Premium Subsidies, Medicaid Expansion & the Mandate: Coverage Impacts of the Affordable Care Act Ben Sommers & Molly Frean Harvard T.H. Chan School of Public Health Jonathan Gruber, MIT and NBER AcademyHealth Annual Research Meeting June 2016
Acknowledgments • This project was supported by grant number K02HS021291 from the Agency for Healthcare Research and Quality (AHRQ). • I currently serve part-time as an advisor in the Office of the Assistant Secretary for Planning and Evaluation, at the U.S. Department of Health and Human Services (HHS). • This paper does not represent the views of HHS or AHRQ.
Background • ACA took a multi-pronged approach expanding health insurance: • Individual mandate & insurance market reforms • Expanded Medicaid (income <138% poverty) • Tax credits for private insurance purchased via Exchanges (138-400% FPL)
Research To Date • Ample evidence from time-series analyses of multiple data sources – ACA implementation associated with dramatic drop in uninsured rate • Analyses of Medicaid show large coverage gains for low- income adults in expanding states, vs. non-expansion • But no research to date has assessed the relative contributions of the ACA’s key policy tools for the nation as a whole Sources: Clemens-Cope et al. Urban Institute 2014; Black & Cohen NHIS 2015 ; Sommers et al. JAMA 2015; Wherry & Miller Annals Int Med 2016; Kaestner et al NBER 2015; Courtemanche et al. NBER 2016
Objective & Overview • Objective : Provide the first comprehensive analysis of ACA coverage impacts from the law’s key features: 1. Medicaid expansion 2. Premium subsidies 3. The individual mandate • Overview: Difference-in-difference-in-difference model comparing changes in coverage over time by income group and by geography
Methods: Data & Sample • Household microdata from American Community Survey (ACS) for 2012-2014 • Information on income, family structure, demographics, and health insurance • Level of analysis is the “health insurance unit” – adult, his/her spouse, and dependent children • All non-elderly adults, 0-64 years of age (>2 million observations per year) • Detailed within-state geography , but can’t go back further than 2012
Policy Measures: Medicaid • Medicaid eligibility from CMS, Kaiser Foundation, and state sources • Based on age, income, disability and family structure • We distinguish between existing Medicaid eligibility as of 2013 and new Medicaid eligibility in 2014 • For secondary analyses, we also distinguish between: • Non-expansion states as of 2014 (n=24) • Early expander states, between 2011-2013 (n=6) • 2014 expansion states (n=21)
Medicaid Eligibility: Children
Medicaid Eligibility: Adults
Policy Measures: Premiums • Marketplace premiums by rating area from Robert Wood Johnson Foundation - mapped onto 2,350 ACS “public use microdata areas” (PUMAs): • Using 2 nd -lowest cost silver plan in each rating area • Age-specific using CMS age-rating curves • Subsidy amount based on ACA provisions: family size, family income (between 100/138% and 400% FPL) • Compare state-based and federal marketplace states
Policy Measures: Premiums % Subsidy per Family
Policy Measures: Mandate • Exempt from Mandate (38%): • Those below tax filing threshold (21% of sample) • Those in Medicaid gap in non-expansion states (6%) • Native Americans (1%) • Affordability exemption, based on lowest cost bronze plan > 8% of family income (10%) • Mandate Penalty for Non-Exempt (62%): • $95 per uninsured adult (half per child) or 1% of taxable income, whichever is greater • Capped at national average bronze premium
Policy Measures: Mandate
Simulated Policy Variables PROBLEM – ‘Observed’ eligibility, mandate, subsidy subject to bias: • Income and eligibility may change in response to ACA • Family-level estimates of policy parameters - imprecisely measured SOLUTION – Use “simulated” measure: • Randomly select 200 families of each of three types – single adults, adult couples, and adults with children • Assign the same 200 families to each PUMA & income band, and compute values for all key policy measures using this standardized • Similar to previous work by Currie & Gruber on Medicaid in the 1990s
Results: Uninsured Variable Coefficient Percent Subsidy * 2014 -0.052*** Family Mandate Penalty * 2014 ($100s) 0.0001* Previously Medicaid-Eligible * 2014 -0.040*** Newly Medicaid-Eligible * 2014 -0.088*** Notes: ***p<0.01, **p<0.05, *p<0.10 Models include demographic controls, state and year fixed effects, with robust SE clustered at PUMA level.
Interpretation • Each additional 10% subsidy reduces uninsured by 5.2 percentage points, or roughly 1.5 million persons • Mandate effect negligible (and wrong-signed): $100 increase in average mandate penalty (~20% of baseline) raises uninsured by 0.01 percentage points • Big effect in 2014 of Medicaid eligibility • 8.8% reduction in uninsured among newly-eligible • 4.0% reduction in uninsured among previously eligible
Decomposing the ACA’s Policy Effects Implied % of ACA- Reduced Population Percentage Related Form ß Mean Point Change Variable (* 2014) Change -0.052 0.162 -0.85% 37% Percent Subsidy 0.0001 4.58 0.05% N/A Family Mandate Penalty Previously Medicaid- -0.040 0.248 -1.00% 44% Eligible Newly Medicaid-Eligible -0.088 0.049 -0.43% 19% Note: Our parameterized policies overall explain a ~2.3 percentage-point drop in uninsured rate in 2014, compared to ~3.4 point drop in raw data.
Results: Type of Coverage Medicaid / “State - Uninsured ESI Non-Group Subsidized Variable (* 2014) Coverage” -0.052*** 0.018*** 0.009*** 0.029*** Percent Subsidy 0.0001* 0.0003*** -0.0002 -0.0003*** Family Mandate Penalty Previously Medicaid- -0.040*** 0.038*** 0.005** 0.003* Eligible Newly Medicaid-Eligible -0.088*** 0.092*** 0.002 0.002 Notes: ***p<0.01, **p<0.05, *p<0.10 Models include demographic controls, state and year fixed effects, with robust SE clustered at PUMA level.
Results: Type of Coverage Medicaid / “State - Uninsured ESI Non-Group Subsidized Variable (* 2014) Coverage” -0.052*** 0.018*** 0.009*** 0.029*** Percent Subsidy 0.0001* 0.0003*** -0.0002 -0.0003*** Family Mandate Penalty Previously Medicaid- -0.040*** 0.038*** 0.005** 0.003* Eligible Newly Medicaid-Eligible -0.088*** 0.092*** 0.002 0.002 Notes: ***p<0.01, **p<0.05, *p<0.10 Models include demographic controls, state and year fixed effects, with robust SE clustered at PUMA level.
Results: Type of Coverage Medicaid / “State - Uninsured ESI Non-Group Subsidized Variable (* 2014) Coverage” -0.052*** 0.018*** 0.009*** 0.029*** Percent Subsidy 0.0001* 0.0003*** -0.0002 -0.0003*** Family Mandate Penalty Previously Medicaid- -0.040*** 0.038*** 0.005** 0.003* Eligible Newly Medicaid-Eligible -0.088*** 0.092*** 0.002 0.002 Notes: ***p<0.01, **p<0.05, *p<0.10 Models include demographic controls, state and year fixed effects, with robust SE clustered at PUMA level.
State Subgroups • Premium subsidy effects were twice as large in states with state-based Exchanges than federal healthcare.gov • PercentSubsidy ß=-0.080 vs. -0.044, both p<0.001 • Pre-ACA Medicaid effect was 3x as large in early expansion states than in 2014 expanders and non- expanders: • PreviousMcaidElig ß=-0.063 vs. -0.028 & -0.021, all p<0.001
Policy Implications • 40% of ACA coverage gains due to Marketplace premium subsidies • State Marketplaces had much higher take-up rates than federal (outreach, functionality, politics?) • Mandate – no specific effects of penalty details, but this misses general “taste for compliance” (Saltzman et al 2015) • Effect may be larger in future years as penalties increase
Policy Implications • 60% of ACA is Medicaid effect, can be roughly divided into thirds: 1. New Medicaid expansion in 2014 • Note that 7 additional states have expanded since then 2. Woodwork or welcome mat effect, which occurred in all states regardless of expansion status • Streamlined application, navigators, and publicity (all states) 3. Early expansion effect – building on 6 states’ early efforts • ‘Priming the pump,’ takes time for awareness to spread • No crowd-out from Medicaid once we model premium subsidies
Questions? Ben Sommers bsommers@hsph.harvard.edu
Recommend
More recommend
