National Poverty Center Working Paper Series #12 – 02 January 2012 Are Interstate Cases Still the ‘Black Hole’ of Child Support Enforcement? Effects of the Uniform Interstate Enforcement Act Elizabeth T. Powers University of Illinois, Urbana ‐ Champaign This paper is available online at the National Poverty Center Working Paper Series index at: http://www.npc.umich.edu/publications/working_papers/ This project was supported by the National Poverty Center using funds received from the U.S. Census Bureau, Housing and Household Economics Statistics Division. The opinions and conclusions expressed herein are solely those of the author(s) and should not be construed as representing the opinions or policy of any agency of the Federal government. Are Interstate Cases Still the ‘Black Hole’ of Child Support Enforcement? Effects of the Uniform Interstate Enforcement Act ELIZABETH T. POWERS University of Illinois at Urbana-Champaign epowers@illinois.edu January 2012 ABSTRACT: So long as child support enforcement was entirely the legal domain of the states, it was nearly impossible to pursue claims across state lines, and interstate claims were characterized as the “black hole” of child support enforcement. The Uniform Interstate Family Support Act (UIFSA) clarified lines of authority, opened state IV-D agencies and courts to interstate claimants, and invented powerful new tools for pursuing cross-state claims. This paper uses Survey of Income and Program Participation data spanning the reform era to assess the success of this policy. The potential endogeneity of interstate moves with the policy regime may bias conventional regression estimates. A conditional difference-in-difference matching estimator is implemented instead. The findings indicate greatly increased administrative enforcement activity for interstate cases subsequent to UIFSA. This activity increased formal support agreements and identified greater amounts of support owed. There is also evidence of increased interstate collections and a closing of the ‘black hole’. Support collections increased especially for welfare-receiving households, but nonwhite households and households with nonmarital births do not appear to be helped by UIFSA. KEYWORDS: Child support policy, child support enforcement, single parent families, welfare, matching estimators ACKNOWLEDGMENTS: Development of this project was supported by a grant from the National Poverty Center at the University of Michigan with funds provided by the U.S. Census Bureau, Housing and Household Economics Statistics Division. The opinions and conclusions expressed herein are solely those of the author and should not be construed as representing the opinions or policy of the National Poverty Center or of any agency of the Federal Government. Maghaisvarei Sellakumaran and Emilie Bagby provided able research assistance. I am grateful to Yunhee Chang, Austin Nichols and other participants at a joint Census-National Poverty Center, Don Fullerton, Petra Todd and economics department seminar participants at Purdue University for comments on various iterations of this research. 1. Introduction Policymakers have long hoped that child support would serve as a major income source for families with children. Over the past several decades, the federal government has attempted to correct widely acknowledged weaknesses in the establishment and enforcement of child support claims through a series of system reforms. One of the first major enforcement reform initiatives, a requirement that states garnish the wages of uncooperative obligors, ocurred in the mid-1980s. From the mid-1990s there was a marked shift in responsibility from the courts to administrative agencies that was integrated with the introduction of important new enforcement tools (see Huang, 2002, Sorensen and Hill (2004), and Pirog and Ziol-Guest, 2006, for overviews). The near-impossibility of pursuing claims through legal or administrative means across state lines prior to the mid-1990s was a major exception to a trend of continually more coercive child support policy. Prior to reform, interstate cases were characterized as the “black hole” of child support enforcement (Lerman, 1993; Haynes, 1996). Improvement of interstate collection was a consequential issue, as interstate cases were estimated to comprise as much as one-quarter to one-third of all support cases in the early 1990s (U.S. GAO, 1992). The Uniform Interstate Family Support Act (UIFSA) was intended to not only put interstate and in-state cases on an equal footing by clarifying lines of authority and priority in both the legal and administrative realms but also to create powerful new tools for pursuing interstate claims (see Haynes, 1996, for details). The 1996 Federal welfare reform essentially mandated states’ universal adoption of UIFSA, and all states did so by 1998. This study uses rich, nationally representative data from multiple panels of the Survey of Income and Program Participation (SIPP) spanning the decade from 1993 to 2003 on child support claims, collections, and claimant contacts with the enforcement system in order to assess how well the objective of closing the black hole of interstate cases was met. UIFSA removed the legal and administrative obstacles to interstate case establishment and enforcement. If UIFSA worked as intended, outcomes of interstate claims should have converged to patterns more typical of instate claims after 1998. There is little research on interstate child support enforcement and none that is nationally representative (for research using administrative data for Illinois, see Chang, 2003). The SIPP is 1 an unusually rich data source but little used for child support research. Peters, Argys, Howard, and Butler (2004) use the SIPP to study whether improved child support collections increase paternal contact with non-resident children. However, their work focuses exclusively on nonmarital cases and does not contain a specific test of UIFSA’s impact on interstate case enforcement. Rather, the UIFSA adoption status of the obligee’s state of residence appeared as one of several instrumental variables for child support collections. In fact, much of the enforcement literature relies on a small number of cases in the National Longitudinal Survey of Youth (NLSY) that are tracked into the early 1990s, prior to the most significant reform efforts. Research on the effect of UIFSA on interstate cases faces a methodological challenge. If moving behavior is endogenous with UIFSA adoption, estimates of the effectiveness of UIFSA derived from a standard difference-in-differences (DID) analysis comparing movers’ and non-movers’ outcomes before and after implementation are biased. The direction of the bias likely depends on the identity of the mover; that is, whether the interstate case status was created by the move of the parent who owed or was owed support (obligee or obligor).1 Specifically, prior to the implementation of UIFSA, uncooperative obligors may move across a state line to evade enforcement. If so, the pre-UIFSA-implementation sample of interstate cases is negatively selected with respect to the latent propensity to cooperate with child support. Selfselection of the least compliant or cooperative obligor-parents as ‘movers’ associates poor enforcement outcomes to the pre-UIFSA period. UIFSA implementation reduces or eliminates the advantage of an out-of-state move to an uncooperative obligor. If moving is strategic, UIFSA implementation reduces negative selection of interstate movers, creating the appearance of improving enforcement under UIFSA in the data. A standard difference-in-difference analysis would overstate the effectiveness of UIFSA policy.2 1 Self-selection of movers generally is not necessarily problematic. For example, movers often have better economic opportunities. If this is not fully captured with the controls, then estimates of the interstate gap prior to UIFSA are biased downward, but the estimated policy effects of UIFSA are unbiased so long as the process governing the beneficial selection of movers is stable over the sample period. 2 Interstate cases associated with obligee parents who move across a state line prior to UIFSA are likely beneficially selected. Only obligees without some confidence that the obligor parent will honor his obligation would risk an interstate move prior to UIFSA. If UIFSA is effective, obligee parents with uncooperative former partners are made no worse off by a move. Therefore, in a sample of interstate cases originating with the obligee parent’s move, the standard DID approach understates the effectiveness of UIFSA implementation. 2 In the absence of compelling instrumental variables or natural experiments to identify the moving decision, I employ a conditional difference-in-differences matching estimator. Under the assumption that the counterfactual difference in outcomes between movers and nonmovers in the absence of UIFSA is time-invariant, this method provides an unbiased estimate of the effect of UIFSA policy on child support outcomes. Matching is performed separately within the subsamples of both interstate and within-state cases, based on the relationships of outcomes to observables in the pre-UIFSA-implementation era. The pre-post differences that are constructed for the interstate and within-state cases are then contrasted, as in a conventional DID analysis. To preview the findings, there is strong evidence that UIFSA implementation increases interstate obligees’ access to administrative enforcement. After UIFSA implementation, interstate obligees who approach state IV-D agencies for help report much higher rates of satisfaction. IV-D intermediation of interstate payments also increases subsequent to UIFSA implementation. There is also strong evidence that UIFSA implementation increases the incidence of formal support arrangements for interstate cases and increases the amount of back support identified as owed. The incidence of payments to households potentially needing child support in interstate situations rises after UIFSA implementation. Under reasonable sample restrictions, the estimated effects of the policy on interstate payments are large enough to completely close the pre-UIFSAera interstate payment black hole. The next section provides essential background on the child support enforcement system and details the problems UIFSA addresses. Section 3 presents the predicted effects of the universal adoption of UIFSA by the states on access and enforcement outcomes. Section 4 explains the empirical strategy for identifying the effects of UIFSA implementation. Section 5 discusses the data source and construction of the variables. The main findings are presented in section 6, while the robustness of the main findings to various changes in samples and specifications is discussed in section 7. Conclusions and a discussion of the findings are presented in section 8. 2. Background on Child Support Enforcement 3 Although some parents are able to amicably and privately agree on support and self-enforce their agreements, obtaining and enforcing child support is more often a protracted, multi-step process involving third parties.3 The necessary steps are roughly as follows. First, legal paternity is established at birth or conceded later. A claim for financial support is made on the potential obligor and an award amount is established by a local court, following guidelines laid out in state law. The basis for the award amount is often a formula that takes into account the numbers and ages of the children whose support is in question, parents’ incomes, custody arrangements, and other factors (see Neelakantan, 2009, Argys, Peters, and Waldman, 2001, and Argys, and Peters, 2003, for discussions of guidelines). The court then issues an order to pay. If the obligor ignores this order, the obligee can seek enforcement, although if the obligor is unemployed or incarcerated, he may be excused from payment.4 At many stages of the process, it is extremely difficult for the case to progress if the obligor’s whereabouts are unknown. Obligees may pursue paternity establishment, awards, and payments with the aid of public and private entities. Welfare recipients are required to assign their child support rights to their state’s child support enforcement (IV-D) agency with the aim of recouping the state’s welfare and medical expenses for the child. Obligees outside the welfare system can either voluntarily obtain assistance from their state’s IV-D agency, possibly incurring a modest fee, or use a private attorney. In addition to the value of having the IV-D agency act as an intermediary in the court system, administrative enforcement authority permits the IV-D agency to take action without consent of the courts or other state entities. Over time IV-D agencies have gained the authority to order paternity tests; intercept obligor’s unemployment insurance payments, income tax refunds, and lottery winnings; suspend driver’s and professional licenses; and deny passport applications. Another important feature of IV-D operations is the requirement since 1990 to collect child support for new or modified support orders through wage withholding. Pirog and Ziol-Guest (2006) provide an overview of IV-D services and characteristics of the IV-D caseload. 3 It is not known whether support arrangements and payments are voluntary in the SIPP. Argys and Peters (2003) estimate that 21 percent of child support agreements in a small NLSY sample are set by parents alone. These agreements appear largely self-enforcing; almost 80 percent of child support amounts promised are reported received when the parents work exclusively together. Another 35 percent of cases use an attorney to arrive at an agreement. It is not possible to determine the share of those agreements that are still essentially cooperative, but the share of promised payments received drops to 61 percent in this group. The remaining 44 percent of agreements come through a court. 4 Throughout, I simplify language by assuming that the father is obliged to pay child support and that the child lives in the mother’s household. While this is not always the case, it is by far the typical case in these data. 4 The difficulty of enforcing child support when parents live in different states has deep roots in the development of U.S. family law, traditionally the jurisdiction of state and local governments. The laws and court orders of each state are inapplicable and unenforceable in another’s. As a consequence, prior to UIFSA, a parent residing in state A who sought to enforce the child support order of state A’s court faced insurmountable obstacles if the obligor moved to state B. State A had no jurisdiction in state B, and was therefore powerless to enforce its order. Even if state B were willing to assist with enforcement, it was unclear whether the award should conform to the laws of State A or State B. While the parent residing in state A would benefit enormously from access to the services of state B’s IV-D agency, as a nonresident she had no claim to these services whatsoever. Even had they been willing to entertain the obligee’s claim in principle, State B’s courts and IV-D agency might have been reluctant to voluntarily enforce state A’s court order on behalf of a nonresident for other reasons like cost. UIFSA was created in 1992 to directly address and resolve conflicting jurisdictional issues and the lack of cooperation and coordination in states’ enforcement efforts.5 UIFSA resolved jurisdictional problems by requiring all states to defer to the child support order originating in the child’s state of residence. This state was assigned a continuing exclusive jurisdiction over the order, prohibiting its alteration by other states’ courts.6 Beyond demarcating the authority for establishing and modifying awards, UIFSA obliged not only state courts, but also state IV-D agencies, to fully recognize the authority of another state’s orders. As a result, Courts and IV-D agencies were compelled to give out-of-state and in-state cases equal priority (see DeMaria, 1999, for a detailed analysis of jurisdictional issues under UIFSA). UIFSA actually grants preferential status to out-of-state cases, because it permits the obligee’s attorney to send a child support order directly to the obligor’s employer, bypassing the court and administrative systems entirely, immediately triggering wage withholding and health insurance 5 Title IV-D of the Social Security Act of 1975 provided federal funding for child support enforcement and paternity establishment but was careful to preserve state autonomy in child support enforcement program design. Prior attempts at coordination of state systems were ineffective, with so-called “uniform” laws applied differently from state to state (particularly with respect to arrears and medical support; OCSE, 2008), and states permitted to adopt only parts of the Acts. 6 During the transition to UIFSA involved states were directed to cooperatively determine a single controlling order for cases with multiple orders. 5 coverage.7 “Enrollment of the child in the health care plan at the employee-obligor’s expense is not dependent on the obligor’s consent, any more than withholding a sum certain from the obligor’s wages is subject to a veto. It is up to the obligor to assert any defense to prevent the employer from abiding by the medical support order (UIFSA, 1996, p.50).” Finally, locating the obligor is a common problem in interstate cases. UIFSA mandates that law enforcement officers in the obligor’s state (usually either the state’s attorney or state attorney general) seek the obligor’s location from an exhaustive list of sources, including state and federal records. Use of the obligor’s Social Security number, as listed in the Act, reveals not only the obligor’s address but also the identity of his employer, expediting collection through wage garnishment. Under welfare reform, the IV-D agencies are mandated to provide locator services. UIFSA was drafted by the National Conference of Commissioners on Uniform State Laws, enacted in 1992 with revisions enacted in 1996 and 2001. The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (“welfare reform”) mandated the adoption and implementation of UIFSA by all states by January 1, 1998. Universal adoption was promptly achieved under the threat of a complete loss of federal child support enforcement funding. In 1993, fewer than 10 percent of at-risk households (those containing a child whose other parent is absent from the household) were under UIFSA. By 1995, 27 percent of these households were covered by voluntary state UIFSA adoption. By 1997, two-thirds of at-risk households were covered. All at-risk households have been covered by UIFSA since 1999.8 3. Predicted Effects of UIFSA Adoption UIFSA’s focus is on legal and administrative systems for child support enforcement, and it is possible that important causes of non-payment are unrelated to these systems. For example, the most common reasons given for the absence of a child support agreement in the SIPP are that the 7 UIFSA requires medical support arrangements to be specified in the child support order. They can be provided either through a cash payment or enrollment as a dependent in a health plan. 8 Author’s computations using 27,865 observations on households eligible for the child support edit (i.e., respondent is parent of child under age 21 whose other parent is absent from the household) from the 1992, 1993, 1996 and 2001 SIPP panels. Statistics are weighted by wave 1 household weights, as recommended by Census Bureau staff. 6 potential obligee did not try to obtain support, that the potential obligee did not want support, and that the potential obligor was impoverished and unable to pay. UIFSA implementation may have little or no effect on these causes of nonsupport. Self-reported reasons are difficult to interpret, and this research focuses on objective measures of activity and outcomes. By opening access to state IV-D agencies, UIFSA dramatically reduces obligee’s transaction costs for pursuing interstate claims. By establishing the authority of the originating state’s order both with the obligor’s state’s own IV-D system and his employer, UIFSA implementation also greatly increases the potential for involuntarily collection from an interstate obligor. Clarification of state lines of authority, access to IV-D services in an era of heightened administrative enforcement powers, and a prioritization of locate services (a problem disproportionately plaguing interstate cases) are all predicted to contribute to a positive overall effect of UIFSA on the incidence and amount of established awards and collections made for all at-risk households. In contrast, the overall impact of UIFSA adoption on the average awards and payment amounts from interstate obligors, conditional on owing any payment, is theoretically ambiguous. By lowering transaction costs, UIFSA rationalizes the pursuit of smaller claims. The average effect on award amounts and payments also depends on how the cases pursued under UIFSA are weighted between “deadbeats” and “turnips” (Mincy and Sorenson, 1998). As one can’t get blood from a turnip, one can’t garnish earnings where there are none. If the problem of preUIFSA interstate enforcement is largely that better-off obligors who can afford to pay do not (the “deadbeats” problem), average awards and payments rise as it becomes possible to pursue these obligors effectively under UIFSA. Alternatively, if the powers of the new interstate enforcement regime are largely directed at low-income “turnips”—those unable to pay –average awards and payments for those identified to provide support are predicted to fall. Mandating that the courts and administrative agencies undertake interstate enforcement, providing easier and faster methods of collection from uncooperative obligors, and lowering barriers to using IV-D agencies by interstate claimants should all increase third-party intermediation in interstate cases. General equilibrium effects of changes to the legal environment also suggest that intermediation increases with UIFSA implementation. Argys and Peters (2003) lay out models of cooperative and noncooperative parental bargaining over child 7 support and show that when state enforcement effectiveness increases, obligees negotiating for support are attracted to the enhanced ‘default’ position of state enforcement under state award guidelines. UIFSA implementation is predicted to increase intermediation by increasing the equilibrium level of non-cooperative, state-enforced, interstate agreements. Finally, given the increase in administrative enforcement and the advantages it offers to obligees around the era of UIFSA adoption, it is seems plausible that intermediation through IV-D agencies dominates intermediation by the courts subsequent to UIFSA adoption. If state IV-D agencies implemented UIFSA as envisioned, obligees should have turned to out-ofstate IV-D agencies (either directly or through the intercession of their own state of residence’s IV-D agency) for help with claims and enforcement. If information about improved IV-D assistance was disseminated to interstate obligees, UIFSA implementation might have generated more help requests in out-of-state cases. Conditional on requesting help, the equal priority of outof-state obligees under UIFSA ought to have increased the frequency of successful resolutions to interstate enforcement problems. Finally, UIFSA requires arrangements for medical support and introduces a powerful new tool for its collection. If effective, these policies should lead to more frequently specified arrangements for medical support in agreements and orders and increase the collection of medical support conditional on its inclusion in the order for interstate cases. 4. Empirical strategy Endogenous self-selection of movers with respect to UIFSA’s nationwide implementation may severely bias simple D-in-D estimates of UIFSA’s effectiveness. A matching approach is an alternative that solves this problem under specific assumptions. 4.1 The conditional differences-in-differences matching estimator It is convenient to introduce some simple notation. Designating the average child support outcome as Y –assumed throughout the discussion to be conditioned on X, a set of observables— child support outcomes in the pre-UIFSA-implementation era are specified asYN = Y and YM = 8 Y + 𝛾 + η. The subscript M indicates a situation where the parents live in different states, while N indicates ‘within-state’ status. The parameter 𝛾 is the influence of sample selection on the expected outcome due to strategic interstate moves.9 In other words, 𝛾 is the influence of the preponderance of ‘uncooperative’ cases in the subsample on the expected outcome Y. The parameter η is the unobservable difference between interstate movers and nonmovers that exists independent of the UIFSA regime. Designating the post-UIFSA-implementation era with prime superscripts, outcomes in the postimplementation era are specified as is = Y+ τ and 𝛾 . The key parameter ’, the change in interstate case outcomes due to the policy change. The parameter τ is a secular time trend, while the parameter 𝛾 is the post-UIFSA-implementation selection bias in the sample of movers. The crux of the identification problem with respect to the endogeneity of moving behavior and UIFSA policy is that 𝛾 and 𝛾 may differ. This poses a problem if differences between movers and nonmovers are changing over time for reasons other than UIFSA policy; i.e., might differ from zero. A within-group matching estimator is implemented by matching interstate cases in the postUIFSA-implementation regime with interstate cases in the pre-UIFSA-implementation regime, using the estimated propensity scores derived from a matching function. This matching function is estimated from the sample of pre-UIFSA-implementation interstate cases (alternatively, instate cases). Let outcomes for the matched sample be indicated ̂ . Without further assumptions, ̂ + τ and ̂ τ. The difference 𝛾 𝛾) –the change in the selection bias due to strategic moving—is eliminated by the matching method, because the selection process is fixed at the pre-UIFSA-implementation regime. Taking the difference-indifference eliminates the secular time trend, τ. The policy effect of UIFSA is clearly not identified unless The assumption that is zero. is zero is readily interpreted within the framework of the matching literature, which emphasizes how the data at hand relate to the unobserved counterfactuals (treatment of untreated groups or non-treatment of treated groups). The difference-in-difference 9 To avoid excessive notation, it is assumed that the share of movers is small (as is empirically true), so that selection bias is an inconsequential issue in the sample of nonmovers. 9 (DID) matching estimator identifies under the assumption that the counterfactual difference between movers’ and nonmovers’ outcomes if UIFSA had never been implemented is independent of the UIFSA implementation regime. If strong differential time trends affect the movers and nonmovers—that is if (η’- η) is not zero—knowledge of UIFSA regime helps predict if an observation belongs to the mover or nonmover group, violating the assumption underpinning matching techniques.10 Note that the assumption =0 is no stronger than those that typically support matching estimators in the literature. This DID conditional matching estimator differs importantly from the one described in Heckman, Ichimura, and Todd (1998), which cannot identify θ in this case. Following Heckman et al.’s (1998) procedure, one would match nonmovers to movers within each (preimplementation/post-implementation) period, using a matching function estimated from each period’s sample of movers. Under the assumption that the counterfactual change in the outcome in the absence of any treatment is orthogonal to treatment group membership (conditional on X), this method controls for unobserved differences in movers and nonmovers that are not attributable to the policy change. However, Heckman et al.’s (1998) method does not control the movers’ sample composition changes due to UIFSA implementation, because the policy change is reflected in each period’s matching function. Matching according to Heckman, et al. (1998) yields, in effect, ̂ 𝛾 and ̂ 𝛾 and a DID estimator 𝛾 𝛾 . This DID estimate of the policy effect is contaminated by the UIFSA-driven composition effect. What this estimator does have in common with other conditional DID estimators is the advantage, touted by Heckman, et al. (1998), of the feasibility of matching on outcomes, as opposed to, say, the propensity to make an interstate move. This is due to the fact that the sample mean of Y is expected to change over time, even conditional on matching, due to random shocks and trends unrelated to the treatment. 4.2 Implementation of the matching estimator and evidence on match quality 10 Note changing selection of the samples due to UIFSA itself does not violate the assumption; the assumption is based on the (counterfactual) difference in outcomes in a world without UIFSA implementation. 10 There are many options for deriving the propensity score used to match the subsamples. Matching on outcomes is appealing because outcomes incorporate the quality of ‘noncooperation’ that drives the endogenous negative selection of interstate moves prior to UIFSA implementation. In quite a few cases, however, the number of positive observations of the outcome are sparse, preventing reliable estimation of a propensity score. Throughout this paper, a variable indicating whether a payment has been made in the past 12 months serves as the basis for the propensity score. This variable captures the degree of cooperation that potentially underlies the troublesome sample selection issues associated with UIFSA implementation. In effect, the matching estimator strives to compare cases that are equally ‘uncooperative’ in the pre and post-UIFSA implementation eras. The exact matching technique implemented uses local linear regression method and restriction over common support. Local linear regression uses a tricube kernel weighting scheme (i.e., locally weighted polynomial regression, also known as LOWESS) to assign weights to all observations based on their similarity according to the matching criterion. Appendix A presents the raw difference-in-difference calculations for the explanatory variables used in the estimation. D-in-D statistics that are insignificantly different from zero indicate that the samples are balanced over the two periods with respect to that factor.11 In general, there is more evidence of balance as the samples are more tightly restricted on observed outcomes, as one expects. Table A also presents recalculated DID statistics employing the weights from the conditional DID matching estimator. There is considerable improvement in balance, although evidence of imbalance is largely but not entirely eliminated by the matching process for the least restricted sample. When households with a written agreement are considered, the weighting scheme is quite successful at rebalancing the sample. For those with a written agreement who are owed a payment in the past 12 months, reweighting is moderately successful at rebalancing; no DID statistics are significantly different from zero at the 95 percent confidence level or above. Finally, when the sample is restricted to those seeking help with child support from a state 11 Under the DID conditional matching strategy, it is innocuous for the mover sample to be imbalanced between periods, so long as the nonmover sample is imbalanced in a similar fashion. Intuitively, similar imbalance across treatment and control groups suggests that the imbalance does not spring from UIFSA implementation policy. What matters is how well the matching process balances trends in the variables. 11 agency, the DID PPM weighting scheme balances the sample quite well. Appendix A provides further details and discussion. 5. Data source and variables The SIPP is a nationally representative survey of U.S. households. The sample for this analysis draws from the 1992, 1993, 1996, 2001, and 2004 panels and spans the calendar years 1993 through 2003. Each panel of households is interviewed every 4 months over an approximately 2-year period. Respondents with a child under age 21 are eligible for a survey module covering matters related to child support (the “child support edit”) if that child has a parent living outside the household.12 This criterion defines the group of households ‘at risk’ for enforcement services. It is unknown whether the surveyed parent is the primary caregiver of the child, who is usually the parent most appropriately awarded financial support. SIPP respondents are asked if a child who “lived with” them had another parent outside the household. Lin, Schaeffer, Seltzer, and Tuschen (2004) provide evidence that parent-respondents interpret the term “live with” (as opposed to ‘stay with’) as a permanent arrangement, meaning it is likely that most SIPP household parents answering the child support questionnaire are the child’s primary caregiver. The child support topical module is usually administered twice during the course of each panel, at a one-year interval. There are four major samples used in the primary analysis, and they are the sample of 22,544 households at risk for child support, a sample of 12,539 at-risk households with a written childsupport agreement, a sample of 10,593 at-risk households with a written child support agreement who are owed a payment over the prior 12 months and a sample of 6,098 at-risk households with a written child-support agreement that request help with child support from a state IV-D or welfare agency. Appendix Table B provides the sample statistics for these variables. Interstate cases created by the move of an obligee are excluded from all samples. The share of interstate cases created by the move of a potential obligor ranges from 16 to 22 percent of the samples. 5.1 Information on child support 12 If the other parent’s absence is military or job-related, the household is ineligible for the child support edit. 12 The SIPP child support module is complex and highly detailed. Parents are asked about the existence of written or verbal agreements for each and every at-risk child. The empirical analysis often focuses on child support cases with a written agreement, a status that restricts the universe of many important questions.13 Most information is not available for each and every at-risk child in a household. Support questions sometimes pertain to either the youngest or oldest child eligible for the child support edit, but most important questions refer to “any or all” agreements for any or all of a household’s children in the child-support edit. The dependent variables reference either any at-risk child in the household or all at-risk children in the household. The key areas of child support information are award establishment, payments owed and made, help-seeking for various problems from IV-D agencies, payment intermediation, and health care arrangements. The key establishment variables are having a written agreement for any child or having a court-ordered agreement for any child. The major award and payment variables used are whether a payment was due during the 12 months preceding the interview date, the amount owed over that period, whether at least one payment was received in the past 12 months, and the amount received over that period (amounts are cumulated over all children in the family with a payment due in the past 12 months under a written agreement). I use the ratio of amounts paid and owed over the past year as a payment yield variable. Arguably, this measures the efficiency of the child support system in harvesting money that is owed. Argys and Peters (2003) also argue that controlling for support owed may be important, and this variable implicitly does that. However, since both support and awards can change, it is also reasonable to evaluate the “success” of enforcement by effects on these variables alone as well. In addition, there is some information on the total amount of back-support owed prior to the previous year, cumulated over an indefinite period, for all households with a payment due in the past 12 months under a written agreement. Households at-risk for enforcement who do not have a written agreement are assigned zeros for payments owed and made and for the payment ratio. Questions about help sought and obtained from a state IV-D agency are put to those who report they were due a payment in the previous 12 months pursuant to a written agreement for any child. While specific questions are also asked about the specific areas of enforcement, order, 13 Paternity establishment questions are only asked of potential obligees with verbal agreements and are not analyzed. 13 locate, modify, paternity establishment, medical support establishment, and other, the universe restriction is unfortunate in this instance, as those who have a written order and are owed support have already achieved a large measure of success in child support establishment. Information is provided on the extent to which payments are intermediated. Respondents in the child support edit who are due a payment in the past 12 months for any child under a written agreement are asked about the source of the payment (the entity from which the payment was received). Permissible responses are directly from the parent, through a court, through a welfare or child support (IV-D) agency, or no payment received. In addition to detailed questions about payments owed and made, the SIPP asks about arrangements for children’s health care. Respondents due a payment in the past 12 months under a written agreement for any child are asked which parent is to provide health insurance, whether there is another provision for medical care, or whether health care is not specified in the support agreement. There is no information on whether these agreements are successfully enforced. State’s UIFSA implementation dates are matched to the SIPP households by state of residence. The key policy variable indicates the in-sample UIFSA regime (pre- or post-implementation). Since most of the child support questions refer to written agreements with a payment due in the past year, implementation of UIFSA at the survey date is the relevant policy variable. Estimates using alternative UIFSA policy variables matched to the date of case origination and the most recent date of modification of the child support agreement differ little and are not discussed. 5.2 Other information from the SIPP The SIPP collects extensive information on all household members. Three broad categories of exogenous variables are defined: characteristics of the youngest child covered in the child support edit, characteristics of the responding parent, and household-level characteristics. Characteristics of the youngest child covered in the child support edit are a binary variable indicating sex, binary variables for the age categories 0-5, 6-12, and 13-17, and a binary variable for race (white or other).14 Characteristics of the responding parent are age (years), a binary 14 Demographic information focuses on the youngest child under the assumption that, in the presence of multiple children covered by the child support edit, the youngest child is likely to have the most activity on its case. 14 variable indicating sex, a binary variable for whether the responding parent was younger than 18 years of age at the birth of their oldest child covered by the child support edit (i.e., early motherhood), a binary variable for education (no high school diploma), and a binary variable for never married. Household characteristics include binary variables indicating the household parent’s total number of own children at home (defined as one or two children and three or more children), binary variables for the number of children 0-17 covered in the child support edit (defined as one, two, three, and more than three), an indicator for the household’s food stamps receipt, an indicator for household cash welfare receipt, a binary variable indicating the parent is presently married, and a binary variable indicating the household’s second appearance in a panel. The SIPP longitudinal framework is short and there is scant information on potential obligorparents. Household parents are asked about the location of the absent parent for all children for whom there is a child support agreement, written or verbal. For children without agreements, the question is asked about the absent parents of both the youngest and oldest household children in the child support edit. I combine these two sources of information to construct an indicator of the absent parent’s location for the entire at-risk sample. If the absent parent of any child in the household is resides in another state, the household is coded as having an interstate case. 5.3 Empirical evidence on the ‘Black Hole’ Cross-sectional difference regression estimates give a sense of the black hole of interstate child support prior to UIFSA. The coefficients on the ‘interstate case’ variable, restricting the sample to households that have not yet been exposed to UIFSA, are presented in Table 1 (the regression specifications also include all the explanators described in section 5.2). In general, these estimates are biased because unobserved differences between movers and nonmovers are uncontrolled. Nevertheless, they may give a rough idea of the magnitude of the problem faced. They are an appropriate benchmark for the conditional matching DID estimates, because the conditional DID matching estimate adjusts the post-UIFSA-implementation samples to match the degree of ‘noncooperation’ of pre-UIFSA-implementation interstate movers. Panel A covers help requested and reports of help received. The first row indicates that at-risk households with interstate cases request help at higher rates than other at-risk households. Although interstate at-risk households report more help received, this is due to their higher 15 incidence of help requests, as the findings in the next row reveal. Conditional on requesting help, interstate-case households are much less likely to report that the IV-D agency helps them with their problem. The findings on specific help requested (row 2) indicate that interstate-cases’ help requests are especially focused on locate services, at a rate double that of other households.15 The last row of panel A conditions the sample on the type of help requested and presents estimates of the rates of help received for specific problems. In the case of locate services, interstate-case households are more likely to report receiving help with their problem. Findings for other individual help categories are not estimated to be significantly different from zero. Panel B presents findings on payment intermediation. Interstate households are more likely to receive payment through an institution (the court or IV-D agency) and less likely to receive payment directly from the child’s absent parent. The estimated difference in the incidence of direct-from-parent payments in interstate cases is large, amounting to more than one-quarter of the sample mean of parent payment incidence for all households. Panel C presents findings for variables that describe whether there is a written agreement and the specific features of existing written agreements. The findings in the first row indicate that interstate-case, at-risk households are less likely to have a written agreement of any kind (including a non-court-ordered written agreement, which may include ‘voluntary’ agreements).16 Among all at-risk households, total liability for child support over the past calendar year is somewhat lower for interstate-case households (the point estimate of -$73.26 is around 6 percent of the sample mean), but the total amount of support owed that has accumulated from preceding years is considerably higher, with a point estimate of $250.07 that approaches 50 percent of the sample mean. Conditional on a written agreement, there is evidence of a lower incidence of voluntary agreements for interstate cases (this is also true when the sample is restricted to those owed a payment last year; the point estimate is around 16 percent of the sample mean in this case). There is not a significant difference in support owed over the past calendar year according to interstate status conditional on a written agreement, but back support identified as owed from past years is much greater for interstate cases. This is consistent with interstate cases not being 15 The point estimate is a 0.131 differential in locate requests for interstate cases in contrast to a rate of locate requests among all help-requesting households of 0.149. 16 These variables are coded to zero for households with no written agreement. 16 enforced over longer periods than other cases, as this liability has time to build. The estimated differences for the samples of those with a written agreement and those owed a payment amount to two-thirds or more of the mean back support liability estimated for the entire sample. The last three columns of Panel C indicate health care arrangements. Interstate-case child support agreements are significantly less likely to specify that the obligor is responsible for the child’s health care expenses (the ‘interstate gap’ relative to the mean for the entire sample is almost 15 percent), and more likely to specify that the obligee is responsible or that an unspecified arrangement has been made. Panel D indicates the payment gaps between in-state and interstate cases. Households with interstate cases are much less likely to report any payment was received in the past year (the point estimate is over 20 percent of the sample mean in absolute magnitude), although the gap in payments amounts over the past year conditional on having a written agreement or being owed a payment is not significantly different from zero. This suggests that the interstate payment gap of $111.48 reported for the entire at-risk sample (approximately 13 percent of the sample mean) is largely driven by the absence of written agreements for interstate cases, rather than nonpayment or underpayment on established agreements. However, if payments are conditioned on award amounts, there is evidence of underpayment in interstate cases, even for those with an award. The payment ‘yield’ (i.e., the ratio of payments made to payments owed during the last year) is significantly lower for interstate cases across all samples. While this is strong evidence of for ineffective collection in interstate cases, the magnitude of the interstate gap in the payment ratio is modest, at about 10 percent of the sample mean. 6. Major Findings This section presents and contrasts the conditional DID matching model estimates with raw DID and DID estimates from simple linear regressions.17 The findings are organized into four domains; (1) help sought and received from state IV-D agencies; (2) payment intermediation; (3) agreements on payments and health care; and (4) cash support received. 17 Bertrand, Duflo, and Mullainathan (2004) criticize D-in-D regressions in longer repeated cross sections and demonstrate that estimated policy effects have an upward bias with downward-biased estimated standard errors. Since the analysis is not focused on DID regression estimates, I choose to present ‘naïve’ versions. 17 6.1 Help from states’ IV-D agencies Table 2 provides estimates of the impact of UIFSA implementation on help asked for and help received from state IV-D agencies. There is no evidence that UIFSA implementation increases the incidence of help-seeking by those with interstate cases (panel A). For particular types of help (panel B), there is some weak evidence of increased requests for help with paternity establishment in interstate cases, and none of the estimates, including the conditional DID matching estimates, for specific help requests are estimated to be significantly different from zero. In sharp contrast, UIFSA implementation raises the probability that help is received by requestors in interstate cases. The point estimates of 0.086 is large enough to close the prior interstate gap in receipt of help conditional on making a request.18 6.2 Payment intermediation The increasing use of IV-D services for interstate cases is hypothesized to increase intermediation of interstate payments and the findings in Table 3 indicate that this is clearly the case. The increase in interstate cases receiving payment through the IV-D system relative to instate cases is highly significant across all estimators. The preferred estimate of a 7.5 percentagepoint increase is an increase of 25% above the mean of the data. There is a corresponding 8.5 percentage-point decline in payments received directly from parents (a one-third decline relative to the mean) and no difference in the pattern of payments made through the courts. In comparison with estimates of the interstate gap in intermediation, findings indicate that UIFSA implementation greatly and further increases the prior disproportionate payment intermediation of interstate cases by institutions, by increasing IV-D agency intermediation. 6.3 Awards and agreements for financial support and health care The preferred estimate (Table 4) indicates that UIFSA implementation raises the share of interstate cases with a written agreement a modest 5 percent above the sample mean. This closes about half of the interstate gap in agreements for all at-risk households that exists prior to UIFSA 18 The improved success of interstate cases in receiving help from IV-D is likely driven by improved locate services. A linear DID regression indicates a 22 percentage-point increase in help received with locate for interstate cases, conditional on requesting this service (not reported in a table; estimates for other types of help received are all insignificant). Unfortunately, the DID matching estimates cannot be obtained for these very small samples. 18 implementation. There is no estimated impact on non-court-ordered (‘voluntary’) agreements, consistent with the hypothesis that UIFSA’s impact works largely through improving access to IV-D services. This finding also suggests that UIFSA implementation does not change the bargaining stance of parents in interstate cases (i.e., voluntary agreements do not decline). UIFSA implementation has little impact on support amounts owed over the past year (rows 3 and 4 of panel A). These estimates of the amount owed last year are negative in sign, suggesting that UIFSA may have encouraged the pursuit of smaller-than-average claims by lowering transaction costs, in the case of samples conditioned on a written agreement, but all are imprecisely estimated. However, UIFSA implementation has a very large positive effect on the total amount of pending ‘back’ support identified in interstate cases. For example, the preferred estimate of $279.69 for the at-risk household sample is 53 percent of the sample mean. The findings are quite consistent across all households, those with a written agreement, and those owed a payment. In all cases, the preferred estimate suggests that UIFSA implementation roughly doubles the existing interstate gap in back support liability. Table 4 also provides evidence on health care arrangements (see Panel B). There is little evidence that UIFSA implementation affects medical support arrangements for children in interstate cases. The magnitudes suggest a possible shift from obligee responsibility to the ‘other’ category but coefficient estimates are not estimated to differ significantly from zero. 6.4 Payment Patterns Table 5 provides the estimates of the effects of UIFSA on actual child support payments made and total amounts collected in the past year. There is consistent evidence that UIFSA increases the probability that any payment is received in the sample of all at-risk households. The underlying mechanism for this appears to be the increase in written agreements under UISFA, since there is no increase in the probability of receiving a payment, conditional on having a written agreement, or conditional on being owed a payment under a written agreement. There is marginal evidence of increases in the amount of payments received over the past year in interstate cases under UIFSA in some samples. There is consistent evidence that the 12-month payment ‘yield’ (the ratio of payments made to payments owed over the past 12 months) on 19 interstate cases for the group of at-risk households rises by about 4 percentage points (about 10 percent of the sample mean) under UIFSA. Coupled with the estimates from the restricted subsamples, the evidence is consistent with a hypothesis that UIFSA implementation increases payments by establishing ‘new’ interstate support claims, which obligees then pay on, and not by increasing payments when support orders are ignored. Recall the evidence from table 3 that the average amount identified as owed is constant or even declines somewhat (although the point estimates are insignificant at standard confidence levels). The small estimated preferred increase in payments of about 10 percent is enough to increase the overall yield of payments on interstate support owed under UIFSA. Many of the point estimates in Table 5 are insignificant and because of their imprecision, it is difficult to reject the hypothesis that UIFSA implementation closes the interstate payment gap in many cases. The (significant) point estimate for ‘any payment’ amounts to half of the magnitude of the estimated interstate payment gap in (see Table 1), while the hypotheses that UIFSA implementation closes the interstate gaps in the amount paid in the past year and the payment yield for all at-risk households cannot be rejected at standard confidence levels. When restricting consideration to households with written agreements (whether owing a payment or not), UIFSA implementation arguably falls short of closing the interstate gap in the incidence of payments made. Again, this is consistent with a story of UIFSA implementation having most of its effect through the increased establishment of orders, rather than through more vigorous enforcement of established orders. While one cannot reject the hypotheses that UIFSA implementation closes existing interstate payment gaps as measured by any payment made and the payment yield, the insignificant conditional DID point estimates preclude a definitive answer. The robustness analysis, presented next, provides more conclusive evidence on this point. 7. Robustness of the Findings This section presents findings from important changes to the specifications and samples made in order to explore the robustness of the main findings and also presents an examination of the heterogeneity of findings for important sub-populations. Issues addressed are the treatment of duplicate observations, the concentration of child support cases in large states, the treatment of likely (as opposed to actual) interstate movers, and the treatment of transitional years in which 20 some but not all states adopt UIFSA. In order to place the findings in the context of the wider child support literature, I also discuss findings for the subpopulations of nonwhites, welfare recipients, and the never married. A table summarizing the original and alternative findings is provided as Appendix C. 7.1 Duplicate observations To this point, the estimation samples include multiple appearances of the same household.19 To investigate the sensitivity of the findings to their inclusion, I drop all second observations of a child support case from the estimation.20 The major findings are overall quite robust with respect to this change (see the second column of Appendix Table C). Some key findings for at-risk households—the probability of a written agreement, the amount of support received in the past year, and the amount of back support identified for those owed a payment—grow larger in magnitude with this modification and are precisely estimated. An exception is that reports of help received by help seekers is no longer estimated to be significantly different from zero. 7.2 Concentration of child support cases in large states One third of the sample households reside in California, Texas, New York, Florida, or Illinois. It is possible that states with very large child support caseloads differ in characteristics, practices and policies. If so, the findings may be dominated by the actions of larger states and may not be broadly applicable across the country. The findings are in fact very robust to removing households in these states from the samples, and some key findings grow sharper. For example, the effect of UIFSA on the probability that an atrisk household with an interstate case has a written agreement rises from 0.029 to 0.051, effectively closing the interstate gap, and the estimated effect on the payment yield for all households rises from 0.043 to 0.070 (eliminating the interstate gap). While the effect on the total amount of support owed for interstate cases under UIFSA is smaller in this subsample (about $80 below the prior estimate), this is likely due to cost-of-living differences among states. 19 In principle, including duplicates could be advantageous, as the matching technique increases the likelihood of a household in the pre-implementation period being matched with itself in the post period. Unfortunately, very few observations straddle a UIFSA implementation in their state, so even this limited longitudinal aspect does not contribute much to the estimation. 20 Two-thirds of the 2,144 affected observations are from 1999 and 2003, the other from 1994 and 1995. 21 If anything, the evidence suggests that UIFSA implementation may have been more effective in lower-population states. 7.3 Treatment of likely interstate movers The difference strategies presented so far rely on the implicit assumption that relatively few households are prone to experience an obligor’s interstate move. However, it is possible that there are households in the nonmovers’ sample who are marginal to making an interstate move, and whose behavior and outcomes might be affected by UIFSA. I explore whether this is a potential problem by estimating a moving propensity based on the pre-UIFSA implementation distribution of interstate cases created by an obligor’s move and dropping households with propensity scores in the top quintile from the group of ‘nonmovers’ prior to estimating the conditional DID matching model. Conditional DID matching estimates, presented in the fourth column of the table, are quite robust with respect to this alternative estimation strategy. UIFSA’s effect on the probability of an at-risk household with an interstate case receiving a payment in the past year drops from an estimate of 0.043 to 0.031 and is significant. The effect on the probability that a help-requesting household receives help from IV-D also drops somewhat (from 0.086 to 0.076) but remains significant. 7.4 Transitional years of UIFSA policy In the early years of UIFSA adoption, the effect of the policy is likely weak, because UIFSA’s initiatives are only in full force if both parents reside in states that have adopted. The SIPP does not reveal where the absent parent lives, so cases cannot be classified according to specific statepair UIFSA adoption policies. As an alternative, ‘pre’ and ‘post’ UIFSA observations are identified more cleanly by dropping the transition years 1995-1997 from the analysis, which is hypothesized to generate sharper estimated effects of UIFSA implementation. Overall, the findings are stable with respect to this change, with some more pronounced estimated effects of UIFSA, as hypothesized. For example, the effect of UIFSA implementation on the payment yield for at-risk households jumps from 0.043 to 0.067 (large enough to offset the interstate gap prior to implementation), and the payment yield conditional on being owed a payment becomes significant and very large (at 0.108), providing some evidence that UIFSA 22 increases payments by improving collections conditional on orders (and not entirely through establishing new orders, as the main findings suggest). The estimated effects on payment intermediation are also larger for the ‘sharper’ sample; the coefficient associated with IV-D intermediation rises 0.075 to 0.097. 7.5 Heterogeneity of key findings by race, welfare receipt, and marital history The child support literature frequently finds varying levels of child support enforcement effectiveness according to race, welfare receipt, and marital history. White parents typically have better child support outcomes than African-Americans (Graham and Beller, 1996). Welfare recipients often face disincentives to cooperate with child support enforcement because of benefit offset policies, particularly prior to welfare reform, and this may explains their inferior outcomes (Cancian, Meyer, and Caspar, 2008; Roff, 2010). Finally, when cases originate with nonmarital births, the enforcement process can be particularly difficult (Graham and Beller, 1996; Freeman and Waldfogel, 2001). While SIPP samples of nonwhite households are fairly small, there is convincing evidence that the effects of UIFSA implementation are far different for this group. Most notably perhaps, nonwhites with interstate cases are not estimated to experience any improvement in help received from IV-D, conditional on a help request, subsequent to UIFSA implementation. Also in stark contrast to the unrestricted sample, while UIFSA implementation has an extremely large impact on payment intermediation for nonwhites (the absolute magnitude of the decline is more than double the original estimate), the decline in payments from parents is offset by increased payments from courts (for whites this decline is entirely accounted for by a shift into IV-D). These two pieces of evidence suggest that UIFSA implementation does not improve administrative access for nonwhites with interstate cases. There is no evidence that UIFSA implementation improved agreements or payments for interstate cases in the nonwhite sample. In fact, for nonwhites with interstate cases, less support is identified and collected after UIFSA implementation. Few estimates are significant for the never-married group and it was infeasible to estimate the DID conditional matching model for many dependent variables. In contrast with prior findings, 23 there is a reduction in help requests from IV-D for those with interstate cases subsequent to UIFSA implementation. In contrast with the findings for nonwhites, estimates for welfare-receiving households suggest that UIFSA implementation is very effective at improving agreements, support identified, and payments for interstate cases. First, note that the estimated effect of UIFSA on help received with a problem from IV-D rises to 0.133 (from 0.086) for welfare households, and while intermediation estimates are imprecise for this smaller sample, they are at least as large in magnitude as the initial estimates and follow a pattern of substitution of IV-D intermediation for parent contact. For at-risk welfare recipients with interstate cases, UIFSA’s impact on establishment of agreements for welfare cases (0.089) is nearly three times as large as the estimate for the entire sample. There are large increases in the amount of support owed over the past year ($237.41), the probability of receiving any payment (0.093), and in the amount actually received in the past year ($213.62). While the payment yield is unchanged, it appears that UIFSA has a large beneficial impact on both total support identified and collections for at-risk, welfarereceiving households. For this group, there are huge effects of UISFA on payments, conditional on being owed a payment. The estimated effect on any payment being made for this group is 0.213, while the estimated payment increase exceeds $500. Finally, estimates for health care agreements are occasionally significant. In several cases 9when duplicates are omitted, when large states are omitted, and for nonwhites), UIFSA implementation significantly increase the share of households with “other” arrangements for health care. 8. Conclusion 8.1 Summary This paper has analyzed the impact of a sweeping reform to the legal and administrative treatment of interstate child support cases on help sought and received from state child support (IV-D) agencies, payment intermediation, agreements on payments and health care, and payments received. Conditional DID matching estimates adjusted for the influence of UIFSA 24 implementation on incentives to move out of state and were shown to substantially improve the balance of the analysis samples with respect to the exogenous variables. The findings from the unrestricted sample indicated no help-seeking response to UIFSA, but a much higher rate of satisfaction with help from IV-D conditional on requesting help. For those owed support, payment mode shifted strongly away from parents to IV-D agencies. UIFSA was estimated to raise the incidence of written agreements, the amount of back support identified but there was little evidence of an impact on health care arrangements. UIFSA was estimated to increase payments received in the past year and the ratio of payments to support owed among all at-risk households with interstate cases. In the general sample, UIFSA appeared to increase support received in large part by addressing a deficit in award establishment for interstate cases. Most findings were robust with respect to sample modifications that addressed the treatment of duplicates, the dominance of large states, and omission of likely interstate movers from the control group. In fact, these sample modifications frequently resulted in estimates suggesting that UIFSA implementation successfully closed the black hole of interstate enforcement. There was evidence of heterogeneity in the responses of various groups to UIFSA implementation. There was a strong effect of UIFSA on payment intermediation on nonwhites, but in contrast with all other groups, the shift was from parent-to-parent to court (not IV-D) intermediation. Nonwhites who requested help from the IV-D agency showed no improvement in help received subsequent to UIFSA implementation. Nonwhites with interstate cases also showed no improvement in the amount of back support identified, payments received, and payment yield subsequent to UIFSA. Like nonwhites, those who never married showed no improvement in the amount of back support identified, payments received, and payment yield (many estimates could not be obtained for this group due to sample size issues). In contrast, welfare recipients with interstate cases appear to have experienced the greatest benefits of UIFSA implementation, with huge increases in the incidence of written agreements, the amount of back support identified, payments made, and payment yield. 8.2 Discussion 25 The overall pattern of findings suggests an explanation of UIFSA’s success. UIFSA ‘leveled’ access of interstate obligees to IV-D services—services that had become quite effective by the welfare reform era due to enhanced administrative enforcement powers. It appears that the greatest area of effectiveness was awards establishment. Significantly more support orders were established for at-risk interstate cases after UIFSA implementation, and these orders were then paid on, very likely on an involuntarily basis. Although the findings are consistent with administrative reforms and initiatives ‘causing’ improved outcomes, caution is needed because award and payment outcomes cannot be tied directly to institutional changes using the data. It is possible to affect intermediation without affecting ‘real’ enforcement (i.e., actual payments) if reforms mostly succeed in formalizing previously voluntary agreements. Since payment information is only available on a consistent basis for cases with a written agreement, it is not possible to estimate the extent to which the increased payments received by the at-risk sample with interstate cases after UIFSA implementation merely substitute for voluntary payments. However, the reputation of the preUIFSA interstate child support arena as a black hole, coupled with research on the negative relationship of support and geographic distance of the obligor from the children (e.g., Seltzer, 1991), suggests that crowd-out of voluntary payments is minimal in this case. The fact that helpseekers report improved help with problems under UIFSA provides further independent evidence that the activities of IV-D agencies go beyond formalization to actual enforcement. That nonwhites with interstate cases failed to experience the improved administrative enforcement effects of UIFSA merits further research. The findings for both help received from IV-D and IV-D agency intermediation suggest the failure stems from lack of contact with IV-D. While there is a strong trend of declining welfare participation and hence IV-D contact (the sample proportion of nonwhites receiving cash welfare plummets from almost half in 1993 to just 20 percent by 2005), this trend is similar for households with inter- and in-state cases. Greater difficulty enforcing interstate cases may mean that the loss of contact with IV-D was more damaging to the interstate group. Enforcement for the subsample of welfare recipients with interstate cases was dramatically improved by UIFSA. The major changes to welfare in this era (increases in the child support 26 pass through and increased negative selection of recipients as reform proceeds) are not expected to necessarily have a differential impact on individuals with interstate situations on welfare. As with nonwhites, a reasonable hypothesis consistent with this strong evidence of the increasing value of IV-D contact over this era is that IV-D contact has greater value for those with interstate cases, post-UIFSA. It is also the case that interstate cases may require, on average, more resources to pursue, in which case welfare recipients with interstate cases in an era of rapidly declining caseloads might benefit from receiving more intense services. 8.3 Conclusions and Directions for Further Research The findings suggest that UIFSA was very successful at leveling the playing field for interstate and in-state child support enforcement, as intended. Its success was manifested in increased awards stablishment in interstate cases. Once awards were established, administrative enforcement (e.g., wage garnishment) was evidently highly effective at collecting payments. There is less evidence that UIFSA was successful at addressing underpayment on existing awards. Why this is the case is an interesting topic for further research and likely requires a more detailed understanding of the child support enforcement process at IV-D offices. The findings for subgroups also indicate that direct contact with IV-D agencies was a very important determinant of UIFSA’s success. Understanding how welfare reform affected access to IV-D services as well as the quality of services received is also an important area for future research. Larger policy questions related to this research merit further study. UIFSA may have encouraged interstate mobility of obligees and discouraged inefficient cross-state moves of obligors. The welfare improvement from increased mobility due to UIFSA to obligees, especially, could plausibly be quite large. Second, UIFSA and other initiatives brought a new focus on health care arrangements to child support enforcement policy, and the importance of health insurance and health care coverage for children have only grown in the intervening years. The effectiveness of the various initiatives to improve health care coverage of children at risk for child support, including tradeoffs with Medicaid and SCHIP coverage, would be an interesting area for further research. Finally, UIFSA is also part of an important trend of ever-more-coercive child support policy. More research is needed on the longer-run effects of such a strategy on child development, family relationships, work effort, and job mobility. 27 Appendix A Table A presents DID statistics computed for the entire sample of at-risk households used in the regressions in the first two columns. In this case, there are significant differences between movers and nonmovers in the interperiod changes in the sex of the youngest child, the incidence of very young children at home, low education of the obligee, and the share of cash-welfarereceiving households. In the case of the sample restricted to households with a written agreement, interperiod changes in the sex of the youngest child, the presence of a very large number of children at-risk for child support, and the share of cash-welfare-receiving households for interstate movers are all relatively larger than the corresponding changes in the nonmovers’ sample, while the interperiod changes in the obligee’s age and the appearance of the household for a second time in the survey for interstate movers are relatively smaller than the corresponding changes in the nonmovers’ sample. When the sample is restricted further to those who are also owed a payment in the past 12 months, interperiod changes in the sex of the youngest child, the presence of a large number of children at-risk for child support, and the share of cash-welfarereceiving households are relatively larger for interstate movers than the corresponding changes in the nonmovers’ sample. Balance is much improved by restricting the sample to households seeking help from a IV-D agency. The interperiod change in the sex of the youngest child is relatively larger than the corresponding change in the nonmovers’ sample, while the DID statistic for very young children is only marginally significantly different from zero. Table A also presents the DID statistics for the explanatory variables when the calculations are re-weighted with the weights generated by the conditional DID matching estimator. The reweighted unrestricted sample is balanced with respect to the presence of young children, cash welfare participation and obligee education (although the latter two are only marginally imbalanced in the first place), although the significant DID statistic associated with the sex of the youngest ‘at-risk’ child persists . There is a marginally significant statistic associated with never married status of the obligee. While there is improvement, not all evidence of imbalance is eliminated. When only those with a written agreement are considered, the weighting scheme is quite successful at rebalancing the sample, although weighting introduces a marginally significant DID 28 statistic for the ‘married’ variable. For those owed a payment in the past 12 months under a written agreement, reweighting is moderately successful at rebalancing; there are no DID statistics significantly different from zero at the 95 percent confidence level or above. Finally, when the sample is restricted to those seeking help with child support from a state agency, the DID PPM weighting scheme balances the sample quite well on the exogenous explanators. Appendix Table A: Difference-in-difference statistics for explanators Youngest child in edit is male At-risk N=22,544 UnWeighted weighted 0.048*** 0.071** (0.018) (0.029) With a written agreement N= 12,539 UnWeighted weighted 0.056** 0.047 (0.027) (0.045) Owed payment in the past 12 months N= 10,593 UnWeighted weighted 0.058** 0.049 (0.029) (0.044) Sought help from a IVD agency N=6,098 UnWeighted weighted 0.073** 0.043 (0.034) (0.045) Youngest child in edit is white 0.012 (0.018) .012 (0.019) 0.008 (0.023) 0.033 (0.028) 0.011 (0.026) 0.033 (0.039) 0.013 (0.030 0.037 (0.049) 2+ at-risk children -0.031 (0.021) -.008 (0.028) -0.011 (0.030) -0.012 (0.037) -0.027 (0.028) -0.015 (0.038) -0.039 (0.034 -0.076* (0.046) 3+ at-risk children -0.006 (0.015) -.020 (0.017) 0.020 (0.020) 0.005 (0.025) 0.019 (0.022) 0.011 (0.031) -0.002 (0 .028 -0.015 (0.030) 4+ at-risk children 0.000 (0.007) -.004 (0.008) 0.020** (0.009) 0.0193 (0.014) 0.023** (0.010) 0.022* (0.012) 0.005 (0.016) 0.015 (0.017) Any children 0-5 0.036*** (0.012) 0.012 (0.020) 0.022 (0.017) 0.007 (0.025) 0.023 (0.018) 0.008 (0.032) 0.046* (0.026) 0.027 (0.038) Any children 512 -0.011 (0.017) -.017 (0.025) -0.011 (0.030) -0.007 (0.038) -0.002 (0.029) 0.005 (0.036) -0.022 (0.031) 0.002 (0.039) Lacks high school diploma 0.025* (0.015) .018 (0.020) 0.025 (0.020) 0.024 (0.024) 0.026 0.022 0.021 (0.028) -0.005 (0.030) -0.033 (0.033) Age -0.534 (0.332) -.029 (0.474) -0.622* (0.374) -0.241 (0.473 -0.512 (0.452) 0.273 (0.652) -0.216 (0.539) -0.189 0.706 Young mother 0.004 (0.011) -0.009 (0.014) 0.001 (0.013) -0.000 (0.019) -0.005 (0.015) -0.006 (0.023) -0.026 (0.019) -0.043 (0.027) 1 child 0.004 (0.012) 0.002 (0.015) -0.005 (0.014) -0.016 (0.020) 0.013 (0.012) 0.018 (0.018) 0.005 (0.012) 0.005 (0.015) 2 children -0.014 (0.017) 0.000 (0.020) -0.024 (0.032) -0.014 (0.036) -0.033 (0 .031) -0.051 (0.038) -0.020 (0.033) 0.014 (0.039) 29 Appendix Table A (continued): Difference-in-difference statistics for explanators Sex At-risk N=22,544 UnWeighted weighted -0.010 -0.001 (0.012) (0.017) With a written agreement N= 12,539 UnWeighted weighted 0.012 0.002 (0.014) (0.027) Owed payment in the past 12 months N= 10,593 UnWeighted weighted 0.002 0.002 (0.018) (0.022) Sought help from a IVD agency N=6,098 UnWeighted weighted 0.008 0.004 (0.015) (0.019) Never married 0.010 (0.018) 0.035* (0.019) -0.001 (0.016) 0.003 (0.024) .008 (0.019) -0.001 (0.0315 -0.028 (0.030) -0.032 (0.032) Currently married -0.004 (0.020) -0.019 (0.030) -0.021 (0.025) -0.066* (0.037) -0.022 (0.029 -0.037 (0.045) 0.014 (0.033) -0.002 (0.033) Receives cash welfare 0.028* (0.016) -0.008 (0.021) 0.049** (0.023) 0.035 (0.029) 0.054** (0.022) 0.063* (0.035) 0.016 (0.040) 0.013 (0.036) Receives Food Stamps 0.012 (0.016) 0.005 (0.023) 0.025 (0.021) 0.020 (0.025) 0.023 (0.026) 0.037 (0.034) -0.007 (0.028) 0.004 (0.047) Repeats in -0.007 -0.040 -0.053** -0.052 -0.031 -0.024 -0.024 -0.026 panel (0.020) (0.025) (0.025) (0.037) (0.029) (0.036) (0.027) (0.043) Notes: Each cell contains the difference-in-difference statistic for the variable with bootstrapped standard error in parentheses beneath. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. 30 Appendix B: Table of sample statistics accompanying estimates At-risk N=22,544 With a written agreement N=12,539 Owed payment, past 12 months N=10,593 Sought help from a IV-D agency N=6,098 0.270 (0.444) 0.356 (0.479) 0.369 (0.483) 1.000 (0.000) Received help from IV-D (0/1) 0.140 (0.347) 0.208 (0.406) 0.218 (0.413) 0.516 (0.500) Written child support agreement (0/1) 0.556 (0.497) 1 0 1.00 (0.00) 0.732 (0.443) Written child support agreement that is not court-ordered (0/1) 0.134 (0.340) 0.240 (0 .427) 0.233 (0.423) 0.089 (0.285) Request help with enforcement (0/1) 0.145 (0.353) 0.225 (0.418) 0.237 (0.425) 0.538 (0.499) Request help establishing order (0/1) 0.107 (0.310) 0.116 (0.321) 0.118 (0.322) 0.397 (0.489) Request help with locate (0/1) 0.040 (0.197) 0.041 (0.199) 0.043 (0.203) 0.149 (0.356) Request help with modifying order (0/1) 0.016 (0.126) 0.025 (0.157) 0.026 (0.159) 0.060 (0.237) Request help with medical support (0/1) 0.016 (0124) 0.0183 (0.134) 0.019 (0.136) 0.058 (0.233) Request help with other (0/1) 0.015 (0.123) 0.019 (0.138) 0.019 (0.138) 0.057 (0.231) Request help with paternity establishment (0/1) 0.013 (0.113) 0.011 (0.105) 0.012 (0.107) 0.048 (0.213) Obligor responsible for health care (0/1) 0.218 (0.413) 0.392 (0.488) 0.412 (0.492) 0.276 (0.447) Obligee responsible for health care (0/1) 0.118 (0.322) 0.212 (0.409) 0.212 (0.409) 0.126 (0.332) Order does not specify health care responsibilities (0/1) 0.157 (0.364) 0.282 (0.450) 0.290 (0.454) 0.243 (0.429) Other arrangements have been made for health care (0/1) 0.041 (0.198) 0.074 (0.261) 0.076 (0.265) 0.057 (0.232) Amount of child support owed over previous 12 months ($82-84=100) 1149.57 (1825.10) 2066.82 (2023.13) 2340.99 (2005.70) 1404.24 (1685.74) 0.391 (0.488) 0.703 (0.457) 0.797 (0.403) 0.494 (0.500) Dependent variables Requested help from IV-D (0/1) Any payment made over previous 12 months (0/1) 31 Appendix B (continued): Table of sample statistics accompanying estimates Amount of child support paid over previous 12 months ($82-84=100) At-risk N=22,544 854.38 (1716.86) With a written agreement N=12,539 1536.10 (2062.14) Owed payment, past 12 months N=10,593 1738.30 (2116.76) Sought help from a IV-D agency N=6,098 836.40 (1521.65) Amount of child support owed over prior 12 months ($82-84=100) 1149.57 (1825.10) 2066.82 (2023.13) 2340.99 (2005.70) 1404.24 (1685.74) 0.351 (0.620) 0.634 (0.718) 0.719 (0.720) 0.594 (0.742) 527.53 (2725.34) 941.19 (3586.48) 1025.00 (3729.95) 1023.19 (3849.50) Payment received directly from parent (0/1) 0.153 (0.360) 0.275 (0.447) 0.306 (0.461) 0.100 (0.300) Payment received directly from court (0/1) 0.196 (0.397) 0.352 (0.478) 0.402 (0.490) 0.295 (0.456) Payment received directly from IV-D (0/1) 0.125 (0.331) 0.225 (0.418) 0.257 (0.437) 0.256 (0.437) Interstate case, post-UIFSA implementation (0/1) 0.135 (0.341) 0.124 (0.329) 0.125 (0.331) 0.155 (0.362) Post-UIFSA implementation (0/1) 0.744 (0.437) 0.748 (0.434) 0.758 (0.429) 0.710 (0.454) Obligor’s move created interstate case (0/1) 0.181 (0.385) 0.163 (0.370) 0.162 (0.368) 0.218 (0.413) 0.512 (0.500) 0.504 (0.500) 0.504 (0.500) 0.509 (0.500) Youngest at-risk child is white (0/1) 0.719 (0.450) 0.780 (0.414) 0.778 (0.416) 0.689 (0.463) Two or more ‘at-risk’ children (0/1) 0.453 (0.498) 0.493 (0.500) 0.512 (0.500) 0.529 (0.499) Three or more ‘at-risk’ children (0/1) 0.146 (0.353) 0.153 (0.360) 0.160 (0.367) 0.196 (0.397) Four or more ‘at-risk’ children (0/1) 0.040 (0.197) 0.040 (0.197) 0.041 (0.199) 0.060 (0.237) Any child 0-5 (0/1) 0.169 (0.375) 0.131 (0.338) 0.132 (0.339) 0.186 (0.389) Any child 5-12 (0/1) 0.334 (0.472) 0.357 (0.479) 0.371 (0.483) 0.380 (0.485) Ratio of child support paid to child support owed over previous 12 months Total amount of back support owed (excluding previous 12 months) ($82-84=100) Explanators Youngest at-risk child is male (0/1) 32 Appendix B (continued): Table of sample statistics accompanying estimates At-risk N=22,544 With a written agreement N=12,539 Owed payment, past 12 months N=10,593 Sought help from a IV-D agency N=6,098 Obligee does not have high school diploma (0/1) 0.164 (0.370) 0.132 (0.338) 0.133 (0.339) 0.179 (0.383) Obligee age (years) 36.5 (8.78) 36.72 8.08 36.35 (7.92) 34.453 (8.151) Obligee under 18 at first birth (0/1) 0.088 (0.283) 0.083 (0.276) 0.079 (0.270) 0.073 (0.260) 0.082 (0.274) 0.054 (0.225) 0.115 (0.319) 0.053 (0.225) Two children present (0/1) 0.682 (0.466) 0.693 (0.461) 0.704 (0.457) 0.660 (0.474) Three children present (0/1) 0.235 (0.424) 0.234 (0.400) 0.243 (0.429) 0.286 (0.452) Obligee is male (0/1) 0.158 (0.365) 0.100 (0.300) 0.089 (0.284) 0.064 (0.245) Obligee never married (0/1) 0.236 (0.424) 0.173 (0.378) 0.176 (0.381) 0.286 (0.452) Obligee currently married (0/1) 0.312 (0.463) 0.331 (0.471) 0.331 (0.471) 0.287 (0.453) Household receives cash welfare (0/1) 0.164 (0.371) 0.135 (0.342) 0.135 (0.342) 0.237 (0.425) Household receives Food Stamps (0/1) 0.218 (0.413) 0.195 (0.396) 0.197 (0.398) 0.325 (0.469) Second appearance of household in panel (0/1) 0.398 (0.489) 0.407 (0.491) 0.384 (0.486) 0.397 (0.489) One child present (0/1) 33 Appendix C: Estimated effects of UIFSA implementation on interstate cases for alternative subsamples Panel A: At-risk households Original sample N=22,544 -0.006 (0.015) Drop duplicates N=14,880 0.022 (0.018) Lowerpopulation states N=15,508 0.018 (0.017) Delete likely movers N=18,854 -0.009 (0.014) Delete transition years N=17,471 -0.012 (0.018) Nonwhite N=5,975 -0.015 (0.041) Receives welfare N=5,886 -0.008 (0.034) Never married N=5,282 -0.079* (0.043) Written agreement 0.029** (0.011) 0.054*** (0 .020) 0.051** (0.023) 0.026** (0.013) 0.027 (0.019) -0.022 (0.029) 0.089*** (0.028) 0.016 (0.050) Written, not courtordered 0.002 (0.009) 0.009 (0.014) 0.009 (0.012) 0.003 (0.010) 0.004 (0.012) -0.016 (0.025) 0.012 (0.016) -0.036 (0.028) Amount owed in past year -16.65 (85.90) 132.75* (72.16) 53.23 (76.81) 21.74 (64.04) -54.22 (89.23) -299.65** (139.14) 237.41*** (77.54) -57.26 (120.28) Total amount of back support owed 279.69*** (78.00) 281.07*** (81.31) 196.65*** (63.84) 235.60*** (53.35) 251.74*** (55.68) 239.75* (145.49) 37.22 (110.34) 170.23 (104.64) Any payment made 0.043** (0.019) 0.055*** (0.015) 0.063*** (0.018) 0.031** (0.014) 0.037* (0.019) -0.026 (0.034) 0.093*** (0.025) -0.014 (0.040) Amount paid in past year 125.81* (72.47) 169.06*** (58.36) 152.38*** (47.96) 123.68** (52.76) 144.94** (70.44) -239.72* (127.02) 213.62*** (62.06) -158.92 (114.58) Payment yield, past year 0.043** (0.022) 0.043 (0.029) 0.070*** (0.025) 0.041** (0.019) 0.067*** (0.025) -0.042 (0.035) 0.036 (0.050) -0.130 (0.093) Panel B: Households owed a payment N=10,593 N=6,804 N= 7,641 N=8,832 N=8,522 N= 2,213 N= 2511 Amount owed in past year -173.38 (125.04) 0.945 (167.53) -148.74 (120.36) -140.77 (146.92) -210.04 (165.50) -602.49 (430.70) 122.42 (188.49) NAa Total amount of back support owed 648.70*** (132.15) 775.66*** (182.40) 533.05*** (149.42) 654.01*** (133.84) 670.57*** (124.29) 778.04** (340.44) 304.92 (332.86) NAa Amount paid in past year 193.93* (105.61) 220.90 (153.87) 176.28 (129.71) 217.52 (149.33) 219.99 (147.49) -119.62 (292.09) 518.94*** (127.63) NAa Requested help Appendix C (continued): Estimated effects of UIFSA implementation on interstate cases for alternative subsamples Original sample 0.056 (0.050) Drop duplicates 0.023 (0.062) Lowerpopulation states 0.064 (0.047) Delete likely movers 0.057 (0.049) Delete transition years 0.108** (0.049) Nonwhite -0.066 (0.182) Receives welfare 0.069 (0.189) Never married NAa Payment from IV-D 0.075*** (0.024) 0.078** (0.036) 0.073** (0.031 0.076*** (0.021) 0.097*** (0.029) 0.048 .0707 0.117 (0.086) NA a Payment from court 0.018 (0.027) 0.016 (0.034) 0.023 (0.029) 0.019 (0.025) 0.008 (0.031) 0.161** (0.0745 -0.054 (0.072) NAa Payment from parent -0.085*** (0.020) -0.093*** (0.032) -0.084*** (0.028) -0.089*** (0.029) -0.094*** (0.028) -0.193*** .0691 -0.092 (0.058) NAa Obligor provides health care 0.009 (0.026) 0.029 (0.033) -0.025 (0.025) 0.005 (0.029) -0.009 (0.030) -0.025 (0.070) -0.048 (0.068) NAa Obligee provides health care -0.019 (0.017) -0.045 (0.030) 0.004 (0.028) -0.007 (0.028) -0.005 (0.028) -0.049 (0.085) -0.034 (0.047) NAa Other arrangement for health care 0.020 (0.014) 0.034** (0.016) 0.026** (0.013) 0.017 (0.012) 0.010 (0.016) 0.049*** (0.016) -0.026 (0.038) NAa No arrangement for health care -0.007 (0.022) -0.016 (0.031) -0.007 (0.028) -0.011 (0.025) 0.001 v0.034) 0.018 (0.079) 0.103 (0.076) NAa N=4,100 N=4,405 N=5,288 N=4,676 N= 1,783 N=2,277 N=1,735 Payment yield, past year Panel C: Households asking IV-D for help N=6,098 0.086*** 0.050 0.083** 0.076*** 0.111** 0.039 0.133*** 0.092 (0.026) (0.032) (0.035) (0.026) (0.043) (0.059) (0.046) (0.074) Notes: Each cell contains the estimated effect of UIFSA implementation on interstate cases, using a DID calculation, DID regression or conditional matching DID estimator as indicated. Bootstrapped standard errors are reported in parentheses beneath the estimate. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. a Subsample sizes of interstate cases are too small to obtain the conditional matching DID estimator. Received help 35 References Argys Laura and H. Elizabeth Peters, 2003. “Can Adequate Child Support Be Legislated? Responses to Guidelines and Enforcement.” Economic Inquiry 41(3, July):463-479. Argys, Laura M., H. Elizabeth Peters, and Donald M. Waldman. 2001. "Can the Family Support Act Put Some Life Back into Deadbeat Dads? An Analysis of Child-Support Guidelines, Award Rates, and Levels." Journal Of Human Resources 36, no. 2: 224-252. Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan, 2004. “How Much Should we Trust Differences-in-Differences Estimates?” The Quarterly Journal of Economics 119 (1,February): 249-275. Cancian, Maria; Meyer, Daniel R.; Caspar, Emma, 2008. “Welfare and Child Support: Complements, Not Substitutes.” Journal of Policy Analysis and Management 27(2, Spring):35475 Chang, Yunhee. 2003. "Essays on Welfare Reform and Child Support Enforcement: Evidence from State Administrative Data." Dissertation. University of Illinois. DeMaria,Mechelene, 1999. “Comment, Jurisdictional Issues under the Uniform Interstate Family Support Act.” Journal of the American Academy of Matrimonial Lawyers. 16(1): 243- 257. Freeman, Richard B., and Jane Waldfogel, 2001. “Dunning Delinquent Dads: The Effects of Child Support Enforcement Policy on Child Support Receipt by Never Married Women.” Journal of Human Resources 36(2, Spring): 207-25. Graham, John W., and Andrea H. Beller, 1996. “Child support in Black and White: Racial Differentials in the Award and Receipt of Child Support during the 1980s.” Social Science Quarterly 77(3, September): 528-42. Haynes, Margaret Campbell, 1996. Child Support across State Lines: The Uniform Interstate Family Support Act: Curriculum for Judicial Educators. Washington, DC: American Bar Association Center on Children and Law. 286 pages. Heckman, James J., Hidehiko Ichimura, and Petra Todd. 1998. "Matching as an Econometric Evaluation Estimator." Review of Economic Studies 65(2): 261-294. Huang, Chien-Chung, 2002. “The Impact of Child Support Enforcement on Nonmarital and Marital Births: Does it Differ by Racial and Age Groups?” Social Service Review 76 (2, June): 275-301. Lerman, R.I., 1993. “Policy Watch: Child Support Policies.” Journal of Economic Perspectives 7: 171-182. Lin, I-Fen, Nora Cate Schaeffer, Judith A. Seltzer, and Kay L. Tuschen, 2004. “Divorced Parents’ Qualitative and Quantitative Reports of Children’s Living Arrangements,” Journal of Marriage and the Family 66(May): 385-97. Neelakantan, Urvi. 2009. "The Impact of Changes in Child Support Policy." Journal Of Population Economics 22(3): 641-663. Office of Child Support Enforcement, 2008b. Essentials for Attorneys in Child Support Enforcement. U.S. Department of Health and Human Services, 3rd edition. Retrieved September 12, 2008 from http://www.acf.hhs.gov/programs/cse/pubs/2002/reports/essentials/index.html . Peters H., Argys L., Howard H., Butler J. Legislating Love: The Effect of Child Support and Welfare Policies on Father-Child Contact. 2004. Review Of Economics Of The Household 2(3, September):255-274. Pirog, Maureen A., and Kathleen M. Ziol-Guest. “Child Support Enforcement: Programs and Policies, Impacts and Questions.” Journal of Policy Analysis and Management 25 (4, Fall): 943990. Roff, Jennifer, 2010. “Welfare, Child Support, and Strategic Behavior: Do High Orders and Low Disregards Discourage Child Support Awards?” Journal of Human Resources 45(1, Winter): 5986. Seltzer, Judy, 1991. “Relationships between Fathers and Children who Live Apart: The Father’s Role after Separation.” Journal of Marriage and the Family 53(1, February): 79-101. Sorensen, E., and Hill, A. 2004. “Single Mothers and their Child-Support Receipt: How Well is Child-Support Enforcement Doing?” Journal of Human Resources 39: 135-154. Uniform Interstate Family Support Act (1992; Revised 1996, 2001). National Conference of Commissioners on Uniform State Laws. U.S. General Accounting Office, 1992. Interstate Child Support: Mothers Report Receiving Less Support from Out-of-State Fathers. GAO/HRD-92-39FS. 37 Table 1: Estimated differences between interstate and other cases in help-seeking, child support awards, and enforcement prior to UIFSA implementation Panel A: Help requested and received Request Receipt *** At-risk 0.078 0.0164** households (0.009) (0.007) N = 16,764 Enforce Establish Locate Paternity Modify Medical Help requested N= 4,330 NA -0.084*** (0.019) -0.037** (0.019) 0.017 (0.018 0.131*** (0.015) 0.023*** (0.008) -0.014* (0.008) 0.017** (0.009) Help received NA NA -0.018 (0.027) N=1,337 -0.012 (0.041) N=832 0.120* (0.062) N=223 -0.123 (0.120) N=95 0.038 (0.086) N=182 0.041 (0.117) N=118 Panel B: Intermediation Households with a written agreement and owed a payment in the past 12 months (N=8025) Panel C: Payment & health care agreements Written, Written not courtagreement ordered *** At-risk -0.071 -0.036*** households (0.010) (0.006) N = 16,764 Payment from IV-D 0.048*** (0.014) Amount owed in past year -73.62** (33.27) Back support owed 250.07*** (67.64) Written agreement N= 9,385 NA -0.043*** (0.011) 78.98 (50.70) 603.10*** (121.82) Written agreement & owed a payment N=8,025 NA -0.037*** (0.012) 45.77 (53.40) 703.62*** (136.94) Payment from court 0.034** (0.015) Payment from parent -0.081*** (0.013) Obligor responsible NA Obligee responsible NA Unspecified NA -0.061*** (0.015) 0.021* (0.013) 0.047*** (0.014) Panel D: Payments At-risk N = 16,764 Written agreement N= 9,385 Any payment made -0.085*** (0.009) Amount paid in past year -111.48*** (31.73) Payment yield, past year -0.068*** (0.012) -0.082*** (0.014) -54.45 (52.51) -0.055*** (0.019) Written agreement & owed a payment -0.106*** -84.86 -0.074*** (N=8,025) (0.014) ( 58.00) (0.021) Notes: Each cell contains the difference in the outcome between interstate and other cases as estimated from a regression that also includes a full set of explanators, described in the narrative. The standard error of the estimate is reported in parentheses beneath the estimate. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. In Panel A, findings of “other” category not reported to economize on space. The coefficient estimates for ‘other help’ are all zero. In Panel C, findings of “other” category not reported to economize on space. The coefficient estimates for ‘other help’ are all zero. 38 Table 2: Estimated effects of UIFSA implementation on help sought and received for interstate cases Unadjusted DID DID Regression Conditional matching DID 0.000 (0.020) 0.002 (0.017) -0.006 (0.015) 0.101*** (0.033) 0.106*** (0.034) 0.086*** (0.026) Enforcement 0.003 (0.0318) 0.002 (0.035) 0.005 (0.025) Establish order -0.043 (0.032) -0.045 (0.033) -0.026 (0.032) Locate 0.014 (0.028) 0.021 (0.023) 0 .004 (0.023) Modify order 0.013 (0.013) 0.027 (0.023) 0.013 (0.010) Medical support 0.026 (0.018) 0.026 (0.020) 0.019 (0.012) Other 0.013 (0.018) 0.014 (0.018) 0.010 (0.012) Paternity establishment 0.027 (0.017) 0.024* (0.016) 0.018 (0.011) Panel A: At-risk households, N=22,544 Requested help Panel B: Help requesters N=6,098 Received help Notes: Each cell contains the estimated effect of UIFSA implementation on interstate cases, using a DID calculation, DID regression or conditional matching DID estimator as indicated. Bootstrapped standard errors are reported in parentheses beneath the estimate. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. a The sample size is too small to obtain the conditional matching DID estimator. 39 Table 3: Estimated effects of UIFSA implementation on payment intermediation for interstate cases. Payment from IV-D Payment from court Unadjusted DID calculation 0.097*** (0.024) Regression-adjusted DID 0.098*** (0.032) Conditional matching DID 0.075*** (0.024) 0.018 (0.035) 0.019 (0.031) 0.018 (0.027) -0.104*** -0.094*** -0.085*** (0.032) (0.025) (0.020) Notes: Each cell contains the estimated effect of UIFSA implementation on interstate cases, using a DID calculation, DID regression or conditional matching DID estimator as indicated. Bootstrapped standard errors are reported in parentheses beneath the estimate. Sample is at-risk households with a written agreement who were owed a payment in the past 12 months. N=10,593. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. a The sample size is too small to obtain the conditional matching DID estimator. Payment from parent 40 Table 4: Estimated effects of UIFSA implementation on agreements and awards for interstate cases. At-risk N=22,544 Unadjusted DID Panel A: Payment agreements Written 0.035* (0.019) Written agreement N=12,539 Regressionadjusted DID Conditional matching DID 0.045** (0.019) Written agreement and owed a payment last year N=10,593 Conditional Unadjusted Regressionmatching DID adjusted DID DID Unadjusted DID Regressionadjusted DID Conditional matching DID 0.029** (0.011) NA NA NA NA NA NA Written, not court-ordered 0.004 (0.013) 0.003 (0.014) 0.002 (0.009) NA NA NA NA NA NA Amount owed in past year -16.65 (78.74) 22.10 (67.49) -16.65 (85.90) -222.67 (148.61) -166.49 (106.67) -103.41 (113.44) -275.96 (187.15) -212.14* (117.41) -173.38 (125.04) Total amount of back support owed 279.69*** (72.43) (N=18,771) 287.90** (120.48) (N=18,771) 279.69*** (78.00) (N18,771) 646.87*** (132.29) (N=10,521) 640.61*** (220.05) (N =10,521) 547.09*** (129.53) (N=10,521) 755.29 (166.92) (N=8,802) 754.89*** 257.96 (N=8,802) 648.70*** 132.15 (N=8,802) NA NA NA NA NA 0.005 (0.035) 0.007 (0.031) 0.009 (0.026) Panel B: Health care agreements NA Obligor responsible *** Obligee responsible NA NA NA NA NA NA -0.024 (0.019) -0.019 (0.024) -0.019 (0.017) Unspecified NA NA NA NA NA NA 0.003 (0.026) -0.006 (0.028) -0.007 (0.022) Other NA NA NA NA NA NA 0.019 0.018 0.020 (0.016) (0.019) (0.014) Notes: Each cell contains the estimated effect of UIFSA implementation on interstate cases, using a DID calculation, DID regression or conditional matching DID estimator as indicated. Bootstrapped standard errors are reported in parentheses beneath the estimate. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. Table 5: Estimated effects of UIFSA implementation on child support collections for interstate cases. Any payment made Amount paid in past year Unadjusted DID 0.043** (0.019) At-risk N=22,544 Regressionadjusted DID 0.051*** (0.018) Conditional matching DID 0.043** (0.019) 125.81** (60.92) 162.21** (64.01) 125.81* (72.47) Written agreement N=12,539 Unadjusted RegressionConditional DID adjusted DID matching DID 0.033 0.052** 0.034 (0.024) (0.024) (0.026) 182.75* (109.13) 115.14 (123.29) 157.05 (101.34) Written agreement and owed a payment last year N=10,593 Unadjusted RegressionConditional DID adjusted DID matching DID 0.033 0.047** 0.031 (0.027) (0.021) (0.023) 179.92 (141.77) 258.67** (124.60) 193.93* (105.61) Payment 0.052** 0.060** 0.043** 0.049 0.060 0.050 0.053 0.067 0.056 yield, past (0.026) (0.024) (0.022) (0.048) (0.040) (0.046) (0 .051) (0.045) (0.050) year Notes: Each cell contains the estimated effect of UIFSA implementation on interstate cases, using a DID calculation, DID regression or conditional matching DID estimator as indicated. Bootstrapped standard errors are reported in parentheses beneath the estimate. */**/*** indicates significantly different from zero at a confidence level exceeding 90%/95%/99%. 42