The Impact of ACA Medicaid Expansions on Applications to Federal Disability Programs

Transcription

1 The Impact of ACA Medicaid Expansions on Applications to Federal Disability Programs Jody Schimmel Hyde Priyanka Anand, Maggie Colby, and Lauren Hula Paul O Leary (SSA) Presented at the Annual DRC Research Meeting Washington, DC August 3, 2016

2 ACA Medicaid Expansions 2014: States had option to expand Medicaid under the Patient Protection and Affordable Care Act (ACA) of 2010 Adults with household income <138 percent of federal poverty level (FPL) 24 states plus DC expanded Medicaid on January 1, additional states expanded later in 2014 or 2015 Expanded access to health insurance coverage for low-income childless adults

3 Study Purpose Estimate impact of ACA Medicaid expansions on applications to Supplemental Security Income (SSI) and Social Security Disability Insurance (SSDI) Previous similar studies Maestas, Mullen, and Strand (2014): Massachusetts health reform Burns and Dague (2016): Pre-ACA Medicaid expansions

4 Potential Effects of Medicaid Expansions on Disability Applications Fewer disability applications if: New Medicaid coverage reduces value of SSI and SSDI Improved access to health care reduces work disability More disability applications if: Newly insured have greater awareness of public supports and seek SSI or SSDI Expansion induced SSDI applications by reducing costs during 5-month waiting period Any effects will be moderated by implementation and awareness of ACA coverage provisions

5 Analysis Overview Matched geographic areas within each expansion state to areas in comparable non-expansion states based on population characteristics in the years prior to 2014 Census Bureau Public Use Microdata Area (PUMA): A geographic area within a state that has a population of at least 100,000 Compared SSI and SSDI application rates around 2014 in expansion PUMAs relative to matched nonexpansion PUMAs

6 Data Sources SSA s Structured Data Repository (SDR): SSI and SSDI application counts Quarterly PUMA-level counts of initial-level applications with a decision from 2010 through 2015 SSI/concurrent and SSDI-only American Community Survey (ACS): PUMA-level population, demographic, and socioeconomic characteristics Bureau of Labor Statistics (BLS): Quarterly unemployment rate

7 Matching Expansion States to Non-Expansion States Limited analysis to 19 states in which ACA Medicaid expansions significantly increased eligibility For each of those, identified a set of potential comparison states based on pre-aca Medicaid policies The linkage between Medicaid and SSI Presence of a medically needy program in Medicaid

8 Categorization of Treatment States and Potential Comparison States Group SSI/ Medicaid linked Medically needy program Treatment states (19) 1 Yes Yes (8): Arkansas, Michigan, New Jersey, Pennsylvania, West Virginia, Kentucky, Rhode Island, Washington 2 Yes No (4): Arizona, Indiana, Colorado, New Mexico 3 No Yes (4): Connecticut, Illinois, New Hampshire, North Dakota 4 No No (3): Alaska, Ohio, Nevada Potential comparison states (20) (6): Florida, Georgia, Louisiana, Montana, North Carolina, Tennessee (6): Alabama, Mississippi, South Carolina, South Dakota, Texas, Wyoming (4): Kansas, Nebraska, Utah, Virginia (3): Idaho, Missouri, Oklahoma

9 Identifying Comparison PUMAs Propensity score model, matching done separately for each of the groups on the last slide Predictors for a PUMA being in an expansion state in the propensity score model: Trend in SSI/SSDI application rates, Population density, working-age population, change in population size ( ) Overall age distribution Cost of living (median rent in 2013) Racial composition of working-age population Health insurance status of working-age population Share of working-age population with income <137% FPL

10 Important Details About Matching Model Up to 4 matches for each expansion PUMA Nearest neighbor matching with replacement Matches within 0.1 standard deviations (SD) of treatment case propensity score Could not find a suitable match for a significant share of expansion state PUMAs Propensity scores trimmed to the (0.1, 0.9) interval Matched PUMAs much closer on observable characteristics Separately in the four groups and overall, standardized bias was reduced to less than 0.25 SD

11 Matching Produced a Reasonable Group of Comparison PUMAs Unmatched Matched Expansion Non-Expansion Expansion Comparison Number of PUMAs Change in unemployment, 2010 to 2013 (%) Working-age (ages 18-64) population 83,049 80,940 81,504 81,331 Population density (per mile) 2,273 1,457 1,236 1,275 Uninsured (%) Employer or private insurance (%) White (%) Black (%) Hispanic (%) Household income less than 137% FPL (%)

12 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 2015Q1 2015Q2 Applications per 1,000 working age adults Trends in SSI Application Rates Before and After Matching Expansion PUMAs- unmatched Non-expansion PUMAs- unmatched Expansion PUMAs- matched Comparison PUMAs- matched 2 Medicaid expansion date 1.5 Note: These data have been smoothed by averaging each value with the previous four quarters to minimize seasonal patterns in applications.

13 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 2015Q1 2015Q2 Applications per 1,000 working age adults Trends in SSI Application Rates Before and After Matching Expansion PUMAs- unmatched Non-expansion PUMAs- unmatched Expansion PUMAs- matched Comparison PUMAs- matched 2 Medicaid expansion date 1.5 Note: These data have been smoothed by averaging each value with the previous four quarters to minimize seasonal patterns in applications.

14 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 2015Q1 2015Q2 Applications per 1,000 working age adults Trends in SSDI Application Rates Before and After Matching Expansion PUMAs- unmatched Non-expansion PUMAs- unmatched Expansion PUMAs- matched Comparison PUMAs- matched Medicaid expansion date 1 Note: These data have been smoothed by averaging each value with the previous four quarters to minimize seasonal patterns in applications.

15 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 2015Q1 2015Q2 Applications per 1,000 working age adults Trends in SSDI Application Rates Before and After Matching Expansion PUMAs- unmatched Non-expansion PUMAs- unmatched Expansion PUMAs- matched Comparison PUMAs- matched Medicaid expansion date 1 Note: These data have been smoothed by averaging each value with the previous four quarters to minimize seasonal patterns in applications.

16 Estimating the Impact of Medicaid Expansions Difference-in differences regression model Change in the application rate in expansion PUMAs relative to the change in application rate in matched comparison PUMAs Model accounted for time-invariant characteristics of PUMAs, calendar effects, and quarterly unemployment Results presented today are aggregated using PUMAs from all states that expanded in January 2014 Masks cross-state heterogeneity in impacts We will produce state-level statistics to include those states that expanded after January 2014

17 Percentage point change in application rates in expansion relative to comparison PUMAs Note: Values shown were derived from the difference-in-differences regression estimates described on the previous slide. Models were run separately for SSI and SSDI. The values shown are the regression coefficients relative to the expansion PUMA application rate in the quarter. Point estimates were not statistically significant, except for SSDI in 2015 Q1. Estimated Impact of Medicaid Expansions on SSI and SSDI-only Applications Q Q Q Q Q Q2-1 SSI SSDI-only

18 Interpretation of Initial Findings Though estimated effects are positive, they are not statistically significant or substantively large No estimated impact of ACA Medicaid expansion on applications to SSI or SSDI using matched PUMAs Additional issues we are investigating: How the estimates change if matching model is refined to more closely align the SSDI pattern in the pre-expansion period Whether aggregate estimates obscure important state-level differences The extent to which matched PUMAs generalize to the experience of the entire expansion state

19 Contact Information Jody Schimmel Hyde Center for Studying Disability Policy Mathematica Policy Research st Street NE, 12 th Floor Washington, DC (202) jschimmel@mathematica-mpr.com