The ACA Medicaid Expansions and Employment of Adults With Disabilities

Abstract

The Patient Protection and Affordable Care Act (ACA) of 2010 substantially expanded the availability of health insurance coverage, particularly for adults with disabilities. One notable change was the option for states to offer Medicaid coverage to adults with household incomes that were below 138% of the federal poverty line; most but not all states expanded Medicaid to this population. This article investigates whether states that expanded Medicaid coverage through the ACA in 2014—the first year that expansion was possible under the ACA, and the year that most states opted to expand—experienced differential changes in the employment rate of adults with disabilities relative to states that did not expand Medicaid. Using nationally representative data from the American Community Survey, we do not find evidence that the postexpansion employment trend in Medicaid expansion states was significantly different from that trend in states that did not expand Medicaid.

Keywords

employment Medicaid health insurance Affordable Care Act

Before passage of the Patient Protection and Affordable Care Act (ACA) in 2010, individuals with disabling conditions interested in work faced limited options for obtaining health insurance. Offers of employer-sponsored insurance were eroding, and many jobs—jobs that often were available or attractive to workers with disabilities—did not offer health insurance coverage, because they were part-time or hourly positions, in small firms, or in industries or occupations that less often offer health insurance to employees (Claxton et al., 2017). As a result, some workers with disabilities may have been “locked” in their jobs, continuing to work to maintain their employer-provided health insurance coverage. Other adults with disabilities did not work at all or restrained their earnings to maintain eligibility for federal disability benefits, such as Supplemental Security Income (SSI), which often provides Medicaid coverage, and Social Security Disability Insurance (SSDI), which offers Medicare coverage after a 24-month waiting period. Anecdotally, many people with disabilities sought federal disability benefits as a means to obtaining insurance coverage, so-called health insurance-motivated disability enrollment (Kennedy & Blodgett, 2012). Outside of those avenues, coverage on the nongroup market was often unavailable or prohibitively costly (Pizer et al., 2009; Sommers, 2006).

Although not targeted to individuals with disabilities specifically, the ACA significantly changed the health insurance landscape for that group in critical ways (Sevak et al., 2017). Uninsurance rates for adults with disabilities have declined significantly, with most of the change occurring in 2014 and later, reflecting the rollout of key ACA provisions (Kennedy et al., 2017). In 2010, the ACA banned preexisting condition exclusions for children and lifetime benefit caps, created a high-risk pool for individuals seeking coverage but were otherwise unable to obtain it, and extended dependent coverage until the age of 26 years (Kaiser Family Foundation, 2018). In 2014, the changes were even more sweeping. The ACA banned preexisting condition exclusions for adults (guaranteed issue), the individual mandate required adults to have health insurance or pay a penalty, health insurance marketplaces facilitated obtaining group (community-rated) coverage, and subsidies were available for those with incomes between 100% and 400% of federal poverty line (FPL) to make such coverage more affordable. In addition, Medicaid expansions allowed states to offer coverage to adults with incomes up to 138% of the FPL (see Note 1).

Unlike other ACA provisions, Medicaid expansions offered states the flexibility to decide whether to implement the policy or not. The federal government incentivized states to expand Medicaid by covering the full cost of the newly eligible from 2014 through 2016, and 90% of the costs in and after 2020 (Rudowitz, 2014). Many, but not all states chose to expand, with 26 states and the District of Columbia doing so by the end of 2014 (see Figure 1). Eleven additional states opted to expand since that time through January 2020, though not all of those states have implemented the policy change yet.

Figure 1.

Categorization of state Medicaid expansion status by 2016.

In this article, we test whether the Medicaid expansions affected the employment of adults with disabilities by comparing the experience in Medicaid expansion states to that in nonexpansion states. The effects of this policy change may contribute critically to the current policy debate, as the federal and state governments consider avenues for limiting access to Medicaid coverage or imposing barriers such as work requirements (Katch et al., 2019). Although the ACA changed coverage options nationwide, states that expanded Medicaid offered additional coverage options to adults that were not available in states that did not expand Medicaid. In Medicaid expansion states, low-income individuals obtained a new, low-cost option for health insurance that similar individuals in nonexpansion states could not access, even with all of the other ACA-based reforms around the same time. If the employment of individuals with disabilities is affected by the availability of Medicaid coverage, then we would expect to observe changes in employment around the time of the Medicaid expansions to be larger in expansion states than nonexpansion states. Of course, states’ decision to expand Medicaid was nonrandom, and may have been shaped by political, cultural, and economic factors within the state that also correlate with supports for working with a disability. To account for this possibility, our analysis attempts to align areas in expansion and nonexpansion states that seemed to have similar environments for working with a disability before the Medicaid expansions were implemented.

Predicting the Effect of Medicaid Expansions on Employment

A great deal of literature has considered a range of outcomes across Medicaid expansion and nonexpansion states. The literature that has focused on employment outcomes of all adults in response to ACA Medicaid expansions, including any employment, hours worked, job lock, and retirement timing; that literature has not found a large effect in response to Medicaid expansions (Abraham & Royalty, 2017). Yet, there is reason to believe that the response to Medicaid expansions might be stronger for adults with disabilities because of their relatively high demand for health services, a low rate of employment (discussed below), and complex relationship between employment, health insurance, and other cash assistance. The relatively high poverty rate among people with disabilities (41% had income at or below 138% of the poverty threshold in 2013) means that the Medicaid expansions broadened eligibility for a larger share of this group than the population without disabilities. Because individuals with disabilities are a small share of all adults, studies of all adults may mask important effects for this group.

The direction of the impact of Medicaid expansion on employment is not known. More likely, expanded access to health insurance through Medicaid might encourage some with disabilities to seek employment as many individuals with disabilities want to work (Ali et al., 2011). If this were true, we would expect to see the employment rates of adults with disabilities in states that expanded Medicaid rise relative to the same rate in states than did not expand Medicaid. Less likely, but still possible, employment could decrease if newly expanded Medicaid coverage reduced “job lock” and allowed workers with disabilities to exit the labor force. Job lock has been documented for workers considering exiting jobs with employer-sponsored health insurance coverage, both in usual job transitions and at retirement, and is more common among those who place a higher value on health insurance (Baker, 2015). If this were true, we would expect to see the employment rate of adults with disabilities in Medicaid expansion states fall relative to the same rate in states that did not expand Medicaid.

Of course, adults with disabilities face many barriers to work, of which health insurance is only one. As such, it may be unsurprising that several studies did not find an effect of newly available health insurance coverage (including ACA Medicaid expansions) on employment for this population (Hall et al., 2017; Michalopolous et al., 2011; Thomas et al., 2018). However, one recent study found that the employment of adults with disabilities in Medicaid expansion states increased from 2013 to 2017 (see Note 2), but the magnitude of the effect was not documented (Hall et al., 2018). Our study contributes to this literature by applying methodological tools to large nationally representative samples of data to rigorously estimate the impact of the Medicaid expansions on employment of adults with disabilities.

Method

Our analysis used data from the American Community Survey (ACS) by the U.S. Census Bureau. The ACS is a nationally representative, cross-sectional survey of the civilian, noninstitutionalized population, designed to collect information from the Census long form at more frequent intervals than the decennial census. The ACS offers the largest nationally representative sample of community-dwelling individuals annually, and its geographic subsamples are also representative at the subnational level. We used annual data from the 1-year 1% public use microdata sample (PUMS) ACS from 2010 through 2017; covering 4 years before ACA Medicaid expansion began and 4 years after the expansions went into effect in 2014.

We limited our analysis to 24 states and the District of Columbia that expanded in 2014 (“expansion states”) and 19 states that did not expand by 2016 (“nonexpansion states”); these states are shown in dark gray and white in Figure 1. We excluded states that expanded in 2015 and 2016 (none expanded in 2017) from our analysis both for simplicity and because the decision to expand later may have made their experience different from the earliest adopting states.

The total sample included 1.52 million individual records of working-age (ages 18–64 years) adults with disabilities reported in the ACS, or about 200,000 per year. We identified individuals with disabilities using the ACS six-question disability series that asks whether a person has an ambulatory, cognitive, hearing, independent living, self-care, or visual impairment (see Note 2). In the years of our study period, individuals with disabilities represented about 12%−13% of the total ACS sample (Kraus, 2018).

In addition to individual data from the ACS PUMS, we derived characteristics reflecting the conditions in the geographic areas where individuals resided. The ACS identifies where respondents live at the level of the public use microdata area (PUMA), which is a within-state geographic area with at least 100,000 residents. In some cases, PUMAs may align to more familiar geographic units such as counties, while in others, they may span multiple counties or not directly align with county borders to achieve the population-based definition. Based on the PUMA definition in place during our study period, our sample had 2,350 PUMAs (see Note 3): 1,439 in Medicaid expansion states and 911 in nonexpansion states. By calculating weighted statistics for all individuals in the survey in a given PUMA, we were able to construct PUMA-level characteristics.

Assessing the Impact of Medicaid Expansion Using Difference-in-Differences (DD)

We assessed the impact of ACA Medicaid expansions using a standard DD regression framework. The DD specification estimates whether the change in employment outcomes for adults with disabilities around the ACA Medicaid expansions was different in expansion states than in nonexpansion states over the same period, controlling for individual and PUMA-level differences. Underlying this model is the hypothesis that the employment trends in expansion and nonexpansion states would have been the same across this period in the absence of the Medicaid expansion. While there were other policy changes that occurred simultaneous to the Medicaid expansions in 2014, the fact that they were implemented nationwide means that the other ACA policy changes do not invalidate our use of this model.

Our regression specification takes the form:

\begin{array}{l} E M P_{i j t} = α + δ E X P_{j} + γ P O S T_{t} + λ E X P_{j} \\ \times P O S T_{t} + β_{1} X_{i j t} + υ_{j} + ε_{i}, \end{array}

EMP_ijt equals one (1) if an ACS participant i residing in PUMA j at time t was employed in the week before the ACS interview and zero (0) if not. EXP_j is an indicator that the PUMA j was in a state that expanded Medicaid in 2014, while POST_t indicates that the observation occurred in a year in which the Medicaid expansion was in effect. Given improvements in the economy in recent years as well as the possibility that other ACA reforms led to increased employment across all states, our expectation is that the coefficient on POST_t will be positive. The coefficient λ on the interaction of EXP_j and POST_t is the DD estimate, and it identifies any differential impact the Medicaid expansion had on employment. Our hypothesis that increases in employment dominated decreases in employment, on average, would mean that this coefficient would be positive.

We also estimated a version of the above-mentioned model where we replaced POST_t with year indicators. Allowing the effect to vary by year could be important if it takes time for individuals to respond to the newly available coverage by changing their employment behavior. As we describe in the “Results” section, the individual year effects also allow us to verify whether our DD framework meets the parallel trends assumption required for unbiased estimates.

In the model, X_ijt contained characteristics of individual i that might correlate with the decision to work, including sex, marital status by sex, race, Hispanic ethnicity, foreign born, educational attainment, veteran’s status, age, and each of six functional limitations used in the ACS disability measure. We did not include individual characteristics such as family income because these are a function of the outcome of interest, employment.

The model also contained PUMA fixed effects, υ_j. This approach is an alternative to the inclusion of observable PUMA-level characteristics such as population density, PUMA size, and local economic conditions in the model directly. In addition to accounting for those factors, fixed effects also account for time-invariant features of the PUMA that might affect employment rates but cannot be measured with available data. For example, if a PUMA (or the state in which a PUMA is based) is generally supportive toward working with a disability, we may capture some of this effect through the level of the employment rate in the PUMA, but the fixed effect would capture additional factors, including employer attitudes or the availability of reliable public transportation, that cannot be easily measured. While the model with PUMA fixed effects was our preferred specification, the substance of our findings was the same when using PUMA covariates.

Are Nonexpansion States a Good Counterfactual for Expansion States?

The validity of the DD model relies on the assumption that employment rates for adults with disabilities in expansion states would have evolved similarly as those in nonexpansion states in the absence of the expansion. Much of the literature considering the effects of the ACA Medicaid expansions has accepted this as a given because the policy was implemented nationally but adopted at the state level. Yet, because states had the option to expand Medicaid, it stands to reason that states that chose to expand may differ from states that did not. Many nonexpansion states are in the South (see Figure 1), an area in which individuals with disabilities often struggle to find work and where receipt of disability benefits is high (Center on Budget and Policy Priorities, 2018). Thus, a comparison of trends in the employment of adults with disabilities in all expansion states compared with all nonexpansion states could yield a biased estimate of the effect of the expansions on employment.

To better align nonexpansion states to expansion states, we refined the comparison group to be more heavily weighted toward individuals residing in geographic areas (PUMAs) with similar characteristics as expansion state PUMAs during the years prior to the policy change. Specifically, we used inverse probability weighting (IPW), in which observations are weighted by the inverse of the probability of being in the treatment group based on a propensity score model. In simpler terms, IPW assigns a higher weight to individuals residing in nonexpansion state PUMAs that are similar to expansion state PUMAs than to individuals residing in nonexpansion PUMAs that are substantially different from expansion PUMAs. This method is intended to produce a comparison sample that is closely aligned to the treatment sample on observed characteristics (Stuart, 2010).

The IPW method estimated a logistic regression of the probability that an individual resided in an expansion state, similar to what would be done in propensity score matching. To align individuals based on the characteristics of where they lived, the logistic model included only characteristics of the PUMA in the years prior to the implementation of the ACA (see Table 1). We selected characteristics to capture the local area labor market conditions as well as the share of the population that might have been affected by the Medicaid expansion had it been available, based on their income eligibility criteria (see Note 4). We estimated this logistic regression using data solely from the years prior to the Medicaid expansions, 2010–2013, as data after 2013 may have partially reflected the effects of Medicaid expansions.

Table 1.

Characteristics of PUMAs in Expansion and Nonexpansion States, Without and With IPW.

Variable included in IPW model	Without IPW				With IPW
Variable included in IPW model	Expansion	Nonexpansion	Standardized difference	p value	Expansion	Nonexpansion	Standardized difference	p value
Percentage with income <138% FPL	0.307	0.325	−0.17	.034	0.315	0.315	−0.001	.776
Population density 2010
Very low	0.273	0.371	−0.211	.001	0.320	0.317	0.005	.988
Low	0.218	0.306	−0.200	.001	0.263	0.261	0.005	.988
Medium	0.201	0.211	−0.025	.001	0.199	0.204	−0.012	.988
High	0.307	0.112	0.495	.001	0.218	0.217	0.001	.988
Unemployment
2010 (%)	0.116	0.107	0.241	.303	0.111	0.111	0.024	.988
2011 (%)	0.111	0.102	0.245	.229	0.107	0.106	0.032	.945
2012 (%)	0.103	0.095	0.215	.253	0.099	0.098	0.025	.977
2013 (%)	0.092	0.087	0.172	.304	0.089	0.089	0.022	.999
Employment rate of PWD, 2010–2013 (%)	0.323	0.320	0.037	.830	0.322	0.324	−0.023	.964

Note. Authors’ calculations using the 2010–2016 ACS. Variables not included in the IPW model were measured as averages over the 2010–2013 period. The standardized difference is calculated as the difference in mean values between both groups divided by the pooled standard deviation across groups. The p values correspond to a test of the difference in means, using a t test for continuous variables and chi-square test for dichotomous variables. PUMA = public use microdata area; IPW = inverse probability weighting; FPL = federal poverty line; ACS = American Community Survey.

Using the coefficients of the IPW, we calculated weights for all individuals in each year of the ACS data (2010–2017) to construct a weighted comparison group of nonexpansion states that was more similar to expansion states before the ACA (see Note 5). Our results were estimated accounting for the weight derived from the IPW as well as the ACS sampling weight to produce nationally representative estimates. Although we describe the IPW and regression estimation separately for ease of exposition, procedurally, the steps were implemented jointly so that the standard errors in the employment regression account for the sampling variation in the IPW. Standard errors in the regression models were clustered at the state level to account for the fact that the expansion varied at the state rather than individual level.

Results

Before presenting estimates of the impact of ACA Medicaid expansions on the employment of working-age adults with disabilities, we discuss our success in using IPW to better align the expansion and nonexpansion PUMAs on observable characteristics.

Assessing Post-IPW Sample Balance

The IPW approach can yield unbiased treatment effect estimates if the resulting treatment and comparison samples do not have systematic differences on baseline characteristics after applying the weights from the model (Austin & Stuart, 2015). Table 1 reports mean characteristics in the expansion and nonexpansion states, without and with IPW. The statistics without IPW correspond to a simple comparison of all expansion and nonexpansion states, as has been done in other studies. We assessed balance using the standardized difference across the two groups—an accepted diagnostic tool in the matching literature, where differences lower than 3%–5% satisfy the strictest definition of “good balance” (Caliendo & Kopeinig, 2005; Institute of Education Sciences, 2014).

The statistics with IPW show that, across all of the measures included in the logistic regression, expansion and nonexpansion PUMAs were well-balanced. Most notably, the distribution of population density in expansion and nonexpansion states without IPW was quite different, and job opportunities for adults with disabilities can be quite different in urban and rural areas (Research and Training Center on Disability in Rural Communities, 2019). The standardized difference with IPW was smaller than without IPW, meaning that the IPW-weighted treatment and comparison samples were more closely aligned, and the comparison group was a better counterfactual for what would have happened in expansion states in the absence of the policy change. This was further confirmed by the fact that statistically significant differences in means observed between expansion and nonexpansion states without IPW were insignificant after IPW.

Overall Effect of Medicaid Expansions on Employment of Adults With Disabilities

Trends in employment rates in IPW-weighted expansion and nonexpansion states from 2010 through 2017 are shown in Figure 2, demonstrating that the state groups had similar employment trends before the ACA. The figure shows parallel trends prior to the policy change in 2014, are required for the DD model to produce an unbiased estimate of the policy change. In the version without IPW, the trends mostly look parallel between nonexpansion and expansion states as well, though the employment rate lines do cross between 2010–2011 and 2011–2012.

Figure 2.

Employment rate of adults with disabilities in expansion and nonexpansion states.

Figure 2 also shows that the unadjusted employment rate in expansion states rose less rapidly than in nonexpansion states from 2014 through 2016, before converging in 2017. This early divergence may have partially reflected a differential response to the postrecession recovery; some of the post-2014 increase among nonexpansion states was driven by a substantial decline in employment between 2013 and 2014 in nonexpansion states, which was not present in expansion states. Our multivariate model allowed us to better account for the broader economy.

Table 2 shows the overall DD regression estimates, comparing expansion states to nonexpansion states, building from Figure 2 and including individual controls and PUMA fixed effects. Columns A and B in Table 2 present results from Equation (1), with a single term for the postexpansion period that equals one in years 2014–2017 and an interaction of the postexpansion term with the expansion term that equals one in states that expanded Medicaid; Column A presents ordinary least squares (OLS) estimates while Column B presents results from our IPW estimation. Consistent with Figure 2, the positive coefficient on the post term in Column A illustrates the increasing employment rates over this time period in both expansion and nonexpansion states. This likely reflects the economic recovery following the recession but may also reflect a broader response to provisions of the ACA that were rolled out nationwide during this time.

Table 2.

Difference-in-Differences Estimates of the Effect of Medicaid Expansion on Employment of Working-Age Adults With Disabilities.

Variable	(A) OLS with PUMA fixed effects	(B) IPW with PUMA fixed effects	(C) IPW with PUMA fixed effects^a
Expansion × Post	0.005 (0.003)	−0.004 (0.012)	—
Expansion	—	—	—
Post	0.012** (0.002)	0.014 (0.009)	—
Expansion × 2010	—	—	(Omitted)
Expansion × 2011	—	—	−0.018 (0.020)
Expansion × 2012	—	—	0.007 (0.021)
Expansion × 2013			0.004 (0.021)
Expansion × 2014	—	—	0.007 (0.019)
Expansion × 2015	—	—	0.003 (0.022)
Expansion × 2016	—	—	−0.014 (0.025)
Expansion × 2017			0.012 (0.018)
2010	—	—	(Omitted)
2011	—	—	0.007 (0.014)
2012	—	—	−0.001 (0.016)
2013	—	—	−0.004 (0.016)
2014	—	—	−0.005 (0.013)
2015	—	—	0.010 (0.017)
2016	—	—	0.017 (0.019)
2017			0.011 (0.011)

Note. Authors’ calculations using the 2010–2017 ACS. All models contain 1.52 million observations. Results include ACS sampling weights to produce nationally representative statistics as well as IPW as described in the text. Standard errors (in parentheses) account for clustering at the state level. OLS = ordinary least squares, PUMA = public use microdata area; IPW = inverse probability weighting; ACS = American Community Survey.

Variant of the model in Equation (1).

**p = .01.

We did not find evidence that the ACA Medicaid expansions led to an increase in employment in the postexpansion period relative to nonexpansion states (see Table 2). The coefficient on Expansion × Post is close to zero, and not significant in either the standard OLS or IPW specification. Even if the estimated effects were statistically significant, they would be small changes of roughly 1%–2% of the baseline employment rates of adults with disabilities around that time. Interestingly the sign of the estimated impact is different in the IPW specification relative to the OLS specification, suggesting that rebalancing the expansion and nonexpansion samples so they are similar is important and can potentially lead to different results.

We also did not find that the effects on employment changed over time. Column C presents a variant of the model in Equation (1), where the POST term was replaced with individual year indicators (with 2010 omitted). This allowed estimated effects to differ by year to capture the fact that responses to the Medicaid expansion could have been stronger immediately or could have taken time to materialize. While the estimated coefficients varied from year to year, sometimes with different signs, the imprecision of the estimates mean that we could not conclusively say that there was or was not a differential effect in any of the years.

The model with annual effects also allowed us to assess whether the parallel trends assumption required for DD estimates to be valid held true, by considering the pattern on the Expansion × Year covariates in the pre-2014 period. Column C shows that these coefficients were not significant, suggesting that we did not violate the parallel trends assumption. A test of joint significance (not shown) confirmed that the combination of the coefficients on the interaction terms for 2011–2013 are not significantly different from zero.

Effect of Medicaid Expansions on Employment of Adults With Disabilities Subgroups

One possible reason for the lack of an effect of Medicaid expansions on employment is that the national estimates masked offsetting differences across states or individual characteristics. For example, it is possible that the expansions may have had a greater impact in some states. In particular some states had relatively generous Medicaid coverage prior to the expansions and as such, the relative size and resulting impact of the expansions may have been smaller in those states. In addition, it is possible that the expansions may be have had more of an impact on employment of young workers or workers with less severe disabilities. To explore these possibilities, we conducted a series of subgroup analyses. In each case, we reestimated the IPW and DD model to be limited to the subgroup of interest, to assure a strong comparison sample for each subgroup.

We first estimated the impact on employment when we excluded four states—California, Hawaii, Vermont, Wisconsin—and the District of Columbia because those states expanded Medicaid in a substantial way prior to 2014, providing coverage to childless adults with incomes below 100% or 133% of the poverty level by 2014. The second bar of Figure 3 presents this estimate relative to the estimate for the full set of expansion states; the dark bars plotting the magnitude of the coefficient estimates and the lines on either side of the bars representing the confidence intervals. While the direction of the estimates was different, both were close to zero and we could not reject the hypothesis that the effect size is the same. We considered a variant of our model where the binary expansion term was replaced by the “size” of the expansion relative to the FPL to get at this state-level heterogeneity in another way, and similarly did not find an effect (results not shown).

Figure 3.

Estimated difference-in-differences estimate for selected subgroups.

We also estimated impacts across the full group of expansion states, but for additional subgroups based on gender, age, educational attainment, number of disabilities, pre-ACA PUMA poverty rate, pre-ACA PUMA uninsurance rate, and whether the state had a Medicaid 1634 waiver we did not find a significant impact on employment (see Figure 3). However, the fluctuation in magnitude and direction of the estimated coefficients is notable, with the largest negative estimate for men and the largest positive estimates for individuals with three or more disabilities.

Discussion

We did not find evidence that Medicaid expansions significantly affected the employment of adults with disabilities at the national level, either overall or for most subgroups of states or individuals. Our estimates were imprecise, so we cannot rule out the possibility that the expansions had an effect on employment, but the magnitude of any such effect would have been quite modest based on our estimates. Our findings are consistent with the broader literature that has considered the impact of the ACA on work and wages (Abraham & Royalty, 2017; Kaestner et al., 2017), and with findings based on interviews with adults with disabilities, who reported having difficulty obtaining accurate and timely information about their benefits and access to care after ACA reforms were implemented (Hall et al., 2019).

It is important to note that the ACA was not designed to affect employment, and thus little or no effect on employment does not in any way invalidate the importance of increased access to health insurance. If anything, the lack of an effect on employment adds to a growing body of evidence that concerns about the ACA leading to reductions in work have not borne out for the population with disabilities (Congressional Budget Office, 2014). This may be especially important as policies to limit the availability of Medicaid coverage are considered; the fact that there were not significant negative effects on employment should provide confidence that at least on this measure, there were no unintended effects of expanded health insurance coverage on weakened connections to the labor force.

It is also important to recognize that the Medicaid expansions may have affected the employment of some individuals with disabilities in ways that we cannot capture with the ACS measure of disability. The ACS six-question series is the official measure used for statistics related to employment for people with disabilities in the United States (Burkhauser et al., 2014) and has been shown to present a generally unbiased picture of the population with disabilities (Altman et al., 2017). But we cannot identify those on the margin of working, for whom the policy might have been most salient, we are only able to estimate the average affect across the full sample with disabilities. Some have raised concerns that the population captured by the ACS disability series is not reflective of the full population with disabilities (Henderson, 2011). The relatively large, albeit not statistically significant, estimated impact for the subgroup with three or more disabilities in our analysis suggests that impacts may differ by the nature of disability, potentially related to one’s eligibility for Medicaid through other channels including SSI. It also suggests that how disability is measured will affect the impact estimates. Our results only apply to individuals captured by the ACS disability questions. Studies based on other data, including those of Hall et al. (2018), may yield different estimates.

It is worth considering why we might not have observed an effect on employment, beyond the simple answer that one simply did not exist or resulted from limitations in the measure of disability. First, it could be the case that the expansions led some individuals to leave jobs while it led others to enter jobs; no aggregate impact may reflect that these populations are similarly sized and therefore offsetting in national data. For example, if health insurance–motivated disability enrollment was important for a subgroup of individuals with disabilities, the Medicaid expansions might reduce program participation and increase employment. On the contrary, job lock might have kept a different subgroup of individuals working at jobs that offered health insurance coverage and who could now reduce their employment with Medicaid coverage. The ACS data are repeated cross sections and was therefore not suited to consider these types of changes, which could be better accounted for using longitudinal data.

Second, it could be the case that responses to the Medicaid expansions will continue to change over time. This is consistent with Hall et al. (2018) and Hall et al. (2017), which documented no employment effect of the Medicaid expansions through 2016 but found a positive effect on employment in Medicaid expansion states in 2017. Discernible effects, as measured with the ACS data, may take even more time to manifest. Future studies should continue to follow the evolution of employment for adults with disabilities and may reach different conclusions. However, the more time that passes from the initial policy implementation, the more difficult it will be to argue that the cause is attributable to Medicaid expansions.

Third, the lack of an effect on employment may reflect uncertainty about the future of Medicaid expansions and a weariness to alter employment when the availability of coverage continues to change. Since the ACA was passed, there has been nearly continual talk about repealing or replacing its policies, with a number of provisions under attack that would negatively affect workers with disabilities (Sevak et al., 2017). In 2017, the annual open enrollment period for obtaining new coverage was shortened, funding for outreach during open enrollment was cut, and penalties related to the individual mandate were no longer enforced (“With the Federal Individual Mandate Gone, . . .,” 2018). Since that time, proposed changes have sought to limit Medicaid access specifically, ranging from a complete repeal of the expansions to changing federal Medicaid funding to states to allowing states to implement work requirements to maintain coverage.

Finally, is also important to note that national estimates may obscure important within-state and cross-state variation, with offsetting effects resulting in a net effect of zero at the national level. The proportional increase in the Medicaid population following the 2014 expansions varied widely across states (Centers for Medicare & Medicaid Services, 2018), as did the reduction in numbers of individuals uninsured (Courtemanche et al., 2017). States have a fair amount of flexibility when designing their Medicaid programs, which may encourage or discourage enrollment (Artiga et al., 2017). For individuals with disabilities, the magnitude of the impact of the ACA might be explained by the availability of Medicaid in the state prior to the ACA (e.g., the income limit relative to 138% of poverty, caps on enrollment, or the linkage between Medicaid and SSI as discussed in Rupp & Riley, 2016), the generosity of Medicaid coverage (e.g., whether fee-for-service or managed care, limits on coverage, copayments, and deductibles), or the extent to which the state promoted ACA-related changes (state-based) versus federally facilitated health insurance marketplace for enrolling in coverage, the success of the selected marketplace, or the outreach efforts conducted in the state. As such, several state-level studies have documented employment changes resulting from Medicaid expansion (Antonisse et al., 2018).

Footnotes

Acknowledgements

The authors wish to thank Joseph Mastrianni, a programmer at Mathematica, for excellent programming and analysis on this project as well as Michael Levere of Mathematica’s Center for Studying Disability Policy for insightful comments on the draft manuscript.

Authors’ Note

The contents do not necessarily represent the policy of DHHS and you should not assume endorsement by the federal government (EDGAR, 75.620(b)).

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this study was provided by the Research and Training Center on Employment Policy and Measurement at the University of New Hampshire, which is funded by the National Institute for Disability, Independent Living, and Rehabilitation Research, in the Administration for Community Living, at the U.S. Department of Health and Human Services (DHHS) under cooperative agreement 9ORT5037-02-00.

Notes

References

Abraham

Royalty

A. B.

(2017). How has the Affordable Care Act affected work and wages? Leonard Davis Institute Issue Brief, 21(3), 1–5. https://ldi.upenn.edu/brief/how-has-affordable-care-act-affected-work-and-wages

Ali

Schur

Blanck

(2011). What types of jobs do people with disabilities want? Journal of Occupational Rehabilitation, 21, 199–210.

Altman

B. M.

Madans

Weeks

J. D.

(2017). An evaluation of the American Community Survey indicators of disability. Disability and Health Journal, 10(4), 485–491.

Antonisse

Garfield

Rudowitz

Artiga

(2018). The effects of Medicaid expansion under the ACA: Updated findings from a literature review. https://www.kff.org/report-section/the-effects-of-medicaid-expansion-under-the-aca-updated-findings-from-a-literature-review-report/

Artiga

Hinton

Rudowitz

Musumeci

(2017). Current flexibility in Medicaid: An overview of federal standards and state options. https://www.kff.org/medicaid/issue-brief/current-flexibility-in-medicaid-an-overview-of-federal-standards-and-state-options/

Austin

P. C.

Stuart

E. A.

(2015). Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Statistics in Medicine, 34, 3661–3679.

Baker

(2015). Job lock and employer provided health insurance: Evidence from the literature [AARP Future of Work at 50+ Research Report]. https://www.aarp.org/ppi/info-2015/job-lock-and-employer-provided-healthcare.html

Burkhauser

R. V.

Houtenville

A. J.

Tennant

J. R.

(2014). Capturing the elusive working-age population with disabilities: Reconciling conflicting social success estimates from the Current Population Survey and American Community Survey. Journal of Disability Policy Studies, 24, 195–205.

Caliendo

Kopeinig

(2005). IZA DP No. 1588: Some practical guidance for the implementation of propensity score matching [Discussion paper]. https://www.iza.org/publications/dp/1588/some-practical-guidance-for-the-implementation-of-propensity-score-matching

10.

Center on Budget and Policy Priorities. (2018). Chart book: Social Security disability insurance. Center of Budget and Policy Priorities. https://www.cbpp.org/sites/default/files/atoms/files/7-21-14socsec-chartbook.pdf

11.

Centers for Medicare & Medicaid Services. (2018, August 31). Medicaid & CHIP monthly applications, eligibility determinations, and enrollment reports: January 2014–February 2018 (preliminary), as of April 30, 2018. https://www.medicaid.gov/medicaid/program-information/medicaid-and-chip-enrollment-data/monthly-reports/index.html

12.

Claxton

Rae

Long

Damico

Foster

Whitmore

(2017). Employer health benefits: 2017 annual survey. https://www.kff.org/health-costs/report/2017-employer-health-benefits-survey/

13.

Congressional Budget Office. (2014). The budget and economic outlook: 2014 to 2024, Appendix C. https://www.cbo.gov/publication/45010

14.

Courtemanche

Marton

Ukert

Yelowitz

Zapata

. (2017). Early impacts of the Affordable Care Act on health insurance coverage in Medicaid expansion and non-expansion states. Journal of Policy Analysis and Management, 36, 178–210.

15.

Hall

J. P.

Kurth

N. K.

Gimm

Smith

(2019). Perspectives of adults with disabilities on access to health care after the ACA: Qualitative findings. Disability and Health Journal, 12, 350–358.

16.

Hall

J. P.

Shartzer

Kurth

N. K.

Thomas

K. C.

(2017). Effect of Medicaid expansion on workforce participation for people with disabilities. American Journal of Public Health, 107, 262–264. https://doi.org/10.2105/AJPH.2016.303543

17.

Hall

J. P.

Shartzer

Kurth

N. K.

Thomas

K. C.

(2018). Medicaid expansion as an employment incentive program for people with disabilities. American Journal of Public Health, 108, 1235–1237. https://doi.org/10.2105/AJPH.2018.304536

18.

Henderson

(2011, August 1). Letter to Secretary Sebilius regarding proposed data collection standards for race, ethnicity, primary language, sex, and disability status required by Section 4302 of the Affordable Care Act. Disability Rights Education and Defense Fund. https://dredf.org/2011/08/01/proposed-aca-data-collection-standards/

19.

Institute of Education Sciences. (2014). What Works Clearing-house: Procedures and standards handbook, Version 3.0. U.S. Department of Education, Institute of Education Sciences. http://ies.ed.gov/ncee/wwc/DocumentSum.aspx?sid=19

20.

Kaestner

Garrett

Chen

Gangopadhyaya

Fleming

(2017). Effects of ACA Medicaid expansions on health insurance coverage and labor supply. Journal of Policy Analysis and Management, 36(3), 608–642. https://doi.org/10.1002/pam.21993

21.

Kaiser Family Foundation. (2018). Current status of health insurance marketplace and Medicaid expansion decisions [PowerPoint slides]. http://kff.org/health-reform/slide/current-status-of-health-insurance-marketplace-and-medicaid-expansion-decisions/

22.

Katch

Straw

Schubel

Broaddus

Gonzales

(2019). Health proposals in president’s budget would reduce health insurance coverage and access to care. Center of Budget and Policy Priorities. https://www.cbpp.org/research/health/health-proposals-in-presidents-2020-budget-would-reduce-health-insurance-coverage

23.

Kennedy

Blodgett

(2012). Health insurance-motivated disability enrollment and ACA. New England Journal of Medicine, 367(12), Article e16.

24.

Kennedy

Wood

E. G.

Frieden

(2017). Disparities in insurance coverage, health services use, and access following implementation of the Affordable Care Act: A comparison of disabled and nondisabled working-age adults. Inquiry: A Journal of Medical Care Organization, Provision and Financing, 54. http://doi.org/10.1177/0046958017734031

25.

Kraus

(2018). 2017 Disability statistics annual report. University of New Hampshire. https://disabilitycompendium.org/sites/default/files/user-uploads/2017_AnnualReport_2017_FINAL.pdf

26.

Long

S. K.

Kenney

G. M.

Zuckerman

Goin

D. E.

Wissoker

Blavin

Blumberg

L. J.

Clemans-Cope

Holahan

Hempstead

(2014). The health reform monitoring survey: Addressing data gaps to provide timely insights into the affordable care act. Health Affairs, 33(1): 161–167.

27.

Michalopolous

Wittenburg

Israel

D. A. R.

Schore

Warren

Zutshi

Freedman

Schwartz

(2011). The Accelerated Benefits demonstration and evaluation project: Impacts on health and employment at twelve months. MDRC. https://www.mdrc.org/sites/default/files/full_528.pdf

28.

Pizer

S. D.

Frakt

A. B.

Iezzoni

L. I.

(2009). Uninsured adults with chronic conditions or disabilities: Gaps in public insurance programs. Health Affairs, 28, 1141–1150.

29.

Research and Training Center on Disability in Rural Communities. (2019, January). Employment disparity grows for rural Americans with disability. The University of Montana, Rural Institute for Inclusive Communities.

30.

Rudowitz

(2014). Understanding how states access the ACA enhanced Medicaid match rates. Kaiser Commission on Medicaid and the Uninsured.

31.

Rupp

Riley

G. F.

(2016). State Medicaid eligibility and enrollment policies and rates of participation among disabled Supplemental Security Income recipients. Social Security Bulletin, 76(3), 17–40. https://www.ssa.gov/policy/docs/ssb/index.html

32.

Sevak

Schimmel Hyde

Kehn

(2017). Repealing or replacing the ACA’s provisions: How would adults with disabilities fare? [Advancing Policy Supports for People with Disabilities (Issue Brief No. 4)]. http://www.roosevelthouse.hunter.cuny.edu/?forum-post=repealing-replacing-acas-provisions-adults-disabilities-fare

33.

Sommers

A. S.

(2006). Access to health insurance, barriers to care, and service use among adults with disabilities. Inquiry: The Journal of Health Care Organization, Provision, and Financing, 43, 393–405. https://doi.org/10.5034/inquiryjrnl_43.4.393

34.

Stuart

E. A.

(2010). Matching methods for causal inference: A review and a look forward. Statistical Science: A Review Journal of the Institute of Mathematical Statistics, 25(1), 1–21. http://doi.org/10.1214/09-STS313

35.

Thomas

K. C.

Shartzer

Kurth

N. K.

Hall

J. P.

(2018). Impact of ACA health reforms for people with mental health conditions. Psychiatric Services, 69(2), 231–234.

36.

With the federal individual mandate gone, states might step up: Lessons from Massachusetts. (2018, January 16). Health Affairs Blog. https://www.healthaffairs.org/do/10.1377/hblog20180108.464274/full/