ENGAGING ABSENT FATHERS: LESSONS FROM PATERNITY ESTABLISHMENT PROGRAMS* Maya Rossin-Slater Columbia University, Department of Economics November 2011 Abstract This paper provides the first comprehensive analysis of the causal effects of in-hospital voluntary paternity establishment (IHVPE) programs on paternity establishment rates and consequent family structure and behavior. Using variation in the timing of IHVPE program initiation across states and years, I first show that IHVPE programs increase paternity establishment rates by 40 percent. Then, using data from the March and April Current Population Survey Supplements, I show that IHVPE programs reduce the likelihood of marriage post- childbirth. The decrease in marriage leads to an increase in the average characteristics of both married and unmarried fathers. Accounting for selection out of marriage, private health insurance provision for children declines, while maternal labor supply increases. The results from my analysis are consistent with a framework where fathers, who are heterogeneous in quality, must make transfers to mothers in exchange for rights to their children. Maternal utility is more sensitive to father quality in marriage than outside marriage, so a decrease in the cost of paternity establishment induces more mothers to choose higher partial transfers outside marriage over full transfers and interaction with lower-than-desired quality fathers in marriage. I provide evidence that the timing of IHVPE program implementation is uncorrelated with numerous state time- varying characteristics and that the results are not driven by pre-existing trends. My results are robust to the inclusion of numerous controls for maternal, child, and state time-varying characteristics, state and year fixed effects, state-specific time trends, and across several specifications, methods, and data sets. JEL Codes: I00, I38, J12, J13, Z18 * Contact e-mail: [email protected]. I am very grateful for the abundance of support, guidance, and helpful comments from Janet Currie and Wojciech Kopczuk. I thank Pierre-André Chiappori, Jonathan Dingel, Lena Edlund, Jane Fortson, Irwin Garfinkel, Ilyana Kuziemko, Corinne Low, Matt Neidell, Brendan O’Flaherty, Petra Persson, Kiki Pop-Eleches, Debbie Reed, Jonah Rockoff, Robert Santillano, Andrew Slater, Lesley Turner, Miguel Urquiola, Jessica Van Parys, Eric Verhoogen, Till von Wachter, Jane Waldfogel, Reed Walker, Elizabeth Ty Wilde, Afshin Zilanawala, participants at the Columbia University Applied Microeconomics Colloquium, participants at Mathematica Policy Research, Inc. summer 2011 brown bag seminars, and participants at the Fall 2011 APPAM Research Conference for their comments and suggestions. I thank Jean Roth for help with the CPS-CSS data. I also thank Irwin Garfinkel and Lenna Nepomnyaschy for sharing data on child support laws with me. The project described was supported by Award Number R24HD058486 from the Eunice Kennedy Shriver National Institute of Child Health & Human Development. The content is solely the responsibility of the author and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health & Human Development or the National Institutes of Health. All errors are my own. I. Introduction Some of the most disadvantaged children in the United States are in single-mother households. In fact, in 2010, 43% of children in single-mother households lived below the poverty line (U.S. Census Bureau, 2010). As the rate of births by unmarried mothers has been rising over the last several decades – such that in 2007, 40% of all births were out-of-wedlock (National Center for Health Statistics, 2010) – policymakers have become increasingly concerned with alleviating the hardships faced by these families. Since unmarried mothers generally retain physical custody of their children, a central issue is that non-resident fathers are uninvolved with the family and do not provide financial or emotional support for the mother and the child.1 As a result, public policy has sought ways to encourage fathers to fulfill their paternal responsibilities and stay involved with their children. Throughout the 1980s and 1990s, the U.S. government implemented several measures in attempts to help single mothers receive a higher and more stable source of income support by strengthening child support enforcement. An important component of these measures has been a push for increased paternity establishment at birth through in-hospital voluntary paternity establishment (IHVPE) programs, as establishing paternity is a crucial prerequisite to obtaining a legal child support order for unmarried mothers. In fact, the paternity establishment rate among children who need paternity established has increased from 29 percent in 1987 to 74 percent in 2002 (U.S. House of Representatives, 2004). There is some evidence that child support payments constitute a substantial fraction of female-headed households’ family incomes and that increased child support enforcement and payments lead to greater involvement of non-resident fathers with their children (see Garfinkel et al. (1998) for a review). More recently, policymakers have been focused on “Healthy Marriage” initiatives. The Deficit Reduction Act of 2005 provided $150 million in funding every year for “healthy marriage promotion and father involvement” (Administration for Children and Families, 2011). Most programs funded by these initiatives provide relationship education and counseling and conduct public advertising campaigns on “the value of healthy marriages” (Administration for 1 According to data from March Current Population Survey supplements over 1989-2010, only 19% of never- married mothers report receiving any child support income. 1 Children and Families, 2011). Many of these programs are specifically aimed at unmarried pregnant women and expectant fathers. Yet while the goals of these policies and programs arguably seek to address the best interests of some of the most disadvantaged families in the United States, their effectiveness may be hindered due to the complexities of the trade-offs that unmarried parents face in their decisions regarding involvement with each other and with their children. In this paper, I provide the first comprehensive analysis of one of these measures, the implementation of IHVPE programs, and develop a conceptual framework with which to interpret my findings. The empirical analysis in this paper adds to a large literature on the overall effects of child support enforcement (e.g.: Garfinkel et al. (1998); Freeman and Waldfogel (2001); Aizer and McLanahan (2006); Garfinkel and Nepomnyaschy (2007), among others), and improves upon the existing evaluations of IHVPE programs (Turner (2001); Sorensen and Oliver (2002); Mincy, Garfinkel, and Nepomnyaschy (2005)) by using a strategy that can arguably identify the causal effects of IHVPE and by considering a large number of states and several outcomes that impact family well-being in repeated cross-section data spanning more than one decade. The conceptual framework, which draws heavily upon a theoretical literature on the role of paternity rights in marriage largely pioneered and developed by Lena Edlund (see Edlund (2011) for an overview), provides motivation for the empirical findings. Using data from Office of Child Support Enforcement (OCSE) yearly reports on the number of established paternities in each state and year over 1992-2005 along with information on the year of program implementation across states, I first analyze whether IHVPE programs are in fact effective at increasing paternity establishment rates. While OCSE yearly reports suggest that the substantial increase in paternity establishments during the 1990s was due to IHVPE programs, there could be other factors driving the effect. The same decade experienced a drastic increase in the proportion of births by unmarried mothers, thus inducing nontrivial selection into the population of families likely affected by paternity establishment programs. So, if the types of mothers that are more likely to be unmarried in the late 1990s relative to the early 1990s are also more likely to establish paternity (perhaps as out-of-wedlock childbearing and cohabitation become more common), then the observed increase in paternities could be at least in part driven by the compositional shifts in the distribution of births by unmarried mothers. 2 My results suggest that IHVPE programs increase paternity establishment rates by about 40 percent. Further, I provide evidence that the timing of IHVPE program implementation is uncorrelated with numerous state characteristics and that my results are not driven by pre- existing trends in paternity establishment rates. Additionally, since my results are robust across several specifications and data sources, to controls for maternal and child characteristics, state time-varying characteristics, indicators for other child support enforcement laws, and indicators for Aid for Families with Dependent Children (AFDC) waivers and Temporary Assistance to Needy Families (TANF) introduction, as well as to the inclusion of state and year fixed effects and state-specific time trends, I conclude that the relationship is causal and not driven by other factors. I then proceed to analyze the effects of IHVPE programs on several measures of family behavior using data from March/April matched Current Population Survey Child Support Supplements (CPS-CSS) for 1994, 1996, 1998, 2000, 2002, 2004, 2006, and 20082. My results suggest that IHVPE programs have a negative effect on marriage. Specifically, the likelihood that a mother of a child aged 5 years or less is married to the child’s biological father is decreased, while the likelihood that she is never-married is increased by about 13 percent at the sample mean. Importantly, I find no effect on the likelihood that a mother is married at the time of childbirth, which suggests that paternity establishments that result from IHVPE programs influence marriage behavior post-childbirth. Given this finding, any analysis on a sample of mothers eligible to be asked Child Support (CS) supplement questions in the CPS is likely biased because of selection into the sample.3 In fact, I find evidence of positive selection: average measures of paternal demographic characteristics, child support agreements and father involvement improve among both married and unmarried (CS-eligible) fathers. Taken together, these findings imply that IHVPE programs lead to an increase in the “marriage threshold” in father quality. To study the net effects of IHVPE on father involvement, I consider private child health insurance provision, the only measure of involvement available for all children, regardless of 2 Because of changes to the CPS-CSS in the early 1990s, data collected in or after 1994 are not compatible with those from earlier survey years (Freeman and Waldfogel (2001)). 3 All household members aged 15 years or older who are a biological parent of a child in the household from an absent parent are asked CS supplement questions. 3 whether their parents are married or not.4 I find that overall, IHVPE leads to a 4% (at the sample mean) decrease in children’s private health insurance coverage. I also provide some suggestive evidence of negative effects on other measures of father involvement when accounting for selection out of marriage – fathers are marginally less likely to make any or on-time child support payments or to have joint custody, spend fewer days with their children, and are less likely to pay for childcare expenses for their children.5 Finally, using data from Annual Demographic Supplements of the March CPS over 1989- 2010, I show that IHVPE leads to a 3 percent (at the sample mean) increase in maternal labor supply. This result is consistent with the decrease in marriage as married women are generally less likely to be in the labor force than unmarried women (U.S. Census Bureau, 2011). Further, this finding suggests that IHVPE leads to a net decrease in monetary transfers from fathers to mothers and their children, which mothers must compensate for by working. My empirical results imply that paternity establishment might constitute a substitute for marriage post-childbirth, at least for some parents. To understand this mechanism further, I develop a simple model of marriage behavior and parental transfers post-childbirth, which is largely based on the theoretical foundations in Edlund (2011).6 In this model, I make the observation that within marriage, mothers obtain utility both from what the father transfers to the family in financial support and from his parental and partner qualities (which are affected by numerous factors observable by the mother such as his involvement in criminal activity, drug 4 In the CPS-CSS, all other questions regarding father involvement are only asked of CS-eligible mothers. However, private child health insurance provision can be seen as an important measure of father involvement. As I discuss in more detail in Section VII, the decrease in private health insurance coverage is driven entirely by a decrease in health insurance provision by members of the household, and is not compensated by any changes in children’s coverage by members outside the household. 5 This evidence is only suggestive as I must rely on some assumptions regarding the father involvement variables for families that are not in the CS-eligible sample in order to account for selection out of marriage. Specifically, I assume that marriage to the biological father consists of visitation rights, joint legal custody and complete and on- time “child support payments”, and that in marriage the father provides food, clothes, and gifts to the child, covers childcare and medical expenses, and spends the whole year with the child. Section VIII discusses these assumptions in more detail. 6 The key idea that forms the theoretical backbone of this line of work (which was developed in Edlund (1998), and then served as the basis for the conclusions in several papers including Edlund and Korn (2002), Edlund and Pande (2002), Edlund (2005), Edlund and Lagerlöf (2006), Chiappori and Weiss (2007), Chiappori and Orrefice (2008), Saint-Paul (2008), Francesconi, Ghiglino, and Perry (2010), and Bethmann and Kvasnicka (2011)) is the observation that an unmarried mother is a child’s sole parent by default, and therefore marriage serves as a transfer of paternity rights from the mother to the father. In essence, marriage can be seen as a contract for trade in children, where the father must make a positive transfer to the mother in marriage in exchange for rights to his children. Further, rights to children are usually “lumpy”, which allows for the possibility of out-of-wedlock fertility (see Edlund (2011) for a detailed discussion of this argument). 4 use, involvement with other women, and parenting obligations to children from other mothers, among others). On the other hand, as most unmarried mothers retain full custody of their children, a mother’s utility outside marriage is less sensitive to the father’s quality. Thus, for certain (lower-than-desired) levels of father quality, a mother may value a lower level of support outside marriage more than a higher level of support that the father would have transferred to her within marriage. Fathers offer transfers to mothers (and their children) in marriage as they can only obtain full utility from their children within marriage.7 Outside of marriage, fathers can pay a fixed cost to establish paternity, and then gain some partial rights to their children in return for making partial transfers (i.e., child support payments). IHVPE programs provide an easily accessible and very inexpensive mechanism for paternity establishment, as well as increased education of new unmarried parents about the fathers’ rights and responsibilities once paternity is established. Consequently, I model IHVPE introduction as an exogenous reduction in the fixed cost of establishing paternity, which in turn increases the level of partial support that mothers expect to get outside marriage. This results in an increase in the number of parents establishing paternity, which is driven both by parents who would have previously maintained no relationship and parents who would have previously gotten married. The switch out of marriage arises because in a certain range of father quality, it becomes optimal for more mothers to remain unmarried and obtain increased partial transfers instead of obtaining full transfers in marriage but experiencing lower utility from interacting with a lower-than-desired-quality father.8 In effect, the marriage threshold in father quality increases. This prediction is consistent with Edlund’s (2011) conclusions that an increase in the woman’s outside option (for example, her relative wage) should decrease marriage and that mothers may opt to share only partial rights with a low-quality father instead of sharing rights equally in marriage (for example, through cohabitation). The net effect on total transfers from the fathers to the mothers and children is ambiguous in the model. It depends on the relative magnitudes of the decrease in transfers from switchers 7 This element of the model arises because paternity presumption is an important feature in marriage (see Edlund (2006) for evidence from anthropology, sociology, and family law on this topic). 8 For example, it may be the case that a father only provides health insurance coverage for his child if he is married and within the household. However, for low levels of partner quality, the mother may be willing to forego the health insurance provision (i.e., “full support”) and remain unmarried if she now expects to receive at least partial child support outside marriage. 5 out of marriage and the increase in transfers from switchers out of no relationship. Thus, this framework suggests that the question of net impacts of IHVPE on measures of paternal transfers to the family (such as involvement with the child and maternal labor supply) is ultimately an empirical one. The model presented in this paper is a simplified version of Edlund’s theoretical work; it further makes a novel contribution by explicitly addressing parental transfers through analysis of father involvement and maternal labor supply, and by applying the framework to a specific policy instrument – IHVPE programs. In summary, this model predicts an increase in the rate of paternity establishments and a decline in parental marriage post-childbirth. The rise in the marriage threshold leads to an increase in the average characteristics of both married and unmarried fathers. Additionally, as the fathers who do not marry as a result of IHVPE would have provided higher transfers within marriage than they do outside marriage, the effects on net levels of father support and involvement when accounting for selection out of marriage are ambiguous and could potentially be negative. Finally, if there is an overall reduction in paternal transfers, mothers may compensate by increasing their labor supply. The results from the main analysis are consistent with the predictions of this model. The paper proceeds as follows. Section II discusses the IHVPE programs in more detail. Section III reviews the relevant background literature, while section IV presents the conceptual model. Section V discusses the data sources and presents summary statistics, while Section VI discusses the empirical methods. Section VII presents the main results, while Section VIII presents a series of robustness checks. Finally, Section IX concludes. II. In-Hospital Voluntary Paternity Establishment Programs The Omnibus Budget Reconciliation Act (OMBRA) of 1993 required all states to establish IHVPE programs, and these programs were then expanded by the 1996 Personal Responsibility and Work Opportunity Act (PRWORA). As a result, all states have initiated an IHVPE program, and all hospitals and birthing centers are currently required to provide adult unmarried new mothers and fathers with an opportunity to sign a voluntary paternity 6 acknowledgement form.9 State child support agencies are required to make available materials for educating parents, and hospital staff must provide mothers and fathers with both written materials and oral explanations regarding the rights and responsibilities related to paternity establishment. Additionally, in some states, minor parents are either not allowed to participate in the IHVPE programs, require parental consent to acknowledge paternity, or have more lenient rules for rescinding paternity within a short period after childbirth.10 Prior to the federal mandate, most states provided some kind of “voluntary acknowledgement” forms to new unmarried parents, but only in some hospitals. According to a survey of state child support agencies conducted by the Department of Health and Human Services, the forms had “no real legal significance” and many states “only kept forms at the public health office and didn’t promote the idea [of paternity acknowledgement]” (Department of Health and Human Services (1997a)). For the purposes of this study, I consider an IHVPE program initiated in a state only when the in-hospital voluntary acknowledgement of paternity process becomes part of the state’s legal code and/or the state implements a formal program that targets all hospitals and birthing centers in the state and involves education of new parents about the paternity establishment process (for example, the Paternity Affidavit Program in Washington or the Paternity Opportunity Program in New Jersey). Despite the federal mandate, the administration of the in-hospital paternity acknowledgement process is largely under state discretion. The variation in the timing of IHVPE program implementation across states stems largely from the length of time necessary to forge relationships between state child support agencies, vital statistics registries, and hospitals (Department of Health and Human Services (1997a)). By 1997, 37 states reported full implementation of IHVPE, while the rest listed reasons such as “too early for the [office of child support] staff to have contacted every state birthing hospital” to explain the delays (Department of Health and Human Services (1997a)). Since identification of the causal effects of IHVPE programs on paternity establishment and family behavior relies on the assumption that the timing of implementation is uncorrelated with other time-varying determinants of these outcomes, it is important to assess whether the differences in timing are related to other potential confounding 9 Both unmarried parents have to be present at the hospital to participate in IHVPE. According to data from the Fragile Families and Child Well-Being Study, over 1998-2000, 76% of unmarried mothers reported that the child’s father came to the hospital at the time of the child’s birth. 10 Specifically, in 2004, the following states had special provisions restricting participation for minors: CA, DE, IL, KS, KY, TN, TX, UT, VA, WI, and WY (Roberts (2004)). 7 variables. While it may be the case that early IHVPE implementers have more efficient administrative processes and more organized existing networks across state agencies, time invariant differences in these characteristics are absorbed by the inclusion of state fixed effects in my analysis.11 I also show that none of the results in this analysis is driven by any particular state (all regressions are robust to the exclusion of each state, results available upon request). Finally, given that the empirical evidence shows no correlation between IHVPE program initiation and numerous state time-varying characteristics of interest, it seems unlikely unobserved state time- varying omitted variables pose serious issues. Unfortunately, a unified source of information on the timing of IHVPE program implementation across states does not exist. For most states, I obtained information on the year (and month if possible) of program implementation from searches of state legal statutes on LexisNexis Academic, internet searches of state paternity programs, and direct conversations with officials at state child support agencies and IHVPE programs. Additionally, as Nepomnyaschy and Garfinkel (2007) have collected this information for several states, I use their data as well.12 Figure 1 shows the variation in the timing of IHVPE program implementation across states, while Appendix Table 1 presents more details for each of the 44 states in my data.13 Births in these states account for about 96 percent of all births in the United States over the time period of analysis. Until paternity is legally established, unmarried fathers essentially have no rights or obligations with regard to their children. Following paternity establishment, fathers usually have the right to refuse a requested adoption and block foster care placement for the child. They also have the right to request the court for partial custody and visitation rights. However, fathers have no rights regarding many decisions about their children’s well-being (such as consent over medical care) if the mother has sole custody. Finally, fathers are legally obligated to make child support payments once a court order has been established. Prior to the implementation of IHVPE programs, paternity establishment was a relatively uncommon and costly process that occurred through the court system, and most paternities were only established several years after the child’s birth, if ever (Office of Child Support 11 Further, differences in linear trends in such characteristics across states should be accounted for by the inclusion of state-specific time trends. 12 I thank Irwin Garfinkel and Lenna Nepomnyaschy for graciously sharing these data with me. 13 I do not have data for the following states: IA, MT, NH, NM, OK, WV, WY. 8 Enforcement, 1996). Policymakers speculated that IHVPE programs would be effective as they attempt to reach families during the “happy hour” in the hospital following the birth of the child and encourage the father to stay involved in his family’s life (Department of Health and Human Services (1997b)). Figure 2 plots the trend in the total number of paternities established in the United States over 1992-2007, and the substantial increase from about 600,000 to over 1.5 million in the late 1990s coincides with the time when most states implemented IHVPE programs. However, rigorous research on the causal effects of the IHVPE programs on paternity establishment rates is quite sparse. The existing literature is limited to several reports on individual state programs (for example, Pearson and Thoennes (1996); Ovwigho, Head, and Born (2007); Wisconsin Bureau of Child Support (2010)) and analyses of a few states and over short periods of time (Turner (2001); Sorensen and Oliver (2002); Mincy, Garfinkel, and Nepomnyaschy (2005)). To my knowledge, this paper is the first to examine the effectiveness of these programs for a large number of states and years, and to use methods that can arguably identify true causal effects of the programs that are not confounded by unobservable factors like variation in the composition of unmarried births. By uncovering the causal effects of IHVPE programs, this paper can shed light on how paternity establishment at birth may impact the decisions of unmarried parents regarding involvement with each other and their children. Further, analysis of the causal effects of IHVPE on marriage behavior in particular can reveal the trade-offs in parental marriage decisions and thus have important implications for the impacts of the more recent marriage promotion programs. III. Background Literature A. Child Support Enforcement Literature The implementation of numerous child support enforcement measures (which include IHVPE programs, as well as automatic wage withholding, the new hires directory, and license revocation for non-payment among others) throughout the 1980s and 1990s across states created a “natural experiment” for researchers to study their overall effects. As a result, there is a wealth of literature that focuses on the effects of child support enforcement on numerous family and 9