“swapping” families: serial parenting and economic support for children

Journal of Marriage and the Family 62 (February 2000): 111–122 111

WENDY D. MANNING Bowling Green State University

PAMELA J. SMOCK* University of Michigan

l

‘‘Swapping’’ Families: Serial Parenting and Economic

Support for Children

The claim that fathers ‘‘swap’’ families when theyform new ones—that is, they shift allegiances fromnonresident children to new residential children(e.g., Furstenberg, 1995)—has not been directlyevaluated empirically. Drawing on data from thetwo waves of the National Survey of Families andHouseholds, we test Furstenberg’s argument interms of child-support transfers to nonresidentialchildren, and we also test an elaboration of hisapproach that distinguishes between resident bi-ological children and stepchildren. Using static-score models, our findings indicate that fathers doswap families but only when the trade-off is be-tween new biological children living inside fa-thers’ households and existing biological childrenliving outside fathers’ households. Even thoughour analytic sample is small, our findings haveimportant implications for child well-being, child-support policy, and the meaning of fatherhood.

From the perspective of children, families are in-creasingly likely to be spread across more thanone household, with one biological parent, typi-cally the father, living in a separate household. In

Department of Sociology, Bowling Green State University,Bowling Green, OH 43403 ([email protected]).

*Department of Sociology and Population Studies Center,University of Michigan, Ann Arbor, MI 48104.

Key Words: child support, divorce, fatherhood, nonresidentparenting.

1990, at least 22 million children—over one thirdof all children—in the United States did not livewith their biological fathers, and about half of allchildren are expected to live in a single-parentfamily at some point before adulthood (Bumpass& Sweet, 1989a; Norton & Miller, 1992). At thesame time, many fathers enter new unions, oftenacquiring new biological children or stepchildren(Sorensen, 1997).

These patterns raise a general question criticalto understanding both the well-being of childrenand the meaning of fatherhood in the UnitedStates: Do fathers reduce their social and econom-ic investments in children from whom they liveapart as they take on new parenting roles? In thispaper, we focus on the economic dimension of thisquestion. Our aim is to test a notion suggested byFurstenberg and colleagues that nonresident fa-thers ‘‘swap’’ families (e.g., Furstenberg, 1995;Furstenberg & Cherlin, 1991; Furstenberg, Nord,Peterson, & Zill, 1983; Furstenberg & Spanier,1984). Specifically, we examine whether fathersreduce their child-support payments to nonresi-dential children when they have new children. Wealso elaborate Furstenberg’s theory by developingand testing a hypothesis that distinguishes be-tween new biological children and stepchildren.

Our focus on child support is important be-cause social science research has shown that non-residential father involvement benefits children,and particularly that child-support payments fromnonresidential fathers positively affect children’soutcomes above and beyond dollars from other

112 Journal of Marriage and the Family

sources (King, 1994; Knox & Bane, 1994; Mc-Lanahan, Seltzer, Hanson, & Thomson, 1994).Additionally, policy makers have recently focusedon child-support reform as a foundation for im-proving the outcomes of children in the UnitedStates. Understanding the impact of fathers’ newfamily responsibilities is essential for this effort.

BACKGROUND

Patterns of father disengagement from nonresi-dential biological children and the formation ofnew families have been described in terms of‘‘swapping’’ families (e.g., Furstenberg & Nord,1985; Furstenberg & Spanier, 1984). The notionis that fathers trade old parenting responsibilitiesfor new, with fatherhood being a transient state inwhich parental obligations are dictated by resi-dence (Furstenberg, 1988, 1995; Furstenberg &Cherlin, 1991; Seltzer, 1991). The underlying ideais that nonresident fathers face new demandswhen forming new families, leading them to with-draw their ties to nonresident children and to basetheir economic and noneconomic investmentswithin their current households (Ihinger-Tallman& Pasley, 1987). Furstenberg and Cherlin (1991,p. 37) state, ‘‘When fathers invest in new familyresponsibilities, such as having additional chil-dren, the offspring from their first marriage some-times get the short shrift.’’ A part of this processmay include showing allegiance to a new wife orpartner as well as new children.

Trading allegiances from old children to newcan be conceptualized in at least two ways. Definedin terms of social involvement, the swapping hy-pothesis has in fact received some support; pastresearch indicates that children are unlikely to haveboth a biological father and stepfather socially in-volved in their lives on a regular basis (Fursten-berg, 1988; Mott, 1990; Seltzer & Bianchi, 1988).Direct empirical evidence to corroborate Fursten-berg’s theory in terms of the economic dimensionof nonresidential parenting is lacking. However,Furstenberg’s hypothesis is commonly regarded astrue for all domains of father involvement, includ-ing economic support. Given the association be-tween children’s economic well-being, part ofwhich is determined by child support, and chil-dren’s life chances, this issue deserves scrutiny.

Most child-support studies that take into accountfathers’ new family responsibilities have relied ona rough proxy of fathers’ new parenting responsi-bilities: fathers’ new marriages (Furstenberg, 1988;Peters, Argys, Maccoby, & Mnookin, 1993; Seltzer,

1991; Sonenstein & Calhoun, 1992; Teachman,1991; exceptions include Hill, 1992; Veum, 1993).This is problematic for at least two reasons. First,researchers have sometimes, because of data limi-tations, used mothers’ reports of fathers’ new mar-riages (McLanahan et al., 1994; Seltzer, 1991;Teachman, 1991). However, mothers are not al-ways reliable sources of information about fathers’lives. Seltzer and Brandreth (1994), for example,find significant differences in resident mothers’ andnonresident fathers’ reports of fathers’ remarriages.More important, remarriage is not an adequatemeasure of new parenting responsibilities becauseit does not distinguish the formation of new unionsfrom raising new children. To test Furstenberg’s hy-pothesis, the needed measure is one that focuses onchildren within and across households rather thanon ties to an adult via remarriage.

Two approaches have been applied to examinenonresident fathers’ financial disengagement. Oneapproach uses mothers’ reports of child-supportlevels paid at specific years since separation. Thiswork suggests that, as time since separation in-creases, the levels of child support paid decline(Seltzer, 1991; Teachman, 1991). However, thiswork can only be considered suggestive, becausethe aggregate comparisons do not permit analysisof individual fathers’ changes in child support paid.As a result, this type of analysis cannot assess thefactors associated with changes in child supportpaid. The second approach uses longitudinal datato examine changes in child support paid by non-resident fathers. A study using relatively youngnonresident fathers (28–31 years old) from the Na-tional Longitudinal Survey of Youth (Veum, 1993)and another study, based on married couples fromthe original 1968 sample of the Panel Study of In-come Dynamics (PSID) (Hill, 1992), include co-resident children primarily as a control variable anddo not explicitly test the swapping hypothesis.These studies suggest that coresident children donot influence changes in child support paid, butthey do not account for the biological relationshipof these children to the nonresident parent.

HYPOTHESES

Our first hypothesis is drawn from Furstenberg’spremise that fathers trade parenting responsibili-ties, withdrawing from nonresidential childrenwhen they acquire new, coresidential children. AsFurstenberg et al. (1983) state, ‘‘Social parent-hood takes precedence over biological parent-hood’’ (p. 666).

113Swapping Families

Hypothesis 1: The addition of new residentchildren (i.e., either biological children orstepchildren) lowers economic investmentsin nonresidential, biological children.

We operationalize ‘‘economic investments’’ aschild support payments in our analyses because itis the primary means by which nonresident par-ents transfer income to their children. The emo-tional and financial responsibilities resulting fromresiding with children may interfere with fathers’possible relationships with their nonresidential bi-ological children. Nonresident fathers with newfamilies may perceive competing responsibilitiesand loyalties across households (Clingempeel,Coylar, & Hetherington, 1994; Ihinger-Tallman &Pasley, 1987). It also may be easier and lessstressful from a father’s perspective to provide forchildren who are sharing a residence with himthan to do so for children living in other house-holds. Providing economic support often requirescontact with the child’s mother, sometimes caus-ing strains that a father may wish to avoid (Adritti& Kelly, 1994; Ahrons, 1993; Arendell, 1995;Seltzer, McLanahan, & Hanson, 1998). Economictheory also supports this hypothesis. Whereas fa-thers can be certain that all of their economic in-vestments in resident children are received by thechildren, only a portion of the payment to mothersfor nonresident children may be received by thechildren themselves. Thus, as fathers may face re-duced incentives to pay child support (Weiss &Willis, 1985), they may choose to invest in resi-dent children and not in nonresident children.

Hypothesis 1 does not distinguish between newbiological children and stepchildren, reflecting Fur-stenberg’s contention that coresidence is the criticalelement determining fathers’ investments. In oursecond hypothesis, we refine Furstenberg’s theoryby distinguishing between new biological childrenand stepchildren. Conceptual arguments from eco-nomics, evolutionary biology, and sociology implythat nonresident fathers only raising stepchildrenwill not swap family responsibilities; the trade-offis between supporting biological children living in-side and those outside their household.

Economist Gary Becker (1981), for example,suggests that there is an important distinction be-tween stepchildren and biological children. He ar-gues that men invest more heavily in ‘‘marital-specific capital’’—one type of which is biologicalchildren with a new wife or partner—than in chil-dren from a previous union. Because stepchildrendo not represent marital-specific capital according

to Becker, the implication is that their presencewill not influence fathers’ economic contributionsto biological children living elsewhere.

Evolutionary biologists take a different ap-proach, but their argument leads to similar conclu-sions. Trivers (1974) suggests that a conflict of in-terest arises between parents and biologicaloffspring when investment in one child limits in-vestment in future children. In response to such asituation, fathers may divert resources from olderoffspring to younger offspring when older off-spring appear capable of surviving without theirfathers’ resources. In the human context, one signalof capacity to survive might simply be that a childlives with the mother and experiences a moderatelevel of well-being. This argument does not applyto stepchildren; it is only when the future offspringare biological children that the trade-off occurs.

Finally, family sociologists argue that stepfam-ilies are ‘‘incomplete institutions,’’ an arrange-ment in which the roles (i.e., rights and obliga-tions) of family members are not clearly defined(Cherlin, 1978), especially as compared to firstmarriages with only biological children. This im-plies that fathers might not respond to the pres-ence of stepchildren in their decision-makingabout contributions to nonresidential children. Thefollowing empirical evidence supports the distinc-tion between stepchildren and biological children.Fathers’ parenting of children differs dependingon the biological relationship to them (Marsiglio,1992): stepchildren are more often excluded fromfathers’ lists of family constituents than are bio-logical children (Furstenberg & Cherlin, 1991),and nonresident fathers living only with biologicalchildren report more benefits from parenthoodthan nonresidential fathers living with stepchil-dren (Seltzer & Brandreth, 1994).

Hypothesis 2: Only new biological childrenlower fathers’ economic investments innonresidential, biological children; new res-idential stepchildren have no effect on fa-thers’ economic investments in nonresiden-tial, biological children.

This hypothesis argues for separate effects ina particular direction and is essentially an elabo-ration of Hypothesis 1, not an alternative to it. Forexample, the empirical results could support bothHypothesis 1 and 2. This would suggest that, ingeneral, the mechanism driving the relationshipbetween coresident children and child-supportpayments is coresident biological children, indi-


cating the need for a substantive refinement ofFurstenberg’s theory. This would also be the case,of course, if Hypothesis 2 is supported but Hy-pothesis 1 is not, which would imply that col-lapsing different kinds of new children into onecategory obscures differential effects of new bio-logical and stepchildren.

METHOD

DataThe National Survey of Families and Households(NSFH) is a national probability sample of ap-proximately 13,000 individuals. The first wave(NSFH1) was conducted between 1987 and 1988.The second wave (NSFH2) reinterviewed respon-dents and was conducted between 1992 and 1994.The response rate for the first wave of the NSFHwas 74%, and the follow-up response rate wasroughly 82% of the original sample (Sweet, Bum-pass, & Call, 1989).

An important advantage of the NSFH is that itdirectly asks respondents whether they have chil-dren living outside the household. Other datasources require indirect identification of nonresi-dential fathers (e.g., Survey of Income and Pro-gram Participation) or do not attempt to gatherinformation from nonresident fathers (e.g., Cur-rent Population Survey). Another asset of theNSFH is that we can infer causality because thetwo waves of data collection allow us to conductdynamic analyses that link changes in child-sup-port payments over time to changes in coresidentchildren. The NSFH is also suited to this projectbecause it contains rich data on many aspects offamily experiences, such as legal support arrange-ments, income, residential histories, and marriageand cohabitation histories.

The eligible sample was based on male mainrespondents, who reported having an absent ‘‘fo-cal’’ biological child less than 13 years old resid-ing with his or her mother and who provided com-plete child support information (n 5 298) in thefirst wave. Ten percent of the sample was limitedbecause of missing child-support data. A focalchild was randomly selected among all absentchildren less than age 18, and child-support ques-tions referred to a focal child (and his or her co-resident siblings), not to all sets of nonresidentialchildren. Because we focused on changes in child-support payments across the waves, we also re-stricted the sample to children less than age 13 atthe first wave, so that they remain eligible forchild support at the second wave. Fathers, in many

cases, did not owe support or pay for other ex-penses when children reached age 18.

Our final analytical sample was restricted tothose fathers who met the above criteria and whowere also reinterviewed at the second wave; re-mained nonresidential fathers of the same focal,absent child at both waves; reported that the childcontinued to live with his or her mother; and pro-vided data on the amount of child support paid atWave 2. Of the 298 eligible nonresidential fathers,213 were reinterviewed. Next, the sample was re-duced to 170 to assure that the focal child re-mained the same at both waves. A different focalchild was sometimes referenced in the secondwave, and there was no systematic reason for thischange. The sample was further limited to 155cases in which the focal child continued to livewith his or her mother at Wave 2. We use thiscriterion because child-support obligations changeif children move in with their fathers, other rela-tives, or foster parents. Our last criterion was thatcomplete child-support information was provided.

The resulting sample contains 133 cases or onlyabout 45% of the cross-sectional sample. This maybe problematic for two reasons. The first is smallsample size. Although unfortunate, we believe thatthe importance of the research aims, combined withthe strength of the dynamic analyses, outweigh res-ervations about sample size. Our small sample sizewill lower the statistical power of our estimates.Another potential problem is clearly sample selec-tivity. Fathers in our sample may differ in unmea-sured ways from those that did not meet all criteria.If this is so, parameter estimates of effects may bebiased (Heckman, 1979). We address this problemin two ways. The first is simply by comparing thedistributions of independent variables of thosecases that meet all our criteria with those that donot. Second, we estimate a sample selection modelthat jointly predicts falling into our final sampleand child support. By allowing the error terms ofthe two equations to be correlated, this model at-tempts to adjust coefficients for potential observedsample selectivity.

A final limitation of our sample and data is onethat plagues virtually all studies of nonresident fa-thers. Fathers may either not be reporting all oftheir absent children (Cherlin, Griffith, & McCar-thy, 1983; Seltzer & Brandreth, 1994) or not par-ticipating in the survey at all. Unfortunately, with-out the use of administrative records, it is difficultto assess the accuracy of fathers’ reports of non-resident children. The likely direction of the bias isthat the nonresident fathers identified in our study


have lived with their children, probably represent-ing fathers with closer economic and social ties totheir children than the population of nonresidentfathers as a whole (Lin & Schaeffer, 1995; Schaef-fer, Seltzer, & Kalwitter, 1991). Both underrepre-sentation of nonresident fathers and underreportingof their children remain unresolved problems forresearch on fathers, and our selection models donot account for these unobserved biases.

Measures

The dependent variable in our multivariate anal-yses is the logged monthly amount of child sup-port reported paid by the nonresidential father atWave 2. The measure includes payments to thefocal child and to his or her coresident siblings.Some research suggests that nonresident fathersand resident mothers report similar levels of childsupport (Seltzer & Brandreth, 1994; Smock &Manning, 1997), and fathers’ own reports of childsupport paid are important to consider. Our mea-sure of child support captures both formal and in-formal payments. Including both types is impor-tant because some fathers make payments tononresidential children without a child-support or-der, or they pay more than the order (Seltzer,1994). We control for child support paid at Wave1 (measured in a similar manner to the Wave 2variable), resulting in analyses that predict changein child support paid. The child-support amountsare adjusted by the Consumer Price Index to rep-resent constant 1994 dollars.

To operationalize our hypotheses, our key in-dependent variables concern new coresident chil-dren. We construct the following variables: (a)number of new coresident children; (b) number ofnew coresident biological children (i.e., births be-tween interviews); and (c) number of new step-children acquired between interviews. A modelcontaining the first variable tests Hypothesis 1,and a model containing the latter two, in lieu ofthe first, tests Hypothesis 2. To test the secondhypothesis, we also include a control variablemeasuring the number of biological children whomove into a father’s household between the twowaves; these are not the ‘‘new’’ children implicitin the swapping hypothesis, but they need to betaken into account.

Our other control variables are drawn frompast theory and research. Almost all are measuredas of the first wave of the survey and include char-acteristics of the father and the focal child. Theonly variable available in the first wave about the

mother was the father’s proxy report of her currentmarital status; we have excluded this informationbecause of missing data and low expected reli-ability (see Seltzer & Brandreth, 1994). Sociode-mographic and economic characteristics of the fa-ther include age, race, current union status,educational attainment, and earnings. We measureage of the father in years and anticipate that olderfathers have the economic resources to pay theirchild support (Zill, 1996). Race is categorizedbroadly into White and nonwhite because of sam-ple size limitations. Prior work suggests that non-white fathers are less likely to pay child supportthan white fathers, partly because of differencesby race in the prevalence of nonmarital births(Sorensen, 1997). The relationship between re-marriage and child support varies in prior work(e.g., Seltzer, 1991). Current union status is coded‘‘1’’ if the father is either married or cohabitingat Wave 1. We include both union types, becausea substantial minority of individuals, especiallyamong those who have been previously married,enter cohabitations rather than remarriages (e.g.,Bumpass & Sweet, 1989b). We code educationalattainment as a series of four dummy variables.Studies indicate that fathers’ education is posi-tively associated with child-support payments(Braver, Fitzpatrick, & Bay, 1991; Veum, 1993).A father’s economic situation is captured with hisearnings and is coded by dividing earnings by10,000. Prior work indicates that earnings are pos-itively associated with child-support levels (Brav-er et al., 1991; Hill, 1992; Smock & Manning,1997; Teachman 1991). We include a squaredterm for earnings to test whether the variable earn-ings has a nonlinear effect.

We also measure other aspects of fathers’ livesthat have been found to be related to child sup-port. We include a dummy variable indicatingwhether the father reports having a legal child-support agreement. Legal agreements are associ-ated with higher child-support payments (Seltzer,1991). We also incorporate a measure of whetherthe father ever lived with the focal child, becauseresidence with the child may create stronger tiesto the child and thus be associated with higherlevels of payments (Seltzer, 1991). Finally, we in-clude a measure of the years since the father andchild last coresided. Past research indicates thatlonger durations since coresidence have a negativeimpact on levels of child support payments (Selt-zer, 1991). Fathers who never lived with theirchild are assigned the value of the age of the child.

The focal child variables include age, sex, dis-


tance from the father’s current residence, and thenumber of coresident siblings of the focal child.Older children have been found to be more likelyto receive payments than younger children (Fur-stenberg & Harris, 1992; Seltzer, 1991). Age ofthe focal child is coded in years. We include thegender of the focal child as well, although pastresearch yields mixed results about the effects ofthis variable (Furstenberg et al., 1983; Paasch &Teachman, 1993; Seltzer, 1991; Seltzer & Bianchi,1988). Distance between the child’s and father’sresidences is coded as a continuous variable inmiles. Previous research indicates that the greaterthe physical distance, the lower the likelihood offinancial payments (Seltzer, 1991; Sonenstein &Calhoun, 1990). We include the squared value ofdistance to test whether distance has a nonlineareffect. Finally, we include a measure of the num-ber of full siblings less than age 18 living in thefocal child’s household. This is an important vari-able because the number of siblings will be relatedto the amount of child support owed by the father.Generally, the greater the number of children, thelarger the child-support obligation. We includenumber of siblings at both Wave 1 and Wave 2 toallow us to control for child-support obligationsat both interview waves.

Analysis

We conduct multivariate analyses using maximum-likelihood tobit models to analyze changes in childsupport. We use tobits rather than ordinary leastsquares, because the dependent variable is quiteskewed, due to a substantial proportion of fatherswho report that they pay no child support. Tobitestimation assumes that independent variables haveboth linear and nonlinear effects and produces con-sistent estimates of coefficients and standard errors.We employ static-score models to predict changesin child-support payments across the two waves(Finkel, 1995; Kessler & Greenberg, 1981). Thestatic-score model includes a lagged endogenousversion of the dependent variable. In the currentapplication, the lagged variable is the loggedmonthly child support paid at Wave 1.

Because of the possible selectivity of our sam-ple, we estimate a sample selection model to ac-count for observed selection. It jointly estimatestwo equations: a probit predicting whether a caseis included in our sample and a tobit model pre-dicting change in child support. The probit equa-tion uses all potentially eligible cases (n 5 298)and predicts cases included in our sample (n 5

133). The correlation between the equationsshows the direction and magnitude of nonrandomassociations between sample status and changes inchild support. If a model that assumes the corre-lation is zero fits the data well, then our single-equation estimates are unbiased. We estimate themodel using maximum likelihood in HOTZTRAN(R. Avery & J. Hotz, unpublished data).

Our analytic strategy is as follows. We estimatemodels to test each of the two hypotheses. Thefirst model in each case is a baseline model thatonly includes increases in coresident children,along with important controls (child support paidat Wave 1, number of coresident children at Wave1, and number of the focal child’s siblings at bothwaves). Given our sample size constraints, wethen determine whether each of the control vari-ables separately contributes to the fit of the model.We assess the goodness of fit by conducting log-likelihood ratio tests. To test the first hypothesis,we evaluate whether increases in the overall num-ber of coresident children reduce child supportpayments. A negative and significant coefficientfor the increase in coresident children would beconsistent with the notion that fathers swap fam-ilies. To test the second hypothesis, we distinguishbetween coresident stepchildren and biologicalchildren, evaluating whether increases in stepchil-dren or biological children reduce child support.If the coefficient for increase in stepchildren isinsignificant and the coefficient for increase innew biological children is significant and nega-tive, then we accept Hypothesis 2.

RESULTS

Descriptive Statistics

Table 1 shows that two fifths of nonresidential fa-thers coreside with other children at Wave 2, in-dicating that a substantial percentage of nonresi-dential fathers are potentially economicallyresponsible for at least two households. Nearlyone quarter are residing with two or more chil-dren. The next column shows that three tenths ofnonresident fathers are coresiding with biologicalchildren, and the final column indicates that 15%are coresiding with stepchildren. Further analysisconfirms that remarriage is a poor proxy for non-resident fathers’ family responsibilities. Approxi-mately two fifths (41%) of men living in unionsat Wave 2 (39% in formal marriages and 48% ininformal unions) are not coresiding with any chil-dren, whereas 7% of nonresidential fathers not in


TABLE 1. NONRESIDENT FATHERS’ PARENTING

RESPONSIBILITIES

Coresident Children (%)

Total Biological Step

Number at Wave 20121

60.017.023.0

100.0

69.814.915.3

100.0

84.28.57.3

100.0

Number of new coresident children between waves0121

68.520.111.4

100.0

78.814.07.2

100.0

89.55.64.8

100.0

Note: Weighted percentages and unweighted n’s (n 5133).

Source: National Survey of Families and Households,Waves 1 and 2.

TABLE 2. NET CHANGES IN MONTHLY CHILD SUPPORT

PAID BETWEEN WAVES

ChangeProportion of All

Fathers (%) Mean Change

DecreaseNo ChangeIncrease

40.115.244.6

$174.17$0.00

$248.46Total 100.0 $40.70

Note: At Wave 1, 75.0% of fathers paid child support,with a mean monthly amount paid of $235.46; at Wave 2,77.2% of fathers paid, with a mean monthly amount of$276.21. All amounts are converted to 1994 dollars, andweighted values are reported. n 5 133.


unions are raising a child in their household (re-sults not shown).

Table 1 also shows the percentage of nonresi-dent fathers who at Wave 2 are residing with newchildren (i.e., men who are living with children atWave 2 with whom they were not living at Wave1). A considerable share of men coreside with newchildren at Wave 2; almost one third (31.5%) ofnonresident fathers live with new coresident chil-dren at Wave 2, and the majority live with onlyone additional child. The next column shows thatnearly one quarter (22%) of nonresident fatherscoreside with new biological children at Wave 2,and the final column indicates that one tenth arecoresiding with new stepchildren.

On average, about three quarters of nonresiden-tial fathers in our sample report paying at leastsome child support at Wave 1 and at Wave 2 (Table2). This percentage is somewhat higher than thatreported from other data sources, such as Surveyof Income and Program Participation (SIPP) (Sor-ensen, 1997) and PSID (Smock & Manning, 1997),but similar to other work using the NSFH (Meyer,1998; Seltzer & Brandreth, 1994). The meanmonthly payment of child support at Wave 1 was$235 (in 1994 dollars), and the median paymentwas $166 per month. The mean amount of childsupport reported paid at Wave 2 ($276) is slightlyhigher than at Wave 1. These numbers are alsoquite consistent with those based on mothers’ re-ports from the U.S. Census Bureau (1993). In 1991,mothers with a support order who received anychild support received on average $3,011 per year.A comparable sample from our data is based onnonresident fathers at Wave 1 who have legal sup-

port agreements and make some child support pay-ments; fathers in this group report paying on av-erage $3,066 per year (in 1991 dollars).

Table 2 also shows changes in payments bynonresident fathers across interviews. Changes inamounts paid can be the result of changing child-support obligations (e.g., changes in the numberof children owed support) or changing decisionsby fathers about how much they would like topay. The vast majority of fathers (85%) shiftedtheir monthly child support payments, with slight-ly greater percentages increasing than decreasingtheir child-support payments. Overall, the averagechanges in child support paid were positive, withan increase of $41 per month (results not shown).The average decline in payments was lower thanthe average gain in child-support payments be-tween the interviews. If we limit changes to whatmay be considered substantial (plus or minus $50per month), then we find that one quarter (27%)of nonresidential fathers paid at least $50 less permonth at Wave 2 than at Wave 1, and 38% ofnonresidential fathers paid at least $50 more atWave 2 than at Wave 1 (results not shown). Thefathers who did not alter their payments werethose who paid nothing at both interview waves.It is notable that a considerable share of fatherscontinued to support their children and often in-creased the level of payments. The increases ap-pear to stem largely from ‘‘other support,’’ whichmay in part reflect responses to increased financialneeds of children in their adolescent years. Ifchild-support payments alone are considered, theincreases are much more modest.

Table 3 displays descriptive statistics for thevariables used in our multivariate analyses. Thefirst few rows provide information about the coreindependent variables. The mean increase in cores-


TABLE 3. DISTRIBUTION OF INDEPENDENT VARIABLES

Variable Value

Number of new coresident children (mean)All childrenBiological children: birthsBiological children: move inStepchildren

0.520.290.060.19

Wave 1 number of coresident children (mean)All childrenBiological childrenStepchildren

0.540.300.25

Child support paid at Wave 1(mean, in 1994 dollars) $235.46

Father’s characteristicsAge (mean)Race (%)

32.5

NonwhiteWhite

27.972.1

Union Status (%)Not in unionCohabit/married

51.648.4

Education (%),12 years12 years13–15 years161 years

19.045.024.714.9

Earnings (mean, in 1994 dollars) $20,217

Ever lived with child (%)NoYes

28.171.9

Legal agreement (%)NoYes

27.472.6

Years since separation (mean) 4.5

Child’s characteristicsAge (mean) 6.6Gender (%)

FemaleMale

56.443.6

Distance from father (mean, inincrements of 100 miles) 5.51

Number of full siblings in focal child’sresidence, Wave 1 (mean) 0.6

Number of full siblings in focal child’sresidence, Wave 2 (mean) 0.2

Note: Weighted percentages and unweighted n’s (n 5133).


ident children is .52, and most of this is due to newbiological children. The mean increase of biologi-cal children who move into their fathers’ house-holds between the waves is quite small (.06). The

mean number of coresident children at Wave 1 is.54, and the mean values for biological childrenand for stepchildren are .30 and .25, respectively.

Fathers are on average 32 years old, and 72%of them are white. Almost half of them are livingin a union at Wave 1 (30% in marital and 18% incohabiting unions). The majority of fathers haveat least 12 years of education, with only about20% having less than 12 years of education. Meanearnings are $20,217 per year. Also, the majorityof fathers (72%) lived with the focal child at somepoint and report having a legal child supportagreement (73%). The mean number of yearssince separation from the child is 4.5 years. Theaverage amount of child support the fathers reportpaying per month at Wave 1 is $235.

Regarding the focal child, mean age at Wave1 is almost 7 years, and slightly over half of thefocal children are female. The focal child and fa-ther live on average 551 miles from one another,although the median is a much lower 20 miles.The mean number of additional children the fathersupports in the focal child’s household (i.e., fullsiblings of the focal child) decreases as siblingsage out of their status as minors.

Multivariate Results

Table 4 shows results from the static-score mod-els. The table presents two models: the first modeltests Hypothesis 1 (Model 1), and a second modeltests Hypothesis 2 (Model 2). The coefficients aretobit estimates of the effects of variables on thelogged monthly payments to nonresidential chil-dren made at the second wave of the survey. Theonly control variable that contributes to the fit ofour model is distance between residences of thenonresident father and his child.

The results for Model 1 indicate that new co-resident children generally do not affect changesin child-support payments. The coefficient mea-suring the number of new children is statisticallyinsignificant. Not surprisingly, even when we in-clude all of the control variables, the effects ofnumber of new children are about the same (re-sults not shown). As expected, the amount of childsupport paid at Wave 1 has a statistically signifi-cant, positive effect on child support paid at Wave2, but the number of coresident children at Wave1 and nonresident children at Wave 2 does notsignificantly influence levels of child support paidat Wave 2. Our findings lead us to reject Hypoth-esis 1: New resident children do not lower eco-nomic investments in nonresidential children.


TABLE 4. TOBIT ESTIMATES OF CHANGE IN RESIDENT CHILDREN ON MONTHLY CHILD SUPPORT PAID AT WAVE 2

Variable Model 1 Model 2

Number of new coresident childrenCoresident childrenBiological children: birthsBiological children: move inStepchildren

20.3420.89***

0.290.05

Wave 1 number of coresident childrenCoresident childrenBiological childrenStepchildren

20.2820.2620.25

Child support paid Wave 1Distance from father (in increments of 100 miles)

0.73***20.04**

0.73***20.03*

Number of siblingsWave 1Wave 2

20.160.64

20.190.67

Log likelihood 2263.1 2261.1

*p , .10. **p # .05. ***p # .01.Source: National Survey of Families and Households, Waves 1 and 2. n 5 133.

Model 2 distinguishes between biological chil-dren and stepchildren. The coefficient for the num-ber of additional biological children across thewaves is negative and statistically significant at the.01 level. Nonresident fathers who had new chil-dren between the interview waves paid significant-ly less child support than those who experiencedno change in the number of coresident biologicalchildren. The coefficient for the increase in step-children, on the other hand, does not approach sta-tistical significance. Introducing each control vari-able separately does not increase the fit of themodel; one important exception is distance betweenthe nonresident father’s and the child’s residence.Distance does statistically contribute to the fit ofthe model at the p 5 .10 level; fathers living fartheraway pay less support than fathers who live closer.In any case, whether we include each control vari-able separately or the full array of control variablestogether, our initial pattern of findings does notchange. Based on this evidence, we accept Hy-pothesis 2: The addition of children reduces childsupport to nonresidential children only when thenew children are biological children.

Finally, we find little evidence of observed sam-ple selectivity. First, the distributions of variablesare fairly similar across samples (data not shown).The final analytical sample has a significantlygreater percentage of female focal children, Whitenonresident fathers, and fathers with legal child-support agreements than the percentage of fatherswho were excluded from the sample. The sampleselection model results indicate no evidence of se-

lectivity, conditional on observed characteristics(Appendix Table 1). The rho representing the cor-relation between the error terms of the two equa-tions is statistically insignificant, and a model re-stricting the correlation to be zero fits the data aswell as the unrestricted model ( p 5 1.00). Also,the effects of the independent variables maintainsimilar levels of significance and magnitude.

DISCUSSION

In part as a response to changes in family patternsover recent years, the issue of fatherhood and itsmeaning has become a focus of much theoreticaland empirical attention from social scientists. Ourpaper addresses a question with critical implica-tions for these issues. We evaluate Frank Fursten-berg’s notion that fathers swap old children for newand whether social parenthood, defined by cores-idence, takes precedence over biological parent-hood. We focus on the economic dimension of al-legiances to nonresidential children, namely childsupport. We conclude that Furstenberg is right, withone caveat: Fathers do ‘‘swap’’ families, but onlywhen the trade-off is between new biological chil-dren living inside versus existing biological chil-dren living outside the fathers’ households. Fathersdo not appear to actually swap children, becauseon average, they continue to financially supportnonresident children. These results are inconsistentwith the notion that social parenthood, at least asdefined by coresidence, takes general precedenceover biological parenthood. Instead, biological par-


enthood, when combined with coresidence, appearsto take precedence. However, we must interpretthese results somewhat tentatively because of ourlimited sample size.

An important potential concern is that our find-ings could be an artifact of official modificationsin child-support orders due to fathers’ new familyresponsibilities. We do not believe that this is thecase. First, the majority of states have no guidelinesthat take into account ‘‘subsequent’’ families, andonly 13 states have mandatory guidelines of thissort (Takas, 1991). Second, there is wide variationin the application of the guidelines (Office of ChildSupport Enforcement [OCSE], 1996). Third, andmost important, empirical findings show that onlya small proportion of award modifications are infact due to changes in nonresidential fathers’ familysituations (OCSE, 1996).

Our contribution to policy dialogues about childsupport is that fathers appear to be adjusting theirchild support payments, probably often informally,to accommodate the demands of new biologicalchildren. Indeed, fathers themselves report beingconcerned about whether and how their new fam-ilies can influence child-support awards (OCSE,1996). If a primary goal of child-support policy isto ensure compliance, our findings would suggestthat policies need to explicitly resolve the issue ofaward modification when fathers have new chil-dren. Of course, this issue is a difficult one to re-solve, because it engages the question of whichchildren should take precedence (e.g., Takas, 1991).

There are several limitations of our work. Anobvious one is that our sample is quite small. De-spite this shortcoming, our findings are robust toalternative specifications of the model, increasingour confidence in the findings. Nonetheless, ourstudy ought to be replicated on larger sampleswhen appropriate data are available, allowing forevaluations of differences in fathering for racial-ethnic and socioeconomic groups. Second, ourdata do not allow us to assess nonresident fathers’motivations for adjusting their child-support pay-ments. Thus, for example, we cannot discernwhether fathers feel an erosion in their commit-ment to their nonresident children with the arrivalof a new child, whether they are responding to thefinancial constraints they face when they beginsupporting another family, or whether they are re-acting to some other set of factors. It is critical tobuild upon our analyses and existing qualitativework (e.g., Arendell, 1995) by empirically eval-uating the reasons for fathers’ decisions to modifytheir financial support to nonresidential children.

Third, we lack information about complex familysituations. Our data only allow us to examinechild-support payments to a focal child and his orher coresident siblings; some fathers have morethan one set of biological children living outsidetheir households. Fourth, other changes in fathers’and mothers’ circumstance may create shifts inchild-support payments. Our focus here is on test-ing a specific hypothesis about new children. Amore rich data source should be employed to ex-amine the influence of other factors.

Finally, our data very likely underreport non-residential fatherhood. An inherent problem in thisproject is the selectivity based on unobservedcharacteristics of nonresident fathers. Currently,however, no national data set can make the claimof having a representative sample of nonresiden-tial fathers. Although innovative efforts have beenmade to make adjustments via indirect estimationmethods (e.g., Garfinkel, McLanahan, & Hanson,1998; Sorensen, 1997), these still are not ideal,and better sampling and data on fathers have be-come top research priorities (National Institute forChild and Health Development, 1996).

Nonetheless, our work is unique in being, toour knowledge, the first to empirically evaluatethe effects of resident children, whether step orbiological, on economic investments in nonresi-dential children. Despite the small sample size, webelieve three broad implications of our study de-serve special emphasis. First, future efforts aimedat understanding the nature of nonresident father–child ties, or fatherhood more generally, ought towiden their scope to incorporate both resident andnonresident offspring. Second, fathers in this sam-ple often maintain financial ties to their childrenand in some cases increase their investments innonresident children, further efforts should be em-ployed to identify the mechanisms under whichfathers maintain economic investments in theirnonresident children. The third implication emerg-es from our finding that fathers’ financial supportfor nonresident children appears to be conditionalon their future fertility. Ultimately, this is a patternthat probably does not bode well for children’swell-being in the United States, particularly givena societal context of increasing nonmarital fertilityand union instability. Researchers should continueto pursue work on fathering as efforts to encour-age father involvement are continually emergingand may result in new fathering dynamics.

NOTEAn earlier version of this paper was presented at theannual meeting of the Population Association of Amer-


ica, March 1997 in Washington, DC. We thank partici-pants at the Family Studies Seminar at the Survey Re-search Center of the University of Michigan, Yu Xie,Judith A. Seltzer, Alfred DeMaris, and Daniel D. Wieg-mann for constructive suggestions. We are also gratefulto Susan Stewart for excellent research assistance. Com-puter services were provided by Bowling Green StateUniversity. This research was supported by grantsHD34391 and HD34392 from the National Institute ofChild and Health Development.

REFERENCES

Adritti, J., & Kelly, M. (1994). Fathers perspectives oftheir coparental relationships postdivorce. Family Re-lations, 43, 61–67.

Ahrons, C. (1983). Predictors of paternal involvementpost-divorce: Mothers’ and fathers’ perceptions.Journal of Divorce, 6, 55–69.

Arendell, T. (1995). Fathers and divorce. ThousandOaks, CA: Sage.

Becker, G. S. (1981). A treatise on the family. Cam-bridge, MA: Harvard University Press.

Braver, S., Fitzpatrick, P., & Bay, C. (1991). Noncus-todial parent’s report of child support payments. Fam-ily Relations, 40, 180–185.

Bumpass, L. L., & Sweet, J. A. (1989a). Children’s ex-periences in single-parent families: Implications ofcohabitation and marital transitions. Family PlanningPerspectives, 21, 256–260.

Bumpass, L., & Sweet, J. (1989b). National estimatesof cohabitation. Demography, 26, 615–626.

Cherlin, A. J. (1978). Remarriage as an incomplete in-stitution. American Journal of Sociology, 84, 634–650.

Cherlin, A. J., Griffith, J., & McCarthy, J. (1983). Anote on maritally-disrupted men’s reports of childsupport in the June 1980 Current Population Survey.Demography, 20, 385–389.

Clingempeel, W. G., Coylar, J. J., & Hetherington, E.M. (1994). Toward a cognitive dissonance concep-tualization of stepchildren and biological childrenloyalty conflicts. In K. Pasley & M. Ihinger-Tallman(Eds.), Stepparenting: Issues in theory, research, andpractice. Westport, CT: Greenwood Press.

Finkel, S. (1995). Causal analysis with panel data.Newbury Park: Sage Publications.

Furstenberg, F. F. (1988). Marital disruptions, child cus-tody, and visitation. In A. Kahn & S. Kammerman(Eds.), Child support (pp. 277–305). Beverly Hills,CA: Sage.

Furstenberg, F. F. (1995). Changing roles of fathers. InP. L. Chase-Lansdale & J. Brooks-Gunn (Eds.), Es-cape from poverty: What makes a difference for chil-dren? (pp. 189–210). Cambridge: Cambridge Uni-versity Press.

Furstenberg, F. F., Nord, C. W., Peterson, J., & Zill, N.(1983). The life course of children of divorce: Maritaldisruption and parental contact. American Sociolog-ical Review, 48, 656–668.

Furstenberg, F. F., & Spanier, G. B. (1984). Recyclingthe family: Remarriage after divorce. Beverly Hills,CA: Sage.

Furstenberg, F. F., & Nord, C. W. (1985). Parentingapart: Patterns of childrearing after marital disruption.Journal of Marriage and the Family, 47, 893–904.

Furstenberg, F. F., & Cherlin, A. (1991). Divided fami-lies: What happens to children when parents part.Cambridge, MA: Harvard University Press.

Furstenberg, F. F., & Harris, K. (1992). The disappearingAmerican father? Divorce and the waning signifi-cance of biological fatherhood. In S. South & S. Tol-nay (Eds.), The Changing American Family (pp. 197–223). Boulder, CO: Westview Press.

Garfinkel, I., McLanahan, S. S., & Hanson, T. L. (1998).A patchwork portrait of nonresident fathers. In I. Gar-finkel, S. McLanahan, D. Meyer, & J. Seltzer (Eds.),Fathers under fire (p. 31–60). New York: RussellSage Foundation.

Heckman, J. (1979). Sample selection bias as a speci-fication error. Econometrica, 47, 153–161.

Hill, M. (1992). The role of economic resources andremarriage in financial assistance for children of di-vorce. Journal of Family Issues, 13, 158–178.

Ihinger-Tallman, M., & Pasley, K. (1987). Remarriage.Newbury Park, CA: Sage.

Kessler, R., & Greenberg, D. (1981). Linear panel anal-ysis. New York: Academic Press.

King, V. (1994). Nonresident father involvement andchild well-being: Can dads make a difference? Jour-nal of Family Issues, 15, 78–96.

Knox, V. W., & Bane, M. J. (1994). Child support andschooling. In I. Garfinkel, S. McLanahan, & P. Robins(Eds.), Child support and child well-being (pp. 285–316). Washington, DC: Urban Institute.

Lin, I. F., & Schaeffer, N. C. (1995). Using survey par-ticipants to estimate the impact of nonparticipation.Public Opinion Quarterly, 59, 236–258.

McLanahan, S. S., Seltzer, J., Hanson, T., & Thomson,E. (1994). Child support enforcement and child well-being: Greater security or greater conflict? In I. Gar-finkel, S. McLanahan, & P. Robins (Eds.), Child sup-port and child well-being (pp. 239–256). Washington,DC: Urban Institute.

Marsiglio, W. (1992). Stepfathers with minor childrenliving at home. Journal of Family Issues, 13, 195–214.

Meyer, D. (1998). The effect of child support on theeconomic status of nonresident fathers. In I. Garfin-kel, S. McLanahan, D. Meyer, & J. Seltzer (Eds.),Fathers under fire (p. 67–93). New York: RussellSage Foundation.

Mott, F. (1990). When is father really gone? Paternal–child contract in father-absent homes. Demography,27, 499–517.

National Institute for Child and Health Development.(1996, October). Conference on Father Involvement,Bethesda, MD.

Norton, A., & Miller, L. (1992). Marriage, divorce, andremarriage in the 1990’s. Washington, DC: U.S. Bu-reau of the Census.

Office of Child Support Enforcement. (1996). Evalua-tion of child support guidelines. Washington, DC.

Paasch, K. M., & Teachman, J. D. (1993). Gender ofchildren and receipt of assistance from absent parents.Journal of Family Issues, 12, 450–466.

Peters, H. E., Argys, L. M., Maccoby, E. E. & Mnookin,R. H. (1993). Enforcing divorce settlements: Evi-dence from child support compliance and award mod-ifications. Demography, 30, 719–735.

Schaeffer, N., Seltzer, J., & Kalwitter, M. (1991). Esti-mating nonresponse and response bias: Resident and


APPENDIX

PROBIT AND TOBIT EFFECTS FOR SAMPLE SELECTION AND MONTHLY AMOUNT OF CHILD SUPPORT PAID

Variable

Final AnalyticalSample,

Probit Coefficients

Log Monthly ChildSupport,

Tobit Coefficients

Family background intact 20.02 —

Mother’s education,12 years

12 years.12 years

20.08—0.39*

———

Employed full-time 0.40** —

Number of new coresident childrenBiological children: birthsBiological children: move inStepchildren

———

20.89**0.300.05

Wave 1 number of coresident childrenBiological childrenStepchildren

——

20.2620.26

Child support paid Wave 1Distance (in increments of 100 miles)Wave 1 number of siblingsWave 2 number of siblings

————

0.74**20.03**20.19

0.67

rhonLog likelihood

2980.09

2461.1133

*p # .05. **p # .01.Source: National Survey of Families and Households, Waves 1 and 2.

nonresident parents’ reports about child support. So-ciological Methods and Research, 20, 30–59.

Seltzer, J. A. (1991). Relationships between fathers andchildren who live apart: The father’s role after sepa-ration. Journal of Marriage and the Family, 53, 79–101.

Seltzer, J. A. (1994). Consequences of marital dissolu-tion for children. Annual Review of Sociology, 20,235–266.

Seltzer, J. A., & Bianchi, S. M. (1988). Children’s con-tact with absent parents. Journal of Marriage and theFamily, 50, 663–677.

Seltzer, J. A., & Brandreth, Y. (1994). What fathers sayabout involvement with children after separation.Journal of Family Issues, 15, 49–77.

Seltzer, J. A., McLanahan, S., & Hanson, T. L. (1995).Will child support enforcement increase father–childcontact and parental conflict after separation? In I.Garfinkel, S. McLanahan, D. Meyer, & J. Seltzer(Eds.), Fathers under fire (pp. 157–190). New York:Russell Sage Foundation.

Smock, P., & Manning, W. (1997). Nonresidential par-ents’ economic ties to children: New evidence fromthe PSID. Journal of Marriage and the Family, 59,798–808.

Sonenstein, F., & Calhoun, C. 1990 Determinants ofchild support: A pilot survey of absent parents. Con-temporary Policy Issues, 8, 75–94.

Sorensen, E. (1997). A national profile of nonresident

fathers and their ability to pay child support. Journalof Marriage and the Family, 59, 585–191.

Sweet, J. A., Bumpass, L., & Call, V. (1988). The designand content of the National Survey of Families andHouseholds. NSFH Working Paper 1. Madison, WI.Center for Demography and Ecology, University ofWisconsin—Madison.

Takas, M. (1991). The treatment of multiple family casesunder state child support guidelines. Washington,DC: U.S. Department of Health and Human Services.

Teachman, J. D. (1991). Who pays? Receipt of childsupport in the United States. Journal of Marriage andthe Family, 53, 759–772.

Trivers, R. L. (1974). Parent–offspring conflict. Ameri-can Zoologist, 11, 249–264.

U.S. Bureau of the Census. (1993). Child support forcustodial mothers and fathers: 1991. Current Popu-lation Reports, P-60, No. 186. Washington DC: U.S.Government Printing Office.

Veum, J. R. (1993). The relationship between child sup-port and visitation: Evidence from longitudinal data.Social Science Research, 22, 229–244.

Weiss, Y., & Willis, R. J. (1985). Children as collectivegoods and divorce settlements. Journal of Labor Eco-nomics, 3, 268–292.

Zill, N. (1996). Noncustodial parents’ participation intheir children’s lives: Evidence from the survey ofincome and program participation. Washington DC:U.S. Department of Health and Human Services.

“swapping” families: serial parenting and economic support for children

Documents