ebook img

Evidence from the ACA Medicaid Expansion PDF

33 Pages·2017·0.51 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Evidence from the ACA Medicaid Expansion

Moving for Health Insurance: ∗ Evidence from the ACA Medicaid Expansion Patrick S. Turner University of Colorado - Boulder March 2017 Preliminary Draft - Not for Citation Abstract This paper extends the welfare magnets literature by using the 2014 expansion of Medi- caid to test for welfare migration. The Patient Protection and Affordable Care Act (ACA) mandated the expansion of Medicaid. However, a Supreme Court decision overturned this provision and allowed states to opt-out. As a result, childless adults with incomes up to 138% of the federal poverty level are eligible for Medicaid in states that have expanded coverage and are not eligible at any income level in the majority of states that opted to not expand. Using a difference-in-differences strategy which compares the log estimated low-income population of expansion and non-expansion states, before and after Medicaid expansion, I find that the low-income population increased by 1% in expansion states after 2012 relative to what one would expect given pre-existing trends and post-decision population changes in non-expansion states. This result is robust to the use of state-level synthetic control. ∗ Turner: [email protected]. IwouldliketoacknowledgehelpfulcommentsandguidancefromFrancisca Antman, Tania Barham, Brian Cadena, and Terra McKinnish. 1 Introduction The2012SupremeCourtdecisionallowingstatestoopt-outoftheMedicaidexpansioninthePatient Protection and Affordable Care Act (ACA) induced significant cross-state variation in eligibility for Medicaid. As a result, adults without children are eligible for Medicaid up to 138% of the federal poverty level in the 28 states that have expanded coverage. On the other hand, these same adults would not be eligible at any income level in the majority of the non-expansion states.1 An important policy question is then whether low-income households will migrate across state borders in order to receive public health insurance. That is, do individuals in low(no)-eligibility states tend to move to high-eligibility states to receive benefits? Where low-income households locate have important effects on local labor markets. For instance, welfare-induced migration could shift the skill composition of workers, receipts of other welfare programs could increase, and human capital accumulation could be affected through negative spillovers from changes in the socio-economic distribution of students in the public-school system. Using a difference-in-differences strategy which compares the log estimated low-income population of expansion and non-expansion states, before and after Medicaid expansion, I test for evidence of welfare-induced migration in the context of the Medicaid program. The ACA introduced broad changes to health insurance coverage in the United States. An important feature of the legislation was to increase coverage rates for all citizens. The legislation allowed individuals with pre-existing conditions to be covered under new policies, increased the age in which young adults could maintain coverage from their parents, and created tax-based incentives for individuals to acquire coverage and employers to provide coverage. Importantly, the original legislation provided significant coverage expansions to low-income individuals. States were required to extend Medicaid coverage to childless adults, among other household types, up to 138% of the federal poverty line. However, this measure was struck down by the Supreme Court as unconstitutional in 2012, allowing states to opt-out of expanding Medicaid eligibility. 1Even though adults without children are not eligible for Medicaid, the ACA does provide individuals with a subsidy on the purchase of health insurance in the marketplace. An individual above 100% of the poverty level is eligible for a subsidy. 1 Using individual level data from the American Community Survey, I estimate the effect of Med- icaid expansion on household locational choice. Specifically, I construct estimates of low-income population counts at the state level using time-invariant characteristics of individuals. Then, I implement a difference-in-differences strategy which compares population trends in expansion and non-expansion states before and after the expansion of Medicaid. With this strategy, I find re- sults consistent with the welfare magnets hypothesis. I confirm the results from the difference-in- differences strategy with a synthetic control approach which allows the data to construct a set of weights for non-expansion states so that population trends prior to Medicaid expansion match more closely. Finally, I use an approach similar to McKinnish (2005) an compare differences in Medicaid take-up in border regions to interior regions within a state, before and after Medicaid expansion. In all three approaches, I find results consistent with welfare-induced migration. I estimate that the low-income population in expansion is about 1% than what one would expect in the absence of Medicaid expansion. However, back of the envelop calculations suggest that the size of the effect does little to alleviate the gap in coverage caused by the Supreme Court ruling. This paper is related to the broader “welfare magnets” literature which explores migration in- duced by cross-state differences in welfare benefits. Most work focuses on cross-state variation in cash welfare programs - Aid to Family with Dependent Children (AFDC) and Temporary Assis- tance for Needy Families (TANF). In examining migration of natives, McKinnish (2005, 2007) finds evidence that individuals move across borders to receive higher cash benefits. Fiva (2009) finds sim- ilar evidence in Norway. In considering the locational choice of immigrants in particular, Buckley (1996), Borjas (1999), and Dodson III (2001) all find evidence that immigrants disproportionately locatetohighbenefitstates. However, theirworksuffersfromeconometricproblemsassociatedwith relying on cross-sectional data (Buckley, 1996; Dodson III, 2001) or time-series problems (Borjas, 1999) by not being able to control for state-level unobservables or macroeconomic trends changing over time, respectively. Kaushal (2005) overcomes these difficulties by exploiting cross-state vari- ation in welfare eligibility introduced by the Personal Responsibility and Work Opportunity Act (PRWORA). The author finds no evidence for the “welfare magnets” hypothesis with respect to 2 immigrant welfare migration. Despiteawideliteratureexaminingthemagneticincentivesofcashbenefits, littleworkconsiders whether households relocate to gain public health insurance coverage. Chatterji and Li (2016) briefly consider the topic of migration in their paper which considers the effects of early (pre-2014) Medicaid expansions on disability program takeup. 2 My work contributes to the literature by exploring the possible effects of vast eligibility differences induced by the ACA. Thus, this paper fits into the broader literature considering the effects of ACA. Furthermore, whether households decide torelocateis of greater importancewithin the field oflabor economics. Changes in household behavior could affect the skill composition of the labor supply, demographic characteristics of peers in the education system, tax burden of state residents, among others. Furthermore, the ability for households to “vote with their feet” might mean that national take-up rates could be higher than expected despite low eligibility levels in some states. Finally, whether or not individuals have health insurance coverage could affect long-term health status which has implications for human capital accumulation and employment outcomes. The paper proceeds as follows. Section 2 explains Medicaid expansion and introduces the basic theoretical framework for thinking about welfare induced migration. Section 3 discusses the data used for analysis and the empirical strategy. Section 4 presents results. Section 5 provides a concluding discussion. 2 Theoretical Framework 2.1 Medicaid Expansion The U.S. experienced a significant overhaul of its health insurance system with the signing of the ACA in March, 2010. Various provisions of the law sought to increase health insurance coverage across the country. For instance, insurers are no longer able to deny coverage based on pre-existing 2There are some papers that consider migration induced or reduced by non-cash welfare programs. For example, Barham and Kuhn (2014) considers if individuals stay in rural Bangladesh in response to expansion in healthcare supply and Verdugo (2014) looks at the effect of public housing supply on immigrant locational choice in Europe. 3 conditionsandyoungadultsareeligibletoremainontheirparent’sinsuranceplanuntiltheageof26. However, two important mechanisms were put in place to increase health insurance coverage rates: expanding the populations eligible for Medicaid and the creation of health insurance exchanges to improve access to private insurance. The initial intent of the ACA was to expand Medicaid eligibility to all individuals whose house- holdincomeplacesthembelow138%ofthefederalpovertyline. Thefederalgovernmentisresponsi- blefor100%oftheincreaseincostforexpandingeligibilityfortheinitialyearsafterimplementation. Beyond 2020, the federal government will still cover 90% of the costs from expansion. However, many states felt the required expansion to be unconstitutional. In 2012, the Supreme Court ruled in NFIB v. Sibelius that Congress did not have the constitutional right to dictate that states expand Medicaid. As a result, states were given the right to opt-out of that particular provision in the law. The Supreme Court ruling created significant cross-state variation in Medicaid eligibility levels between non-expansion and expansion states. Importantly, childless adults in most non-expansion states are ineligible for Medicaid, regardless of household income. These same individuals would be eligible up to 138% in expansion states. Table 1 specifies the states used in analysis by expansion states. In total, 24 states are included that expanded Medicaid on January 1, 2014 and 21 states chose not to expand Medicaid at any time since. Six states are removed from analysis. Alaska and Hawaii are excluded because they are not in the continental United States. Indiana, Michigan, New Hampshire, and Pennsylvania are excluded because they expanded Medicaid after the initial date. Figure 1 documents the sharp increase in Medicaid take-up in 2014 due to Medicaid expansion. The dashed line plots the share of the adult population with Medicaid in non-expansion states. In the period 2008-2015, this share slightly increases from about 7 percent to 10 percent. The solid line plots the same outcome for the set of states that expanded Medicaid on January 1, 2014. Adults in these states were more likely to be covered by Medicaid prior to expansion. However, this gap increases in 2014 by almost 4 percent. Since there are certain groups that are eligible in some states but not others, individuals may reap important welfare benefits by migrating across state borders. However, we could be concerned that other provisions within ACA that are in effect in non- 4 expansion states might ease the burden of getting health insurance coverage within this population. Indeed, the ACA implemented what is commonly known as the individual mandate requiring in- dividuals to purchase coverage. Specifically, individuals are required by law to maintain health insurance coverage or face a tax penalty. Indeed, as Figure 2 shows, insurance coverage rates im- proved in both expansion and non-expansion states. To ease the ability to find affordable insurance, the law created state-level health insurance exchanges. The exchanges provides a clearer market structure for individuals to shop and compare various plans. Additionally, certain individuals above 100% of the federal poverty line are eligible for subsidies to cover a large portion of the cost of insurance. However, the alteration of the law from its original intent has left a “coverage gap” for many low-income individuals. In this context, Medicaid is the only affordable option for this population, when available. 2.2 Welfare Migration In what follows, I adapt McKinnish (2005) which provides a suitable framework for considering welfare migration. Imagine that there two bordering states each separated into two regions: the interior and the border. The region which borders the neighboring state is the border county and the other region is the interior county. Within each state, Medicaid eligibility is a uniform policy. However, the eligibility level has the potential to change discretely at the state border. States can either set policy to be a “high” eligibility state or a “low” eligibility state. In this setup, four border types could exist: high-high, high-low, low-high, and low-low. For instance, the high-high type is a high eligibility state which borders another high eligibility state. Remaining characteristics of a region which might influence population flows are assumed to be constant across regions within a state. Consider an individual’s decision to migrate. There are various components which impact the optimal location. For instance, wages, local amenities, social networks, and public benefits are all inputs in an individual’s utility in the region. As assumed above, wages and amenities are constant across a state, or at least observable to the econometrician, and given the importance of 5 local labor markets, we should not expect to see large wage differentials across neighboring regions. Furthermore, since public benefits are set at the state level, individual’s receive the same benefit receipt regardless of region choice within a region. Importantly, eligibility has the potential to vary across the state border. In the four border types listed above, high-low and low-high borders could provide an incentive for a household to migrate from the low to the high state. If individuals decide to move, this would be observed as an increase in the population at-risk of being eligible for Medicaid in states that set higher eligibility levels. Similarly, the size of this population would fall in low eligibility states. I will explore this possibility by comparing differences in population trends in non-expansion and expansion state, before and after Medicaid expansion. However, important social networks and information costs might make it more difficult to move across multiple regions or even be aware of policies in neighboring states. For this reason, we should expect to see more migration between border counties relative to interior-interior migration. That is, marginal migrants who find the benefits of Medicaid exceed the cost of migration will choose to relocate to a border county since the cost will be strictly smaller than choosing the interior county in the neighboring state. Although, there may be some migration from the low-eligibility interior county to the high-eligibility border county, I expect this amount to be smaller given increased costs from leaving behind key support networks and information costs of moving a farther distance. Finally, the welfare magnet mechanism should be stronger when the gap in eligibility levels is wider betweenneighboringstates. Thismechanisminformsatripledifferenceapproachwhichcomparesan outcomevariableofinterestbetweenborderandinteriorcountieswithinnon-expansionorexpansion state, before and after Medicaid expansion. 3 Methodology This section discusses the data used in analysis and lays out the empirical strategy to explore the presence welfare magnets in the setting of public health insurance. The approach is two-tiered: (i) a between-state analysis which compares changes in population levels/trends of the at-risk Medicaid 6 population before and after Medicaid expansion between expansion and non-expansion states and (ii)awithin-stateanalysiswhichcomparesdifferencesinMedicaidtake-upbeforeandafterMedicaid expansion between geographies that border a neighboring state to those that do not. 3.1 Data Data come from the 2005-2015 American Community Survey (ACS) compiled by IPUMS-USA (Ruggles et al., 2015). The ACS is an annual household survey collected by the U.S. Census Bureau and represents an annual 1% sample of the entire United States. The survey contains detailed data on economic, demographic, and geographic characteristics of individuals. Specifically, the analysis in this paper relies on an individual’s age, year of birth, location of residence, income, level of education, and type of health insurance coverage. Additionally, I use the individual-level data to construct average wages of particular sectors (manufacturing, retail, and construction) to use as controls. The between-state analysis of this paper relies on population trends. I construct state-level population estimates of the low-income population for 2005-2015.3 I restrict the sample to native- born individuals not living in group quarters and not working in the armed forces. The ACA had a provision which allows individuals to remain on their parent’s health insurance until the age of 26. Additionally, individuals become eligible for Medicare at the age of 65. For these reasons, I further restrict the sample to adults aged 26-64. As noted above, the ACA expanded Medicaid to all individuals with modified adjusted gross income below 138% of the Federal Poverty Level. An individual’s income can vary over time due to economic conditions and these conditions vary across states. These state-specific economic shocks could bias estimation if the shocks are correlated with states that expanded Medicaid. I constructestimatesofthelow-incomepopulationofstatesforeachyearbypredictingtheprobability that an individual’s income falls below 138% FPL using a set of time-invariant characteristics and summing the product of the predicted probability and the ACS person weight. Specifically, I 3Priorto2005,theACSwasalessthana1%sampleofthenationalpopulation. Forthisreason,Irestrictanalysis to 2005 and later. 7 estimate education-specific probabilities using the following model: Pr[Povertyi <= 138] = XβkEducki +ǫi (1) allk where i indexes the individual and k indexes an educational category (high school dropout, high school graduate, some college, college graduate, and advanced degree). The dependent variable is an indicator if individual i has an IPUMS provided poverty measure that is less than or equal to 138 and Educk are a set of indicators for the individual’s highest educational attainment. The i coefficients, β , are the share of each education group that falls below 138 FPL. By allowing the k data to estimate the likelihood that an educational group is at-risk of being eligible for Medicaid, this approach is similar in spirit to earlier work in the welfare magnets literature which constructed comparison groups based on educational attainment (e.g., Gelbach, 2004; McKinnish, 2007). Table A-1 presents the estimated probabilities of being in the at-risk Medicaid population by educational attainment. Column 1 uses data from the post-Medicaid Expansion years, 2014-2015. High-school dropouts are the most likely to fall below the threshold at 45.1 percent. Higher ed- ucational groups are monotonically less likely to fall below the threshold with individuals with a post-graduate degree only being 4.16 percent likely to be at risk of being eligible for Medicaid. Using data from the years 2005-2015 show estimated probabilities that are largely unchanged and the results are robust to using either set of probabilities. With probabilities in hand, I am able to construct state-year population estimates of the low income population. These estimates are created by summing, over all individuals, the product of the predicted probability of falling below the threshold and the individual’s ACS person weight.4 The population estimate places more weight on individuals that dropped out of high school (45.1%) and graduated high school (21.6%), but still allows changes in the college-educated population to influence the estimates. The within-state analysis of this paper separates individuals within a state into different regions based on their public-use micro data area (PUMA). PUMAs are geographically contiguous areas 4Specifically, I construct population estimates in the following way: Popst =Palli(cid:0)pˆkist·ACSperwtist(cid:1) 8 that with a population around 100,000 people. The analysis requires PUMAs to be categorized as border or interior regions. Geographic data on PUMAs come from shape files provided by IPUMS. Since PUMA borders changed beginning with the 2012 ACS, the analysis is restricted to data from 2012-2015. Border regions are defined as those regions which share a border with a neighboring state. Interior regions are defined as those regions which only share a border with other regions within the state or large bodies of water. PUMAs within a metropolitan statistical area (MSA) tend to have a small geographic area. Some regions that are only a few miles from the border would be classified as interior even though they are in the city that borders a neighboring state. An example is St. Louis, MO which borders Illinois. I combine all PUMAs within the same MSA, whether or not that MSA borders a state, so that if a city borders a neighboring state, then all of the PUMAs within that metropolitan area are coded as one border region. I will refer to regions within a state as PUMA-MSAs. The main outcome for the within-state analysis is the share of the adult PUMA-MSA population on Medicaid. The sample is restricted to individuals aged 26-64. The adult Medicaid share is constructed by summing the ACS person weights of individuals with Medicaid and dividing by the sum of the ACS person weights of all adults within the PUMA-MSA. 3.2 Empirical Strategy The difficulty with estimating a causal relationship between state Medicaid policy and low-income migration is that state-level policy choice is plausibly endogenous and eligibility expansion could be a part of a package of other policy changes that are correlated with welfare migration. However in the case of Medicaid expansion, this is less of a concern. First, the ACA was a national-level policy so its timing does not necessarily coincide with state-level policy packages. Furthermore, as discussed above, it was not until years after the law was passed that the Supreme Court, in an unlikely decision, decided to overrule Medicaid expansion while leaving the remainder of the law in effect. Thus, Medicaid expansion is the only part of the law which introduces cross-state differences in policy. However, a careful identification strategy is still necessary to argue for causality. I begin 9

Description:
This paper extends the welfare magnets literature by using the 2014 expansion .. presence welfare magnets in the setting of public health insurance.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.