The Impact of Child Support Enforcement on Fertility ... Impact of Child Support Enforcement on...

The Impact of Child SupportEnforcement on Fertility, ParentalInvestments, and Child Well-Being

Anna AizerSara McLanahan

A B S T R A C T

Increasing the probability of paying child support, in addition to increasing

resources available for investment in children, also may alter the incentives

faced by men to have children out of wedlock. We find that strengthening

child support enforcement leads men to have fewer out-of wedlock births

and among those who do become fathers, to do so with more educated

women and those with a higher propensity to invest in children. Thus, poli-

cies that compel men to pay child support may affect child outcomes through

two pathways: an increase in financial resources and a birth selection

process.

I. Introduction

Between 1970 and 1989, the proportion of children living in poverty

grew by nearly one third, with most of the increase attributable to the increase in sin-

gle-mother families (Lerman 1993). In response to these changes, and in an effort

to compel absent fathers to provide tinancially for their children, federal and state

govemment ofhcials began pursuing policies designed to strengthen child support

enforcement. Not only have child support payments increased during the past 20 years

(Case. Lin. and McLanahan 2003). but income from child support appears to have

beneficial effects on children over and above income from other sources (Knox and

Anna Aizer is a professor afeioiiomics ul Brown University and NBER; Sarah McLtinahan i.i a profes.'inrof economics al Princeton University. The authors would like lo ihank Jeanette Chima. .lane! Currie. PedroDal Bd. Anf>ie hertif>. Len Lopoo, lr\' Garfinkel, Princeton University seminar participants, and the

MavArihiir Netwtthank the Bendh.supported hy graUniversity whichWell-heinf! al Pri

rk on the l-aniily and the Eioiioniv for useful comments. The author.', would also like tom-Thonuin Center for Research on Child Wellheing al Princeton University which ist 5 ROI HD369l6from the NICHD and ihe Office of Population Research at Princetons supported hy grant 5 P30 HD32030 fnim the NICHD and ihe Center for Hetilth andceton University. The data used in this article can he obtained bei>inninji August 2006

through July 2009 from Anna Aizer. Brown University Department of Economics. 64 Waterman Sireei.Providence. RI 02912 or hy emailinR aizer(^hrown.edu.[Subniittei! April 2004; accepted June 2OO5|ISSN 022-I66X E-ISSN 1548-8004 O 2(KM>by the Board of Regetits ofthe University of Wisconsin Systetn

THE JOURNAL OF HUMAN RESOURCES • XLI • /

Aizer and McLanahan 29

Bane 1994; Graham, Beller, and Hernandez 1994; Knox 1996). To account for the lat-ter, researchers hypothesize five potential mechanisms: (1) child stipport income ismore likely to be spent on children, as compared with other income; (2) child supportalters family dynamics (between mothers and fathers) in a positive way; (3) child sup-port reduces mothers" reliance on welfare and increases employment and maniage;(4) mothers invest more in their children as a signal to absent fathers of their com-mitment to the child in order to obtain more chiid support in the future; and finally(5) child support is positively correlated with father involvement and commitment tothe child. The first four mechanisms are typically characterized as causal, while thelast is considered a selection effect (Argys, Peters, Brooks-Gunn. and Smith 1998).

Attempts to distinguish among the hve mechanisms have been hampered by the dif-hculties inherent in empirically identifying the separate effects.' We posit an alter-native explanation that can be identified empirically—that stricter child supportenforcement creates incentives for men to have fewer ehildren outside marriage, andfor those who do, to partner with women who are more likely to invest in their chil-dren independent of child support receipt. Thus unlike previous research that positseither a causal effect or selection, we posit that stricter child support enforcementcauses positive selection on the maternal quality where quality refers to a mother'sown level of human capital and propensity to invest in her children. As such, we mightexpect current policies of stricter child support enforcement to have positive effectson both present and future child outcomes. Positive maternal selection also providestheoretical support for two of the tive previously mentioned mechanisms (that childsupport income is more likely to be spent on the child and that child support mayreduce reliance on welfare) since higher quality mothers are more likely to invest intheir children and less likely to rely on welfare.

In the (irst part of this paper we discuss the incentives generated by stricter childsupport enforcement policies, how they affect the fertility of single women, and howthey change the average underlying characteristics of single mothers. The discussionincorporates the interaction between state policies of stricter child support enforce-ment and the major public program serving single women with children^the AFDCprogram. We predict that under certain assumptions, increasing the probability thatfathers wili be required to pay child support results in (I) fewer children born to moth-ers who are most likely to use AFDC, and (2) more births to women with a higherunderlying propensity to invest in children.

The intuition behind the first prediction is based on the fact that all child supportpayments received by women on welfare are taxed nearly 100 percent by the state.Thus stricter enforcement does not provide single women who are likely to rely onwelfare with incentives to have children. In eontrast. stricter enforeement provides

1. Del Boca and Flinn (1994) find evidence using consumer expenditure dala that the share ol'expenditureson child g<M)ds and services increases as the share of income from child support increases amcing divorcedmothers. Hernandez. Beller. and Graham ()995) identify (he impai:! of income separate trom father invulve-ment by comparing ihe impatl of child supfwm in a period in which lathers' coniribuiions were more likelyvoluniary (because ot la\ enforcement) with a period of strong enforcement in which fathers were morelikely compelled (and therefore more reluctant). They find ihat as enforcement increases, the positive impactof child stipptin on education declines, suggesting ihal previous estitnates of child support relied, in pan.unobserved characteristics ol faihers who voluntarilv coniribuie.

30 The Journal of Human Resources

men with clear disincentives to have children, especially if their payments go to thestate rather than to the tnother of their child. The intuition behind the second predic-tion—that stricter child support enforcement will lead to a change in the underlyingcomposition of single women—is that, given men's inability to control how theirchild support dollars are spent (on the child or the mother), stronger child supportenforcement provides them with an incentive to have children with women who havea greater underlying preference for investing in children.

In the second part of this paper we provide empirical support for the two predictionsof the model and employ an identification strategy that enables us to isolate this par-ticular mechanism empirically. First, we use annual data on state expenditures for chiidsupport enforcement and on natality for the period 1985-98 to estimate the impact ofincreasing the probability that fathers will have to pay child support on nonmaritalchiidbearing and matemai investments in children born outside marriage.' We find thatmore stringent child support enforcement results in fewer births, espeeiaiiy among lesseducated singie women, and, conditionai on education, greater use of eariy prenataicare (a measure of the underiying propensity to invest in children), both of which sug-gest positive selection on matemai quality.^ By focusing on the impact of increases instate expenditures on child support enforcement that were largely driven by federal leg-islation (as opposed to state laws), and by using a within state difference-in-differenceframework that enables us to control for factors that vary at the state-year and thatmight be correlated with both child support enforcement and fertility, we limit thepotential for policy endogeneity bias in the findings. And because our measure ofmatemai investment is prenatal care that is initiated prior to the receipt of any childsupport, any positive impact of child support enforcement on this measure of maternalinvestment is unlikely related to an increase in available financial resources.

The findings that stricter child support enforcement leads to a decline in the numberof out-of-wedlock births and an average incn^ase in both maternal education and childinvestment suggests positive selection on matemai quality as hypothesized. However,it does not rule out altemative mechanisms. Two altemative mechanisms that wouldgive rise to similar empirical findings are (1) fathers who anticipate child support obli-gations in the future will (voluntarily) provide financial support to mothers during theprenatal period in order to increase mothers' own prenatal investments in the child; and(2) mothers who anticipate child support payments in the future will experience anincrease in their permanent income that enables them to increase their prenatal invest-ments by borrowing against future child support payments. These two mechanismsmay be classified as income effects as opposed to a composition or selection effect.

To determine whether either of these two mechanisms is operating, we use datafrom the Fragile Families and Child Wellbeing Study, a longitudinal survey of new

2. We chose 1984 as the starting year because it is the first year in which the CDC natality liles include theuniverse of births, not a sample.3. These finding.'; do not support the theory proposed by Auginbaugh (2001) ihai child achievement signalsa higher propensity for child investment and thus leads to an increase in future child support receipt. Thistheory would predict a decline in maternal prenatal investment as a result ofthe increased probabiiiiy of childsupport receipt as there would be less need to signal heL:ause men would be more likely to be compelled topay child support. That is. under strict child support enforcement, women would not have to rely on volun-tary support and thus would have less need to signal in order to gain support.


mothers that includes infonnation on future child support, receipt of financial assis-tance during the prenatal period, prenatal investments (by mothers and fathers), andmothers' ability to borrow. We find that mothers who receive child suppon at Year 1are more likely to make greater prenatal investments. However, even when we condi-tion on prenatal financial transfers from the father, the positive impact of future childsupport on maternal investment in the prenatal period remains. That is, even amongmothers who receive no prenatal tran.sfers, those with future child support paymentsare still more likely to invest in the prenatal period. This finding is inconsistent withthe first mechanism—that fathers who expect to pay child support in the future willinvest more during the prenatal period which enables mothers to invest more in pre-natal care.

Nor do we find evidence to support lhe second mechanism—that mothers whoreceive child support in the futtire will invest tnore In prenatal care because they areable to borrow against future child support payments. We find no difference in theimpact of future child support receipt on prenatal investment between mothers whoare able to borrow and those who are not. The results from the Fragile Families datasuggest that selection on fertility, rather than an income effect, is driving the findingthat increases in the probability of future child support payments lead to an increasein the average probability of maternal prenatal investment in children.

Our analysis has implications for a broad range of research that examines the effectof policies on child outcomes, insofar as policies affect fertility decisions as well assubsequent parental behavior and child outcomes, failing to take account of the for-mer may lead to incorrect estimates of the true effect of the policies.

The rest ofthe paper is laid out as follows: In Section II, we explore the incentiveeffeets of stricter child support enforcement and their implications in terms of poten-tial selection on the birth mother and review the relevant literature. In Section III wediscuss our identification strategy and present the results from the aggregate state-level analysis of births from vital statistics data. In Section IV we present results fromthe Fragile Families analyses, which explores the mechanism(s) behind the findingthat more stringent child support enforcement leads to increased prenatal investment.Section V concludes by discu.^sing the implications of our findings regarding estima-tion ofthe impact of child support on child wellbeing.

II, Incentive Effects of Stricter Child supportenforcement

A. Impact of stricter Child support enforcement on Fertility—BackgroundLiterature

Theoretical predictions of the impact of stricter child support enforcement (CSE) on thenumber of births or compositioti of parents are ambiguous. On the one hand, stricterCSE raises the financial obligation of the absent father to the mother, raising the cost tosingle fathers of having children. On the other hand, stricter CSF kwers the cost of hav-ing children faced by singie mothers. Although the former should lead to a decline inout-of-wedlock births, the latter should lead to an increase; so that the net efTect ofstricter CSE on fertility is ambiguous a priori.


The empirical evidence with respect to the impact of child support enforcement onfertility suggests that the incentive effect.s of stricter CSE for men outweigh those forwomen: states with stricter child support enforcement have witnessed a decline in out-of-wedlock births.^ These studies rely on changes in state laws governing child sup-port enforcement (paternity establishment, immediate withholding, etc.) to idetitifythe impact of stricter child support enforcement on fertility.^ Examples of this litera-ture include Sonenstein, Pleck, and Ku (1994) who find that adolescent males areaware of paternity establishment policies and modify their sexual behavior andcontraceptive use accordingly. Garfinkel and his colleagues (Garfinkel, Huang,McLanahan, and Gaylin 2003: Plotnick, Garfinkel. McLanahan, and Ku 2004) linkhigher paternity establishment rates to nonmaritai childbearing and find that higherpaternity establishment rates reduce nonmaritai chiid. Though most of these studiesemploy state and year fixed effects, cndogeneity of the chiid support enforcementpoiicies stiii poses a probiem if the timing of the policies within each state is drivenby unobservabie characteristics related to fertility and nonmaritai childbearing thatare changing over time. We address the potential policy endogenelty of stricter childsupport enforcement and describe our identification strategy in Section 111.

Although there is considerable empirical evidence regarding the impact of stricterCSE on fertility, there is no research on how child support affects the cotnposition orquality of parents. We provide the first such empirical estimates of the impact ofstricter CSE on the composition of mothers and thus the potential quality of parent-ing and child outcomes.

B. Impact of Stricter Child support enforcement on Composition of Mothers

A simple model of the decision of an unmarried couple to have a child illustrates howthe composition of mothers may change if chiid support enforcement increases. Insuch a model, fathers' utility (U^) depends on their own consumption (c^,) and thequaiity or capital of their chiidren (q). Q increases with financial resources devoted tothe child—a fraction of both the mother's income ( / J and any child support paymentsreceived (c.v). Fathers choose their consumption level (r,,,) and how much child sup-port to provide (cs). Liltewise, mothers care about their own consumption {cj and qand choose how much of /„, and cs to invest in the child. Some mothers invest a greaterportion of their income in their children than others. We can think of these mothers ashaving a greater preference for chiid quality, denoted a.

4. Another pt)ssihility is thai stricter CSE is not negaiivdy affecting fertiliij', but positively affecting mar-riage if the cost of being ii single father increases rclalLve U> heing married, Hcim l2()()-ll finds no impiici ofchild suppon enforcemenl on divorce rates: Nixon (1''97) finds a negative relationship hetween child sup-pon enforcement and divorce as does Huang (2002). For our purposes, however, it doesn'i mailer whichmechanism is respimsihle. Though il is interesting to note thai if stricter CSE were lowering out-of-wediockhinhs by encouraging marriage (as opposed lo lowering fertihty), we would expect those marrying to comeIrom the upper lail of the distribution of "quality" single mothers—that is. mothers with higher education.As such, the underlying quality/education of single mothers woulJ decline. Thai is not what we find, sug-gesting that stricter child suppon enforcement in not alfecting the composition of single mothers hy increas-ing the incentives for marriage.

5. There also exists a literature linking increased child support enforcement with child support payments(Freeman and Waldtogel 199H; Argys, Peters, and Waldman 1995; Nixon 1994; Miller. Garlinkel, andMcLanahan 1997; Betler and Graham 1991).


Although a noncohabitating father is unabie to ensure that ail child support pay-ments to the mother arc spent on the child and not on her own consumption (Wiliis1999), he may observe a mother's a. Because a determines how much of her incomeand child support she will invest in her child, men always wiii prefer to have childrenwith women with a greater a.

When deciding whether to have a child, a man wiil compare the indirect utiiitygained from not having a chiid (income spent on his consumption only) and having achild with a woman with a given a (income spent on consumption and child support),Acriticai ievel of a exists above which his utiiity from having a chiid exceeds that ofnot having tjne. This critical ievel increases with the probabiiity that he wiil be forcedto pay child support and with the probability that the woman will rely on welfare(AFDC/TANF). The intuition behind the iatter is that, with some exceptions (somestates aliow for a S.'̂ O pass through each month), most child support payments tomothers on welfare are taxed 100 percent by the state. As such, men receive littie ifany additionai benefit (in the form of higher chiid quality) from their child supportpayments if the mother relies on welfare.** Thus when forced to pay child support,men wili want to have chiidren with women with a greater a and when women areiikeiy to use weifare, men wili require an even higher a to have a chiid with her.

Figure 1 illustrates these points, in the figure, the horizontai line U^ (AT) rep-resents a father's utiiity associated with not havitig a chiid (NC). Utility does notchange with a because it is only a function of income (/,„). The dark iine UJC, NE)represents a father's utility from having a child with no child support enforcement,and the line UjC, E) represents his utiiity from having a chiid with chiid supportenforcement, a*vt represents the minimum a required for the utiiity of having a chiidto exceed that of not having one ((/,,,(C, NE)> UJNC)) under no enforcement. Asis evident from the graph, this critical value is iower than that under enforcement a* vf< a*^. The intuition behind this is simpiy that the constraint introduced by childsupport enforcement (if it is binding) lowers a father's utility ibr any given a. Thefourth iine ((/,„ {C,E,W)) represents the indirect utility associated with having a childwith a woman on welfare under a regime of strict chiid support enforcement. Thecriticai vaiue a*fn, above which men will choose to have chiidren. is higher stiii.

III. Child support enforcement. Prenatal Investmentsand Birth Outcomes: Evideuce from Vital StatisticsData

Our analysis of the impact of stricter child support enforcement onbirth rates, prenatal investment and birth outcomes differs in important ways fromprevious research. Previous studies have examined the impact of state child supportpoiicies on the nutnber of births among single tiiothers (Case 1998; Garfinkel. Huang,

6. This argument presumes thai welfare use has no independent effect on child investment or well-being.That is. a mother with a given a and income will invest the sameamouni in her child regardless of the sourceof income (earnings, child support, or welfare). Alternatively, one could argue that relying on welfare is anegative investment in the child and therefore negatively related to cc. This would amplify the results, pro-viding an additional incentive for men to select with women with a high a.


Figure 1Stricter Child support enforcement and the Composition of Mothers

McLanahan. and Gaylin 2003). We are the first to examine the impact of stricter childsupport enforcement on the composition/quahty of mothers (as measured by tbeireducational attainment) and the prenatal investments they make in their children asmeasured by their early initiation of prenatal care and number of prenatal visits.

Another important distinction lies in our empirical methods. A major challenge ofestimating Ihe impact of stricter child support enforcement on fertility and maternalinvestment is to identify the effect of child support enforcement separate from otherpotentially unobservable factors that might affect both the state's child support poli-cies and fertility or investment in children. Previous studies of the impact of child sup-port enforcement policies on fertility attempt to control for unobserved influencesusing a variety of methods. Most compar;; changes in fertility over time betweenstates that adopt stricter child support enforcement and those that do not (also referredto as a state fixed effect approach). This provides some control for policy endogene-ity by controlling for state-level factors that are fixed and do not vary over time, butnot factors that might vary within state over time.^ ln this paper we adopt an alterna-tive approach that controls for time-varying unobservable factors within a state that

7. Case (1998) is a notable exception. She fxpliL-itly lecognizes this possibility in her invesligalion of iheimpuc! of laws establlshitig stricter child support ctitbrcement on the number of otii-of-wedlock births. In asimple OLS, she iinds ihat the laws predict iin increase iti oti(-ot-wedlock bitlhs. After instrumenting !br thestate laws using churacieristies of the Male legislalurcs. she finds ihat stricter child support enforcement doeslead to a decline in t)ut-of-wedlock births, though ihe results are not significant. Her results underscore theimportance of controlling for policy endogencity.


might affecl child support enforcement and fertility/investment. Examples of suchfactors might include changes in society values and norms or changes in the avail-ability and cost of contraception.

To do this we obtain within state difference-in-differences estimates ofthe effect ofstricter child support enforcement. This estimate is obtained by defining two groups:a treatment or experimental group and a control group. Because state child supportenforcement expenditures target single women, we define single women as the treat-ment and married women as the control group. We therefore construct our estimatesofthe impact of stricter child support enforcement by comparing its impact on the fer-tility/investment decisions of single mothers relative to married mothers residing inthe same state. The assumption of this identification strategy is that changes over timein the behavior of men and women that are unrelated to child support enforcement arecommon to single and married women. It also assumes that married women are unaf-fected by child support enforcement. While this assumption is restrictive, it is prefer-able to simply comparing fertility/investment decisions under strict child supportenforcement regimes with those under more lenient regimes because it allows one toidentify the effect of stricter child support enforcement separate from broader changesin social norms, the cost of investing in children or other factors as measured by thebehavior of married women.

Because these single-married fertility/investment differentials can be computedwithin cells defined by state and year, our approach implicitly controls for all time-varying state specific factors by allowing us to include not only individual state andyear fixed effects buy state*year fixed effects (as discussed further in the next section).The stale*year fixed effects subsume the individual state and year main effects so thatthey need not be included separately. In other words, the vector of .siate*year fixedeffects includes individual state and year fixed effects as well as an interactionbetween them.

A. Data and Estimation

Data on all births in the United States come from the 1985-99 Natality Detail Filescollected by the CDC.*̂ Summary statistics are provided in Appendix Tables AI andA2. Rather than perform the analysis on a data set of all individual births, we aggre-gated the data to the group level with groups defined by year, state, age group, race,education, and marital status.'' We matched these data with data on child supportenforcement by the year in which the child was conceived (calculated using gesta-tiona! age and date of birth). Our measure of child support enforcement is expendi-tures on child support that come from the annual reports ofthe Office of Child supportenforcement, U.S. Department of Health and Human Services. This represents adeparture from other work that has focused on either specific policies/laws withrespect to child support enforcement and/or child support expenditures. We focus on

8. Beginning in I9H5. ull births were included as part of the tile. Prior to 1985. only a random sample ofbinhs were included lor some states. The linal dafy sei Includes all births conceived 1985-98. However, weneeded to use nataliiy data friim 1999 lo calculate ihe lotal number of cbildren (bai were conceived in 1998.9. An example of a single group or cell would be all births lo single while moibcrs age 20-29 in Montanain 1990.


expenditures for the following reasons: (1) Previous work has found the data on childsupport policies to be "incomplete and in some places inconsistent" (Freeman andWaldlogel 1998); and (2) multiple policies and laws require selection of an arguablyarbitrary subset. However, we do repeat the analysis with a single policy—whetherchild support is withheld from father's wages; we discuss those results in the follow-ing section.

The equation estimated takes the form:

(I) y»mra.- = P/ siugle,,, + % \niCSE)*single,,,^,.^ + p , Race, + p^ Education,+ P5 Age,, + pft state*yearly., + e,,,,,,,,̂

In the above equation CS£ refers to the average annuai expenditures in the state in thethree years prior to the year of conception. We do this because we believe that theeffects of increased child support enforcement are not immediate but are lagged, andalso to further limit the potential for policy-endogeneity bias. Single is an indicatorequal to one if the birth mother is single (representing the treatment group). Thus thecoefficient on the term \nCSE*single{^2^ represents the within-state difference in-dif-ferences estimate: the impact of the child support enforcement on single women rel-ative to married women in states that experience large increases in CSE relative tothose in states with small changes in CSE. The vector Race includes the dummy vari-ables for the race ofthe mother (black, white. Hispanic. Asian, and other is the omit-ted group), the vector Age includes dummies for the age of the mother (teen, age2(}-29, age 30-39. and age 40-5,*^ is the omitted group) and Education include.s theeducational status of the mother (less than HS, HS. HS+). For this analysi.s. wedropped college-educated mothers as single college educated mothers are quite rareand unrepresentative. Recall that because the data are aggregated to the eell levelbased on state, year. race, marital status, age. and education, only binary indicators forthese measures, not continuous variables, cai be inciuded.

As noted previously, the inclusion of state*year fixed effects subsumes individualstate and year fixed effects, so they are not included separately in the analysis (in otherwords, they are included in the state*year fixed effects)."^ Including state*year fixedeffects also controls for all factors that may vary within states ovev time, such as unem-ployment rates and AFDC benefits that may be correlated with both the state's childsupport policies and birth outcomes as well as any other factors that vary within stateover time that affect the whole population." In addition, the vector of state*year fixedeffects subsumes the main CSE term (since it varies by state and year) but not the term]n(CSE)* single—enabling identification ofthe impact of CSE on birth outcomes.

The outcomes (K) are the (I) log ol' the number of births. (2) the proportion ofmothers in the cell who initiated prenatal care in the first trimester, (3) the averagenumber of prenatal visits, and (4) the proportion of low birth weight (LBW) births.

10. State^year fixed effects include individual state and year fixed effects. This is can be shown with the fol-lowing short example. Consider two years and two stales. If slate*year fixed effects are included, ihere wouldhe three dummies included: ,v(«(c/*mir/, state I'^yearl and stttte2*year! {stute2*year2 is theomiited or rel-erenue category i which is lhe equivalent of two main effi;cts and an interaction (we are grateful to the ediiorfor suggesting this illustrative example].\ 1. Expan.sions in the Medicaid program represent a change in policy lhat might he of particular concernwith respect to itnprovenient in birth outcomes over this period. Since Medicaid expansions were all imple-menfed by 1486 we run the regressions a second time for the period 1987-98 and get the same results.


For the first of these, we also include the log ofthe number of women in the cell (thedenominator) as a covariate and weight by the size of the denominator.'- For the sec-ond, third, and fourth we weight by the number of births in the cell.

B. Results

Estimates of the impact of CSE on the number of births are presented in Table I. Thetirst column displays estimates of the impact of stricter CSE on the number of births.Consistent with previous work, we find that stricter child support enforcement leadsto a significant decline in births to single women relative to married women. The esti-mated coefficient ^^ (-0.033) can be interpreted as the elasticity of the birth rate withrespect to changes in child support expenditures. That is, a I percent increase inexpenditures leads to a decline in single (nonmaritai) fertility relative to marital fer-tility of 0.03 percent. In the next three columns of Table I. we examine whetherstricter C5£ differentially affects the fertility of some single women more than others,thereby potentially altering the average underlying characteristics of single mothersand their children. We previously predicted that the fertility of those single mothersmost likely to use the AFDC program would decline relative to others. Though we donot have an exact measure of a woman's propensity to use welfare, we do have meas-ures of maternal education, which is highly negatively correlated with welfare use. InColumns 2-4 we present the results stratified by maternal education. Consistent withour prediction, stricter CSE appears to have the greatest negative impact on the fertil-ity of the women who have the least education and therefore are most likely lo usewelfare. The estimated elasticity is -0.094 for women without a high school degreeand -0.039 for women with a high school degree. In contrast, for women with somecollege, the elasticity is slightly positive (0.023).

In Table 2 we present esti tiiates of the impact of stricter CSE on two measures ofprenatal investments of mothers—early initiation of prenatal care and the numberof prenatal visits. Column 1 presents estimates of the impact of increasing childsupport expenditures on the proportion of single mothers in the state that initiatecare in the first trimester, controlling for age, education and race. We find that sin-gle women in states with stricter child support enforcement are more likely to initi-ate care in the first trimester. Over the period 1984-98, the share of single womeninitiating care in the first trimester increased from 0.548 to 0.708 (see AppendixTable 2). Given that CS'E increased on average 300 percent over this period, the esti-mated coefficient of 0.018 suggests that half of this increase can be attributed to theincrease in child support enforcement.'^ Similarly in Column 2 we present resultswhich suggest that the increase in child support enforcement also led to a positiveand significant increase in the number of prenatal visits among single women andwas responsible for 6 percent of the observed increase in prenatal visits over thisperiod. Finally, in Column 3. we present results for LBW. The proportion of single

12. The detiominator (the tiumher of women of a given race, education level, age and marital status in agiven state) was calculated from the 1990 Census.13. We tihtain this hy estimating the impact of the average increase in child support expenditures 1984-98(roughly .100 percent) on the increase in prenatal care as a percent of the total increase in prenatal careobserved over this period.

38 The Joumal of Human Resources

Table 1Impact o/Z7i(Expenditures) on /./i(Births). VS. Natality Data I985-9H

Ln(expenditures)*single

Less than high school

High school

Some college

Black

WWte

Hispanic

Asian

Single

Teenager

Age 20-29

Age 30-39

LiUpoptilation)

ObservalionsR-squarcd

All

- O.O32.'i.'i[0.005551

-0.0525[0.01075]0.0188

[0.008851-0.42096[0.0089210.57623

[0.03475]-0.2756[0,03571[0.16181

[0.03509[-0.85988[0.03728]

-1.47044[0.00946 i3.59184

10.01 123J4.05495

[0.00795]3.27318

[O.(H)759]0.90661

[0.00385]62.0260.91

Less ThanHigh School

- 0.094[0.0101

0.802[0.048]0.012

[0.049]0.695

[0.049]-1.278[0.053]

-0.513[0.018].1.535

[0.018]4.673

[0.017[3,236

!().017[0.758

[0.007[16,6800.93

High School

-0.039[0.0t)7[

0.514[0.041]

-0.088[0.043 [0.068

[0.041[-0.616[0.045 [

-0.904[0.011]4.259

[0.016]4.667

[O.(M)9]

3.504[0.008]0.902

[0.006]17,3930.97

Some College

0.023.[0.007]

0.71[0.047]

-0.098[0.052 [0.085

[0.048]-0.602[0.0501

-1.739[0.012]2.689

[0.019]3.83

[O.OIO]

3.139[0.010]0.896

[0.007]16.0410.96

Standard errors in brackets. A vector of 714 siaie*year fixed effects is included in all regressions.Regressions based on individual hirths aggregated to state, year. ruce. age, and education cells.

women with LBW babies declined from 0.112 to 0.099 over this period. The resultssuggest that the increase in child support enforcement over this period explains 13percent ofthis decline.

In Table 3. we present estimates of the impact of stricter child support enforce-ment on fertility, prenatal investment, and birth outcomes stratified by race. For non-black mothers (Columns 1^). increasing expenditures on child support leads to areduction in the number of births, an increase in the proportion of women initiatingearly prenatal care and the number of visits, and a decline in the proportion of LBWbirths. For black mothers, however, while the number of births decline and the pro-portion initiating prenatal care early increases as does the number of prenatal visits.


Table 2Impact of LiHcxpcndhures) on Frenatal Investments and Birth Outcomes, U.S.Nalalitv Data I9S5-9H

]jc\(expenditures)* single

Less than high school

High school

Some coiiege

Black

White

Hispanic

Asian

Single

Teenager

Age 20-29

Age 30-39

ObservationsR-squared

1st Trimester

0.01847[0.00032]

-0.17752[0.00056]

-0.08365[0.00046]

-0.03683[0.00049]0.01189

[0.00154]0.09584

[0.00150]0.01465

[0.00155]0.0148

[0.00179]-0.13702[0.00060 [

-0.02757[0.00135]0.02047

[0.00127]0.04158

[0.00128]80,1800.91

Number Of Visits

0.08071[0.00344]

-1.71384[0.00600]

-0.64377[0.00488]

-0.18678[0.00521]0.10134

[0.01641]1.15917

[0.01595]0.16596

[0.01648]0.19501

[0.01900]-1.03312[O.(K)638]0.01165

[0.01441]0.01246

[0.01354]0.05192

[0.01362]80,0380.89

LBW

-0.00154[0.00012]0.02879

[0.00021]0.01509

[0.00017]0.00647

[0.00018]0.06202

[0.00058]0.0066

[0.00056]-0.00111[0.00058]0.01646

[0.00067]0.02737

[0.00022]-0.03487[0.00051]

-0.03367[0.00047]

-0.02104[O.(X)O48]80,2330.79

.Standard errors in hracket.s. A vector ot" 714 state*year lixed effects is included in all regressions.Regressions based on individtial births aggregated to .state, year, race, age. and education cells.

birth outcomes do not appear to improve. For black mothers, we find no significanteffect of child support enforcement on LBW.

To provide additional evidence in support of this finding, we also estimate a dis-crete time duration model ofthe impact of CSE on time until first birth, using data onindividual women from the NLSY79. The NLSY79 consists of a nationally represen-tative sample of 12,686 young men and women age 14-22 in 1979 interviewed annu-ally from 1979 to 1994 and then biennially until 2000. The NLSY includes 6,283women of whom 5.762 are included in this analysis (521 are excluded because theyalready had at least one child by 1979). Individual data allow us to include additional


00OS

4=

-o

^1

CO

d c d d d

v f-~. i/"d ^ t f̂ 4 O fN fN

-. C — — ^ _ _ H —-d C r- q i/~. C r-a o

— c o q c q o

• C q q — q — q q p

?S? o d d d d

d d d d d d d d d c d d d d d d d d d d d d

d — q — ^ T t — r^,

q q q q q ""dO in c — O f N q — q ' " j q 3 ; q q

. f^d i/"d — U~d 00 OC - ^ >£> f d • *— r-. — i^d—'Ooqr---o — C —O — O — ' Q r ^ d O r J O ' l l Q 2 Q[;3O — c o S o o c o o c q q —

r i — O O" OC C

g ii

<


controls that might affect fertility such as AFQT score (the score on the Armed ForcesQualifying Exam, a measure of cognitive ability), whether her mother worked whenshe was 14 years old, and the highest grade completed by her mother.

Though sample sizes prevent us from obtaining precise estimates, the hazard mod-els do suggest that the probability that single women wiil have a child in a given year,conditional on not having had a child to date, is lower in states that spend more onchild support enforcement. Increasing expenditures on child support enforcement twostandard deviations around the mean decreases the probability of birth for a singlewoman from 4.5 to 1.7 percent.'•" We also tind that though the likelihood ai' having achild decreases for single women when child support enforcement increases, it doesso at a slower rate for those with more education. This is consistent with our findingsbased on aggregate birth certificate data: stricter child support enforcement has thegreatest negative impact on the fertility of women most likely to use AFDC—thosewith the least education.

Our choice of expenditures as the measure of child support enforcement is subjectto the criticism that it might be confounded with efficiency or with the difficulty ofthe caseload of delinquent fathers. If, for example, expenditures were confoundedwith efliciency—^that is. greater expenditures in states with less efficient child supportcollection systems—then the results would be biased downward. Similarly, if expen-ditures were confounded with caseload difficulty (that is, states that increase theirexpenditures are those that already collect from those least reluctant to contribute andare attempting to collect child support from more reluctant fathers), thi.s might alsobias our estimates downward. On the other hand, if states with more efficient or lessdifficult caseloads spend more because the marginal benefit of expenditures is greater,then our results would be biased upward.

As a check, we estimate the impact of an alternative measure of child supportenforcement—an indicator for enactment of a law that automatically withholds childsupport payments from the wages of absent fathers. While the withhold represents analternative measure, it is stili positively correlated with child support expenditures (asare other child support iaws). We tind that, for example, adoption of [he automaticwithhold would explain 36 percent of the increase in first trimester prenatal care (lessthan the 50 percent explained by the increase in child support enforcement expendi-tures). While the results are qualitatively similar, they suggest a possible small upwardbias to the results based on child support expenditures.

IV. Child support and Prenatal Investments: Evidencefrom the Fragile Families Study

In this section we investigate potential mechanisms behind otir find-ing that increased child support enforcement leads to an increase in prenatal invest-ment. In particular, we wish to know whether this increase is due to positive .selectionas we hypothesize or some alternative mechanism. Two possibie alternative mecha-nisms include: (I) increases in paternal income transfers during the prenatal period (in

14. Results Livailabic upon requeM from the authors.


anticipation of future support obligations); and (2) increases in mother's ability to bor-row against future child support payments dtiring the prenatal period (as a result of apositive permanent income effect). We use data from the Fragile Families Study todetermine to what extent the increase may be attributed to either of these two alter-native mechanisms. Lack of evidence for either of these mechanisms would provideyet further support for our hypothesis of positive selection.

A. The Fragile Families Data

The Fragile Famiiies and Child Well Being Study interviews approximately 3.7(X)new unwed couples residing in 20 cities in i5 states shortly after the birth of theirchild. Parents are interviewed again when the child is between 12 and 18 months old.About 1,900 of these parents were not cohabiting at the time of the follow-up inter-view and therefore were eligible for child support and inclusion in our sample.A unique aspect of the Fragile Families data is the extensive information (taken fromboth maternai and paternal surveys) on noncohabiting fathers.

These data inciude information on whetlier a formai or informal child supportagreement exists at 12-18 months (34 percent). These data also include informationon child investments during the prenatal period: the date when the mother initiatedprenatal care and whether the father provided financial assistance to her during thisperiod. Lastly, the data contain information on whether the mothers can borrowmoney (at least $2(X)) if they need to—a measure of their credit constraint.

B. Results from Fragile Families

In the first two columns of Table 4 we present evidence regarding the impact offuture child support receipt and prenatal transfers on prenatal investments. In thefirst column we explore the relationship between future child support receipt andeariy initiation of prenatai care. We find that those mothers who receive futurechild support are more likely to initiate prenatal care early (0.085). as expected.Even when we control for whether prenatal transfers were received from the fatherin Column 2, the positive effect of future child support receipt remains (0.080),suggesting that the receipt of prenatal transfers cannot explain the positive impactof future child support receipt on prenatai investments. In results not presented, wefind that the same pattern emerges for LBW: Future child support receipt leads toa decline in LBW and the effect remains e\en when we controi for prenatai trans-fers, but the effects are impreciseiy estimated, most iiicely due to small sampiesizes.

To rule out the possibility that in anticipation of future chiid support payments,mothers are able to borrow additional money during the prenatai period (a permanentincome effect), we stratify the sample based on whether the mother is credit con-strained using the above-mentioned measure of credit constraint. The results are pre-sented in Columns 3 and 4. If it were the case that future child support receiptpositively affects maternal prenatal investment because the mother is able to borrowagainst future child support payments, then the positive impact of future child supportreceipt would be limited to (or at least stronger for) mothers who are not credit con-


Table 4

Relationship Bemeen Future Child Support Agreement, Prenatal Transfers andMaternal Investment

Future child support agreemenl

Prenatal transfers from father

Maternal Age

Mother loss than high school

Mother high school

Father less than high school

Father high school

Black

White

Hispatiic

Numher of children

Poverty level

ObservatiotisR-squared

All

0.08510.0221

0.0()3IO.(H)2|

-0.128[0.0701

-0.092[0.067]0.017

10.02810.033

10.02510.049

fO.O69]0.082

[0.07510.118

[0.073]-0.035[0.009]0.001

[0.009]1,6980.04

All

0.08[0.022 [0.056

[0.025]0.004

[0.002]-0.109[0.071]

-0.079[0.067 [0-003

[0.029]0.018

[0.027]0.058

[0.070)0.094

[0.076]0.12

[0.074[-0.036[O.(X)9[O.(H)3

[0.1K)9[1,6570.05

Constrained

0.101[0.078]

0.007[0.007 [0.055

[0.400]-0.018[ 0.389 [

-0.014[0.091]

-0.037[0.088]

-0.451[0.278[

-0.5[0.297]

-0.507[0.296]

- 0 . 0 6[0.025]

- 0 . 0 0 7[0.047]

2030.1

UticotiNtrained

0.088[0.023 [

0.003[0.003]

-0 .128[0.072]

- 0.078[0.068]0.02

[0.030]0.039

[0.027]0.092

[0.072]0.132

[0.078]0.175

[0.076]- 0.032

[0.010[0.001

[0.009]1,4890.05

Standard errors in brackets. Dependent variable is whether miiiher iniliuled prenitlal care in ihe first trimester.All regressions include stale fixed effects.

Strained. The empirical evidence from the Fragile Fatnilies does tiot support this, but

rather rules out the idea that the empirical findings based on the birth certificate data

were simp]y due to increased resources, not positive .selection.

V. Discussion

This paper is the first to examine hovv' the composition of motherschanges in response to an increase in the probability of future child support paymentsbrought about by stricter state child support enforcement. We find that the average

education of mothers increases, as does the level of maternal investments in children


(as measured by early investment in prenata] care), with stronger child supportenforcement. What is notable about these lesults is that the increase in maternalinvestment occurs prior to the receipt of any child support payments, suggesting thatstricter child support enforcement, in addition to increasing the material resourcesavailable to children, may benefit children through a birth selection process. That is.stricter child support enforcement reduces fenility, especially among mothers who areless likely to invest in their children, regardless of actual child support receipt. As aresult, we might expect current policies of child support enforcement to affect bothcurrent and future child outcomes. Existing research on the impact of child supportreceipt on child outcomes has not considered this mechanism. To fully understand theimpact of any changes in state policies affecting child support enforcement on childoutcomes, it is necessary to understand how policies might affect both the birth selec-tion process as well as the receipt of financial resources, conditional on the birth ofthe child.

Appendix Table AlSummary Statistics From Birth Certificate Data 1985-98

SinglePrenatal care initiated in 1st trimesterNumber of prenatal visitsShare LBWLess than high schoo]High schoo]Some collegeWhiteBlackHispanic

All

0.3810.755

10.90.0770.2780.4580.2530.6140.1840.163

Single

0.6349.90.1060.4390.4130.1360.40.3750.192

Married

0.8211.50.0620.1910.4820.3120.730.0810.147

Appendix Table A2Trends in CSF. Prenatal Investment and Birth Outcomes for Single Women

1985 1998 1998-85 Percent Change

Prenatal care initiated in 1st trimesterNumber of prenatal visilsShare LBWLn( Cliikl stipiHirt enforcement)

0.5489.10.1120.329

0.70810.670.0991.404

0.161.57

-0.0131.074

29.2%17.3%

-11.6%


References

Argys. Laura. Elizabeth Peters. Jeanne Brooks-Gunn. and Judith Sniilh. 1998. "The Impact ofChild Support on Cognitive Outcomes of Young Children." Demographw 35(2): 159-73.

Argys. Laura, Elizabeth Peters, and Donald Waldman.1995. "Can the Family Support Act PutSome Lite Back Into [>eadheat Dads? An Analysis of Child Support Guidelines. AwardRales, and Levels." The Journal of Human Resources 36(2):226-52.

Aiiginbaugh, Alison. 2001. "Signals of Child Achievement as Determinants of ChildSupport." American Economic Review Papers and Proceedings 91(2); 140-44.

Belief. Andrea, and John Graham. 1991. "The Effect of Child Support Enforcement on ChildSupport Payments." Populatiim Research and Policy Review 10:91-116.

Case. Anne. 1998. "The Effects of Stronger Child Support Enforcement on Non-MaritalEerlility." In Fathers under Fire: The Revolution in Child Support Enforcement, ed. IrwinGarfinkel. Sura McLanahan. Daniel Meyer, and Judith Seltzer, 191-25. New York: RussellSage Foundation.

Case, Anne. I-Fen Lin. and Sara McLanahan. 2003. "Explaining Trends in Child Support:Economic, Demographic, and Policy Effects." Demography 401 \y. 17]-^9.

Del Boca. Danieia, and Christopher Elinn. 1994. "Expenditure Decisions of Divorced Mothersand Income Composition." Journal of Human Resources 29(3):742-61,

Freeman. Richard, and Jane Waldfogel, 1998. "Dunning Delinquent Dads: The Effect of ChildSupport Enforcement Poiic-y on Child Support Receipt by Never Married Women." NBERworking paper 6664.

Garfinkei, Irwin. Daniel Gaylin. Chung Huang, and Sara McLanahan. 2(K)3, "Will Child SupponEnforcement Reduce Nonmarital ChW^hevinn-^'' Joumal of Population Economics. 16:55-70.

Hernandez, Pedro. Andrea Beller. and John Graham. 1995. "Changes in the RelationshipBetween Child Support Payment and Educational Attainment of Offspring, 1979-1988."DcHirf̂ ra/j/̂ .v 32(2):249-59.

Heim. Bradley. 2003. "Does Child Suppon Enforcement Reduce Divorce Rales? AReexamination." Journal of Human Resources 38(4):773-91.

Huang, Chien-Chung. 2002. "The Impact ol Child Support Enforcement on Nunmarital andMarital Births: Does It Differ hy Racial and Age Groups?" Social Service Review 76(2):275-301.

Knox, Virgina. and Mary Jo Bane. 1991. "Child Support and Schooling." In Child Supportand Child Weil-Being, eds. Irwin Grafinkel, Sara McLanahan, and Philip Robins. 285-316.Washington D.C: The Urban Institute Press.

Knox. Virginia. 1996. "The Effects of Child Support Payments on Developmental Outcomesfor Elementary School-Age Children." The Journal of Human Resources 31(4):816-^O.

Lerman. Robert. 1993. "Policy Watch: Child Support Policies." Journal of EconomicPerspeaive\.l{3)\]7\-S2.

Miller. Cynthia, Irwin Garfinkel. and Sara McLanahan. 1997. "Child Support in the U.S.: CanFathers Aftbrd to Pay More?" Review of Income and WeaUh 43(3):26]-81.

Nixon, Lucia. 1997. "The Effect of Child support enforcement on Marital Dissolution."Journal of Human Resources 32( 1): 159-81.

Plotnick. Robert, irwin Garfinkel, Sara Mcl-anahan, and Inhoe Ku. 2(H.)4. "Better ChildSupport Enforcement: Can it Reduce Teenage Childbearing?" Journal of Family l.t.sues25{5):634-57.

Sonenstein. Ereya. Joseph Pleck. and Leighton Ku. 1994. "Child Suppon Obligations and YoungMen's Contraceptive Behavior: What do Young Men Know? Does it Matter?" Paper Presentedat tbe Annual Meeting ofthe Population Association of American. Miami May 7. 1994.

Willis. Robert. 1999. "A Theory of Out-of-Wed lock Childbearing." Journal of PoliticalEconomy 107: S33-S64.

The Impact of Child Support Enforcement on Fertility ... Impact of Child Support Enforcement on...

Documents

Transcript of The Impact of Child Support Enforcement on Fertility ... Impact of Child Support Enforcement on...