I. Introduction           

Fatherhood is now a well established field of scholarly study (Goldberg, Tan et al. 2009).  Two outgrowths of gender role change have received the bulk of scrutiny by researchers interested in fatherhood (Edwards, Doucet et al. 2009).  The first is the increase in the amount of time relatively affluent children in two parent families spend with their fathers (Bianchi 2000). The second trend is the increase in the number of relatively poor children who live apart from their fathers (Garfinkel, McLanahan et al. 1998).  These two trends, taken together, have given rise to concern that change in paternal behavior is one important reason for increasing disparities among children in the quality of the investment they receive from their parents (McLanahan 2004)

 In this paper we focus on a very specific group of fathers with contested paternity.  These men have been the targets of public policies concerned with improving the support for children of unmarried parents. Seventy-five percent of the fathers of the 1.5 million children born to parents who are not married voluntarily agree to sign affidavits of paternity, nearly 900,000 of which are signed in the hospital at the time of delivery (Mincy, Garfinkel et al. 2005).  That leaves about 375,000 children whose fathers did not acknowledge them at the time of their birth. Among these, there are roughly 340,000 who subsequently undergo genetic testing to determine paternity (American Association of Blood Banks 2002). 

A large motivation for public policies that mandate the identification of putative fathers and compulsory genetic testing is to save government outlays for cash assistance to children.  Therefore, one major criteria for determining the success or failure of such programs is the amount of such savings (Astone 1997). In fact, paternity establishment has proven to be beneficial to state budgets by offsetting the need to publicly finance welfare payments to single mothers after child support orders are obtained (Farrel, Glosser et al. 2003).

Nevertheless, there are sound theoretical reasons from both sociology and economics to expect that undergoing such a test and receiving the results might result in a change in the putative fathers' behavior toward their children.  Discovering whether or not policies do change fathers' behavior can illuminate some issues in the scholarship of fathering, as well as provide information to child advocates who are interested in promoting public policy that fosters child development.   Evidence that paternity establishment is beneficial to children is quite limited, however.

State level child-support agencies in the US are responsible for most of the 340,000 paternity tests conducted annually in the US at a cost of $34 million.  There is a pressing need to learn more about the lives of the children whose unmarried parents have paternity testing.  The hypothesis that paternity testing is beneficial to the children involved as well the state's budget, remains untested.  Although examination of paternity tests exclude the man's paternity in 28% of cases, but there have been few studies to illuminate the pre-test behaviors of men who were excluded compared to those who were not (Bishai, Astone et al. 2006). 

The goal of this paper is to assess the scope for genetic paternity tests to improve the well-being of children by reporting on a survey of 245 respondents who were undergoing genetic testing in child-support agencies in Maryland in 2004.  This study attempts to measure how well the men and women in these cases could predict the test results based on information available to them.  If the test results are not entirely anticipated by the participants before the test, which we show below is a plausible hypothesis, then paternity test results are less likely to have large impacts on behavior.  The study also documents the degree of contact between the men and the child being tested prior to a paternity test.  If contact is already high before the test, especially among men or the parents of men who are later confirmed as fathers then one would expect the test to have little positive impact on the men's behavior.  If contact prior to the test is substantial among men who are later proven not to be the biological fathers then the tests could be harmful by potentially severing relationships between children of unmarried women and the men who are serving in the role of father.

II. Background

 

(top of page)

US Policies on Paternity Establishment and their Implementation

 

(top of page)

            The role of state governments in paternity establishment has expanded as part of state intervention to enforce child support.  In 1975, Title IV-D of the Social Security Act required each state to establish offices of child-support enforcement (OCSEs). Nine years later the Child Support Enforcement Amendments of 1984 required all states to initiate wage withholding and to extend the age for paternity establishment to age 18.  The greatest growth in the use of genetic tests for paternity establishment followed the 1988 Family Support Act which required all states to establish paternity in 50% of out of wedlock cases or to increase paternity establishment by 3% annually for cases on AFDC and to offer genetic tests at the request of any party when paternity is contests. The 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) required states to have a voluntary paternity acknowledgement process and set goals for states to achieve 90% paternity establishment (Miller and Garfinkel 1999).  PRWORA allowed each state to set its own policy on the pass through of child-support funds from father to mother (Farrel, Glosser et al. 2003). 

            The primary rationale for policies to improve paternity establishment was that paternity establishment is a necessary condition for the establishment of a legal child-support order for a child born to an unmarried woman.  States have a financial interest in obtaining child-support orders for TANF beneficiaries because these child-support payments remit directly to the state except for 19 states where an average of $50 per month of the father's child-support payment is passed through to the custodial parent (Farrel, Glosser et al. 2003). 

Because child-support payments offset states' TANF outlays there is a powerful fiscal incentive for states to pursue measures like better paternity establishment that can increase both the number of child-support orders and compliance.  Some worry that state policy makers' focus on achieving fiscal relief via better child-support enforcement has led them to overlook both additional benefits as well as potential harm resulting from both paternity establishment and child-support enforcement. Overlooked benefits to children from paternity establishment include inheritance and eligibility for dependent benefits from Social Security in the event of the fathers' death (Wattenberg 1987; Nichols-Casebolt 1988).  There may also be cognitive and psychological benefits to children from both child-support payments and indirectly if child-support payments increase contact (Argys, Peters et al. 1998).  Of course, there may be negative consequences of paternity establishment as well.  These include the possibility that paternity establishment can bring women into greater contact with men who subject them to domestic violence, and who could be more easily avoided in the absence of paternity establishment (Fertig, McLanahan et al. 2003). 

            There is also concern about the potential for genetic paternity tests to disrupt emotional and care-giving bonds between a man and his children in cases where a man who believes he is the biological father is later confronted with laboratory evidence to the contrary (Bellis, Hughes et al. 2005).  Because, to date, nothing is known about the frequency of contact between men and the children who require genetic paternity tests it is difficult to evaluate this potential harm of paternity disruption. 

            There are a variety of reasons for a man to prefer not to have a child-support order, and one need not presume a lack of paternal concern for the child's well-being.  Aside from the bureaucratic hassles of a child-support order and possible sanctions for defaulting, a child-support order for a child on TANF does little to improve the financial resources of the child or mother.  Under current state policies no more than $50 of a monthly child-support payment to a TANF beneficiary is ever passed through from the father to the mother.  Parents may prefer to arrange a stream of informal payments to the mother.  For example, suppose a father gave a mother on TANF $75 a month.  If that mother lived in a state with no pass through, this would be $75 more dollars than she would receive under a child-support order, and in states with $50 pass through, it would be $25 more.  Compared to a $100 child-support order, a father who reliably makes monthly $75 informal payments to a mother on TANF offers a situation that is better for mother, child, and father.  Many mothers seek a child support order only as a last resort after informal payments from the father become unreliable.  

Federal regulations, however, reduce the TANF benefits of an unmarried mother by 25% unless she cooperates with the state in identifying the father of her child.  It is easy to see how, under certain circumstances, a woman may name someone she knows NOT to be the father, in order to satisfy the demands of the regulations, while maintaining the informal arrangement which she perceives as best for her and her child, and most likely to keep the money from the father coming. Qualitative data suggest that unmarried fathers and mothers in poverty often do consider the strategic implications of paternity establishment to avoid the inflexibility of the formal child-support system (Edin 1995). 

 Prior Studies of Paternity Establishment and Child-support Enforcement

 

(top of page)

            Current knowledge about the impact on families of paternity establishment and child support is drawn primarily from cases where men voluntarily acknowledge paternity.  These men are probably quite different, in unobserved ways, from those who have to be coerced. There is substantial evidence that child-support payments are more likely when paternity is established  (Miller and Garfinkel 1999; Argys, Peters et al. 2001; Freeman and Waldfogel 2001).  The existence of the child-support order and paternity confirmation is associated with greater contact between child and father (Seltzer, McLanahan et al. 1998; Argys and Peters 2001 ).  Child-support awards are also associated with increased influence of fathers over children (Seltzer, McLanahan & Hanson, 1998).  There is evidence that child educational outcomes (Graham, Beller et al. 1994)  and scores on cognitive tests (King 1994; Argys, Peters et al. 1998) are associated with child-support awards. Additional dollars of child support, however, have been found to have stronger effects on child outcomes than additional dollars of mother's earnings of family income (Aughinbaugh, 2001; Argys, Peters, Brooks-Gunn & Smith, 1998).  This suggests either positive selection into child-support payment or that higher child achievement actually triggers higher child-support payments by strategically inclined fathers (Aughinbaugh, 2001).  This is also consistent with the idea that mothers cooperatively agree to spend child-support payments disproportionately on children to encourage future payments (Argys and Peters 2003).

             Because existing literature focuses on general samples of children born to unmarried parents, the effects of paternity establishment are typically confounded with the unobservable factors that have led men to voluntarily acknowledge paternity.  Even in studies where instrumental variable methods are used to infer the causal effects of paternity establishment, these effects apply to the typical or "marginal" man whose paternity is established.    Because so many of the cases of paternity establishment in existing studies occurred voluntarily, it is not valid to extrapolate these observations to the minority of men who demanded a paternity test or to those for whom paternity has not been established.  There have been no prior studies of families in which men do not voluntarily acknowledge paternity, so this study will contribute by offering a better profile of how these families differ from other unmarried families.  Because this study interviewed child-support clients on the day of their paternity test in four Maryland counties and is linked to the subsequent results of the test it can shed light on the extent to which men and women correctly anticipate what the genetic tests will show.

Conceptual Framework and Expected Results of Covariates

 

(top of page)

            The conceptual framework for this study sets in opposition contrasting explanations for why men might fail to voluntarily acknowledge paternity and how this would alter the effects of paternity testing.  Deciding which explanation is correct will have important policy implications because in one case genetic testing helps men do the right thing, in the other case genetic testing wastes resources and interferes needlessly.

            One state of affairs where paternity testing would be of value would be a situation of genuine paternity uncertainty.  This would imply a non-marital union in which the mother had multiple sexual partnerships prior to a child's birth and neither she nor the potential father/s can be certain of paternity.  Men's willingness to acknowledge paternity is simply because of genuine doubt and not because of disinterest.   Resolution of the doubt is both necessary and sufficient to convert men like this into supportive fathers.  The assumption is that once men are certain of their paternity, they have every reason to be just as solicitous of a child's welfare as any other blood relative.  

            For genuinely uncertain fathers, confirmation of paternity may lead to more positive fathering by signaling them to take this role.  Forste and her colleagues point first to the diversity of role expectations, behavioral schemas and cultural constructions of fatherhood that have emerged over the past decades and second to variation among men in their own relationships with their fathers (Forste, Bartkowski et al. 2009).  Given this complexity, they argue that men serve as agents in negotiating fathering in their own lives.  Under complex circumstances, signal life events can serve to galvanize men to organize their own fathering behavior according to a certain individual pattern.  If this is so, one can imagine that, even among men who were fairly certain that they were the biological fathers of a child, confirming paternity might be such a signal event that it galvanizes them to act more positively towards their children.

            Another reason to hypothesize that confirmation of paternity might result in more positive fathering concerns its potential effects on the relationship between the mother and father.  For fifteen years co-parenting has been a central construct in studies of child development (Cowan and McHale 1996). Early researchers focused on co-parenting in intact families (Belsky and Crnic 1995).  Soon scholars moved on to study co-parenting in divorced families (Amato 2005). Life events like subsequent marriages and births have marked effects on contact rates between unmarried men and their children (Tach, Mincy et al. 2010).  Other studies of co-parenting among couples who have experienced extreme challenges such as abuse and incarceration illustrate the importance of the couples relationship over time (Roy and Dyson 2005; Hardesty and Ganong 2006; Schoppe-Sullivan, Brown et al. 2008). It is possible that confirmation of paternity has a salutatory effect on the relationship between the mother and father such that co-parenting is facilitated.  At a minimum, it might eliminate conflict about the fact of paternity itself. 

            Another model might be named the "selective shirking" model.  Proponents of this view argue that there are two types of men: those with and those without an interest in the role of fatherhood.  Those who are interested in fatherhood know that this role is enhanced by a strong bond with the mother that includes co-residence (Willis 1999). Men with a sincere interest in fatherhood are unlikely to father children without co-residence and a stable relationship.   A stable relationship and co-residence with mother and children offer a man better control over the amount of time he can spend with the children how his money is spent on the children.  If this theory is true, then the children of women who do not live with the fathers of the children (virtually 100% of those who present for mandatory paternity tests) are far more likely to be the offspring of a subset of men who have little or no wish to invest in their children.

            The selective shirking hypothesis we just elaborated was framed in the language of unobserved differences in people often employed by economists.  One might base a similar hypothesis on the research on fathering by Doucet, a sociologist who uses ethnographic methods and a different theoretical frame (Doucet 2006; Doucet 2009; Doucet 2009).  She argues, based on extensive ethnography with men who are working exclusively as full-time caregivers for their children, that providing for the physical and emotional needs of children is still extremely stigmatized behavior for men.  She finds that those men who have in the past held well-paid or high prestige jobs can surmount this stigma, but it is extremely difficult for economically unsuccessful men, to do so. These men face the "double jeopardy" of being judged both failed (because they do not fulfill the traditional male role of breadwinner) and deviant (because they are doing women's work (Doucet 2006).  These findings suggest that men whose breadwinner role must be coerced are unlikely to be accompanied by high levels of caretaking, before or after confirmation of paternity.

            Existing studies cannot measure the extent to which unmarried fathers-especially those who have not acknowledged paternity--can be classified as having genuine paternity uncertainty vs being selective shirkers.  Deciding the extent to which each explanation prevails is important for policy purposes because in a situation of genuine paternity uncertainty imposing a child-support order based on paternity testing has the potential to improve paternal care behavior. Under the "selective shirking" hypothesis, men who are not living with their children and the mothers of their children are really not the sort of men who are cut out for fatherhood.  Selective shirkers are more likely to fail to make child support payments and to spend time with children even after paternity tests.             For men in the selective shirking model all pregnancies are unintentional and unwanted, but there is not necessarily any uncertainty about whether they are fathers.  These men will have virtually zero contact with their children and the paternity test is unlikely to alter this.  In contrast men with genuine paternity uncertainty will be unable to accurately predict what their paternity test.  Their subjective belief in their paternity prior to a test would be intermediate between 0 and 100 and thus their rates of contact with children would be reduced relative to men who were certain of paternity. Contact rates for uncertain men would not be zero.

Deciding whether data support either of these two explanations requires one to observe how much never married non-cohabiting fathers are voluntarily investing in their children.  Certainly the observation that 69% of unmarried fathers in the Fragile Families study voluntarily acknowledge paternity around the time of a birth suggests that most unmarried fathers are not intending to shirk responsibility (Mincy, Garfinkel et al. 2005).  But one cannot extrapolate from the Fragile Families sample to men who do not voluntarily acknowledge paternity?  One requires observations of paternal behavior among this selected group of men.

 

 Figure 1Enlarge this image

Figure 1.  Conceptual framework.  The selective shirking model predicts that men in group B1 will be those who wish to have out of wedlock children with little concern for child well-being and little desire to transfer resources to them unless compelled.  The selective shirking model predicts there will be no difference between group B1 and group B2 in terms of relationship quality and paternal behavior prior to paternity establishment. Men in both B1 and B2 will have low levels of child contact and equally poor relationships with the child's mother.             The "genuine uncertainty" model predicts that men in group B will have high rates of child contact. And men in group B1 will have more child contact and better relationships with the mother than men in group B2.  In this model, the reason men in group B1 are not acknowledging paternity is valid uncertainty.

 

Figure 1 lays out the conceptual framework and predictions in more detail and helps to focus our analysis on two sets of issues: 1) comparing the backgrounds of men who are confirmed vs. excluded as biological fathers by genetic testing; and 2) identifying the determinants of father-child contact among men who did not voluntarily acknowledge paternity.  Our conceptual framework suggests the following hypotheses:

Hypothesis 1)  Among men who have not voluntarily acknowledged paternity, there will be systematic differences between men whose genetic paternity is later confirmed and those whose genetic paternity is excluded by laboratory testing.

Hypothesis 2) Men who are later confirmed as genetic fathers will have no contact with their children prior to the test.

            If we accept hypothesis 1 and 2 it would be consistent with the selective shirking explanation of non-marital paternity.  Hypothesis 1 is based on the presumption that with there would be systematic differences would appear that would differentiate shirkers from men interested in fatherhood.  Differences in educational attainment, income, and the relationships with women would be expected. Men with less income and education would be more likely to fall into the predicament of non-marital fatherhood and to shirk their responsibility by not voluntarily acknowledging paternity.  Hypothesis 2 is implied only by the selective shirking theory's stipulation that men who father births outside of marriage are inherently disinterested in contact with their children.  Thus the biological fathers, as confirmed by paternity, tests will have no contact with their children. 

            In contrast, with genuine paternity uncertainty, socioeconomic background variables will not be correlated with the outcomes of a paternity test.  Furthermore, until the test results are back, men with genuine uncertainty of paternity would form a prior belief about the test that was intermediate between 0 and 100.  On the basis of this belief, there could be some men who have contact with the children prior to the test results coming back.

  The conceptual framework helps to guide selection of variables.

Which variables might predict whether paternity is confirmed genetically?

From the conceptual framework the genuine uncertainty theory predicts significantly higher rates of paternity establishment among men with three categories of features 1) a stronger relationship to the mother;  2) an indicated desire to be involved with the child; 3) an expression of bonding to the child.  Plausible control variables would include the child's age and sex and the man's household income, all three of which are known to affect fathering (Carlson, McLanahan et al. 2005; Cheadle, Amato et al. 2010).  

Which variables might confound tests for correlation between paternity and father-child contact?

If the selective shirking explanation is correct, we expect to observe negligible rates of child contact among men who have not voluntarily acknowledged paternity.  Indeed under the selective shirking explanation men whose denial of paternity leads to a genetic test ought to have low rates of child contact whether or not they are later confirmed as biological fathers. Conducting an analysis to demonstrate that biological paternity had no effect on child contact would have to control for confounding factors that could plausibly correlate with both biological fatherhood and child contact.  There is limited literature identifying variables that correlate with confirmation of biological paternity. 

From prior literature on unmarried fathers, we expect four groups of variables to  be potentially correlated to father-child contact (Carlson, McLanahan et al. 2005).  These are relationship quality, attitudinal characteristics,  child's characteristics, and father's human capital.  To this list we add 4 other plausible determinants of contact: longer distance from the father's house, and indicators of father-child bonding including whether the man was present at the birth, says the child looks like him, and whether the child has the man's last name. 

Method

Data

 

(top of page)

We obtained permission from the Maryland Office of Child Support Enforcement (OCSE) to enroll research volunteers from OCSE facilities in Ann Arundel, Baltimore, Montgomery, and Prince George's County from October 24, 2003 until August 12, 2004.  Legal guardians, mothers, and men were all eligible if older than 18 regardless of the participation status of the other party in their triad.  Legal guardians can request child support in Maryland and were considered to have valid information on the amount of contact the children were having with the men being tested.  All interviews were conducted on the day of the genetic test on the premises of the OCSE facilities.  Test results were not known at the time of the interview.  During the enrollment period, a total of 412 eligible individuals attended the four sites on days when our research staff was present.  Our staff were able to invite 395 (96%) of these to participate in the study.  The 30 minute survey was self-administered in English while participants were in the waiting room of the OCSE facilities.  There were 17 who could not be offered enrollment prior to their departure because research staff was occupied with multiple ongoing interviews.  Of the 395 invited participants, 250 (63%) agreed to participate and 245 (110 men and 135 women) completed the survey. There were 57 men and 57 women who participated in the survey and whose data could link them as couples, which implies that of the 245 responses there were only 188 in which only one member of a couple was surveyed and 57 in which both were surveyed. 

  Enrollment rates were 71% in Ann Arundel County, 70% in Baltimore County, 31% in Montgomery County, and 55% in Prince George's County.  Enrollment rates by race/ethnicity were 63% among blacks, 28% of Hispanics, and 69% of whites. Low Hispanic enrollment may have been due to the requirement for English fluency. We also suspect, but have no confirmation, that some of the eligible Hispanic candidates who did not participate may have had misplaced concern that research participation would expose them to undesired attention by immigration authorities. Women (76%) were more likely to enroll than men (50%).  This study was approved by the Johns Hopkins Committee on Human Research.

Measurement

 

(top of page)

Men received slightly different surveys than mothers and guardians. The different phrasing was so that mothers could answer about father's parenting behavior in third person and fathers could answer in first person.  For father-child contact we asked, "During the last 30 days how many days have you [has the father] seen the child/ren?".  For relationship quality we used as indicators: the number of years the man had known the partner, whether they were ever married, categorical indicators for why the relationship faltered (distance, drug problem, relationship reasons, money/other), and relationship intensity coded on a 1 through 5 scale based on a question that read: . 

"Which of the following statements best describes your current relationship with the child (ren)' s mother? 1) We are romantically involved on a steady basis 2) We are involved in an on-again off-again relationship 3) We are just friends  4)We hardly ever talk to each other 5) We never talk to each other"  

              Indicators of parental involvement were statements in response to two questions, "Do you want to be involved in raising your child (ren) in the coming years?" with categorical responses "Yes", "No" or "Don't Know". and to a similar question "Does the child (ren)'s mother want you to be involved in raising your child in the coming years?"  also coded categorically "Yes", "No" or "Don't Know".  Finally there was a dichotomous question, "During the child (ren)'s mother's pregnancy, did you give her money to buy things?"  We counted dichotomous responses for whether the man was present at the child's birth, whether the child looked like the man, and whether the child had the man's last name as expressions of bonding.

            Table 1 presents the mean values of variables used in the analysis for the sample of men and the sample of women.

 

(top of page)

Table 1.  Sample Characteristics:  Percent Distributions.
	All Men (N=110)			All Women (N=135) [1]
	M	SD	Range	M	SD	Range
Paternity confirmed by genetic test	0.56	0.02	0-1	0.90	0.01	0-1
Putative father saw focal child in last 30 days	0.44	0.02	0-1	0.39	0.02	0-1
Putative paternal grandparents saw focal child in last 30 days	0.32	0.02	0-1	0.29	0.02	0-1
Focal child's age
Younger than 1 year	0.32	0.02	0-1	0.23	0.02	0-1
1 to 3 years	0.32	0.02	0-1	0.41	0.02	0-1
4 or older	0.36	0.02	0-1	0.36	0.02	0-1
Mean father's age (SD)	30.8	9.90	18-65	31.2	8.90	18-61
Mother has custody of child	0.87	0.01	0-1	0.91	0.01	0-1
Perceived odds of a positive match
50-50 or less	0.69	0.02	0-1	0.15	0.01	0-1
A pretty good chance	0.16	0.01	0-1	0.14	0.01	0-1
An almost certain chance	0.15	0.01	0-1	0.71	0.02	0-1
African American	0.67	0.02	0-1	0.62	0.02	0-1
Household income
Less than ,000	0.17	0.01	0-1	0.25	0.02	0-1
,000 to 29,999	0.29	0.02	0-1	0.22	0.01	0-1
, 000 or more	0.32	0.02	0-1	0.29	0.02	0-1
Do not Know	0.22	0.02	0-1	0.24	0.02	0-1
Education
High school diploma or more	0.89	0.01	0-1	0.82	0.01	0-1
Less than high school	0.11	0.01	0-1	0.18	0.01	0-1
Does putative father do anything to help mother?
Nothing	0.53	0.02	0-1	0.77	0.02	0-1
Something	0.47	0.02	0-1	0.23	0.02	0-1
Number of children
None	0.2	0.02	0-1	0	0.00	0-1
One	0.32	0.02	0-1	0.34	0.02	0-1
Two	0.25	0.02	0-1	0.31	0.02	0-1
Three or more	0.23	0.02	0-1	0.35	0.02	0-1
Mean mother's  age (SD)				27.4	6.50	18-44
Only one sexual partner the month focal child was conceived				0.58	0.02	0-1
Relationship quality
Romantically involved and steady	0.8	0.02	0-1
On again, off again	0.12	0.01	0-1
Just friends	0.18	0.01	0-1
Hardly ever talk	0.25	0.02	0-1
Never talk	0.37	0.02	0-1

 

(top of page)

 

[1] Missing responses led to the following final sample sizes for each variable by gender for men (m) and women (w): paternity 110 m, 135w; child's age 99 m, 132w; father's age 107m, 125w; custody 107m, 135w; perceived odds 110m, 133w; father sees, 108m, 133w, grandparent sees 108m, 130w; race 107m, 133w; income 108m, 130w;  education 108m, 131w; father helps 108m, 126w; number of children 107m, 116w; mother's age 131w; sex partners 118w;  relationship 106m.

             The extent of missing responses is listed in the footnote to the table.  Missing values were not imputed and so respondents with missing values are not included in our analyses.  Genetic tests revealed that 56% of the sample of men and 90% of the partners of the sample of women were biological fathers of the child in the triad.  This suggests a self-selection bias that led fewer biological fathers and more female partners of biological fathers to enroll into the study.   Both mothers and men reported non-negligible amounts of child contact with 39-44% of men seeing the child within the last 30 days and 29-32% reporting paternal grandparent contact during the same period.  These frequencies are expressed as ranges because of disagreement for the linked couples. The lower estimate corresponds to contact rates as reported by women only and the higher estimate reflects contact rates reported by men only.  Prior to the test, the sample of men reported less confidence than women that the genetic test would confirm paternity with 69% of men reporting odds less than 50-50 and 71% of women reporting an almost certain chance.  Only women were asked about sexual activity the month the child was conceived, and 58% reported only one partner. Given the divergence of the actual lab results between the male sample and the female sample, divergent anticipation of test results is partially to be expected.  More of the female sample (77%) than the male sample (53%) reported that the man did nothing to help the mother.  To check whether these divergences reflected biases in reporting, Table 2 shows responses by gender in the matched sample of 57 cases in which both the man and the woman were enrolled.

Table 2: Concordance between mother's and father's reports of variables for matched couples [1]
	Men in Couples			Women in Couples			P-Value
Perceived Odds of a Positive Match	M	SD	Range	M	SD	Range
50-50 or less	0.66	0.03	0-1	0.2	0.02	0-1	<0.05
A pretty good chance	0.23	0.02	0-1	0.25	0.03	0-1
An almost certain chance	0.11	0.01	0-1	0.55	0.03	0-1	<0.01
Putative father saw focal child in last 30 days	0.54	0.03	0-1	0.57	0.03	0-1
Putative paternal grandparents saw focal child in last 30 days	0.36	0.03	0-1	0.34	0.03	0-1
Does putative father do anything to help mother?
Nothing	0.47	0.03	0-1	0.63	0.03	0-1	<0.05
Something	0.53	0.03	0-1	0.37	0.03	0-1	<0.05

 

(top of page)

              [1] Missing responses led to the following sample sizes for each question in the sample of matched couples: perceived odds n=56 couples, fathers sees n=54, grandparent sees n=50, father helps n=51 

  Even in the couples sample, women were more likely than their linked partners to report an almost certain chance of paternity confirmation and less likely to report a 50-50 chance or less.  Men and women showed good concordance in reporting rates of paternal contact with the child.  They diverged in reporting whether the man does anything to help the mother with 63 % of women and 47% of men reporting doing anything to help the mother.

Analytic Approach

 

(top of page)

            Our analysis proceeds in four steps.  Because there is no prior literature identifying variables that predict biological paternity in disputed cases we first undertake exploratory tabulations to identify any relationships between background variables and the genetic confirmation of paternity.  Each background variable is cross tabulated against the results of the paternity test and differences between the sample of cases where paternity is biologically confirmed and the sample where paternity is excluded are assessed using students t-tests for continuous variables and z-tests for dichotomous variables.

            The second analytical step is a multivariate logistic regression of the odds that paternity is confirmed. We confined this analysis to only the sample of men because only the men were asked details of romantic and parenting intentions. Because cases where both the man and the woman enrolled in the study may exhibit selection bias we included a dummy variable to flag these cases.  To minimize the effects of collinearity, we entered groups of variables suggested by past literature and our conceptual framework blockwise for multivariate analysis in the categories of 1) relationship quality, 2)attitudinal characteristics, 3)bonding, and 4)child variables.  These headings are in Table 4 and 5, abbreviated respectively  as  "Rel Qual" "Attitude" "Bonding" and "Child". The regression analysis applied Huber-White standard errors, also known as "sandwich" estimators to adjust the standard errors for the non-independence of the 57 observations from linked couples.  A dummy variable marking out observations that came from participants whose partner also participated was also included.

Third we used multivariate logistic regression to estimate odds ratios of predictors of whether the man had any contact with the child in the last 30 days. Each of these child contact regressions followed the same blockwise strategy as before.  Because of the potential for collinearity among variables in any single theoretical group, theories are best assessed using the F-statistic testing joint significance for all variables in a group.   Also, we estimated an atheoretical model identified by backward stepwise regression to achieve parsimony.  This model is labeled "Stepwise" and included as the final column in Tables 4 and 5.

Results

Determinants of Paternity Establishment

 

(top of page)

            Table 3 displays descriptive data tabulated by whether paternity was confirmed in the genetic test.  The confirmation rate overall was 76%  (=183/245), which is very similar to the 72% confirmation rate in a nationally representative sample (Bishai, Astone et al. 2006). 

 

(top of page)

Table 3. Socio-Demographic Background by Later Paternity Confirmation Status
	Paternity Excluded  (N=62)				Paternity Confirmed (N=183)			P-Value
	M	SD	Range		M	SD	Range
Father's Age	30.46	10.15	19-65		31.25	8.99	18-61
Father saw child last 30 days.	0.42	0.03		0-1	0.41	0.02	0-1
Pat. Grand parents saw child last 30 days.	0.32	0.03		0-1	0.30	0.02	0-1
Child resembles father	0.12	0.01		0-1	0.16	0.01	0-1
Child has father's last name	0.30	0.03		0-1	0.42	0.02	0-1	<0.05
Father at delivery	0.29	0.03		0-1	0.40	0.02	0-1	<0.05
Man's belief test will confirm paternity 1-5	3.12	0.1		1-5	3.08	1.29	1-5
Child is a boy	0.45	0.03		0-1	0.51	0.02	0-1
Child's age (years)	4.48	4.8		0.3-15	4.06	4.37	0.03-23
Fathers fertility (Children ever fathered)	1.32	1.24		0-5	1.9	1.37	0-5	<0.05
Odds of marrying mother (1-5)	1.49	1.14		1-5	1.42	0.91	1-5
Promised mother financial support	0.47	0.03		0-1	0.59	0.02	0-1	<0.05
Couples' relationship duration (yrs)	7.49	6.39		1-30	7.07	5.75	0-30
Both partners enrolled in study	0.73	0.03		0-1	0.37	0.02	0-1	<0.01

 

(top of page)

There were significant differences in other characteristics between biologically excluded and confirmed cases, including father's attendance at the delivery, father's fertility, promising financial support and having a child with the same last name.   Respondents' subjective predictions of confirmation were weakly correlated with the test outcome, but only for women. Women who stated that there was "no chance" that the man would be confirmed as the father had a 70% paternity confirmation rate which was significantly different (p<0.01) from the 94% confirmation rate of all other women (Figure 2).  What men predicted the test results would be had no correlation with paternity confirmation as is illustrated in Figure 2.

 

(top of page)

Figure 2Enlarge this image

Magnitude of Paternity Disruption

 

(top of page)

            There were 230 men with non-missing reports of contact with the child. Of the 98 men who reported contact with the child in the last 30 days, 25 (25%) had their paternity excluded by the test and this was not statistically significantly different from the average for all men.  There was evidence that many of the 25 men who would later have their paternity refuted had spent some quality time with the child.   In 50% of these cases the men reported reading stories, playing, visiting relatives, or hugging the child.  In 16% of the cases where paternity was refuted the man was co-residing with the mother and child.  In this sample of 25 paternity exclusions where the man was in contact with the child, 37% of the children in these cases were above the age of 2.  On the other hand, of the 132 men who reported no contact with the child in the last 30 days, 104 (79%) were later confirmed as biological fathers. 

 

(top of page)

 

 

(top of page)

Determinants of Paternity Confirmation

 

(top of page)

             Multivariate determinants of the rate of paternity confirmation are shown in Table 4.  

Table 4Enlarge this image 

The specification in the first column includes all 5 categories of variables simultaneously and is subject to sample size reduction due to missing variables.  Subsequent columns of the table focus on more limited subsets of independent variables.   Overall the table shows few systematic predictors of paternity confirmation.  Paternity confirmation appears to have negligible correlation with the man's intention to be more involved with the woman.  Men who stated that they did not know if the mother wanted them more involved with the child had lower odds of paternity confirmation than men who were certain that the mother wanted them involved.  The F-test for the full model rejects the joint significance of the full set of predictor variables, but the F-test for the variables in the stepwise model in the final column is significant.  As shown in Figure 2, men's reported predictions of the genetic test result had no correlation with their actual  paternity confirmation.

Determinants of Child Contact

 

(top of page)

            Results from multivariate analyses of men's contact with the child, reported in Table 5, show that the separate blocks of variables reflecting aspects of the romantic relationship, parental intentions towards the child, and the pre-existing bond were associated with the odds of contact between the men and the focal child. The final row of the table shows that being a father who would later be confirmed by a genetic test did not significantly predict contact with the child, although the association is positive.  This variable also did not emerge as significant in the step wise regression.  F-tests at the bottom of the table show that, considered jointly, attitudinal characteristics were significant determinants of contact as were relationship quality and indicators of father-child bonding.

Table 5Enlarge this image

 The test that all variables in the full model were significant was rejected, but with only 72 complete observations and 19 degrees of freedom this test has low power.  Promising the mother financial support and knowing that the mother wants involvement were two of the strongest predictors of father-child contact.

Discussion

            Our overall paternity confirmation rate of 76% is similar to results obtained in a national sample of paternity tests in the U.S. which documented an overall 72% rate of confirmation of paternity which had almost no systematic variation with the age, race, and ethnic background of triad members(Bishai, Astone et al. 2006) .  To our knowledge this is the very first study of parenting behavior among men immediately prior to the resolution of disputed paternity. 

              Unadjusted rates of baseline father-child contact were uncorrelated with the genetic test results (Table 3).  Multivariate regression found no correlation between genetic paternity and contact with fathers (Table 5).  The selective shirking explanation predicts that rates of child contact among men who did not acknowledge paternity would be negligible and no higher for men who were genetically confirmed as fathers.   Our data show that 41-42% of men had had contact with the child in the last 30 days regardless of their genetic paternity.  This rate of contact is not negligible, though it is substantially lower than the 75% observed for all unmarried men from the Fragile Families study (Carlson, McLanahan et al. 2005).  Although men who do not voluntarily acknowledge paternity do not have as much contact with their children as the average unmarried father, their rates are higher than would be expected if they had no inclination towards the role of fatherhood.  This picture does not look like shirking.  It looks more like men who think they might be fathers but who are unsure.

            Remarkably, men who are ultimately excluded as biological fathers are as likely as confirmed biological fathers to be in contact with the children alleged to be theirs prior to the genetic test.  Part of the explanation lies in the true uncertainty that these men have about their biological paternity.  The multivariate models of the predictors of paternity confirmation were poor and men's subjective rating of confidence in their paternity had no correlation with the actual test results. 

            There are important limitations in our analysis, in particular the small sample of respondents.  Furthermore our evidence shows that men who enrolled when their partners did not were more likely to have their paternity disproven, and when women enrolled alone tests were more likely to confirm paternity.  This selective enrollment occurred despite the joint presence of both members of the couple at the testing site and our practice of inviting both partners to participate in the survey on enrollment days.  Readers should be careful about extrapolating the results of this analysis to the broader population of all families where men have not voluntarily acknowledged paternity. 

            Our results have important implications for the refinement of child-support policy.  We found that roughly 25% of the non-resident men and their parents in our sample who were actively engaged with children would learn through genetic testing that they are not biologically related to the child.  Whether these individuals sever their connections as a consequence of the genetic test information is still unknown.  We found that 37% of the children in these cases are above the age of 2, so strong emotional bonds may have been established, raising the possibility of psychological harm.  There have been no systematic evaluations of the impact of paternity disconfirmation on the children (or adults) who have formed a relationship with a man in a paternal role.  On the other hand, there is a large contingent of previously uninvolved men who will receive child-support orders on the basis of the genetic tests.  This will, in some states and in some cases, potentially improve the financial resources available for these children.  It will certainly reduce the financial burden of supporting these children for the state.  Future research will determine the impact of paternity confirmation on the parenting behavior of previously uninvolved men and paternal kin.