The Impact of Affordable Care Act Medicaid Expansions on Applications to Federal Disability Programs

Priyanka Anand; Jody Schimmel Hyde; Maggie Colby; Paul O’Leary

doi:10.1515/fhep-2018-0001

Published by De Gruyter February 23, 2019

The Impact of Affordable Care Act Medicaid Expansions on Applications to Federal Disability Programs

Priyanka Anand , Jody Schimmel Hyde , Maggie Colby and Paul O’Leary

From the journal Forum for Health Economics and Policy

https://doi.org/10.1515/fhep-2018-0001

Showing a limited preview of this publication:

Abstract

In this paper, we estimate the impact of Medicaid expansions via the Patient Protection and Affordable Care Act (ACA) on applications to federal disability programs in 14 states that expanded Medicaid in January 2014. We use a difference-in-differences regression model to compare disability application rates in geographic areas within states that expanded Medicaid to rates in areas of non-expansion states that were carefully selected using a matching approach that accounts for state Medicaid policies pre-ACA as well as demographic and socioeconomic characteristics that might influence disability application rates. We find a slower decrease in Supplemental Security Income (SSI) application rates after Medicaid expansions in expansion states relative to non-expansion states, with application rates declining in both state groups from 2014 through 2016. Our analysis of the impact of the Medicaid expansions on Social Security Disability Insurance (SSDI) application rates was inconclusive for reasons we discuss in the paper.

Keywords: Affordable Care Act; disability; medicaid

Funding source: U.S. Social Security Administration

Award Identifier / Grant number: Disability Research Consortium, 1-DRC12000001-01-0

Funding statement: U.S. Social Security Administration, Funder Id: 10.13039/100005225, Grant Number: Disability Research Consortium, 1-DRC12000001-01-0.

Acknowledgements

The authors would like to thank Lauren Hula, Andrew McGuirk, Swaati Bangalore, and Charles Hanley at Mathematica Policy Research for their efforts in the data analysis and Neil McCray at George Mason University for research assistance. We also acknowledge valuable input received from Randall Brown and David Stapleton at Mathematica on the research design and manuscript, as well as comments about the study design received from Jack Gettens, Kosali Simon, and participants at the 2016 Disability Research Consortium Annual Research Meeting. The research reported herein was performed pursuant to a grant from the U.S. Social Security Administration (SSA) funded as part of the Disability Research Consortium. The opinions and conclusions expressed are solely those of the author(s) and do not represent the opinions or policy of SSA or any agency of the Federal Government.

Methods Appendix: Matching Model Details

We estimated a separate propensity score model at the PUMA level for each group of states shown in Table 1. Each propensity score model contained a mutually exclusive set of states; neither treatment nor comparison PUMAs were included in more than one model. We used the same set of covariates in each of the state group models because we did not have a strong reason to believe that the set of predictors should vary by group. The reason that we developed the propensity score model separately by state group is that we believed that the relationship between Medicaid expansions and SSI/SSDI may have varied across those groups, and individual models allowed us the best matched comparison sample for each group of expansion states. The variables included in the propensity score model and estimated coefficients for each of the four models are provided in Appendix Table 3.

After running these models separately, a propensity score was calculated for each PUMA. With those scores in hand, we used nearest-neighbor matching with replacement, allowing for up to four matches per expansion PUMA, with a caliper of 0.1 standard deviations of the overall estimated propensity score. Traditionally, the standard in the matching literature has been to use a caliper of 0.2 standard deviations (Rosenbaum and Rubin 1985; Stuart 2010); we used an even smaller caliper of 0.1 standard deviations to ensure high quality matches. We limited the number of matches to four, given findings that increasing the number of matches beyond five tends to increase the bias in the estimated treatment effect (Austin 2010).

To yield better match quality, we excluded treatment and comparison PUMAs that had an estimated propensity score smaller than 0.1 or larger than 0.9, as has been suggested in the literature (Crump et al. 2009). While excluding cases with propensity scores outside this interval improves match quality of the remaining cases, in our case, it significantly reduced the number of matched PUMAs. Before trimming, 487 of 547 expansion PUMAs had at least one match, and 237 of 861 potential comparison PUMAs were matched to one or more expansion PUMAs; after trimming the number of matched expansion PUMAs fell from 487 to 257. Appendix Table 4 highlights the number of PUMAs that were matched in each state. Because we exclude about half of the PUMAs and populations in each state from our analysis, it is important to recognize that the impact estimates based on the trimmed samples may not be generalizable to all PUMAs. Nonetheless, our inspection of the untrimmed results (not shown) highlighted that our overall estimates were not significantly different from the trimmed version and in the state-level estimates, so our general conclusions would have remained unchanged. These results are available to interested readers upon request.

Table 3:

Coefficients from Propensity Score Logit Model, by State Group.

	State group 1	State group 2	State group 3	State group 4
Variable	*Coefficient [standard error]*
Percentage change in the unemployment rate, 2010–2013	0.147***	−0.021	0.271***	0.001
	[0.02]	[0.04]	[0.06]	[0.04]
Size of working-age population, 2013	−0.000	−0.000	0.000	0.000
	[0.00]	[0.00]	[0.00]	[0.00]
Percentage change in the working-age population, 2010–2013	−0.241***	−0.091*	−0.002	0.028
	[0.05]	[0.05]	[0.08]	[0.07]
Population density	0.001***	0.000**	0.001***	0.001***
	[0.00]	[0.00]	[0.00]	[0.00]
Percentage uninsured	−0.164***	−0.584***	−0.443***	−0.641***
	[0.06]	[0.09]	[0.14]	[0.15]
Percentage privately insured	−0.067**	−0.405***	0.017	−0.227***
	[0.03]	[0.07]	[0.07]	[0.08]
Percentage white	−0.193***	−0.033	−0.009	0.063
	[0.04]	[0.03]	[0.07]	[0.08]
Percentage black	−0.307***	−0.446***	0.001	0.046
	[0.04]	[0.08]	[0.07]	[0.08]
Percentage Hispanic	−0.081***	−0.039**	0.234***	0.267***
	[0.03]	[0.02]	[0.07]	[0.07]
Percentage with income < 137% FPL	−0.056	0.257***	0.093	0.119
	[0.05]	[0.07]	[0.10]	[0.09]
Median gross rent, 2013	−0.001	0.012***	−0.006***	−0.001
	[0.00]	[0.00]	[0.00]	[0.00]
Percentage of the population ages 25–34	−0.041	0.258**	−0.179	−0.243*
	[0.07]	[0.10]	[0.12]	[0.14]
Percentage of the population ages 35–64	−0.028	0.162**	0.087	0.230**
	[0.05]	[0.07]	[0.08]	[0.09]
Percentage of the population over age 64	−0.089*	0.026	0.009	−0.194
	[0.05]	[0.10]	[0.12]	[0.12]
Percentage change in SSDI-only applications, 2010–2011	−0.049**	−0.025	−0.056*	0.011
	[0.02]	[0.02]	[0.03]	[0.03]
Percentage change in SSDI-only applications, 2011–2012	−0.094***	−0.031	−0.015	0.029
	[0.02]	[0.03]	[0.03]	[0.03]
Percentage change in SSDI-only applications, 2012–2013	−0.034	−0.035	−0.035	0.079***
	[0.02]	[0.03]	[0.03]	[0.03]
Percentage change in SSI applications, 2010–2011	−0.031	0.094***	0.006	0.044
	[0.02]	[0.03]	[0.03]	[0.03]
Percentage change in SSI applications, 2011–2012	−0.084***	0.028	0.016	−0.034
	[0.02]	[0.03]	[0.03]	[0.03]
Percentage change in SSI applications, 2012–2013	−0.064***	0.006	−0.017	−0.046
	[0.02]	[0.03]	[0.03]	[0.03]
R²	0.568	0.703	0.614	0.426
Number of observations	553	422	233	200

Source: Authors’ calculations, using data derived from SSA’s SDR, the ACS, and BLS unemployment statistics.
*Denotes p-value < 0.10, **denotes p-value < 0.05, and ***denotes p-value < 0.01. Values are derived from the three-year ACS estimates from 2010–2012 unless otherwise noted. Demographic and socioeconomic characteristics were calculated among working-age adults.

Table 4:

Number of PUMAs in State-level Regressions and 2013 application counts.

	Unmatched PUMAs		Matched PUMAs
	Expansion	Non-Expansion	Expansion	Non-Expansion	SSI unmatched regression sample size	SSI matched regression sample size	SSDI unmatched regression sample size	SSDI matched regression sample size
Overall	547	861	257	187	39,416	12,430	39,424	12,432
Arkansas	20	350	16	48	24,664	1,792	24,668	1,792
Kentucky	34	350	27	41	25,056	1,904	25,060	1,904
New Jersey	73	350	27	32	26,146	1,652	26,152	1,652
Rhode Island	7	350	3	10	24,300	364	24,304	364
Washington	56	350	26	51	25,672	2,156	25,676	2,156
West Virginia	13	350	11	30	24,468	1,148	24,472	1,148
Arizona	54	308	16	21	25,616	1,036	25,620	1,036
Colorado	42	308	22	32	25,278	1,511	25,284	1,512
New Mexico	18	308	9	17	24,608	728	24,612	728
Connecticut	26	114	8	13	24,832	588	24,836	588
Illinois	88	114	24	21	26,568	1,260	26,572	1,260
North Dakota	5	114	1	4	24,244	139	24,248	140
Nevada	18	89	10	18	24,608	784	24,612	784
Ohio	93	89	57	41	26,708	2,744	26,712	2,744

Source: Data derived from authors’ calculations using SSA’s SDR, the ACS, and BLS unemployment statistics.
Notes: In the unmatched models, the 547 PUMAs in each expansion state are compared to all 861 PUMAs in non-expansion states, not limited to the PUMAs within the state group. In the matched models, PUMAs from expansion states could only match to PUMAs from states in the same state group. The number of observations in each regression include the number of matched PUMAs, multiplied by the number of quarters of data available (as many as 28 quarters).

Table 5:

Coefficients from Difference-in-Differences Model.

	SSI						SSDI-only
	Post		Quarterly		Quarterly with leads		Post		Quarterly		Quarterly with leads
Variable	Match	Unmatch	Match	Unmatch	Match	Unmatch	Match	Unmatch	Match	Unmatch	Match	Unmatch
Expanded	0.108	0.113*					0.044**	0.035*
	[0.12]	[0.06]					[0.02]	[0.02]
Expanded* (q)			−0.009	0.022	−0.075	−0.041			0.004	0.020	−0.050	−0.013
			[0.09]	[0.06]	[0.05]	[0.03]			[0.03]	[0.02]	[0.05]	[0.02]
Expanded* (q + 1)			0.130	0.061	0.064*	−0.003			0.064**	0.029*	0.009	−0.004
			[0.10]	[0.05]	[0.03]	[0.03]			[0.03]	[0.02]	[0.04]	[0.02]
Expanded* (q + 2)			0.127	0.055	0.061	−0.009			0.047	0.012	−0.007	−0.021
			[0.09]	[0.06]	[0.06]	[0.06]			[0.04]	[0.02]	[0.04]	[0.03]
Expanded* (q + 3)			0.140	0.158**	0.075**	0.094***			0.045	0.040*	−0.010	0.007
			[0.12]	[0.07]	[0.03]	[0.03]			[0.03]	[0.02]	[0.04]	[0.02]
Expanded* (q + 4)			0.150	0.116*	0.084*	0.052			0.051**	0.053**	−0.004	0.020
			[0.13]	[0.06]	[0.04]	[0.03]			[0.02]	[0.02]	[0.04]	[0.02]
Expanded* (q + 5)			0.162	0.140**	0.096*	0.076*			−0.004	0.023	−0.058*	−0.009
			[0.11]	[0.06]	[0.05]	[0.04]			[0.03]	[0.02]	[0.03]	[0.03]
Expanded* (q + 6)			0.108	0.103	0.042	0.039			0.065**	0.022	0.010	−0.011
			[0.11]	[0.06]	[0.05]	[0.05]			[0.02]	[0.03]	[0.05]	[0.04]
Expanded* (q + 7)			0.102	0.180**	0.036	0.116**			0.048	0.060*	−0.006	0.028
			[0.15]	[0.09]	[0.05]	[0.05]			[0.04]	[0.03]	[0.03]	[0.03]
Expanded* (q + 8)			0.061	0.134	−0.005	0.071			0.057**	0.062**	0.003	0.030
			[0.13]	[0.08]	[0.05]	[0.05]			[0.02]	[0.03]	[0.04]	[0.04]
Expanded* (q + 9)			0.070	0.112	0.004	0.048			0.082**	0.035	0.028	0.003
			[0.11]	[0.07]	[0.05]	[0.05]			[0.03]	[0.02]	[0.06]	[0.03]
Expanded* (q + 10)			0.130	0.109	0.064	0.045			0.035	0.015	−0.019	−0.018
			[0.13]	[0.07]	[0.06]	[0.06]			[0.03]	[0.03]	[0.06]	[0.04]
Expanded* (q + 11)			0.137	0.174*	0.071	0.109*			0.028	0.042	−0.026	0.010
			[0.17]	[0.10]	[0.07]	[0.06]			[0.04]	[0.03]	[0.04]	[0.03]
Expanded* (q − 16)					−0.110	−0.064					−0.066	−0.034
					[0.15]	[0.10]					[0.06]	[0.03]
Expanded* (q − 15)					−0.093	−0.133					−0.083	−0.043
					[0.16]	[0.11]					[0.06]	[0.03]
Expanded* (q − 14)					−0.140	−0.171					−0.022	−0.030
					[0.20]	[0.12]					[0.08]	[0.04]
Expanded* (q − 13)					−0.083	−0.015					−0.043	0.009
					[0.11]	[0.08]					[0.04]	[0.02]
Expanded* (q − 12)					−0.081	−0.024					−0.036	−0.017
					[0.14]	[0.08]					[0.06]	[0.03]
Expanded* (q − 11)					−0.009	−0.083					−0.039	−0.040*
					[0.14]	[0.09]					[0.04]	[0.02]
Expanded* (q − 10)					−0.068	−0.118					−0.047	−0.056**
					[0.14]	[0.08]					[0.05]	[0.02]
Expanded* (q − 9)					−0.049	−0.000					−0.058	−0.019
					[0.09]	[0.06]					[0.04]	[0.02]
Expanded* (q − 8)					−0.050	−0.013					−0.065	−0.028
					[0.13]	[0.07]					[0.06]	[0.02]
Expanded* (q − 7)					−0.037	−0.087					−0.035	−0.057**
					[0.13]	[0.08]					[0.04]	[0.02]
Expanded* (q − 6)					−0.090	−0.096					−0.060	−0.049*
					[0.16]	[0.09]					[0.05]	[0.02]
Expanded* (q − 5)					−0.104	−0.022					−0.094***	−0.034
					[0.11]	[0.06]					[0.03]	[0.02]
Expanded* (q − 4)					−0.039	−0.023					−0.103***	−0.037*
					[0.11]	[0.06]					[0.04]	[0.02]
Expanded* (q − 3)					−0.040	−0.059					−0.081*	−0.038
					[0.09]	[0.06]					[0.04]	[0.02]
Expanded* (q − 2)					−0.066	−0.112*					−0.046	−0.051*
					[0.09]	[0.06]					[0.04]	[0.03]
R²	0.948	0.949	0.948	0.949	0.948	0.950	0.901	0.899	0.901	0.899	0.902	0.899
F-statistic (p-value) of test that pre-period quarters jointly equal to zero	–	–	–	–	1.99 [0.06]	2.15 [0.04]	–	–	–	–	4.71 [0.00]	5.33 [0.00]
Number of observations	12,430	39,416	12,430	39,416	12,430	39,416	12,432	39,424	12,432	39,424	12,432	39,424
Number of applications per 1000 working-age adults in Q4 2013 in comparison PUMAs	1.7	1.9	1.7	1.9	1.7	1.9	1.1	1.1	1.1	1.1	1.1	1.1

Source: Authors’ calculations, using data derived from SSA’s SDR, the ACS, and BLS unemployment statistics.
Notes: *Denotes p-value < 0.10, **denotes p-value < 0.05, and ***denotes p-value < 0.01. The model includes quarterly unemployment, quarterly calendar indicators, and PUMA fixed effects. Heteroskedasticity-robust standard errors are clustered at the PUMA level, with weights accounting for the PUMA's population and the matching algorithm as described in the text. Time period q refers to quarter 1, 2014.

References

Abadie, A., A. Diamond, and J. Hainmueller. 2010. “Synthetic Control Methods for Comparative Case Studies: Estimating the Effect of California’s Tobacco Control Program.” Journal of the American Statistical Association 105 (490): 493–505.10.1198/jasa.2009.ap08746Search in Google Scholar

Austin, P. C. 2010. “Statistical Criteria for Selecting the Optimal Number of Untreated Subjects Matched to Each Treated Subject When Using Many-to-One Matching on the Propensity Score.” American Journal of Epidemiology 172: 1092–1097.10.1093/aje/kwq224Search in Google Scholar PubMed

Burns, M., and L. Dague. 2017. “The Effect of Expanding Medicaid Eligibility on Supplemental Security Income Program Participation.” Journal of Public Economics 140: 20–34.10.1016/j.jpubeco.2017.03.004Search in Google Scholar

Caliendo, M., and S. Koepenig. 2005. “Some Practical Guidance for the Implementation of Propensity Score Matching.” Discussion Paper 1588. Bonn, Germany: Institute for the Study of Labor (IZA).10.2139/ssrn.721907Search in Google Scholar

Chatterji, P., and Y. Li. 2016. “Early Effects of the 2010 Affordable Care Act Medicaid Expansions on Federal Disability Program Participation.” Working Paper No. 22531. Cambridge, MA: National Bureau of Economic Research (NBER).10.3386/w22531Search in Google Scholar

Crump, R. K., V. J. Hotz, G. W. Imbens, and O. A. Mitnik. 2009. “Dealing with Limited Overlap in Estimation of Average Treatment Effects.” Biometrika 96: 187–199.10.1093/biomet/asn055Search in Google Scholar

de Vocht, F., R. Campbell, A. Brennan, J. Mooney, C. Angus, M. Hickman. 2016. “Propensity score matching for selection of local areas as controls for evaluation of effects of alcohol policies in case series and quasi case–control designs.” Public Health 132: 40–49.10.1016/j.puhe.2015.10.033Search in Google Scholar PubMed

Garthwaite, C., T. Gross, and M. J. Notowodigdo. 2014. “Public Health Insurance, Labor Supply, and Employment Lock.” The Quarterly Journal of Economics 129: 653–696.10.1093/qje/qju005Search in Google Scholar

Gettens, J., P. P. Lei, and A. Henry. 2016. “Accounting for Geographic Variation in DI and SSI Participation.” DRC Working Paper 2016-03. Washington, DC: Mathematica Policy Research.Search in Google Scholar

Hall, J. P., A. Shartzer, N. K. Kurth, and K. C. Thomas. 2017. “Effect of Medicaid Expansion on Workforce Participation for People with Disabilities.” American Journal of Public Health 107: 262–264.10.2105/AJPH.2016.303543Search in Google Scholar PubMed

Ho, D. E., K. Imai, G. King, and E. A. Stuart. 2007. “Matching as a Nonparametric Preprocessing for Reducing Model Dependence in Parametric Causal Inference.” Political Analysis 15: 199–236.10.1093/pan/mpl013Search in Google Scholar

Kaiser Family Foundation. 2016. “Current Status of Health Insurance Marketplace and Medicaid Expansion Decisions.” Retrieved October 23, 2016. Available at http://kff.org/health-reform/slide/current-status-of-health-insurance-marketplace-and-medicaid-expansion-decisions/.Search in Google Scholar

Kaiser Family Foundation. 2017. “The Effects of Medicaid Expansion under the ACA: Updated Findings from a Literature Review.” (September 2017). Retrieved January 16, 2018. Available at http://files.kff.org/attachment/Issue-Brief-The-Effects-of-Medicaid-Expansion-Under-the-ACA-Updated-Findings-from-a-Literature-Review.Search in Google Scholar

Kennedy, J., and E. Blodgett. 2012. “Health Insurance-Motivated Disability Enrollment and the ACA.” New England Journal of Medicine 367: e16.10.1056/NEJMp1208212Search in Google Scholar

Maestas, N., K. J. Mullen, and A. Strand. 2014. “Disability Insurance and Health Insurance Reform: Evidence from Massachusetts.” American Economic Review: Papers & Proceedings 104: 329–335.10.1257/aer.104.5.329Search in Google Scholar

Maestas, N., K. J. Mullen, and A. Strand. 2015. “Disability Insurance and the Great Recession.” American Economic Review 105: 177–182.10.1257/aer.p20151089Search in Google Scholar

Pizer, S. D., A. B. Frakt, and L. I. Iezzoni. 2009. “Uninsured Adults with Chronic Conditions or Disabilities: Gaps in Public Insurance Programs.” Health Affairs 28: 1141–1150.10.1377/hlthaff.28.6.w1141Search in Google Scholar

Rosenbaum, P. R., and D. B. Rubin. 1985. “Constructing a Control Group using Multivariate Matched Sampling Methods that Incorporate the Propensity Score.” The American Statistician 39: 33–38.10.1017/CBO9780511810725.019Search in Google Scholar

Rupp, K., and G. F. Riley. 2016. “State Medicaid Eligibility and Enrollment Policies and Rates of Participation Among Disabled Supplemental Security Income Recipients.” Social Security Bulletin 76: 17–40.Search in Google Scholar

Social Security Advisory Board. 2018. Accessed at http://www.ssab.gov/Facts-and-Figures/Did-You-Know-Charts/Disability-Processing-Time on February 13, 2018.Search in Google Scholar

Social Security Administration. 2017. Annual Statistical Report on the Social Security Disability Insurance Program, 2016. Baltimore, MD: SSA.Search in Google Scholar

Solon, G., S. J. Haider, and J. M. Wooldridge. 2015. “What Are We Weighting For?” Journal of Human Resources 50: 301–316.10.3368/jhr.50.2.301Search in Google Scholar

Sommers, A. S. 2006–2007. “Access to Health Insurance, Barriers to Care, and Service Use Among Adults with Disabilities.” Inquiry 43: 393–405.10.5034/inquiryjrnl_43.4.393Search in Google Scholar

Soni, A., M. E. Burns, L. Dague, and K. I. Simon. 2017. “Medicaid Expansion and State Trends in Supplemental Security Income Program Participation.” Health Affairs 36: 1485–1488.10.1377/hlthaff.2016.1632Search in Google Scholar

Stuart, E. A. 2010. “Matching Methods for Causal Inference: A Review and a Look Forward.” Statistical Science 25: 1–21.10.1214/09-STS313Search in Google Scholar

Published Online: 2019-02-23

The Impact of Affordable Care Act Medicaid Expansions on Applications to Federal Disability Programs

Abstract

Acknowledgements

Methods Appendix: Matching Model Details

References

Journal and Issue

Articles in the same Issue