Abstract
This paper investigates the effect of the Affordable Care Act young adult provision on the propensity to marry. The young adult provision expanded options for obtaining insurance coverage outside of marriage. Young adults affected by the provision might have less incentive to marry since one avenue for obtaining health insurance coverage is through marriage. This paper examines this question by applying difference-in-differences-type methods using the 2008–2013 American Community Survey. Results suggest that the provision is associated with decreases in the likelihood of marrying, cohabitation, and spousal health insurance coverage and an increase in the probability of divorce.
I. Introduction
In 2010, Congress enacted the Patient Protection and Affordable Care Act (ACA) young adult provision to address the low health insurance rates of young adults. This provision allowed 19–25-year-olds to be covered by their parents’ private health insurance plans, regardless of marital status, student status, and whether they have children. Prior to the implementation of the provision, young adults who were not enrolled in school generally became ineligible for coverage under their parents’ private health insurance plans when they turned 19 years old while students generally aged out when they turned 24 years old. Several papers find that the provision was effective in increasing health insurance coverage of young adults (Cantor et al. 2012; Sommers and Kronick 2012; Akosa Antwi, Moriya, and Simon 2013; Mulcahy et al. 2013; O’Hara and Brault 2013; Sommers et al. 2013; Barbaresco, Courtemanche, and Qi 2015). Other research finds effects on noninsurance outcomes. For example, there is suggestive evidence that the provision decreased labor market participation (Akosa Antwi, Moriya, and Simon 2013), improved health outcomes (Barbaresco, Courtemanche, and Qi 2015), and expanded access to care (Sommers et al. 2013; Akosa Antwi, Moriya, and Simon 2015). This paper considers how the ACA young adult provision influences the decision to marry.
An extensive literature has documented that policies that affect the costs and benefits of marriage do influence individuals’ marriage choices and family structure. Some of these include tax and safety net policies such as progressive tax programs based on family income (Alm and Wittington 1995, 1997, 1999), the Aid to Families with Dependent Children program (Moffitt 1990; Moffitt, Reville, and Winker 1998), and Medicaid (Decker 2000; Yelowitz 1998). A related literature has established that marriage is affected by health-related policies including the legalization of abortion and the increased availability of oral contraception (Akerlof, Yellen, and Katz 1996; Goldin and Katz 2002; Myers 2012) and state health insurance mandates for coverage of assisted reproductive technology (Abramowitz 2014, 2016).
Recent survey and anecdotal evidence suggests a relationship between health insurance coverage and marriage. In April 2008, the Kaiser Family Foundation conducted the Election 2008 Kaiser Health Tracking Survey and reported that 7 percent of respondents affirmed that in the last year they or someone in their household decided to get married mainly so one spouse could have access to the other’s health care benefits (Kaiser Family Foundation 2008b). While a later discussion considered the plausibility of that statistic, the figure suggests that Americans perceive that people they know are making major life decisions partly based on health care concerns (Kaiser Family Foundation 2008a). Media reporting of the survey results reinforced this message with pieces such as, “I Married for Health Insurance,” (Goodman 2008) amongst others (for example, Goldstein 2008; Sack 2008), and coverage of one woman’s YouTube campaign to find a husband to obtain health insurance: “She’s searching for ‘the one’ with the lowest co-pays,” (CBS News 2010).
There is reason to believe that the prospect of obtaining health insurance coverage could be an important consideration for individuals’ decisions about whether to marry. This is because one possible avenue for obtaining health insurance is to marry and become a dependent on a spouse’s insurance plan.1 This may not be an important channel for individuals who obtain their own coverage through their employers or government programs like Medicaid and Medicare. However, many individuals are ineligible to obtain coverage through these sources. These individuals may purchase a suboptimal amount of coverage on their own, and the prospect of obtaining better coverage may induce them to marry when they otherwise would not. This may come about through single individuals searching more intensively for a spouse or coupled but unmarried individuals deciding to marry. If young adults consider health insurance coverage in their marriage decisions, then the ACA provision’s increased access to insurance outside of marriage may result in a decrease in young adults’ probability of marrying.
Although the ACA young adult provision may directly affect young adults’ decisions to marry by facilitating the obtainment of health insurance coverage outside of marriage, the provision also may indirectly affect marriage outcomes. For example, young adults may opt for obtaining coverage through parents rather than their own employers. As a result, individuals may choose to work fewer hours or not at all, may choose different types or locations of jobs, may accept different wages, or may choose to enroll in school and invest in human capital. These outcomes could in turn affect their propensities to marry. Indeed, Akosa Antwi, Moriya, and Simon (2013) find suggestive evidence that the ACA young adult provision led young adults to reduce their working hours and be less likely to work full-time.
This paper considers the relationship between the ACA young adult provision and marriage using pooled 2008 through 2013 one-year American Community Survey (ACS) data. I perform difference-in-differences-type analyses following the approach outlined by Hahn and Yang (2013) and Akosa Antwi, Moriya, and Simon (2013). This approach exploits variation across age groups and over time. This analysis primarily examines the effect of the provision on the likelihood of an individual to marry. I further consider heterogeneous effects of the provision by student status and sex and estimate effects of the provision on outcomes related to marriage including cohabitation and divorce. To consider potential channels driving effects of the provision on marriage, the paper further examines the effect of the provision on spousal health insurance coverage using the 2008 Survey of Income and Program Participation (SIPP) Panel.
The findings of the analysis suggest the enactment and implementation of the ACA young adult provision are associated with decreases in the likelihood of marrying for young adults at ages covered by the provision. Specifically, provision enactment and implementation are associated with decreases in the probability of marrying of 0.53 percentage points and 0.56 percentage points, respectively. These correspond to 8.8 and 9.3 percent decreases in marriage rates as compared to before provision enactment. Results are robust to the choice of treatment and control groups of alternative ages, addressing Slusky’s (2014) concerns that the definition of the treatment and control groups by particular age groups might be driving the results, as well as the inclusion of various controls for economic conditions. The results also show an increase in the probability of divorce and decreases in cohabitation and spousal health insurance coverage associated with the young adult provision. These findings are consistent with the proposition that some individuals marry in order to obtain coverage and that the provision serves to correct inefficiencies in the insurance and marriage markets.
The rest of the paper is organized as follows. The next sections provide background on the ACA young adult provision and offer a theoretical framework illustrating the relationship between insurance coverage and marriage. Subsequent sections outline the data and methods used in the paper’s main analysis of marriage rates and marriage-related outcomes, present findings of this analysis, and consider their robustness. The following section considers the effect of the provision on rates of spousal health insurance coverage. The last section discusses these findings and concludes.
II. Background on the ACA Young Adult Provision
In order to address the low health insurance rates of young adults, the ACA young adult provision allowed 19–25-year-olds to be covered by their parents’ private health insurance plans regardless of their marital status, student status, and whether they have children. Under this provision, young adults were given additional opportunities to enroll in their parents’ private health insurance plans for 30 days following the first day of the first plan or policy year beginning on or after September 23, 2010. Many insurance companies began covering these individuals voluntarily before this date, as early as May 2010 (U.S. Department of Labor 2013). Prior to the ACA provision, young adults who were not enrolled in school generally became ineligible for coverage under their parents’ private health insurance plans when they turned 19 years old while students generally aged out when they turned 24 years old. As a result, many of these young adults not enrolled in school became uninsured upon reaching these age limits. To address this issue, many states passed laws increasing the eligibility age or relaxing eligibility requirements for obtaining coverage under parents’ private health insurance plans (Monheit et al. 2011) but these state mandates were not as comprehensive or well-known as the ACA provision.
Recent work finds the ACA provision has been effective in increasing health insurance coverage for young adults and has in turn affected their health-related outcomes. With regard to insurance, these findings include an increase in the number of young adults with any health insurance (Sommers and Kronick 2012; Akosa Antwi, Moriya, and Simon 2013; Sommers et al. 2013; Barbaresco, Courtemanche, and Qi 2015) and with private health insurance (O’Hara and Brault 2013); increases in access to care and in the share of young adults with dependent coverage and a reduction in their uninsured rate (Cantor et al. 2012), as well as an increase in rates of health insurance coverage for young adults seeking emergency care (Mulcahy et al. 2013) and a decrease in the prevalence of uninsurance among the hospitalized (Akosa Antwi, Moriya, and Simon 2015). Further work finds effects of the provision on health-related outcomes beyond the realm of insurance. For example, Barbaresco, Courtemanche, and Qi (2015) find the provision increased probabilities of having a primary care doctor and excellent self-assessed health as well as decreased body mass index, and Akosa Antwi, Moriya, and Simon (2015) find the provision increased inpatient hospitalizations, particularly those related to mental illness.
III. A Theoretical Framework for the Relationship Between Insurance Coverage and Marriage
There are several channels through which the prospect of obtaining health insurance coverage might influence marriage. Obtaining coverage through an employer or through a spouse are some of the main avenues for procuring private insurance coverage. If an individual sufficiently values insurance coverage and is unable to obtain coverage through other channels, he or she may be willing to choose a less compatible employment arrangement or spouse in order to obtain insurance coverage. This framework considers both how the prospect of obtaining coverage could have a direct effect on marriage through the choice of a spouse as well as an indirect effect on marriage through the choice of employment.
A. Insurance Coverage, the Choice of a Spouse, and the Effect on Marriage
The prospect of obtaining insurance coverage through a spouse could induce individuals to marry by increasing the benefit of marriage. As outlined by Becker (1973, 1974), a necessary condition for marriage is that the total output of the marriage exceeds the sum of the maximum outputs of the single individuals. The total output of the marriage is a function of market goods and services and time spent in home production and is determined by the complementarity of the partners. An individual who values health insurance coverage but is not eligible to obtain coverage through an employer or government programs may be able to obtain coverage through marriage. For each possible covered spouse that this individual could choose, the output of marriage relative to the output of being single is raised by the amount that the individual values insurance coverage. This person would be willing to marry so long as their share of the output associated with marrying a given person exceeds their output associated with being single as well as their expected share of the output associated with waiting for another offer.
The benefit of obtaining insurance coverage through marriage could affect both coupled and single individuals. For individuals who already have found a partner, the benefit of insurance coverage could induce them to marry. For individuals who are single, by making marriage more attractive, the benefit of insurance coverage could induce them to search more for a spouse, either at the intensive or extensive margin. Becker (1974) outlines that this search may take the form of trial living together, consensual unions, or prolonged dating.
B. Insurance Coverage, the Choice of Employment, and the Effect on Marriage
The prospect of obtaining insurance coverage through an employer could indirectly affect individuals’ propensities to marry through effects on employment: to obtain insurance coverage, an individual may choose employment that is otherwise undesirable along dimensions such as wages, job responsibilities, or hours. Changing one’s employment decisions in order to obtain insurance could in turn affect marriage but the direction of the effect is ambiguous. Becker (1973) outlines how income and wages affect the total output from a marriage and, it follows, individuals’ propensity to enter into marriage. If individuals take lower wage jobs to obtain insurance, they may be less desirable on the marriage market. However, if they are also more willing to accept a lower offer from a potential spouse, they may have more incentive to marry. Alternatively, individuals may take jobs with equal or higher wages that are less desirable otherwise—for example, jobs that may be less personally fulfilling—which could make these individuals either more or less desirable on the marriage market. In addition, if individuals work longer than desired hours in order to obtain coverage, they may become more desirable in terms of income but may have less leisure time available to search for a spouse. Likewise, employment effects that in turn affect educational pursuits also could make an individual either more or less desirable on the marriage market.
C. Testable Implications of the Framework with Respect to the ACA Young Adult Provision
The ACA young adult provision may affect young adults’ decisions to marry by facilitating obtaining health insurance coverage outside of marriage and employment. This could come about directly: If young adults consider health insurance coverage in their marriage decisions and become more eligible to obtain coverage outside of marriage following the ACA provision, the provision would be associated with a decrease in their probability of marrying a spouse whom they would not choose in the absence of the prospect of obtaining coverage. This also could come about indirectly: If young adults consider health insurance coverage in their employment decisions and become more eligible to obtain insurance outside of their employment arrangements following the ACA provision, the provision would be associated with changes in types of employment, hours worked and income, and wages, which in turn could increase or decrease their probability of marrying. If the provision affects marriage, larger magnitudes of effects would be expected for those not likely to have had parental coverage prior to the enactment of the provision, such as those not enrolled in school or ineligible for coverage under a state-specific mandate. Further, the provision also may affect outcomes related to marriage, like divorce and cohabitation, as well as the proportion of individuals obtaining spousal health insurance coverage.
IV. Data
Examining the effects of the ACA on an individual’s probability of marrying requires data covering the periods before and after provision enactment and implementation for treated young adults and a control group. An ideal data set would have a large sample size and include information on a respondent’s marital history rather than only information on his or her marital status. Given these considerations, the ACS is used for this analysis. To examine the effects of the ACA on an individual’s probability of marrying, I pool the 2008 through 2013 one-year internal ACS data. The ACS is a nationwide survey conducted continuously throughout each year. It is conducted in all U.S. counties and Puerto Rico municipios. About 3 million housing unit addresses are sampled annually, in addition to a sample of individuals living in group quarters such as college dormitories. The internal ACS data provide a larger sample than public use data (U.S. Census Bureau 2015) and include additional variables such as the date the respondent completed the survey. The large sample size serves as an important benefit of the ACS for this analysis. Another benefit is that, although the ACS is a cross-sectional sample, the survey includes questions on marital timing providing insight into changes in an individual’s marital status over time.2
The primary outcome of interest examined in this analysis is whether an individual married in the calendar year prior to the survey year.3 The analysis uses an outcome reflecting entry into marriage rather than the stock of married people to more precisely capture the effect of the policy change on behavior, as in Lichter, McLaughlin, and Ribar (2002), Bitler et al. (2004), and Schaller (2013), and to minimize specification bias, as noted by Klerman and Haider (2004). Using the stock of married people as an outcome could reflect current as well as past period conditions affecting the decision to marry, cannot specifically identify the characteristics of people who choose to marry, and does not account for changes in immigration and emigration that may affect the stock of married individuals but not marriage rates, introducing additional noise to the analysis. In addition, using the stock of married people as an outcome could underestimate the magnitude of the effects of the provision given that change in the stock of married individuals may be small even if there is a large effect on the number of individuals who marry in a given period, which could bias the magnitude of the estimates downward. Validating these concerns, Abramowitz and Dillender (2016) find that comparing changes in the stock of married individuals over time to reported marriage rates yields inconsistent results. Accordingly, an outcome reflecting entry into marriage rather than the stock of marriages is used in this analysis.
Information on the year of last marriage used to identify whether an individual married in the calendar year prior to the survey year, the outcome of interest, is available in the ACS beginning in 2008 when the survey began asking questions that are more detailed about individuals’ marriage histories.4 Since the ACS collects data over the course of the calendar year, the outcome measure is defined as the prior rather than the current calendar year in order to capture all marriages occurring throughout the year. Using this outcome measure provides a means for identifying when an individual married in a way that is consistent across years and between individuals surveyed at different times of the year. It also permits excluding married individuals who entered into marriage before the period of interest whose decisions to marry, by definition, cannot be affected by the policy. Because the outcome of interest is whether an individual married in the calendar year prior to the survey year, only individuals who were unmarried in the calendar year prior to the survey year and those who married in the calendar year prior to the survey year are included in these analyses. The sample includes individuals from all U.S. states and the District of Columbia living in households or college dormitories. Active duty military are excluded from the sample. Individuals aged 23–25 and 28–30 at the time of survey response are included in the analysis. Table 1 presents summary statistics for the sample used in the main analyses, all individuals ages 23–25 and 28–30 who were either unmarried in the calendar year prior to the survey year or who married in the calendar year prior to the survey year. Twenty-three-25-year-olds are less likely to report having married in the prior calendar year than 28–30-year-olds (5.5 percent versus 8.0 percent) but are more likely to report being enrolled in school (27.6 percent versus 14.3 percent). Both age groups have similar proportions of individuals by sex and race/ethnicity categories.
Summary Statistics
Data from the 2008–2013 survey years are used to examine the periods before, during, and after provision enactment and implementation. The enactment period covers the time after the provision was enacted, but before insurance companies were required to implement it, during which some companies voluntarily extended coverage to young adults; the implementation period covers the time insurance companies were required to implement the provision. Because the outcome of interest is marriage in the calendar year prior to the survey year, provision enactment and implementation periods are extrapolated to cover corresponding calendar years. For the purposes of this analysis, the period before provision enactment covers the 2008–2010 survey years (2007–2009 calendar years), the enactment period covers the 2011 survey year (2010 calendar year); and the implementation period covers the 2012 and 2013 survey years (2011 and 2012 calendar years).5
V. Methods
To estimate the effects of the young adult provision on marriage, it is necessary to identify the effects of the provision separately from other factors that might affect marriage occurring over the same period. Because the provision applies only to young adults ages 19–25, this analysis compares outcomes for individuals in the treatment age group to outcomes for similarly aged individuals at ages not covered by the provision before and after provision enactment and implementation. This approach assumes that individuals in the control group face similar trends in marriage rates as the treatment group and thus will account for time-varying factors that would have led the treatment group to marry at different rates after the enactment and implementation of the provision.
Because the outcome of interest is marriage in the calendar year prior to the survey year, the analysis compares marriage outcomes for a treatment group of young adults ages 23–25 to a control group of slightly older individuals ages 28–30. Individuals aged 26 and 27 were excluded to identify clear effects for the treatment age group of 23–25. This sample restriction is particularly beneficial to minimize the number of individuals who appear in the sample in the older control group but may have been affected by the provision when they were younger given that the sample covers two years of provision implementation. The inclusion of these individuals might bias the results toward finding a negative effect of the provision on marriage rates if individuals delay marriage from the affected ages to older ages. Individuals younger than those at ages affected by the provision, ages 16–18, were not included in the control group. This is because a control group of individuals in their twenties may better reflect the circumstances of the young adults at ages affected by the provision with regard to making choices about marriage, insurance, and employment. The treatment group includes only individuals aged 23–25 because these individuals are closest in age to the control group, and in analysis by Slusky (2014), the use of this age group led to more robust placebo test results for insurance and labor market outcomes. To evaluate the choice of treatment and control age groups, I conduct robustness checks testing the sensitivity of results to different age cutoffs.
To identify effects of the provision, it is necessary for both the treatment and the control groups to have exhibited similar trends in marriage rates before the enactment of the provision. To examine the extent to which treatment- and control-group marriage rate trends were similar prior to the enactment of the provision, Figure 1 presents the proportion of the single population that married in the calendar year prior to the survey year by age group over the periods before, during, and after ACA young adult provision enactment and implementation. The figure shows that the trends in marriage rates are similar prior to provision enactment. From the period before provision enactment over the 2007–2009 calendar years to the enactment period and period after implementation over the 2010–2012 calendar years, marriage rates fell for both the treatment and control age groups; however, it is only for treatment 23–25-year-olds that rates fell over the enactment and implementation periods. The decline for the treatment age group is distinct from the patterns for the control age group, for which rates did not decrease over the same period.
Proportion of Single Population Married in the Calendar Year Prior to Survey Year by Age Group and Calendar Year
Source: 2008–2013 1-year ACS data.
To formally test for equality of trends, I conduct a falsification test using 2008–2010 survey year data covering the pre-enactment period. I estimate the main analysis model, described subsequently, which predicts the probability of having married in the calendar year prior to the survey year, except the key variable of interest here is an interaction between the linear time trend and the treatment age group dummy. Unreported results suggest that, conditional on control variables included in the model, there is no statistically significant difference in the control and treatment age groups’ trends in marriage rates prior to provision enactment.6
To identify effects of the provision on the treatment age group as compared to the control age group, it is also necessary that there are sufficiently small rates of marriage across age groups. Marriage across age groups would bias the results toward zero. In the extreme case that all individuals in the treatment age group married individuals in the control age group, even if the provision only had a large impact on marriage decisions for individuals in the treatment age group, the provision would result in identical effects on marriage rates across age groups. However, the data reveal that the effect of this potential contamination is likely to be small. Examining the proportions of individuals marrying across treatment and control groups in 2008–2013 ACS data shows that the overwhelming majority of individuals either marry a spouse within their age group or at an age not included in the analysis. Only 15.2 percent of married 23–25-year-olds are married to individuals in the control age group while 39.8 percent are married to individuals in the same age group, and 6.7 percent of married 28–30 year-olds are married to individuals in the treatment age group while 38.5 percent are married to individuals in the same age group. Accordingly, while some contamination across age groups may bias results toward zero, it is not a major concern for this analysis.
It is also important to identify the effect of the provision separate from lingering effects of the Great Recession, which lasted from December 2007 through June 2009, during the period covered in this analysis. To address this concern, in the main analysis, age-specific state-year unemployment rates are calculated using the ACS and are used to control for labor market conditions.7 Age-specific rates are used to capture differential effects of the recession by age group because younger adults were particularly hard-hit by the recession. These age-specific rates are further interacted with the control for age group in the analysis to accommodate differential effects of economic conditions on marital choices by age group.
While the patterns evident in Figure 1 are suggestive of a relationship between the provision and marriage, it is important to control for other factors that might affect this relationship. Thus, a more in-depth analysis of this question is warranted. The following difference-in-differences-type regression model estimates these effects accounting for such observable heterogeneity:
(1)
where yiajt is a dummy variable equal to one if an individual i of age group a living in state j during survey year t + 1 reports having married in the prior calendar year t and is equal to zero otherwise. The coefficients on the interaction terms β3 and β4 capture the associated difference in the probability of having married in the calendar year prior to the survey year for a person in the treatment age group during the provision enactment and implementation periods, respectively, relative to a person in the control age group, or a person in the treatment age group in the pre-enactment period, holding other characteristics constant. Vectors of parameters are included to control for state fixed effects (β5), year fixed effects (β6), and treatment age-group fixed effects (β7).8 The X vector (β8) controls for individual demographic characteristics including dummy variables for age, sex, race/ethnicity, and student status.9,10 The Z vector (β9) controls for state-year-level conditions including the age-specific state-year sex ratio and unemployment rate,11,12 the interaction of the unemployment rate with the control for the treatment age group, as well as a national linear calendar year trend and state-calendar year trends. The error term is represented by ε. In all tables and regressions, the data are weighted to be population-representative using population weights. As per Bertrand et al. (2004), in all regressions, White robust standard errors clustered by state are used to control for serial correlation among the outcomes and the policy changes of interest.13
VI. Results
A. Main Results
Results of the main analysis, presented in Table 2, show a significant negative effect of the ACA young adult provision on 23–25-year-olds’ propensity to marry. For the full sample, provision enactment and implementation are associated with decreases in the probability of marrying of 0.53 percentage points and 0.56 percentage points, respectively. These represent 8.8 percent and 9.3 percent decreases in the marriage rates of 23–25-year-olds as compared to before provision enactment.14
Regression Results for Probability of Marrying
To put these results in context, I use the estimated effects of the provision on marriage rates to calculate the number of individuals affected by the provision as proportions of the populations of interest. The decreases in marriage rates of single young adults correspond to 0.5 percent decreases in the number of married young adults aged 23–25 associated with each year of provision enactment and implementation representing approximately 60,000 individuals in the enactment year and approximately 65,000 individuals in each year of implementation.15 However, the effect of the provision on all 23–25-year-olds does not represent the effect of the policy on the treated since not all young adults at ages covered by the provision were actually affected by the provision. For example, some were already eligible for coverage under parents’ plans, some were unable to take up parental coverage if they did not have a parent with a candidate plan, and others maintained insurance coverage from other sources. To estimate the extent of the marriage effects of the provision on treated young adults, defined as those gaining dependent coverage through parental employer-sponsored insurance as a result of the provision, I consider the effect of the provision on marriage in light of the effects of the provision on insurance coverage found in the literature. In a given year, approximately 975,000 23–25-year-olds gained insurance coverage as dependents on parents’ employer-sponsored insurance plans because of provision implementation.16 Considering this estimate as the treated population, the results suggest that, of treated individuals, the number married decreased by 6.6 percent, much larger than the 0.5 percent decrease considering the full population of married 23–25-year-olds presented earlier.
The main results find an overall decrease in marriage rates associated with the provision but it may be the case that the provision has differential effects on marriage for particular subsets of the population. Since young adults were generally eligible for health insurance coverage through their parents’ private plans through age 23 or even older if they were a student, students are less likely to be affected by the provision than nonstudents. Men and women also may experience differential effects of the provision on marriage because they tend to marry at different ages and may have differential access to and demand for health services and health insurance.
Results estimated separately by student status, presented next in Table 2, show no significant effect of the provision for students. However, the results do show a significant decrease in the marriage rates of nonstudents of 0.72 percentage points for enactment and 0.65 percentage points for implementation, corresponding to 10.8 and 9.8 percent decreases, respectively, as compared to before provision enactment.17 As expected, the results suggest that there did appear to be an effect on marriage rates for nonstudents who likely were not eligible for parental coverage prior to the provision but for students, who were unlikely to be affected by expanded access to parental private health insurance coverage, the effects are smaller in magnitude and are not statistically significant. These findings support the proposition that it is the provision that is driving the effects on marriage rather than other concurrent phenomena.18
Results by sex are presented subsequently in Table 2. For both men and women, significant negative effects of the provision on 23–25-year-olds’ likelihood of marrying are found for the provision enactment period. Negative effects also are found for provision implementation for both men and women. While the result is only significant for women, the estimates for men and women are not statistically significantly different from each other.
This analysis examines the effect of expanded access to parental private health insurance coverage through the ACA young adult provision, but several states had implemented mandates to expand eligibility for this coverage prior to the federal provision. These mandates expanded eligibility based on specific criteria related to age, disability status, student status, marital status, whether an individual has children, and whether an individual lives with a parent, and these criteria varied considerably across states. Although not all previously eligible individuals may have known about the mandates, one would expect to see larger effects of the ACA provision on individuals who are newly eligible. To examine this question, I consider only the 23–25-year-old age group and first compare effects across states with prior mandates based on any criteria to states with no prior mandates following the classification of state policies in Monheit et al. (2011). However, this methodology identifies young adults as newly eligible only at the state level and, as a result, does not reflect actual newly eligible status for young adults based on the state-specific mandate criteria described above. Accordingly, it may not be appropriate to simply identify newly eligible status at the state level. To account for heterogeneity within states, the analysis next compares outcomes across individuals previously eligible and those newly eligible based on state-specific mandate criteria.
Table 3 presents results examining differential effects of the provision on marriage rates for states with mandates for parental insurance coverage of young adults in place prior to the ACA provision. Surprisingly, comparing the effect of provision enactment and implementation for individuals in newly eligible states to those in states with prior mandates, results in the first column of Table 3 suggest an increase in the likelihood of marrying associated with both provision enactment and implementation for individuals living in states that had not implemented any reforms as compared to those that did. This would suggest that new eligibility is associated with an increased propensity to marry, counter to the results of the main analysis. However, in the second column of Table 3, looking more closely at new eligibility at the individual level based on state-specific mandate criteria yields results consistent with the main findings: New eligibility is found to be associated with a decrease in the likelihood of marrying. These findings may reflect larger effects of the provision on marriage in states with prior mandates due to a greater number of individuals gaining eligibility for parental insurance coverage because of the provision.
Regression Results for Probability of Marrying for 23–25-Year-Olds by Change in Eligibility
B. Further Results
Having found an effect of the provision on the marriage rates of young adults, the provision also may influence other marriage-related outcomes such as divorce and cohabitation. The provision may affect divorce since if individuals are less likely to enter marriage, they may also be more likely to leave marriage. The provision also may affect cohabitation because if individuals are less likely to marry because of the provision, they may opt to cohabit instead; alternatively, if individuals cohabit with the intent to marry and they are now less likely to marry, they may also opt out of cohabitation.
For the analysis of divorce, the outcome variable is whether an individual divorced in the 12-month period prior to survey response. Given the period of the outcome measure, it is necessary to classify provision-related time periods based on response date, available in the internal ACS data used in this analysis. The period before provision enactment covers responses from January 1, 2008 through April 30, 2010. The transition period covers responses during provision enactment and the first year of provision implementation, during which it is not possible to identify whether they occurred prior to provision enactment, during enactment, or during implementation including responses from May 1, 2010 through September 22, 2011. The implementation period covers responses from September 23, 2011 through December 31, 2013. The analysis follows the methodology used in the main analysis except that the analysis sample is restricted to individuals currently married and those divorced in the prior 12 months.
For the analysis of cohabitation, the outcome variable covers cohabitation at the time of the survey.19 As in the divorce analysis, provision-related time periods are also classified based on response date but can be more accurately attributed to the appropriate period. The period before provision enactment covers responses from January 1, 2008 through April 30, 2010. The enactment period covers responses from May 1, 2010 through September 22, 2010. The first year of the implementation period covers responses from September 23, 2010 through September 22, 2011. The second and later years of the implementation period cover responses from September 23, 2011 through December 31, 2013. The analysis follows the methodology used in the main analysis except that the analysis sample is restricted to unmarried individuals and controls for year fixed effects and trends are replaced by calendar-month fixed effects and trends because the outcome measure can be estimated on a monthly basis.
Figure 2 and Figure 3 show the proportion of individuals divorced in the past year and cohabiting by the treatment and control age groups over time. Figure 2 shows an increase in 2011 in the proportion divorced in the prior 12 months for the treatment age group only, coincident with provision enactment and implementation. Whereas the proportion decreased over 2012 and 2013, it remained higher for the treatment age group than the control age group. Figure 3 shows an increase in the proportion cohabiting among the treatment age group over the provision enactment and implementation periods but shows a decrease for the control age group over the same period. The treatment and control age groups appear to exhibit similar trends for both divorce and cohabitation over the period prior to provision enactment.20
Proportion Divorced in the Past Year by Treatment and Control Groups
Source: 2008 through 2013 1-year ACS data.
Proportion Currently Cohabiting by Treatment and Control Groups
Source: 2008 through 2013 1-year ACS data.
Table 4 presents results for the effects of the provision on divorce and cohabitation. Results for divorce suggest provision implementation is associated with a marginally significant increase in the likelihood of having divorced over the past year. The results for divorce might suggest that married young adults have less incentive to stay married if they have a greater opportunity to obtain health insurance coverage outside of marriage. Results for cohabitation suggest that unmarried individuals are less likely to live with a partner for both the enactment and implementation periods. This result is consistent with the proposition that the provision is associated with young adults being less interested in pursuing relationships and suggests that young adults are not substituting from marriage to cohabitation but instead are opting out of both because of the provision.
Regression Results for Marriage-Related Outcomes
VII. Robustness
A. Alternative Age Groups
To address the concerns presented by Slusky (2014) that the definition of the treatment and control groups by the particular age groups used in the main analysis might be driving the results, I rerun all previously presented analyses considering individuals ages 21–25 and 28–30 with the 21–25 age group defined as the treatment group as well as considering individuals ages 16–18, 21–25, and 28–30 with the 21–25 age group defined as the treatment group. Unreported results are generally similar in magnitude and significance.21
B. Controls for Labor Market Conditions
One concern for this analysis is that lingering effects of the Great Recession, which occurred during the analysis period, also could affect marriage rates at the time of provision enactment and implementation. In an attempt to identify effects of the recession, in the main analysis I use age-specific state-prior-year unemployment rates calculated using ACS data to control for labor market conditions. Other proxies for labor market conditions, including the Bureau of Labor Statistics Local Area Unemployment Statistics county-year unemployment rates and state-year employment-to-population ratios for the prior year, are considered as alternative controls for labor market conditions. Using these alternative controls yields qualitatively similar results.22
C. Placebo Tests
It may be the case that the results of the main analysis simply reflect spurious effects arising from the structure of the data or the model specification used in the analysis. To consider this concern, I follow the approach of Barbaresco, Courtemanche, and Qi (2015) using the same data and specification as in the main analysis to estimate placebo analyses for which I should not see significant effects of the provision. First, I consider a placebo mandate for alternative years prior to provision enactment. Ideally, I would consider placebo mandates over several periods prior to the enactment of the provision. However, 2007 is the earliest year for which the outcome variable for having married in the calendar year prior to the survey year can be constructed. For this test, the pre-enactment period covers the 2007 calendar year, and I construct the placebo provision enactment period to cover the 2008 calendar year and the placebo provision implementation period to cover the 2009 calendar year. Next, I consider a placebo treatment group over the period used in the main analysis that covers before and after the enactment and implementation of the provision. In this placebo test, I include only individuals ages 28–30 and 33–35, who are outside of the age range affected by the provision and define individuals ages 28–30 as the placebo treatment group. Table 5 presents results from these analyses. Results show no significant effects of the placebo mandate enactment or implementation and no significant effects of provision enactment or implementation for the placebo treatment group.
Placebo Regression Results for Probability of Marrying
VIII. Effect on Spousal Health Insurance Coverage
If, because of the provision, individuals are less likely to marry in order to obtain coverage as a dependent on a spouse’s plan, then the provision should be associated with a decrease in spousal health insurance coverage for individuals at ages affected by the provision relative to those that are not. However, it is possible to see a large effect on marriage without seeing a large effect on spousal coverage if the provision affects marriage primarily through indirect channels, such as changes in employment or other nonmarriage choices, and young adults that would have married in the absence of the provision may have been uninsured or had own employer-sponsored insurance coverage rather than spousal coverage.23
To consider whether the provision affects marriage through the direct channel, the analysis examines the effect of the provision on spousal coverage generally following methodology used in the literature examining the impact of the provision on insurance coverage (Akosa Antwi, Moriya, and Simon 2013). The analysis uses all waves of the 2008 SIPP Panel covering August 2008 through November 2013.24
Individuals ages 16–18, 20–25, and 27–29 are included in the analysis regardless of marital status. Individuals aged 19 and 26 were excluded to identify clear effects for the treatment age group of 20–25. The treatment and control age groups used in the analysis of the effects of the provision on spousal coverage were chosen for several reasons. First, this analysis examines an insurance outcome rather than a marriage outcome, and using these age groups is more consistent with the methodology used in the literature examining the effect of the provision on insurance outcomes. Second, more broadly defined age groups allow the analysis to benefit from a larger sample size given the already smaller sample size of the SIPP as compared to the ACS. Third, whereas the main analysis examines marriage rates, this analysis is dependent on the stock of married individuals, and effects of the provision may be harder to detect given that 23–25-year-olds are more likely to have already been married prior to the enactment of the provision as compared to younger individuals. Further, including both younger and older individuals in the treatment age group warrants including both younger and older individuals in the control age group to better reflect the circumstances of all individuals in the treatment group. In addition, since the outcome measure used in this analysis covers the current month rather than the prior year, the unsuitability of a younger control group is of less concern here.25
It is important to note that comparisons of the proportion of individuals with spousal insurance coverage reflect proportions insured that are based on the stock of individuals eligible for spousal coverage: An individual only can obtain spousal coverage if he or she is both married and has a spouse with coverage and the ability to add dependents. This outcome reflects both current and past period conditions affecting the decision to marry as well as current period conditions affecting the ability and decision to take up spousal coverage. The methodology used in this analysis could underestimate the magnitude of the effects of the provision given that the dependent variable is a binary indicator of having spousal coverage and thus a change in the stock of individuals receiving coverage may be small even if there is a large effect on the number of individuals who marry in a given period, which could bias the magnitude of the estimates downward.
Figure 4 compares the proportion of individuals with spousal coverage for all individuals regardless of marital status at ages affected by the provision (ages 20–25) and those at similar ages not affected by the provision (ages 16–18 and 27–29) before and after provision implementation. Despite the limitations described above, Figure 4 shows a dramatic decrease in spousal coverage for the treatment age group relative to the age control group over the provision enactment and implementation periods.
Proportion with Spousal Health Insurance Coverage by Treatment and Control Groups
Source: 2008 SIPP Panel data.
To formally test the effect of the provision on spousal coverage, I employ the sample criteria outlined above and generally follow the specification outlined by Akosa Antwi, Moriya, and Simon (2013) used to examine the impact of the provision on insurance coverage. I include the same independent variables used by Akosa Antwi, Moriya, and Simon (2013) with the exception of marital status because it is through the decision to marry that the provision may affect on spousal coverage.26 Results show a weakly significant negative effect of provision implementation on spousal coverage suggesting a decrease in spousal coverage of 0.6 percentage points corresponding to 15.5 percent as compared to the pre-provision mean,27 representing approximately 80,000 fewer individuals covered in a given year as a dependent on a spouse’s insurance.28,29,30
The effect of the provision on spousal health insurance coverage is small compared to that found by Akosa Antwi, Moriya, and Simon (2013) for having employer own coverage, a decrease of 3.1 percentage points but is similar in magnitude to the effect for individually purchased insurance in one’s own name, a decrease of 0.8 percentage points. Of the approximately 80,000 fewer individuals with spousal coverage as a result of the provision, the decrease in marriage can account for approximately 80 percent of the decrease in spousal coverage.31 Finding a negative effect of the provision on spousal health insurance coverage is consistent with the proposition that young adults may be less likely to marry in order to obtain spousal coverage because of the provision. Given these findings, an effect of the provision on marriage through the direct channel cannot be ruled out.
IX. Discussion and Conclusions
This paper examined the relationship between the ACA young adult provision and marriage. The findings suggest that the enactment and implementation of the provision decrease the likelihood of marrying for young adults at ages covered by the provision. This evidence is robust to the choice of treatment and control groups of alternative ages and the inclusion of various controls for economic conditions. For 23–25-year-olds, the probability of marrying fell by 0.53 percentage points and 0.56 percentage points during provision enactment and implementation, respectively. These represent 8.8 and 9.3 percent decreases in the marriage rates of 23–25-year-olds as compared to before provision enactment. Results suggest provision implementation is associated with decreases in the number of married individuals of 0.5 percent of married young adults and 6.6 percent of young adults gaining parental coverage as a result of provision implementation in a given year. The provision also is associated with an increase in the probability of divorce and decreases in cohabitation and spousal health insurance coverage. Thus, introducing a new avenue for obtaining health insurance is associated with decreased marriage rates. This is consistent with the proposition that some individuals marry in order to obtain health insurance coverage and that the provision serves to correct inefficiencies in the insurance and marriage markets.
The magnitudes of the effects on marriage of the ACA young adult provision are comparable to those found in the literature for other policies. Yelowitz (1998) finds Medicaid expansion that extended eligibility to two-parent families was associated with an increase in the probability of being married of 2.3 percent for women between the ages of 18 and 55 with at least one child younger than age 15 present and 2.8 percent after controlling for outflows from marriage.32 Goldin and Katz (2002) find that nonrestrictive birth control laws decreased the fraction married by age 23 by 5.0 percent33 while Myers (2012) finds that the legalization of abortion is estimated to have led to a reduction of 16.5 percent in the fraction of women who married prior to age 19, which increases to a reduction of 22.4 percent if minors could consent to abortion34. Abramowitz (2014) finds that, for white women, state health insurance mandates for assisted reproductive technology are associated with 2.5 percent and 1.2 percent decreases in the proportions married of 30–34 and 35–39 year-olds, respectively.35 The effects found in this analysis of ACA young adult provision implementation on marriage of 0.5 percent overall and 6.6 percent on the treated are similar in magnitude.
The results of this analysis have implications for considering the effects on marriage of full implementation of the ACA as well as future policies related to health insurance. In 2014, ACA state Medicaid expansions, insurance exchanges and subsidies, and the individual mandate to obtain health insurance coverage began to be implemented. State Medicaid expansions provide health insurance coverage to many low-income individuals otherwise unable to obtain it. New health insurance exchanges and subsidies make it easier for individuals to purchase coverage directly and for small businesses to offer coverage to their employees. The individual mandate requires that most Americans obtain health insurance by 2014 or pay a tax penalty.
Like the ACA young adult provision, both state Medicaid expansion and insurance exchanges and subsidies have the potential to decrease the propensity to marry directly by providing new avenues for obtaining health insurance outside of marriage but they also may have ambiguous effects on marriage through indirect channels related to employment and other nonmarriage-related choices. For both policies, short-term effects may be smaller than those found for the ACA young adult provision given that individuals at older ages than the young adults affected by the provision are more likely to already be married. Even among young adults, the populations eligible to obtain coverage through a parent’s employer-sponsored insurance, eligible for Medicaid expansion, and eligible for coverage and subsidies through exchanges exhibit different demands for health services and health insurance and different propensities to marry. These disparities may result in divergent effects of the policies on marriage. In particular, insurance exchanges and subsidies may have less of an effect on marriage than the young adult provision: If the prices of exchange plans are high, this aspect of the ACA may not have a large effect on marriage directly although it may still impact marriage indirectly if employers change the extent to which they offer coverage, both to employees as well as their dependents.
Separate from these policies, the individual mandate could affect marriage by increasing the demand for health insurance coverage. However, it may prove difficult to identify the effect of the individual mandate apart from the other new policies introduced by the ACA. Given that this analysis documents an effect of the ACA young adult provision on marriage, future work examining the effects on marriage of the policies introduced by the ACA is warranted.
The results of this analysis find a decrease in marriage for the treatment age group but it is not clear whether the effect is one of marriage delayed or marriage forgone. The provision may be associated with young adults delaying marriage if it enables them to make marriage decisions without the added consideration of insurance coverage. Treated individuals might then search longer to find spouses who are better matches and form more stable marriages. In this case, the provision also would be associated with a decrease in divorce over the longer term. Alternatively, the provision may result in young adults delaying the search for a spouse until after aging out of their parents’ health insurance plans. This would result in delayed marriage but no better matches and thereby no longer-term effect on divorce. The provision only would be associated with a decrease in the propensity to ever marry if individuals are better able to obtain coverage outside of marriage after aging out of their parents’ coverage under the provision.
The effects of the provision on marriage also indirectly may affect fertility.36 If individuals only choose to have children within marriage, marital delay could result in a similar delay in childbearing. However, if the provision results in a decrease in marriage over the longer term or leads to long-term marital delay, completed fertility would decline. Alternatively, if individuals delay or opt out of marriage because of the provision but decide to have children out of wedlock, the provision may be associated with an increase in single-parent households. On the other hand, if the provision results in marital delay and thereby better marriages over the longer term, the provision may be associated with a decrease in divorce and single-parent households.
This study makes several main contributions. First, the results shed light on the relationship between the prospect of obtaining health insurance coverage and the propensity to marry: The findings of this analysis suggest that some individuals marry in order to obtain insurance coverage. Further, the results increase understanding of how policies can influence marriage choices. Other papers have found that tax, benefit eligibility, and health-related policies affect individuals’ propensities to marry. This paper shows that policies related to health insurance coverage also affect marriage. Finally, the results add to body of research measuring the effects of the ACA young adult provision. While other papers have found effects of the provision on insurance, labor market, health outcomes, and access to care, this research finds that the provision also influences individuals’ propensity to marry.
Footnotes
Joelle Abramowitz is an assistant research scientist at the University of Michigan’s Institute for Social Research. The author is grateful to Brett O’Hara, Jamie Taber, Mark Klee, Josh Mitchell, Catherine Massey, Robert Plotnick, and two anonymous referees for their invaluable comments. She would also like to thank seminar participants at the 2014 Association for Public Policy Analysis and Management Fall Research Conference, the 2014 Annual Research Data Center Research Conference, as well as the Census Health Research Group for their helpful feedback. The data used in this paper are internal Census data that cannot be shared. The author is willing to direct researchers to the public use version of the data or to explain the process for accessing the internal data.
↵1. In recent years, cohabiting partners increasingly have been able to obtain dependent coverage through employers; however, the majority still cannot. According to Schaefer (2009), in 2008, 36 percent of private employers offered cohabiting couples access to dependent coverage, an increase from seven percent in 1997.
↵2. One alternative data set considered for this analysis was the Current Population Survey. It is possible to longitudinally link years of the Current Population Survey allowing for identification of individuals that married in the past year. However, only data for individuals that remain in the same residence can be linked in this way. Because marriage is often associated with changing residences, a concern is that marriage-related attrition would bias the results. The SIPP also was considered for this analysis but due to its relatively small sample size and complications related to sample attrition, the ACS was deemed most appropriate for this analysis.
↵3. Prior to 2013, individuals reporting being married to a same-sex partner were classified as “unmarried partners.” Beginning in 2013, ACS data and data products include same-sex married couples along with all married couples. Analyses retaining the marital status of individuals reporting being married to a same-sex partner for the 2012 survey year and earlier yield quantitatively similar results.
↵4. An alternative outcome measure considered for this analysis was marriage in the past 12 months for which data is also available in the ACS beginning in 2008. A drawback to using this variable is that, given that the ACS is conducted throughout the year, it is not possible to clearly identify individuals’ date of marriage precisely enough to identify whether they married during the provision pre-enactment, enactment, or implementation periods. However, using this outcome measure and more broad provision periods yields quantitatively similar results.
↵5. Excluding the enactment period (2011 survey year/2010 calendar year) yields quantitatively similar results.
↵6. Results are available from the author by request.
↵7. Age-specific unemployment rates are estimated separately for five-year age groups of 21–25-year-olds, 26–30-year-olds, and 31–35-year-olds.
↵8. Specifications used in this analysis include state fixed effects to account for differences across states. These include differences in state mandates for young adult insurance coverage on parents’ private health insurance plans prior to the ACA young adult provision as well as other policies and characteristics. Specifications including a control for whether the individual would have been eligible for dependent coverage on a parent’s private health insurance plan prior to the ACA young adult provision following the classification of state policies in Monheit et al. (2011) yielded qualitatively similar results.
↵9. Including controls for family income as a percentage of the federal poverty line and its squared term resulted in qualitatively similar results.
↵10. Incorporating information on parents’ insurance coverage into the analysis to identify individuals eligible for coverage through parents as a result of the provision was not used in this analysis investigating marriage because this methodology would limit the sample to only individuals living in the same household as their parents.
↵11. The sex ratio and unemployment rate correspond to the calendar year used in the analysis.
↵12. Alternative controls for labor market conditions, including Bureau of Labor Statistics Local Area Unemployment Statistics county-year unemployment rates and state-year employment-to-population ratios for the prior year, were considered. Analyses using these alternative controls, discussed in Section VII.B, yielded qualitatively similar results.
↵13. Linear probability models are used in all regressions rather than probit models for ease of interpretation. Regressions using probit models yielded qualitatively similar results.
↵14. The marriage rate for unmarried 23–25-year-olds before provision enactment was 6.0 percent.
↵15. Estimates use population numbers from the U.S. Census Bureau (2014): the population aged 23–25 is estimated to be 12.7 million, 12.8 million, and 13.1 million on July 1st of 2010, 2011, and 2012, respectively.
↵16. Calculated using the estimate of an increase in dependent coverage through parental employer-sponsored insurance of 7.6 percent for 23–25-year-olds for provision implementation from Akosa Antwi, Moriya, and Simon (2013) and population estimates of 12.8 million 23–25-year-olds as of July 1, 2011 from the U.S. Census Bureau (2014).
↵17. The marriage rate for unmarried 23–25-year-old nonstudents before provision enactment was 6.4 percent.
↵18. An alternative explanation might suggest that students are less likely to marry in general, and accordingly, may be less affected by the policy.
↵19. The ACS only identifies cohabiting partners of the person designated as the head of the household for each household interviewed, which may consist of several families. Accordingly, the interpretation of these results is limited to reflect only changes in the probability of cohabiting as a household head or the partner of a household head.
↵20. Using the same methodology to formally test for the presence of preprovision enactment trends as in the main analysis discussed in Section V, I do not find significant differential trends in divorce or cohabitation for the treatment age group as compared to the control age group conditional on the controls included in the analysis. Results of these analyses are available from the author by request.
↵21. Results are available from the author by request.
↵22. Results are available from the author by request.
↵23. The provision still may be associated with a decrease in spousal coverage if only indirect channels are driving the marriage result since if fewer individuals marry, they will be less likely to have spousal coverage. However, finding no effect of the provision on spousal health insurance would suggest direct channels are not driving the effects on marriage.
↵24. I limit the sample to include only the most recent month in each reference period. SIPP data on health insurance status is available for each month, and the unit of analysis is the person-month. I classify spousal coverage by identifying individuals with dependent coverage, identifying the policyholder for the health insurance units of those individuals, and identifying whether the policyholder is the individual’s spouse.
↵25. Restricting this analysis of spousal coverage to 23–25-year-olds and 28–30-year-olds as in the main analysis of marriage rates yields insignificant results.
↵26. These include controls for the provision enactment and implementation periods, a control for the treatment age group, and interactions between the controls for the enactment and implementation periods and the treatment age group as well as indicators for age, sex, race/ethnicity, and student status; household income as a share of federal poverty line and its squared term, monthly linear national and state-specific time trends, the monthly state unemployment rate, and an interaction of the treatment dummy variable and the state unemployment rate, and month and state fixed effects. The data are weighted to be population-representative using provided weights, and standard errors are clustered at the state level.
↵27. Prior to provision enactment, 3.9 percent of 20–25 year-olds had spousal coverage.
↵28. Calculated using population estimates of approximately 12.8 million and 13.1 million 23–25-year-olds in 2011 and 2012 (U.S. Census Bureau 2014).
↵29. It may appear from Figure 4 that trends in spousal coverage differed between the treatment and control groups prior to provision enactment. Using the same methodology to formally test for the presence of preprovision enactment trends as in the main analysis discussed in Section V, though, I do not find a significant differential trend in spousal coverage for the treatment age group as compared to the control age group conditional on the controls included in the analysis.
↵30. Results for the analysis of the effect of the provision on spousal coverage and the analysis of pre-enactment period trends in spousal coverage are available from the author by request.
↵31. Given approximately 65,000 fewer marriages per year of provision implementation estimated in Section VI.A.
↵32. Corresponding to 1.7 and 2.0 percentage points.
↵33. Corresponding to 2.1 percentage points.
↵34. Corresponding to 4.1 and 5.5 percentage points.
↵35. Corresponding to 2.3 and 1.1 percentage points.
↵36. The provision may also affect fertility directly through increased access to contraception and childbirth-related care.
- Received September 2014.
- Accepted July 2015.
References
In this issue
Jump to section
- Article
- Abstract
- I. Introduction
- II. Background on the ACA Young Adult Provision
- III. A Theoretical Framework for the Relationship Between Insurance Coverage and Marriage
- IV. Data
- V. Methods
- VI. Results
- VII. Robustness
- VIII. Effect on Spousal Health Insurance Coverage
- IX. Discussion and Conclusions
- Footnotes
- References
- Figures & Data
- Info & Metrics
- References