been shifting dramatically well before COVID-19 began.

During open discussion, Karraker asked whether the associations described by Ne’eman changed over time. Specifically, she wondered whether the COVID-19 era produced an increase in the association between employment and higher education if people who are more educated were more likely to have access to jobs with telework. Ne’eman responded that changes in the estimates over time is a common issue in the context of

decomposition analysis; they examined the issue, but it did not substantially change the results, and the compositional effects were relatively constant.

HOW DO EMPLOYER DISABILITY PRACTICES AFFECT DISABILITY EMPLOYMENT OUTCOMES?

Douglas Kruse (Rutgers University) spoke about recent employer practices and their possible impacts on outcomes for workers with disabilities. Summarizing previous research, he reported results of field experiments showing that people with disabilities often face obstacles to getting hired. In particular, employers are less likely to respond to applications from people with disabilities, even when their disability would not affect productivity for the job in question. Once employed, people with disabilities experience lower pay levels and higher layoff rates than otherwise similar people without disabilities. Kruse stated that an important question is where employer practices make a difference in terms of these outcomes.

Almost all existing evidence that Kruse and other researchers have been able to draw from comes from cross-sectional surveys, raising causality concerns. However, using a random sample of human resource managers interviewed at two points in time, one longitudinal study of managers’ attitudes, intentions, and the hiring of people with disabilities (Araten-Bergman, 2016) found that a written employer policy was better than subjectively assessed management intentions in predicting disability hiring six months later. Other studies noted by Kruse found that disability training was associated with decreases in work injuries and costs and with improvements in disability knowledge and attitudes.

The central part of Kruse’s presentation reported findings based on data from detailed annual employer surveys of members of the National Organization on Disability (NOD), an organization of more than 200 (mostly large) companies interested in improving opportunities for employees with disabilities. NOD collected data on 49 disability employment practices over the period 2019 to 2023; 235 companies filled out the survey in at least two of these years, and 91 did so in all five years. The total employment of respondents was shown to be close to 8.5 million in each year, which represents greater than 5 percent of U.S. nonfarm payroll employment. There was a potential for selection bias in the data, since the companies all chose to participate in the NOD survey; however, the selection issues were reduced through use of a pre-post comparison design that measures change. Also, Kruse stated that the companies joined NOD because they were interested in experimenting and learning about best practices.

The 49 employment practices covered in the survey, not all of which were measured each year, were organized in the NOD survey into eight categories:

• Recruitment: e.g., “My company works directly with disability-focused personnel at the colleges/universities.”
• Disability inclusion: e.g., “My company has senior leaders who are open about their disabilities.”
• Self-ID encouragement: e.g., “My company has run a communications campaign to encourage employees to self-identify as having a disability.”
• Accessibility: e.g., “My company has included accessibility features beyond ADA requirements in our facilities.”
• Promotion of disability initiatives: e.g., “In the past year, my senior leaders have promoted disability initiatives through company town hall meetings about disability.”
• Disability training and education: e.g., “Managers/supervisors are required to take part in annual disability training.”
• Accommodations: e.g., “My company explains the accommodation process to new hires.”
• Support for disability groups and mentoring: e.g., “My company has a formal mentoring program specifically for employees with disabilities.”

The analysis by Kruse and colleagues focused on three outcomes: disability hires, disability promotions, and disability exits. It used a random-effects model at the firm level, measuring both within and between firm variation. They used as predictors both practices in the current year and practices in the previous year while controlling for company size. They used a Mundlak test to estimate whether the random effects were correlated with the variables of interest, and a correlated random-effects model to remove the correlation when found. The following are some of the estimates:

• The lagged value of disability recruitment practices predicts 14.9 percent higher disability hires in the current year.
• The lagged value of disability recruitment practices also predicts a 15.3 percent higher number of disability promotions.
• There were highly positive effects for disability inclusion practices, but they were too large to be plausible as causal effects.
• Promotion of disability initiatives is linked to greater hiring of people with disabilities in the following year.
• Disability training and education of managers and coworkers is linked to more disability hiring and disability promotions.
• Accommodation practices are linked to more disability hires and to lower disability exits.
• The support for disability groups and mentoring seems to be linked to lower disability exits as well.

Kruse mentioned that it is difficult to isolate which specific practices performed best since they are so highly correlated. Within the category of disability recruitment practices, the one practice with the highest unique variance is having a careers page for job seekers with disabilities. The highest specific disability inclusion practice with unique variance is employees with disabilities speaking at company events. It seems to be very important that, when accommodations are made, they do not fall on a department budget, in which case a department manager may be very reluctant to make an expensive accommodation; rather, it appears best for the accommodations to be handled at the company level.

In the discussion that followed, Agree asked Kruse about the extent to which employer policies differentiate between the hiring and retention of workers, and especially the onset of new disabilities among workers who previously might not have qualified for any accommodations. Kruse responded that the NOD survey started collecting data on the number of workers with accommodations only recently. The NOD employers actually have an interest in encouraging their employees to reveal their disabilities; two-thirds of the NOD companies are federal contractors, and they are given the aspirational goal that at least 7 percent of their employees have disabilities.

EMPLOYMENT EFFECTS OF THE AMERICANS WITH DISABILITIES ACT: A FRESH LOOK

The ADA was enacted in July 1990 with the goal of “removing barriers to employment, transportation, public accommodations, public services, and telecommunications” that “have undermined efforts by individuals with disabilities to receive an education, become employed, and be contributing members of society.”¹ Luigi Pistaferri (Stanford University) closed out the session by presenting results from a paper examining the consequences of the ADA for the outcomes of workers with disabilities. The definition of disability adopted for the ADA is considerably broader than the one used by the SSA for determining who is eligible for SSDI benefits. The ADA defines disabilities as “physical or mental impairments that limit substantially one or more major life activities, including (but not limited to) working.”

To facilitate employment of people with disabilities, the ADA introduced three main components: a mandate requiring employers to provide reasonable accommodations for employees and applicants with disabilities; a non-discrimination clause specifying that people with disabilities be treated with an “equal pay for equal positions” criterion; and various tax

___________________

¹ https://www.govinfo.gov/content/pkg/BILLS-111hres1504ih/html/BILLS-111hres1504ih.htm

credits and tax deductions to at least partly cover the cost of accommodating workers with disabilities.

Pistaferri cited papers from the early 2000s—such as those written by Acemoglu and Angrist (2001) and DeLeire (2000)—which found that implementation of the ADA had negative employment consequences for people with disabilities. The explanation for this finding is essentially that the cost of hiring workers with disabilities was higher because the mandate for accommodation was only partially subsidized by tax subsidies. The ADA may also have created higher firing costs because of the risk of litigation due to discrimination. However, Pistaferri noted that these studies may have used an overly narrow definition of disability, whether people have a physical, mental, or other health condition limiting the amount of work they could do (a standard measure in many surveys). However, recall that the ADA has a broader definition, which includes disabilities that might not necessarily prevent work. This means that individuals falling into the broader category may have been treated as part of the controlled population in the previous studies despite being covered by the ADA. Figure 6-2 shows how the proportion of different types of people with disability changed over time.

Pistaferri’s own research confirms the previous research when disability is narrowly defined, finding a negative employment effect, with the impact possibly being even stronger in the long run than in the short run. However, individuals with non-work disabilities (i.e., individuals who have disabilities that impact different spheres of daily living but not necessarily their ability to work) experienced positive employment changes with modestly negative effects on wages. As an example, mobility impairment would prevent engagement in some types of work, such as being a construction worker, but not others, such as being an operator in a call center. There still may be impediments to work, such as barriers in using transportation to commute. And discrimination may still occur if the impairment is highly visible. Pistaferri’s conclusion is that the ADA may both reduce the fixed cost of work and at the same time make it unlawful to discriminate against people with disabilities at the point of hiring and wages; both may push people into work, especially people with what are called non-work disabilities.

Pistaferri found that people with non-work disabilities are more likely than those with work disabilities to be employed (60% vs. 46%) and less likely to be in blue-collar occupations. They are more likely to be college-educated and to be female. A 12-percentage-point decline in employment was found for people categorized as having a work disability, and a 12-percentage-point increase in employment was found for people with non-work disabilities. Combining both groups together results in a 6.9 percentage-point decline in employment.

In the discussion that followed, Agree asked Pistaferri how the discrimination effect is different from the accommodation effect, and whether his research had considered how the 2008 ADA Amendments Act (which broadened the definition of disability) might affect the trends. Pistaferri agreed that discrimination and accommodation are very different across cases, giving the example of employing a lawyer who, on the one hand, might require accommodations for court appearances or, on the other, may be hired to perform research and only need access to broadband internet. Concerning the 2008 amendment, Pistaferri has separate research showing a positive employment effect after 2010, though it is not clear that there is a causal relationship. The analysis discussed in the workshop was intentionally limited to the time period before the 2008 amendment.

Ne’eman asked Pistaferri whether he had accounted for a pre-existing non-discrimination mandate for any entity that received federal funds under Section 504 of the Rehabilitation Act, which would have encompassed pretty much the entire health care sector, the great majority of the education sector, and a significant component of industry as well. Pistaferri responded that it is possible to account for industry effects in the wage regressions, but not for employment because then the industry is not observed. For future work, he suggested that there might be ways to creatively impute industry to non-workers based on occupational history.

This page intentionally left blank.