Health Insurance and College Enrollment: Evidence from a Natural Experiment of the Affordable Care Act Dependent Coverage Mandate

Transcription

1 Health Insurance and College Enrollment: Evidence from a Natural Experiment of the Affordable Care Act Dependent Coverage Mandate Yajuan Li 1 and Marco A. Palma 2 1 Corresponding author. Postdoctoral Research Associate, Department of Agricultural Economics, Texas A&M University TAMU, College Station, TX, Tel: (979) liyajuan@tamu.edu 2 Associate Professor and Extension Economist, Department of Agricultural Economics, Texas A&M University TAMU, College Station, TX, Tel: (979) mapalma@tamu.edu Selected Paper prepared for presentation at the 2017 Agricultural & Applied Economics Association Annual Meeting, Chicago, Illinois, July 30-August 1 Copyright 2017 by Li and Palma. All rights reserved. Readers may make verbatim copies of this document for non-commercial purposes by any means, provided that this copyright notice appears on all such copies. 1

2 Health Insurance and College Enrollment: Evidence from a Natural Experiment of the Affordable Care Act Dependent Coverage Mandate Abstract This study investigates the policy effect of insurance provision on college enrollment by exploiting a natural experiment of the Affordable Care Act dependent coverage mandate. With the purpose of increasing health insurance coverage for young adults, the ACA dependent mandate provides more generous coverage by allowing young adults (19-25 years old) to enroll or remain on their parent s health insurance plan regardless of their student status or marital status. This policy potentially lowers the cost of attending college through a reduction in health insurance cost. Thus, it increases incentives for young adults to attend college. This paper uses a difference-indifferences approach to examine the potential policy effect, and the results suggest that the ACA dependent coverage mandate is associated with increases in the likelihood of enrolling in college by 0.8 percentage points at the enactment stage and 0.9 percentage points at the implementation stage. Keywords: College enrollment; ACA dependent coverage mandate JEL Codes: I13, I20 2

3 1. Introduction Around one in three young adults aged in the United States has no health insurance (Monheit et al. 2011). The eligibility of social health insurance, such as the State Children's Health Insurance Program (SCHIP), is usually phased out when young adults turn 19 years old. Young adults are vulnerable due to illness, injury from accidents and consequences of unhealthy lifestyle and habits (MedicalBillingAndCoding.org 2017). Moreover, although young adults normally have better health relative to other age groups, they tend to need more mental health care and services (Antwi et al. 2015). The long term impacts of lacking health insurance are probably associated with deteriorated health status, attending fewer school days, less access to health care and utilization and higher financial burden in case of serious illness or injury (Monheit et al. 2011). With the purpose of increasing health insurance coverage for this vulnerable group, the dependent coverage mandate was signed into law by the President Barack Obama as an early implementation of the Affordable Care Act (ACA) in March 2010 (Cantor et al. 2012, Obama 2016). Formally, this provision allows young adults aged to remain on their parent s insurance plan. The purpose of the federal dependent coverage mandate is not only to increase the coverage for young adults, but also to help the transition to college or to allow young adults to spend time searching for a job after graduating from high school. Current research topics related to the ACA go beyond insurance takeup and healthcare utilization (Barbaresco, Courtemanche, and Qi 2015, Cantor et al. 2012, Antwi et al. 2015). Recently, researchers are more interested in measuring the policy outcomes from a socioeconomic perspective (Depew and Bailey 2015, Bailey and Chorniy 2016). For example, Depew and Bailey (2015) discuss how insurance premiums are reallocated as a result of the coverage expansion of young adults given that insurance premiums would increase if a young adult is added to the existing family insurance plan. They find that no more than 20% of premiums transfered to employees by employers. Bailey and Chorniy (2016) focus on examining the impacts of the ACA insurance mandate on job mobility to investigate whether young adults are reluctant to change jobs for the sake of health insurance after the ACA reform. Their results suggest that there is no evidence of job lock among young adults. Several studies discuss the ACA Medicaid expansion and its impacts on work effort and labor supply (Kaestner et al. 2017, Bowen and Kaestner 2014, Garrett, Kaestner, and Gangopadhyaya 2016). These studies show very little impact of the ACA Medicaid expansion on work effort or employment. 3

4 Clearly, a lot of attention has been given to labor force activities such as job lock and work effort. There is yet sparse research discussing the effects of the ACA mandate on education (i.e. college enrollment). There are two related studies exploring the relationship between health insurance provision and college enrollment before the ACA mandate (Jung, Hall, and Rhoads 2013, Yaskewich 2015). Specifically, Jung, Hall, and Rhoads (2013) investigate the impacts of health insurance provision through parents on the young adults decisions to attend college as full-time students exploiting state insurance mandates. Their results indicate that access to a dependent insurance plan through a parent increases the probability of enrolling as a full-time student by 5.5%. Yaskewich (2015) studies the case of Pennsylvania and New Jersey to examine the effects of state dependent insurance mandates on college enrollment. He argues that these two cities constitute a good control-treatment group comparison given the fact that New Jersey started the insurance mandate in 2006 while at the same time the neighboring state of Pennsylvania did not implement similar policies. The results show that the insurance mandate in New Jersey significantly decreases college enrollment relative to Pennsylvania. However, a possible concern in terms of the comparability of these two neighboring cities is a lack of a parallel assumption check, the external validity and implications of the results for the ACA reform. It is well known that over the past decade colleges have increased tuition and fees (Schoen 2015, Douglas-Gabriel 2016). One possible avenue for obtaining health insurance coverage for young students in college is through their parent s employer sponsored insurance plans (GAO 2008). Relative to private insurance, the advantages of being covered through a parent s employer based insurance plans is the lower cost from a family perspective and access to better coverage (MedicalBillingAndCoding.org 2017). Economically, we assume that insurance is taken into account as one consideration along with other factors in the utility function when young adults decide whether to attend college or not. With a reduction in the cost, the same utility could be achieved with a lower level of the budget. It is reasonable to expect an increase in college enrollment after the implementation of the ACA dependent coverage mandate. Building upon previous research efforts, we intend to answer the question of how the ACA dependent coverage mandate affects college enrollment for young adults. Earning a college degree is important for human capital accumulation, so it is necessary to understand how young adults make decisions conditional on the availability of health insurance. We exploit the ACA dependent insurance mandate as a natural experiment and employ a difference-in-differences (DID) strategy for the analysis. The clear policy context (i.e. the unique eligibility and policy start time) provides a plausible research design for the DID method. The results suggest that the ACA insurance mandate is associated with increases in the likelihood 4

5 of enrolling in college. Specifically, it increases college enrollment by 0.8 percentage points at the enactment stage and 0.9 percentage points at the implementation stage, respectively. The results are robust to different model specifications. Furthermore, we study the effects of the dependent health insurance on young adults decision to register as a full-time student and compare the effects before and after the ACA mandate. We find that young adults with dependent insurance have a higher probability to register as fulltime students after the ACA mandate. To the best of the authors knowledge, this is the first study evaluating the ACA dependent coverage mandate on educational outcomes. Understanding the ACA policy effects helps fine tune future expansions. The rest of the paper proceeds as follows: Section 2 provides a review of the policy background. Section 3 outlines one potential mechanism for insurance coverage provision to influence college enrollment. Section 4 describes the data and the empirical strategy. Section 5 presents the results and falsification tests. Section 6 concludes. 2. Policy Background The legislation of the ACA dependent insurance mandate was passed to reduce the gap of health insurance coverage during the transitional stage from high school to college or the first job for young adults. Young adults are allowed to enroll in their parent s insurance plan 30 days after the new policy year effective from September 23, Some insurers and employers adopted the new policy immediately after the passage of the law; while others waited until the new start of the next plan year. It was estimated that around 1.2 million eligible young adults were covered under the ACA mandate in 2010 with around half of them being previously uninsured (Collins and Nicholson 2010, Bulletin 2010). One year after the ACA mandate implementation, the statistics released by the National Center for Health Statistics (NCHS) show that there are over 2.5 million young adults aged covered by their parent s health insurance plans from September 2010 to June 2011 (Goldman 2013). At a later time, the commonwealth fund report mentions that from March 2012 to March 2013 about 15 million young adults have been covered by their parent s health insurance plan with about half becoming eligible as a result of the ACA mandate (Collins et al. 2014, Green-Weir and Stevenson 2014). The ACA dependent coverage mandate was built based on the provisions and implementation of previous state mandates from the mid-1990s; however, it sought to further reduce uninsured rates among young adults. There are several differences between state dependent insurance mandates and the ACA dependent coverage mandate. First, under the ACA mandate, full-insured plans (i.e. health insurance 5

6 purchased by employers for employees from an insurance company), self-insured plans (i.e. direct pay benefits to employees from employers), and privately purchased insurance plans of parents are all included (Collins and Nicholson 2010, Cantor et al. 2012). Self-insured and privately purchased insurance plans are usually exempted from state mandates. Prior to the ACA mandate, over half (around 55%) of the employer sponsored insurance plans were self-insured plan. Thus, the ACA mandate leads to a big expansion on eligibility of parent s employer sponsored insurance plans. Second, the ACA dependent coverage mandate has more flexible eligibility criteria compared to the strict eligibility requirements of state mandates. The ACA mandate eligibility only depends on the age of the individual rather than their residence status, student status, marital status, living arrangements or financial dependency (Green-Weir and Stevenson 2014). Given this, more young people would be covered and potentially benefit from the ACA provision. 1 Third, prior to the ACA mandate, only health insurance contributions of dependents younger than 19 or full-time students younger than 24 could be exempted from taxable income. After 2010, the tax exemption was extended to health insurance contributions to dependents younger than 27 (Antwi, Moriya, and Simon 2013, IRS 2010). 3. One Potential Mechanism for Insurance Coverage Provision to Influence College Enrollment We pay special attention to the question of whether the ACA mandate affects the college enrollment decisions of young adults or not. However, it is not clear how the ACA mandate reshapes the educational decisions of young people. The Government Accountability Office report reveals that dependent health insurance represents an important source of insurance provision for college students (GAO 2008). For example, about 67% of college students were covered through dependent insurance plans in Given the large increase in the number of eligible young adults as a result of the ACA mandate, we expect a potential increase in college enrollment. We use a simple utility function to demonstrate potential considerations in the individual decision-making process for attending college. Suppose the utility function takes the form U = (Cost of attending, School quality, Program variety, Job expectaion, Location, Big cities around ), constrained by financial resources: family disposable income per young adult. Among these considerations, the cost of attending school is important and related to the budget line. 1 More details of state mandates can be found in Levine, McKnight, and Heep (2011), Depew (2015), Amuedo-Dorantes and Yaya (2016), Legislatures (2016) and Initiatives (2007). 6

7 Any policies that influence the cost of education affect people s decisions to attend college. Previous literature discusses policies and programs influencing marginal benefits and the cost of obtaining a college education, such as the Pell Grant (Seftor and Turner 2002, Rubin 2011, Kane 2003), financial aid and assistance programs (Cornwell, Mustard, and Sridhar 2006, Dynarski 2003, Kane 2007) and insurance provision (Cohodes et al. 2016, Miller and Wherry 2016). These studies provide evidence that increasing exposure to the information of assistantships and expanding eligibility of assistantship programs improves college enrollment. We assume health insurance works in a similar way to financial aid and assistantship programs by reducing the cost of the family to attend college. 4. Empirical Strategy and Survey Data 4.1 Models and data used for the analysis To examine the effects of the ACA dependent insurance mandate, we employ a DID research design by comparing eligible and non-eligible individuals before and after the reform. One primary data source is the American Community Survey (ACS). Previous related research uses Survey of Income and Program Participation (SIPP) data (Jung, Hall, and Rhoads 2013). Compared to the SIPP, the benefit of using the ACS data lies in its larger sample size on an annual basis from 2006 to The ACS is a nationally representative survey conducted annually by the U.S. Census Bureau, which covers questions related to demographics, education, employment, housing, health insurance, etc. For example, respondents report their highest educational attainment and whether they are attending college when the interview is conducted. We combine these two questions to construct an indicator variable for college enrollment. This indicator variable helps us to obtain an accurate estimate and minimize specification bias. We keep observations from 2006 to 2015 to make sure that there are enough observations before and after the policy enactment date. In order to fully capture the effects of different stages of the reform, we divide the timeline into a pre-reform and post-reform stage, with the 2011 survey year (2010 calendar year) as the cutoff. Therefore, the pre-reform period covers survey year ( calendar year). 2 Considering the potential differential impacts of the reform, we further separate the post-reform 2 The survey year means the year of the survey being conducted. The survey year data reflects the status of respondents one year before, which is the calendar year. 7

8 period into the enactment stage, 2011 survey year (2010 calendar year) and the implementation stage, from 2012 to 2016 survey year ( calendar year). The model used is: y iast = b 0 + b 1 Enact t + b 2 Implement t + b 3 Eligibility a + b 4 (Enact t Eligibility a ) + b 5 (Implement t Eligibility a ) + b 6 X iast + b 7 Unemployment st + b 8 AgespecificUnemployment ast + state s + year t + ε iast, (1) where the outcome variable y iast is college enrollment, defined as one if individual i in the eligibility group a is enrolled in college in state s from year t when the interview is conducted and already graduated from high school; Enact t is a dummy variable for the initial post-reform stage, defined as one if the survey year equals 2011 and zero otherwise; Implement t is a dummy variable for the full implementation of the post-reform stage, defined as one if the survey year falls into and zero otherwise; Eligibility a is the age eligibility dummy variable, defined as one for the age group and zero for the age group 26-30; x iast is a set of individual control variables including age, gender, marital status, race, ethnicity, high school graduation, employment, number of siblings and poverty level. 3 The demographic variables are constructed directly using the ACS survey questions. Due to the fact that the analysis spans the Great Recession during the late 2000s and early 2010s, we control for the impacts from the adverse macroeconomic environment by adding state-year unemployment rates (Antwi et al. 2015). In addition, we incorporate age-specific state-year unemployment rates to account for the confounding effects of macroeconomic conditions which potentially differ by age groups. The state-year unemployment rates are obtained from the Bureau of Labor Statistics (BLS) and merged with the main ACS dataset by the state and year identifier. State and year fixed effects are also accounted for in the model. The standard errors are corrected for state-level clustering (Bertrand, Duflo, and Mullainathan 2004) Parallel trend assumption In order to validate the DID method, we first show evidence for the parallel assumption by presenting college enrollment graphically between the control and treatment groups before and after the reform in Figure 1. A visual inspection suggests that before 2011, the lines of college enrollment for the treatment group and control groups are parallel. The two groups followed the same trend/pattern prior to the reform. Starting around 2011, the college enrollment of the treatment group shows an upward trend. Comparatively, the college enrollment of the control group remains steady relative to the pre-reform stage. 3 Poverty level is calculated as the ratio of total family income to the poverty threshold by family size. 8

9 4.1.2 First stage estimation of the effects of the ACA reform on dependent health insurance take-up In this section, we present the analysis of the effects of the ACA mandate on dependent health insurance coverage for targeted young adults as the first stage evidence. The underlying assumption is that the effects on college enrollments for young adults work because after the reform more young adults enroll in their parent s insurance plan. The ACS survey data include health insurance information since 2008, but it does not specify the source of insurance as detailed as in the Current population survey (CPS). Thus, we use the Annual Social and Economic Supplement (ASEC) of the CPS for this part of the analysis. The ASEC survey data is collected annually and has rich information regarding family income, poverty, employment and health insurance. Specifically, health insurance is characterized by insurance sources and types. For example, respondents report whether they purchased a private insurance plan, or whether they were covered by employment based insurance, military insurance, or as a dependent of a private insurance plan, or as a dependent of an employment based insurance plan. We follow the same method used for the main analysis, and revise the model accordingly. One major change is that we do not differentiate between the enactment and implementation stages as in the main analysis. We believe the dependent insurance enrollment would increase after the policy enactment, but it is not necessary to differentiate between these two stages. As a robustness check, we also estimate the model differentiating between the two stages as in equation 1. The results are consistent and discussed in Section 4.1. y iast = b 0 + b 1 Eligibility a + b 2 Post reform t + b 3 (Eligibility a Post reform t ) + b 4 X iast + b 5 Unemployment st + b 6 AgespecificUnemployment ast + state s + year t + ε iast, (2) where the outcome variable y iast is the dependent insurance enrollment dummy, defined as one if individual i in the eligibility group a is enrolled in his/her parent s insurance plan in state s from year t, including both private insurance plan and employer sponsored plan. For explanatory variables, Eligibility a is the age eligibility dummy variable, defined as one for the age group and zero for the age group 26-30; Post reform t is a time dummy variable, defined as one if the survey year falls into and zero otherwise; x iast is a set of individual and family control variables, which include age and dummies for gender, marital status, race, ethnicity, employment, poverty level, family size, parent s age, parent s education and number of siblings. For the same concerns of the macroeconomic 9

10 environment, we also include in the model state-year unemployment rates and age-specific state-year unemployment rates. State and year fixed effects are also controlled for in the model. The standard errors are corrected for state-level clustering (Bertrand, Duflo, and Mullainathan 2004). 4.2 Heckman two-stage model Given that full-time students have a higher probability of graduating and more opportunities to attend campus activities and interact with faculty (Waiwaiole 2017a, b), we are also interested in understanding the effects of parental insurance coverage provision on decisions to enroll as full-time students. Similar to the case discussed in the original work in Heckman (1979), we assume that there are two steps when young adults make the decision to enroll as a full-time student (Jung, Hall, and Rhoads 2013). First, they choose whether to attend college or not. Conditioning on the first stage decision, they decide to enroll as a full-time student or a part time student. We can only observe individuals in the second stage if they choose to attend college in the first stage. The first step is the selection process, and the Heckman selection model is introduced to control for the potential selection bias if two error terms in the selection equation and outcome equation are correlated. Otherwise, the two stage equations could be estimated separately. Specifically, in this study, we include dependent insurance, age, age squared, gender, race, ethnicity, employment, family size, number of siblings, parent s age, parent s education, parent s insurance types and family income as control variables in the selection process. For the second stage, a set of control variables including dependent insurance, age, age squared, gender, race, ethnicity, family size, number of siblings, parent s age, parent s education, and family income constitute explanatory variables. Again, we use the ASEC data from the CPS for the analysis. It is worth noting that ASEC data has information of the family unit and interrelations within a family. Therefore, we fully utilize the data structure and arrange the information of individual young adults with the information of their parents and siblings. 5. Results and Discussion 5.1 Results of first stage estimation of the effects of the ACA mandate on dependent health insurance take-up The final sample of the ASEC used for the analysis includes 121,355 observations with an average age of 23. Females account for 46.7% of the sample. The majority of the sample are non-hispanic white individuals (52.29%), followed by Hispanics (22.94%) and non-hispanic black individuals (13.69%). We 10

11 present the summary statistics of the treatment group and control group before and after the reform in Table 1. The demographic and racial/ethnic compositions of each group remain similar before and after the reform. The parent s insurance information is summarized in the lower part of the table. The changes for each insurance type between the control and treatment group follow the same trend. After the reform, the percentage of parents purchasing private insurance plans increases, but the percentage of parents having a group insurance plan decreases. Meanwhile, the percentage of parents having a family plan via private insurance increases, but the percentage of parents having a family plan via group insurance decreases. The estimates of the effects on dependent insurance take-up are shown in Table 2. We estimate three models sequentially by adding more control variables for robustness. The results, shown in columns 2-4, are robust to different model specifications. With the full specification (shown in column 4), we find that the ACA dependent insurance mandate increases the dependent insurance coverage by 3.7 percentage points. We also employ the same method of separating the post reform stage into enactment and implementation and re-estimate the model. The results are shown in Appendix TableA1 and suggest strong positive effects on dependent insurance take-up at the implementation stage (i.e. 4.1 percentage points with the full model specification). 5.2 Main results of the effects of the ACA mandate on college enrollment using the ACS data The final sample from the ACS data used in the analysis consists of 4,370,786 observations. The average age of the sample is 25 years. Approximately half (49.5%) of the sample are females. There is a high proportion of non-hispanic white individuals (61.75%), followed by Hispanics (17.65%) and non- Hispanic black individuals (11.58%) in the sample. We separate the sample into the treatment group (aged 19-25) and control group (aged 26-30), and present the summary statistics of the two groups before and after the ACA policy in Table 3. For each group, the demographic and socioeconomic compositions before and after the reform are fairly consistent. There is a higher proportion of married individuals and more uniformly distributed poverty status for the treatment group; intuitively it is expected that as respondents get older, they are more likely to be married and to have higher earnings. The similarities in the demographic and socioeconomic compositions before and after the reform lay a good foundation for the DID method. The rate of college enrollment for both groups is shown at the bottom of Table 3. We conduct a Chi-squared test to examine the equality of college enrollment before and after the reform. The test statistic shows that the percentage of college enrollment 11

12 for the treatment group is significantly different (P<0.001) before and after the reform; meanwhile the percentage of college enrollment for the control group is not statistically different (P=0.815) before and after the reform. The estimates of the effects of the ACA mandate on college enrollment are shown in Table 4. The coefficients of interest are shown in the upper part of the table. The results indicate a positive effect of the ACA mandate on college enrollment following the policy enactment with the effect carrying through the implementation stage. More specifically, the ACA dependent coverage mandate increases college enrollment by 0.8 percentage points during the enactment stage, and by 0.9 percentage points during the implementation stage. We also consider that the effects associated with the reform might be a result of specific impacts to a subpopulation, so further estimation based on certain subgroups might deliver some insights. To explore potential heterogeneity in the policy impacts, we estimate the model in equation (1) using subsamples by gender, race and age groups. We demonstrate the validity of analyzing subsamples by showing the parallel trend of these subgroups in Figure 2. Clearly, the parallel trend check suggests that comparisons between gender, race, and age groups are valid. The results of heterogeneous effects from subgroups are shown in Table 5. Compared to males, there are strong positive effects at the enactment and implementation stages for females. For the two racial groups, there are larger effects at both stages for non-hispanic black individuals compared to non- Hispanic white individuals. We further separate the age treatment group into and and compare these two age groups to the control group respectively. We find that for the age group, the ACA policy increases college enrollment by 2.5 and 2.8 percentage points at the enactment and implementation stage respectively. However, the effect for the treatment age group is only marginally significant at the implementation stage by 0.3 percentage points. 5.3 Results from the Heckman two-stage model We estimate the Heckman two-stage model with subsamples of survey years and survey years , which allows us to observe differences in the effects before and after the ACA reform. The results are presented in Table 6. The Wald tests of independent equations shown at the bottom of Table 6 justify the Hackman two-stage process in this study (P<0.05). Columns 2-3 and 4-5 show the pre-reform and post-reform estimates respectively. Both coefficient estimates and marginal effects (dy dx) are shown in the table. We find that young adults are 0.6% more likely to enroll as a full-time student if they have 12

13 dependent insurance coverage after the ACA dependent coverage mandate. There is no significant effect of dependent insurance coverage on full-time student registration before the ACA reform. In addition, non-hispanic white individuals are 0.9% more likely to enroll as a full-time student. Young adults from large families are 0.4% less likely to enroll as a full-time student and this relationship exists before the ACA reform as well. Further, we also find that young adults with high family income are 0.9% more likely to enroll as a full-time student. 5.4 Falsifications tests Some states had already adopted a statewide young dependent insurance mandate before the implementation of the ACA mandate, but we do not differentiate between states with and without previous adoption. A falsification test is conducted to investigate the possibility that the effects we find in the main analysis come from the states with previous mandates before the ACA policy. Therefore, we estimate the model in equation (1) with two subgroups based on previous state mandate adoption. The results are shown in Table 7. There are significant positive effects of the ACA dependent coverage mandate on college enrollment in the states with previous state mandates before the ACA (i.e. 0.9 percentage points at the enactment stage and 0.9 percentage points at the implementation stage). We also find a significant positive effect of the ACA mandate on college enrollment in the states without previous state mandate (i.e. 0.8 percentage points at the implementation stage). It is possible that residents living in states with previous mandates before the ACA have higher awareness of the ACA policy and lower information cost to learn the new stipulation. Thus, we can see the effects at the enactment and implementation stages for states with previous mandates before the ACA, but only at the implementation stage for states without previous mandates before the ACA. Moreover, we conduct another falsification test by changing the treatment age from to 26-28, which is actually part of the control age group in the above analysis. The new control group for this analysis only consists of individuals aged By assigning the new treatment age group and reestimating the model, we do not expect to see any effects from the ACA policy due to the nature of the treatment group. The results indicate there are no significant effects of the ACA mandate on college enrollment for the false treatment age group (Table 8). 6. Conclusions This study evaluates the effects of the ACA dependent coverage mandate on college enrollment taking advantage of a natural experiment context. Our findings suggest a positive effect of dependent health 13

14 insurance coverage on college enrollment and document heterogeneous effects by gender, race and age groups. We further show that young adults with dependent insurance coverage are more likely to enroll as full-time college students. Both the government and individuals benefit from the ACA dependent mandate as extending the health insurance coverage to more young dependents. Specifically, for the government, higher insurance coverage rates for the target population come from relatively low public spending (Antwi, Moriya, and Simon 2013, Jensen and Morrisey 1999). For individuals, insurance coverage is associated with access to health care and utilization and prevention of potential financial losses. Moreover, considering the human capital accumulation from attending college and the future benefits in terms of job security and wellbeing, as well as the advantages of registering as full-time students, a formal cost benefit analysis of the ACA dependent mandate is required for the next step in research. 14

15 References Amuedo-Dorantes, Catalina, and Mehmet E Yaya "The Impact of the ACA's Extension of Coverage to Dependents on Young Adults Access to Care and Prescription Drugs." Southern Economic Journal 83 (1): Antwi, Yaa Akosa, Asako S Moriya, and Kosali Simon "Effects of Federal Policy to Insure Young Adults: Evidence from the 2010 Affordable Care Act's Dependent- Coverage Mandate." American Economic Journal: Economic Policy 5 (4):1-28. Antwi, Yaa Akosa, Asako S Moriya, Kosali Simon, and Benjamin D Sommers "Changes in Emergency Department Use among Young Adults after the Patient Protection and Affordable Care Act s Dependent Coverage Provision." Annals of Emergency Medicine 65 (6): e2. Bailey, James, and Anna Chorniy "Employer Provided Health Insurance and Job Mobility: Did the Affordable Care Act Reduce Job Lock?" Contemporary Economic Policy 34 (1): Barbaresco, Silvia, Charles J Courtemanche, and Yanling Qi "Impacts of the Affordable Care Act Dependent Coverage Provision on Health-Related Outcomes of Young Adults." Journal of Health Economics 40: Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan "How Much Should We Trust Differences-in-Differences Estimates?" Quarterly Journal of Economics 119 (1): Bowen, Garrett, and Robert Kaestner "The Best Evidence Suggests the Effects of the ACA on Employment Will Be Small." Urban Institute, Policy Briefs/Timely Analysis of Health Policy Issues. Bulletin, Internal Revenue Interim Final Rules for Group Health Plans and Health Insurance Issuers Relating to Dependent Coverage of Children to Age 26 under the Patient Protection and Affordable Care Act. Department of Treasury, Department of Labor, Department of Health and Human Services. Cantor, Joel C, Alan C Monheit, Derek DeLia, and Kristen Lloyd "Early Impact of the Affordable Care Act on Health Insurance Coverage of Young Adults." Health Services Research 47 (5): Cohodes, Sarah R, Daniel S Grossman, Samuel A Kleiner, and Michael F Lovenheim "The Effect of Child Health Insurance Access on Schooling: Evidence from Public Insurance Expansions." Journal of Human Resources 51 (3): Collins, S. R., P. W. Rasmussen, T. Garber, and M. M. Doty "Covering Young Adults Under the Affordable Care Act: The Importance of Outreach and Medicaid Expansion." The Commonwealth Fund, accessed May 15th. Young-Adults-Under-the-Affordable-Care-Act.aspx. Collins, Sara R., and Jennifer L. Nicholson Rite of Passage: Young Adults and the Affordable Care Act of The Commonwealth Fund. Cornwell, Christopher, David B Mustard, and Deepa J Sridhar "The Enrollment Effects of Merit-based Financial Aid: Evidence from Georgia s HOPE Program." Journal of Labor Economics 24 (4): Depew, Briggs "The Effect of State Dependent Mandate Laws on the Labor Supply Decisions of Young Adults." Journal of Health Economics 39:

16 Depew, Briggs, and James Bailey "Did the Affordable Care Act's Dependent Coverage Mandate Increase Premiums?" Journal of Health Economics 41:1-14. Douglas-Gabriel, Danielle "College Costs Rising Faster than Financial Aid." The Washington Post, accessed May Dynarski, Susan M "Does Aid Matter? Measuring the Effect of Student Aid on College Attendance and Completion." American Economic Review 93 (1): GAO Most College Students Are Covered through Employer-Sponsored Plans, and Some Colleges and States Are Taking Steps to Increase Coverage. Government Accountability Office. Garrett, A Bowen, Robert Kaestner, and Anuj Gangopadhyaya "Recent Evidence on the ACA and Employment: Has the ACA Been a Job Killer? 2016 Update." The Urban Institute, ACA Implementation Monitoring and Tracking, doi: Available at SSRN: Goldman, T.R "Progress Report: The Affordable Care Act s Extended Dependent Coverage Provision." Health Affairs Blog, May 15th. Green-Weir, Robbya R., and Tamara N. Stevenson "Healthy Decisions and Unintended Consequences: Implications of the Protection and Affordable Care Act (ACA) for Higher Education." In The Obama Administration and Educational Reform, Heckman, James J "Sample Selection Bias as A Specification Error." Econometrica 47 (1): Initiatives, State Coverage "States Enact New Laws to Increase Dependent Coverage." Robert Wood Johnson Foundation, accessed May IRS "Tax-Free Employer-Provided Health Coverage Now Available for Children under Age 27." Internal Revenue Service, accessed April Jensen, Gail A, and Michael A Morrisey "Employer Sponsored Health Insurance and Mandated Benefit Laws." The Milbank Quarterly 77 (4): Jung, Juergen, Diane M Harnek Hall, and Thomas Rhoads "Does the Availability of Parental Health Insurance Affect the College Enrollment Decision of Young Americans?" Economics of Education Review 32: Kaestner, Robert, Bowen Garrett, Jiajia Chen, Anuj Gangopadhyaya, and Caitlyn Fleming "Effects of ACA Medicaid Expansions on Health Insurance Coverage and Labor Supply." Journal of Policy Analysis and Management. doi: doi: /pam Kane, Thomas J "Evaluating the Impact of the DC Tuition Assistance Grant Program." Journal of Human resources 42 (3): Kane, Thomas J "A Quasi-Experimental Estimate of the Impact of Financial Aid on College-Going." National Bureau of Economic Research Working Paper Series No doi: /w9703. Legislatures, National Conference of State "Dependent Health Coverage and Age for Healthcare Benefits." accessed May

17 Levine, Phillip B, Robin McKnight, and Samantha Heep "How Effective are Public Policies to Increase Health Insurance Coverage Among Young Adults?" American Economic Journal: Economic Policy: MedicalBillingAndCoding.org "Everything You Need to Get Started in Medical Billing and Coding." accessed May Miller, Sarah, and Laura R. Wherry The Long-Term Effects of Early Life Medicaid Coverage. Monheit, Alan C, Joel C Cantor, Derek DeLia, and Dina Belloff "How Have State Policies to Expand Dependent Coverage Affected the Health Insurance Status of Young Adults?" Health Services Research 46 (1p2): Obama, Barack "United states Health Care Reform: Progress to Date and Next Steps." JAMA 316 (5): doi: /jama Rubin, Rachel B "The Pell and the Poor: A Regression-Discontinuity Analysis of On- Time College Enrollment." Research in Higher Education 52 (7): Schoen, John W "Why Does a College Degree Cost So Much?". CNBC, accessed May Seftor, Neil S., and Sarah E. Turner "Back to School: Federal Student Aid Policy and Adult College Enrollment." Journal of Human Resources 37 (2): Waiwaiole, Evelyn. 2017a. Full-Time Enrollment and Student Success. edited by Center for Community College Student Engagement: University of Texas at Austin. Waiwaiole, Evelyn. 2017b. "Students who go to college full-time are more likely to graduate." USATodayCollege, accessed May Yaskewich, David "Dependent Health Insurance Laws and College Enrollment: Is There Evidence of College Lock?" Journal of Family and Economic Issues 36 (4): doi: /s

18 Tables Table 1 Summary Statistics of the CPS Data Treatment group (aged 19-25) Control group (aged 26-30) Pre-reform Post-reform Pre-reform Post-reform Observations 47,841 50,431 10,499 12,584 Gender Average Age Female 46.86% 46.94% 45.95% 45.77% Race/Ethnicity non-hispanic Whites 57.53% 50.54% 46.85% 43.90% non-hispanic Blacks 12.46% 13.86% 15.98% 15.77% Hispanics 20.14% 24.85% 23.40% 25.52% Socioeconomic Status Marital status (married=1) 2.35% 2.09% 10.72% 9.54% High school graduation 87.79% 89.47% 87.20% 88.95% Unemployment 8.70% 9.53% 10.09% 10.43% Poverty (<150% FPL) 14.74% 18.56% 15.31% 16.51% Poverty ( % FPL) 24.43% 26.12% 25.28% 26.71% Poverty ( % FPL) 28.81% 26.87% 27.86% 26.67% Poverty (>500% FPL) 32.02% 28.45% 31.56% 30.12% Family Characteristics Father age Father education (HS graduation=1) 85.82% 85.82% 80.57% 82.92% Father private insurance 5.46% 8.38% 5.81% 9.05% Father group insurance 58.76% 53.44% 53.97% 48.68% Father private insurance type (family plan=1) 4.07% 6.63% 2.51% 4.58% Father group insurance type (family plan=1) 49.96% Mother age Mother education (HS graduation=1) 86.16% 86.46% 80.46% 84.38% Mother private insurance 3.85% 6.59% 4.93% 7.62% Mother group insurance 41.32% 38.19% 42.50% 40.30% Mother private insurance type (family plan=1) 2.31% 4.58% 1.26% 2.94% Mother group insurance type (family plan=1) 29.09% 28.56% 18.75% 19.51% 18

19 Table 2 Results of the First Stage Estimation of the ACA mandates on Dependent Health Insurance Take-up Treat effect 0.044*** 0.032*** 0.037*** (0.008) (0.011) (0.012) Post reform 0.151*** 0.168*** 0.159*** (0.010) (0.012) (0.011) Eligibility *** ** (0.009) (0.010) (0.027) Age *** *** *** (0.002) (0.002) (0.002) Gender (female=1) 0.040*** 0.042*** 0.042*** (0.004) (0.004) (0.004) Race (white=1) 0.057*** 0.047*** 0.046*** (0.009) (0.010) (0.010) Race (Blacks=1) (0.010) (0.011) (0.010) Ethnicity (Hispanics=1) *** *** *** (0.008) (0.007) (0.007) Martial (married=1) *** 0.051*** (0.009) (0.008) (0.008) Employment *** *** *** (0.004) (0.005) (0.005) Family size 0.018*** *** *** (0.002) (0.006) (0.006) Number of siblings 0.014*** 0.036*** 0.034*** (0.003) (0.007) (0.007) Poverty ( % FPL) 0.184*** 0.128*** 0.128*** (0.007) (0.010) (0.010) Poverty ( % FPL) 0.337*** 0.217*** 0.219*** (0.008) (0.010) (0.009) Poverty (>500% FPL) 0.426*** 0.232*** 0.232*** (0.009) (0.011) (0.011) Father age 0.002*** 0.002*** (0.000) (0.000) Father education (HS graduation=1) 0.051*** 0.052*** (0.005) (0.005) Mother age (0.000) (0.000) Mother education (HS graduation=1) 0.083*** 0.082*** (0.005) (0.005) Father insurance contribution 0.000*** 0.000*** (0.000) (0.000) State-year unemployment rates (0.003) Age specific state-year * unemployment rates (0.003) State and year fixed effect Yes Yes Yes Family characteristics Yes Yes State and age-specific unemployment rates Yes N 81,039 56,688 47,681 R-squared Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

20 Table 3 Summary Statistics of the ACS Data Treatment group (aged 19-25) Control group (aged 26-30) Pre-reform Post-reform Pre-reform Post-reform Observations 1,230,556 1,365, , ,810 Average age Gender Female 49.15% 48.69% 50.73% 49.78% Race/Ethnicity Non-Hispanic Whites 63.17% 59.47% 63.53% 61.59% Non-Hispanic Blacks 11.48% 12.67% 10.54% 11.07% Hispanics 17.25% 18.12% 17.53% 17.60% Socioeconomic Status Marital status (married=1) 15.29% 11.42% 46.83% 40.46% High school graduation 75.39% 89.98% 78.03% 89.48% Unemployment 9.76% 9.69% 6.48% 6.80% Poverty (<150% FPL) 36.36% 44.67% 23.91% 27.91% Poverty ( % FPL) 25.08% 23.24% 26.59% 26.71% Poverty ( % FPL) 21.12% 17.64% 26.19% 24.15% Poverty (>500% FPL) 17.43% 14.45% 23.31% 31.13% College enrollment 40.50% 43.35% 9.7% 9.81% 20

21 Table 4 Main Results of the Effects of the ACA mandate on College Enrollment Enact * Eligibility 0.008* (0.004) Implementation* Eligibility 0.009*** (0.002) Enact (0.004) Implement *** (0.002) Eligibility *** (0.008) Age *** (0.002) Gender (female=1) 0.053*** (0.002) Race (non-hispanic White=1) *** (0.005) Race (non-hispanic Black=1) *** Ethnicity (Hispanics=1) *** (0.005) Number of siblings 0.011*** (0.001) Martial (married=1) *** (0.003) Employment *** (0.006) Poverty ( % FPL) *** (0.010) Poverty ( % FPL) *** (0.011) Poverty (>500% FPL) *** (0.010) State-year unemployment rates (0.001) Age specific state-year unemployment rates (0.001) State and year fixed effect Yes N 3,065,262 R-squared Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

22 Table 5 Heterogeneous Effects of the ACA mandate on College Enrollment Non- Hispanic Whites Non- Hispanic Blacks Age Age Male Female Enact* Eligibility ** ** 0.025*** (0.005) (0.005) (0.005) (0.006) (0.007) (0.004) Implementation* Eligibility 0.006*** 0.012*** *** 0.028*** 0.003* (0.002) (0.002) (0.002) (0.005) (0.003) (0.002) State and year fixed effect Yes Yes Yes Yes Yes Yes Individual control variables Yes Yes Yes Yes Yes Yes N 1,559,587 1,505,675 2,030, ,012 1,795,141 2,609,889 R-squared Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

23 Table 6 Estimates of the effects of parental insurance plan provision on full-time student registration from the Heckman Two-Stage Model Pre-Policy Coefficient estimate Marginal ( y x) Post-Policy Coefficient estimate Marginal ( y x) ACA Dependent * 0.006* insurance mandate (0.081) (0.005) (0.054) (0.003) Age (0.491) (0.027) (0.322) (0.019) Age-squared (0.012) (0.001) (0.008) (0.000) Gender (female=1) (0.049) (0.002) (0.044) (0.003) Race ** 0.009** (non-hispanic Whites=1) (0.086) (0.004) (0.066) (0.004) Race (non-hispanic Blacks=1) (0.11) (0.006) (0.091) (0.005) Ethnicity (Hispanics=1) * * (0.094) (0.005) (0.075) (0.005) Family size *** ** ** * (0.044) (0.003) (0.033) 0.002) Number of siblings 0.162*** 0.008** (0.052) (0.004) (0.041) (0.003) Household income 0.154*** 0.007*** 0.151*** 0.009*** (0.029) (0.002) (0.03) (0.002) Father age (0.004) (0.000) (0.003) (0.000) Father education (HS graduation=1) (0.064) (0.003) (0.053) (0.003) Mother age (0.004) (0.000) (0.003) (0.000) Mother education (HS graduation=1) (0.061) (0.003) (0.054) (0.003) N 23,813 24,296 Wald Test (P value) Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

24 Table 7 Falsification Test 1: Estimates by Separating the States into without or with Previous Insurance Mandates Without Previous Mandates With Previous Mandates Enact * Eligibility ** (0.007) (0.004) Implementation* Eligibility 0.008*** 0.009*** (0.002) (0.002) Enact (0.007) (0.005) Implement *** *** (0.003) (0.002) Eligibility *** *** (0.011) (0.012) Age *** *** (0.002) (0.003) Gender (female=1) 0.059*** 0.048*** (0.003) (0.003) Race (non-hispanic White=1) * *** (0.010) (0.004) Race (non-hispanic Black=1) *** (0.009) (0.005) Ethnicity (Hispanics=1) *** *** (0.005) (0.006) Number of siblings 0.011*** 0.012*** (0.002) (0.001) Martial (married=1) *** *** (0.002) (0.005) Employment *** *** (0.009) (0.005) Poverty ( % FPL) *** *** (0.011) (0.013) Poverty ( % FPL) *** *** (0.011) (0.013) Poverty (>500% FPL) *** *** (0.008) (0.013) State-year unemployment rates (0.002) (0.001) Age specific state-year unemployment 0.003* rates (0.002) (0.001) State and year fixed effect Yes Yes N 1,117,242 1,948,020 R-squared Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

25 Table 8 Falsification Test 2: Estimates by Changing the Treatment Age Group to Enact * Eligibility (0.002) Implementation* Eligibility (0.001) Enact 0.008** (0.004) Implement *** Eligibility (0.004) (0.003) Age *** (0.001) Gender (female=1) 0.020*** (0.001) Race (non-hispanic White=1) ** (0.002) Race (non-hispanic Black=1) 0.039*** (0.003) Ethnicity (Hispanics=1) 0.003* (0.002) Number of siblings 0.011*** (0.001) Martial (married=1) *** (0.001) Employment 0.017*** (0.003) Poverty ( % FPL) *** (0.003) Poverty ( % FPL) *** (0.003) Poverty (>500% FPL) *** (0.003) State-year unemployment rates (0.001) Age specific state-year unemployment rates (0.000) State and year fixed effect Yes N 1,075,821 R-squared Notes: Numbers in parentheses are standard errors. * p<0.1, ** p<0.05, *** p<

26 Figures Figure 1 Parallel Assumption Check of College Enrollments from 2006 to

27 Figure 2 Parallel Assumption Check of College Enrollments for Subgroups by Gender, Race and Two Age Treatment Groups 27