More about the Dads: Exploring Associations between Nonresident Father Involvement and Child Welfare Case Outcomes
Prepared for: U.S. Department of Health and Human Services Office of the Assistant Secretary for Planning and Evaluation Office of Human Services Policy and Administration for Children and Families Administration on Children, Youth, and Families Children's Bureau
This report is available on the Internet at:http://aspe.hhs.gov/hsp/08/MoreAboutDads/index.htm
Printer Friendly Version in PDF format (30 pages)
- Summary of the Original Sample and Findings
In 2006, the U.S. Department of Health and Human Services (HHS) published the results of a study regarding child welfare agencies efforts to identify, locate, and involve nonresident fathers of children in foster care. That study, conducted by the Urban Institute under contract to HHSs Office of the Assistant Secretary for Planning and Evaluation and in cooperation with the Childrens Bureau in the Administration for Children and Families, was based on telephone interviews with caseworkers in four states (Arizona, Massachusetts, Minnesota, and Tennessee) about specific children in their caseloads. The findings of that study are described briefly below and are presented in more detail in the studys full report, What about the Dads? Child Welfare Agencies Efforts to Identify Locate and Involve Nonresident Fathers. Those findings are primarily descriptive in nature. Because all children in the sample were in foster care at the time the caseworker interviews were conducted, the original study could not examine the relationship between nonresident father involvement and case outcomes. By design, none of the cases had outcomes when the original data collection occurred.
This report, using administrative data supplied by each state that participated in the original study, examines case outcomes for the children whose caseworkers were interviewed previously. At the time data were extracted for this follow-up analysis, approximately two years had passed since the original interviews, and most of the children had exited foster care. These analyses use information from the original survey about whether the father had been identified and contacted by the child welfare agency and about the fathers level of involvement with their children, combined with administrative data on case outcomes two years following the interviews, to explore three research questions: (1) Is nonresident father involvement associated with case length? (2) Is nonresident father involvement associated with foster care discharge outcomes? and (3) Is nonresident father involvement associated with subsequent child maltreatment allegations? Before presenting current findings, we provide a brief overview of the study sample and findings from the prior caseworker survey, and describe the current studys methodology and limitations.
For the original What about the Dads study, data on 1,958 children in foster care as of October 2004 in four study states were collected through telephone interviews with 1,222 caseworkers between October 2004 and February 2005. To be eligible for the study, each child had to have been in foster care for the first time, had to have been in foster care for at least 3 months but no more than 36 months at the time of sample selection, and each childs biological father had to be alive but not living in the home from which the child was removed. The surveys response rate was 83 percent.
Demographic characteristics of identified nonresident fathers were similar to those of the childrens mothers, though, as expected, caseworkers knew less about the nonresident fathers than they knew about the mothers of the children on their caseloads. While most caseworkers knew the identity of the fathers in the studys sample, paternity had been established for not quite two-thirds of the children and contact had been made with just over half of the fathers. Contact with the nonresident father was likely only if his identity and location were both known immediately at the time of case opening. Figure 1 summarizes the levels of father engagement with the child welfare system, among the 1,958 cases in the sample.
Results of multivariate analyses examining factors associated with identified father cases, cases in which the nonresident father is reported to be visiting with his child, and cases in which the nonresident father is reported to be providing either financial or nonfinancial support to his child establish important context for understanding the current results involving case outcomes. Results from logistic regression models examining factors associated with father identification, father visitation, and father support show several independent variables to be related to these factors:
- Father identification. Older children were more likely than younger ones to have identified fathers. Compared to white children, African American and Hispanic children were less likely to have identified fathers. Cases involving male caseworkers and African American caseworkers (compared to white caseworkers) were also less likely to have identified fathers.
- Father visitation. Cases with fathers who were older, who had expressed an interest in the child living with him, or who had provided support to the child (financial or nonfinancial) were all more likely to have a father who visited his child.
- Father support. Fathers more likely to provide support to their children were those who had once been married to the mother, who had expressed an interest in having the child live with him, who had visited the child at least once, and those with whom the agency had considered placing the child.
These findings helped provide a framework for the current analyses, which are described in more detail in the next section.
[ Go to Contents ]
Starting in June 2006, Urban Institute researchers contacted the four original study states (Arizona, Massachusetts, Minnesota and Tennessee) to determine whether they would provide updated administrative data on each of the sampled children. All four states agreed to participate in this follow-up component of the study and provided the requested data. Administrative data on case outcomes (case length, discharge reasons and status, and subsequent maltreatment allegations) were extracted between December 2006 and February 2007, making the elapsed time between the date of caseworker survey responses and outcomes data approximately two years.
Several criteria had to be met for a child to be part of the original studys caseworker survey sample. As noted above, children had to have been in out-of-home care for the first time (first placement episode), and had to have been in care between 3 and 36 months as of October 2004. Thus, very short-term placements were not included in our sample. In addition, each child had to have a living father who did not live in the home from which the child was removed.
For the caseworker survey, a total of 1,222 workers were interviewed about 1,958 foster children. In 1,721 cases (88 percent), workers reported that the father had been identified, and in 1,071 (55 percent) of the cases the father had been contacted by the worker or agency at least once. Questions targeting father involvement whether the father visited the child and whether the father provided financial or nonfinancial support were only asked of caseworkers who had made contact with the father. Caseworkers who reported no contact with the nonresident father would presumably have had no information with which to respond to these questions.
Nonresident father involvement was defined two ways. First, we examined involvement in terms of agency-father interaction was the father unknown to the child welfare agency, was he identified by name in the childs case file but not contacted, or was he both identified in the case file and contacted by the agency. Second, we examined father-child involvement by whether the worker reported that the father had visited the child, and whether he provided financial or nonfinancial support to the child.
The sample included 237 cases with unknown fathers, 590 with fathers who had been identified but not contacted and 1,071 with fathers who had been both identified and contacted by the child welfare agency. Information on father-child involvement, defined as a father-child visit or the father providing financial and/or nonfinancial support to the child, was only collected for cases in which the father had been both identified and contacted.
We use three categories of father-child involvement in the analyses that follow (see figure 2). Fathers classified as not involved provided no financial or nonfinancial support and did not visit their child. Involved fathers comprise the largest group of cases, those in which caseworkers reported fathers who had done one or two of the following: provided financial support, provided nonfinancial support, or visited their child at least once since case opening. The smallest group of cases included highly involved fathers who provided both financial and nonfinancial support to their child and had visited their child.
Our research questions focus on three primary outcomes case length, subsequent child maltreatment allegations, and discharge status. All three are included as dependent variables in the analyses conducted.
- Case length was the time period between the date of first entry into foster care and the date of discharge. If the case was still open, the date of the data extraction was used.
- Discharge status was the status of the case as of the time administrative data were extracted in each state (ranging from December 2006 through February 2007). Extracted administrative case data were obtained at least 30 months after the child had originally entered care. Discharge status included four categories adoption, reunification, other (e.g. guardianship, discharged to and living with a relative, and emancipation), and still open cases.
- A subsequent allegation was defined as any allegation that occurred since the time of the original sample identification.
Because of the exploratory nature of the original study, a large number of independent variables were incorporated into the logistic regression models. For the multivariate analyses conducted for the follow-up study we continue to use a large number of independent variables. Similar to the earlier models, variables are categorized into case (child), father and mother, family, and caseworker characteristics.
Case and child characteristics. Case characteristics include state, case length, the primary type of maltreatment, perpetrator, and permanency goal. In addition, we used the zip code for each local child welfare office to construct a new case variable measuring rurality using the Urban Influence classification (large metro, small metro, and several less urban categories) established by the U.S. Department of Agriculture. Child characteristics include gender, age, and race.
Father and mother characteristics. Father and mother characteristics include age of mothers and fathers, whether the father was the legal father, and mothers education level. For both fathers and mothers, information was collected from caseworkers on circumstances that could be barriers to agency contact and problems the father and mother might have faced that would make reunification difficult. We created a multi-circumstance variable for barriers. A case is a multi-circumstance case if three or more circumstances were reported. Similar to the multi-circumstance variable, three or more problem areas were categorized as a multi-problem case. The multi-circumstance and multi-problem case categories were mutually exclusive. A resident mother variable was also included.
Family characteristics. A relationship quality variable and a relationship status variable were included as family characteristics. The relationship quality variable was coded from caseworker responses in the original survey to a question about the affective relationship between the nonresident father and birth mother of the child. The relationship status variable identified whether the father was ever married to the mother.
Caseworker characteristics. Caseworker characteristics include gender, race, and whether the worker received training on how to identify, locate, and involve nonresident fathers. We also included data on workers opinions of nonresident father involvement. Worker opinion is coded as positive if the worker agreed or strongly agreed that Involvement of nonresident fathers enhances a childs well-being and they disagreed or strongly disagreed that working with nonresident fathers is more trouble than its worth. Conversely, worker opinion is coded as neutral/negative if the worker neither agreed nor disagreed, disagreed, or strongly disagreed that the involvement of nonresident fathers enhances a childs well-being and neither agreed nor disagreed, agreed, or strongly agreed that working with nonresident fathers is more trouble than its worth.
To examine our research questions, we conducted a series of bivariate and multivariate analyses. For each of the three research topics case length, discharge status, and subsequent maltreatment allegations we conducted only bivariate analyses to explore differences in level of father-agency interaction. Multivariate analysis was not possible regarding father-agency interaction because we lacked independent variables for fathers who had not been contacted by the agency. We also conducted bivariate analyses to explore descriptive differences between fathers who had different levels of involvement with their children (i.e., not involved, involved, or highly involved).
To examine differences in case length, we conducted bivariate analyses on whether the father was unknown, identified, or contacted and his level of father-child involvement. For the last research question regarding differences in discharge status, bivariate analyses were again used to explore differences across groupings of fathers. Multinomial logit models were developed to determine the predicted probability of being adopted, reunified, discharged for other reasons, or remaining in foster care. Other variables of interest added to the models included whether or not the father provided financial or nonfinancial support to the child, and whether or not the father had visited the child since case opening. This technique enabled us, for instance, to determine the probability of a child being adopted if his/her father was highly involved, assuming mean values for other variables in the model.
To explore relationships between subsequent maltreatment allegations and level of father-agency interaction we ran bivariate analyses to determine whether significant differences existed. We also ran bivariate analyses to determine differences in subsequent maltreatment allegations across types of father involvement.
[ Go to Contents ]
The study findings have several limitations. First, due to the different sources of information, we are unable to determine whether father involvement changed over time or assess the fathers level of involvement at the time of discharge. The data were collected at two different points in time the caseworker survey data was collected in early 2005, while administrative data was extracted late in 2006. In addition, there are two data sources caseworkers responded to a telephone survey to provide what we refer to as the survey data, while the data on case length, subsequent allegations, and discharge outcomes, were obtained from states administrative data systems. While the administrative data provided are linked directly back to the caseworker survey response data, the findings reflect the relationship between early measures of father involvement and outcomes almost two years later. Relationships could have changed or other events could have intervened that affected the case outcomes.
Second, the information on the nonresident fathers of the sample children was obtained from the childs caseworker and the surveyed caseworkers had varying degrees of knowledge of the parents. For example, when we discuss nonresident fathers payment of financial support to the child, the information we use was obtained from the childs caseworker, but the caseworker may not know about payments made through the child support system. It is also possible that a father could potentially visit his child without the caseworkers knowledge. Because caseworkers cannot report on information they do not have, the strength of the father-child involvement could be understated in the analyses.
Lastly, the caseworker survey instrument was designed so that certain questions were asked only if the nonresident father had been identified, and still more questions were only asked if the father had been contacted by the child welfare agency. For example, questions about problems the nonresident father may have been experiencing (such as unemployment, criminal justice involvement, etc.) which are used to create the multi-problem variable, were only asked of cases in which the caseworker had reported contact with the father. Likewise, only in these cases were caseworkers asked whether the father visited the child, and whether the father was providing financial or nonfinancial support to the child. This information is not available on the substantial group of cases with unknown fathers and fathers with whom the worker or agency never made contact.
It is important to emphasize that the findings presented in the sections that follow are the result of exploratory analyses that cannot determine causality. The cross-sectional survey dataset does not allow for the examination of the timing of events. Without knowledge of the timing of caseworker decisions and other events such as termination of parental rights, causality cannot be determined for associated factors. While the study findings indicate that father-child involvement is associated with different discharge outcomes for children in foster care, a reader cannot conclude that the father-child involvement directly caused the different outcomes.
[ Go to Contents ]
This section begins with a general description of case outcomes for the study sample. Case length, discharge outcomes and status, and subsequent allegations are described from the administrative data. Following this description of outcomes overall, each of the three research questions are examined, relating fathers involvement in the case to outcomes experienced by their children.
Just over 75 percent of the children in the study sample had been discharged by the time of the follow-up (see figure 3). The average (mean) case length including all cases (both discharged and still open) was 31.0 months. Discharged cases had an average case length of 25.5 months and children with open cases had been in foster care an average of 44.6 months. Average case length for children discharged to adoption was 29.4 months, while the average for children discharged to reunification was 20.4 months. The proportion of cases with a subsequent maltreatment allegation varied by type of discharge. Consistent with other research, cases discharged to reunification were far more likely than cases discharged to adoption to include a subsequent maltreatment allegation (15.2 vs. 2.1 percent, p<.001).
Because most of the sample children had been discharged from foster care at the time the administrative data were extracted for the follow-up analysis, we were able to perform analyses of discharge outcomes for these cases (see figure 4). Adoption was the outcome for 41 percent of discharged children while reunification was the outcome for 30 percent. Guardianship, emancipation from foster care, and being placed with relatives were other possible discharge outcomes.
Our sample varies from national data (Adoption and Foster Care Analysis and Reporting System) on outcomes and time spent in foster care. On average, children who left the foster care system in FY 2005 spent 21 months in care, which is nine months fewer than children in the sample. Given the sample criteria, that is, children in foster care a minimum of 3 months and a maximum of 36 months at the time the sample was identified, the shorter period in foster care seen in AFCARS data is likely due to the exclusion from our sample of children who experience very short stays in care (less than 3 months).
Discharge outcomes for the children in our sample are substantially different than those shown in AFCARS data for the same period. AFCARS data show 54 percent of children exiting foster care in 2005 exited to reunification while only 30 percent of our sample did so. Percentages of children exiting to adoption also differ greatly between AFCARS and our sample. While only 18 percent of AFCARS cases exited to adoption, 41 percent of our sample did so. As with the data on case length, our criteria for minimum and maximum months in foster care may be the cause of these differences. Children who stay in care less than three months and quickly exit to reunification were systematically excluded from our study sample. In addition, children who were in care for more than three years at the time of sample selection were excluded from the sample. These long-term foster care cases may be children living with relatives and thus, less likely to exit to adoption. State policy differences are also likely to have affected the discharge percentages.
Finally, our sample included only foster care cases in which the child had been removed from homes in which their biological fathers did not reside. This selection criterion may also have produced a sample of children who, when compared to national data, are more likely to exit to an adoption than cases in which the father was living in the home.
We examined the first research question Is nonresident father involvement associated with case length? in two ways. First, we examined case length by fathers level of contact with the child welfare agency. That is, did the case length vary if the father was unknown, identified-only, or had been contacted by the child welfare agency? Bivariate analyses were conducted to examine this question.
The average case length was 31.1 months for cases with unknown fathers, 30.8 months for cases with identified-only fathers, and 29.4 months for cases with contacted fathers. The difference between cases with identified and contacted fathers was marginally significant (p<.10) while the difference between unknown father cases and contacted father cases was not significant. Among discharged cases, the mean case length was 25.5 months overall and differences in case length among these groups was not statistically significant.
We examined differences in case length across different discharge outcomes. For unknown father cases, there was not a significant difference in case length between cases discharged to adoption and cases discharged to reunification (27.3 months and 24.5 months, respectively). This pattern is unusual for the child welfare system because typically adoptions take much longer than reunifications. That this group of cases has relatively speedy adoption outcomes may be related to levels of maternal dysfunction that make an adoption decision easier, or that a diligent search for an unknown father is a shorter process than when the father is known. Shorter case lengths could also be due to better casework practice and organizational functioning which translates into enhanced efforts to identify, locate, and contact nonresident fathers.
For cases with identified-only fathers and contacted fathers there was a significant difference in case length for the two types of discharges. Reunification occurs more quickly than adoption in both groups of cases (identified-only father cases p<.001, contacted father cases p<.001). The average case length for children discharged to reunification was 20.4 months for cases with identified-only fathers and 20.2 months for cases with contacted fathers, while the case length for adoption discharges was 30.8 and 29.3 months, respectively. The differences in case length between cases with identified-only and contacted fathers were not statistically significant.
Contrary to often reported caseworker concerns, contacting fathers does not appear to slow case proceedings or complicate the case in a way that delays permanency for the child.
When we examined all cases (both discharged and still open), we found significant differences in case length between cases with highly involved fathers and cases with involved fathers, with mean case lengths of 26.2 months and 30.6 months (p<.01), respectively. The difference in case length between cases with involved fathers and cases with not-involved fathers was marginally significant (p<.1). However, the mean case length is actually longer for involved fathers than not-involved fathers (30.6 vs. 28.6 months). There was no significant difference in case length between cases with highly involved fathers and not-involved fathers. Notably, case length was shortest for cases with highly involved fathers.
We also compared case lengths of discharged cases with different levels of father involvement (highly involved, involved, not involved). Examining only discharged cases, we again found that cases with highly involved fathers were shorter, on average, than cases with involved fathers (21.4 vs. 26.3 months, p<.001) (figure 5). Also, cases with highly involved fathers were shorter, on average, than cases with fathers who were not involved (21.4 vs. 25.3 months, p<.001). Similar to the all case comparisons above, among all discharged cases, case length for the involved fathers was longer than the case length for either highly involved or not-involved fathers.
Figure 5. Average Case Length for Discharged Cases by Level of Nonresident Father Involvement
We examined differences in case length by different levels of father involvement across discharge outcomes. For adoption discharges, there was a significant difference in case length between involved and not-involved father cases, with involved-father cases taking longer to reach discharge than those in which fathers were not involved. The average case length to adoption was 27.8 months for cases in which the father was not involved, 30.5 months for cases with involved fathers, and 31.1 months for cases with highly involved fathers (30.5 vs. 27.8, p<.1) (31.1 vs. 27.8, p <.1).
The reverse was true for reunification cases. Cases with involved fathers (both involved and highly involved) took less time to discharge than cases with fathers who were not involved. The average reunification case length was 24.4 months for cases with no father involvement, 19.7 months for cases with involved fathers, and 19.0 months for cases with highly involved fathers (24.4 vs. 19.7, p<.1; 24.4 vs. 19.0, p<.05). The greater likelihood of reunification for cases with highly involved fathers is likely affecting the overall lower average case length shown in figure 5. More information on the association between nonresident father involvement and different discharge outcomes is presented below.
We examined the second research question Is father involvement associated with type and status of the childs discharge? along similar dimensions as the first research question. Using bivariate analysis, we first examined whether the three levels of nonresident father involvement were associated with different discharge outcomes. Second, we examined discharge status by level of father involvement. As noted earlier, we included several possible activities in our definition of father involvement (see methodology section, page 5).
There were significant differences in the distribution of discharge outcomes and status across the three groups of cases (see figure 6). Cases with unknown fathers were almost twice as likely to result in adoption as an outcome than cases with an identified-only father (p<.001). Cases with contacted fathers were less likely than cases with unknown fathers to have adoption as an outcome (p<.05), however, they were more likely than cases with identified-only fathers (p<.001) to have this outcome. With regards to reunification as an outcome, cases with unknown fathers were less likely to be reunified with a parent than cases with an identified father (p<.01) or cases with contacted fathers (p<.001). Cases with identified-only fathers were more likely to be discharged to other than both cases with unknown fathers (p<.001) and contacted fathers (p<.01). There were no significant differences in the likelihood of the childs case remaining open across groups.
* Other exit or discharge options include guardianship, emancipation from foster care, and being placed with relatives.
Unknown father cases may more frequently conclude with the childs adoption for a variety of reasons. For instance, judges may be more willing to terminate parental rights for an unknown father. In addition, there may be important characteristics of the mother associated with not identifying the childs father. The high percent of contacted father cases that also discharge to adoption may mean that, once identified, some fathers are being contacted primarily to terminate their parental rights. The relatively small percent of unknown father cases that discharge to reunification may reflect issues regarding the mother. That is, mothers who are unable or unwilling to identify their childs father may also have other characteristics that are related to the likelihood of an adoption outcome. In this way, the fact that the fathers identify is unknown may be serving as a marker for broader family dysfunction. The higher percent of identified-only father cases with other discharges may be the result of guardianships with relatives. If an initial placement with a relative is stable, the caseworker may have less motivation to seek contact with an identified father.
Using similar bivariate analyses, we considered whether discharge status varied by different levels of father involvement. Figure 7 presents the discharge outcomes for each level of father involvement used in the analysis. Definitions of the levels of involvement are described in the methodology section earlier in this report. As shown below, cases with fathers who were not involved were more likely than other groups to be discharged and far more likely to be discharged to adoption than discharged to reunification. Alternatively, the cases of children with highly involved nonresident fathers were far more likely to end in reunification than adoption.
Similar to the earlier findings on the smaller percentage of identified and contacted father cases discharging to adoption compared to unknown father cases, the degrees of father involvement also appear to be associated with different discharge outcomes. As involvement increases (from not involved to involved to highly involved), the percent of cases discharged to adoption decreases and reunification increases. When disaggregated by level of father involvement, children whose fathers were contacted but are not involved (in figure 7) experience outcomes that appear similar those of children whose fathers are unknown (figure 6). This suggests that simply contacting fathers is unlikely to affect outcomes for children, but that contact should support fathers engagement or re-engagement in their childrens lives.
* Other exit or discharge options include guardianship, emancipation from foster care, and being placed with relatives.
Finding significant differences in discharge status across father involvement types using bivariate analyses led us to develop a multinomial logit model using four outcomes adoption, reunification, other exit, and still open. The multinomial logit model enabled us to further examine different types of father support and discharge outcomes while holding a number of other factors constant, such as childs characteristics, mothers and fathers characteristics, and other case factors.
This examination of specific categories of support, i.e., financial versus nonfinancial, was intended to improve our understanding of the association between types of support and child welfare outcomes, specifically how a fathers support may facilitate reunification. In the multinomial logit models, we use all possible combinations of financial and nonfinancial support. Visits with the child are included in the model as a control and are combined with the support variables to show predicted probabilities of the different discharge outcomes based on varying levels of father involvement.
In Tables 1 and 2 we present the results of the multinomial logit model in two ways. Table 1 holds visitation constant for all categories of support in order to examine the relative associations between financial and nonfinancial support and discharge outcomes. Table 2 provides the predicted probabilities when we include visit or no visit variables and is intended to examine whether findings on type of father support are different when visitation is taken into account.
Table 1 presents the predicted probabilities of adoption, reunification, other exits, and still open cases, given different levels of father support. The predicted probability of a reunification discharge is most likely for a father providing financial and/or nonfinancial support and least likely for cases where the father provides no support. Specifically, controlling for other factors, a child whose father provides both financial and nonfinancial support has a 47 percent predicted likelihood of reunification and a 13 percent predicted likelihood of adoption. On the other hand, a child whose father provides no support of either type has only a 14 percent predicted likelihood of reunification and a 43 percent predicted likelihood of adoption.
|Father Support||Discharge Status|
|Both Types of Support (N=210)||0.1279||0.4715||0.3642||0.0364|
|Financial Support Only (N=98)||0.3370||0.5068||0.1195||0.0367|
|Nonfinancial Support Only (N=109)||0.0978||0.6323||0.2430||0.0269|
|No Support (N=467)||0.4313||0.1429||0.3744||0.0515|
|Note: Probabilities come from multinomial logit model and are based on means of allother covariates in the model.|
While reunification has the highest probability of occurring across all categories of positive support (versus no support), there are larger differences related to the type of support in the other discharge categories (e.g., guardianship, emancipation from foster care, and placed with relatives). For example, children whose fathers provided nonfinancial support, with or without financial support, are less likely to be discharged to adoption than are children whose fathers provide only financial support. In addition, children with fathers providing nonfinancial support are more likely to experience other discharges than children with fathers providing only financial support.
Table 2 presents results when visitation is considered together with the various types of support. Visiting, in this survey, was defined as the alleged/birth father visiting the child at least once at any point in the case. As shown below, the overall discharge patterns do not change when we consider visitation, indicating that visitation seems to play a less important role in influencing discharge outcomes than does the provision of financial or nonfinancial support. Consistent across all categories of support, cases with fathers not visiting have a lower probability of still being open or being discharged to adoption and a slightly higher probability of reunification and other discharges. However, small sample sizes in the no visiting case categories makes reliable interpretation problematic.
|Father Support and Visitation*||Discharge Status|
|Both Types of Support|
|Did Not Visit (N=37)||0.1126||0.4825||0.3889||0.0160|
|Financial Support Only|
|Did Not Visit (N=41)||0.3095||0.5407||0.1330||0.0168|
|Nonfinancial Support Only|
|Did Not Visit (N=19)||0.0857||0.6442||0.2583||0.0118|
|Did Not Visit (N=253)||0.4005||0.1542||0.4215||0.0238|
|Note: Probabilities comefrom multinomial logit model and are based on means of all other covariates inthe model. * N's do not equal N's in Table 1 due to missing visitation information.|
Note that caseworker knowledge of father-child visits likely varies systematically across different placement types, further complicating the interpretation of results regarding visitation. Placements in non-kin foster homes require more formalized visits with birth parents, making caseworker knowledge of visitation more likely; placements with kin may result in less formal birthparent visitation and subsequently, less worker knowledge and intervention. This may bias the relationship of visitation away from other outcomes which includes guardianships, the majority of which are placements with relatives.
The final research question concerns whether nonresident father involvement is associated with subsequent abuse or neglect allegations. Similar to the earlier research questions, we examined this issue along two dimensions father agency interaction levels (i.e., whether the father was unknown, identified-only, or contacted) and father involvement. We controlled for discharge status in the analyses given the greater likelihood of a subsequent allegation for reunified cases (see page 9). Note that in these analyses we did not differentiate among potential perpetrators; the perpetrator in the allegation could have been the father, the mother, or someone else.
In terms of father agency interaction, the only significant differences occur for adoption cases and this is likely due to having no subsequent allegations within the unknown father group of cases. None of the unknown father cases had a subsequent allegation while only 1.1 percent of identified-only father cases (p<.01) had an allegation. The difference between identified-only father cases and contacted father cases was significant, though only marginally so (1.1 vs. 5.0 percent, p<.05).
The finding of more relevance to practice is that within reunification cases, there are no differences in the likelihood of a subsequent allegation across the three types of fathers (unknown, identified, contacted). At least in the aggregate, contacting fathers does not appear to increase maltreatment risk to the child. Individual risks must, of course, be assessed on a case by case basis.
The only significant differences found when examining subsequent allegations by father involvement were for subsequent allegations in reunification cases. We found a significant difference between cases with not involved and involved fathers. Cases in which the father was not involved were more likely to have a subsequent maltreatment allegation compared to cases with involved fathers (32.0 vs. 12.2 percent, p<.05). It appears that more involvement on the part of the nonresident father is not associated with subsequent allegations of abuse or neglect. In fact, in cases with reunification discharges, involvement of the nonresident father appears associated with a substantially lower likelihood of a subsequent allegation.
In order to better understand pathways that might influence how the primary explanatory variables father-agency interaction and father-child involvement might affect differences in childrens discharge outcomes, we examined the relationship between the primary explanatory variables and several of the other independent variables including basic descriptive information (e.g., age, marital status), quality of the mother-father relationship, and termination of parental rights.
Description of Involved Fathers. As discussed earlier, we defined father involvement based on the available survey data. That is, the survey asked caseworkers to respond to questions about whether the nonresident father had ever visited with the child in foster care or provided either financial or nonfinancial support to the child. If a case had no identified and contacted father, the caseworker was not asked these questions about that father. To provide a more detailed description of highly involved fathers, we compared these cases with fathers who were less involved or not involved on a number of characteristics including whether the mother and father had ever been married, the current marital status and age of the nonresident fathers.
Highly involved fathers (that is, those who provided both types of support and visited) are older, on average, than their less involved counterparts (38.0 vs. 35.9 years old, p<.05) and their not involved counterparts (38.0 vs. 35.0 years, p<.05). There were no significant differences in fathers age between involved fathers and not-involved fathers.
Highly involved and involved fathers are more likely to have been married to the childs mother (either at time of childs birth or sometime after) than not involved fathers (44.7 and 34.5 vs. 19.2 percent, p<.001). There were no significant differences in whether the father had ever been or was still married to the childs mother between cases with highly involved fathers and involved fathers. This suggests that the marital relationship between the father and mother is related to the quantity, and perhaps, quality of father involvement. However, more than 50 percent of highly involved fathers were never married to their childs mother although nearly one-fifth of not involved fathers were married. Thus marriage may facilitate, but does not guarantee, a fathers post-separation involvement with his child.
Parent Relationship. We examined whether there were differences in whether or not the nonresident father was reported to have a positive relationship with the childs mother by level of father involvement. There were significant differences across levels of father involvement. Both involved and highly involved fathers were more likely to have a positive relationship with the childs mother, as perceived by the caseworker, as compared to fathers who were not involved (p<.001 for both comparisons). Positive mother-father relationships were reported for 33.8 percent of highly involved fathers and 31.1 percent of involved fathers but only for 14.4 percent of fathers who were not involved. There was no significant difference in reported positive mother-father relationships between cases with highly involved fathers and involved fathers. Conversely, these findings also hold for negative relationships between the childs mother and father as reported by the caseworker. Specifically, involved and highly involved fathers were less likely to have a negative relationship with the childs mother, as perceived by the caseworker, compared to fathers who were not involved (p<.001 for both comparisons). This is consistent with findings from other studies that a mothers relationship with the father can affect the fathers relationship with his child.
Termination of Parental Rights (TPR). First we examined the relationship between the level of interaction between the father and the child welfare agencies and TPR for both mothers and fathers. The frequency with which the fathers rights had been terminated varied across the three groups of fathers unknown fathers (20.4 percent); identified-only fathers (18.7 percent); and contacted fathers (26.0 percent). Statistically significant differences in fathers TPR existed between unknown and contacted fathers (p<.10) and identified-only and contacted fathers (p<.01). The frequency of the childs mother having her parental rights terminated also varied across the three groups of cases those with unknown fathers (34.1 percent); identified-only fathers (22.7 percent); and contacted fathers (30.6 percent). Statistically significant differences in mothers TPR existed between unknown and identified-only fathers (p<.01) and identified-only and contacted fathers (p<.01). As noted, most, but not all, of the differences in mothers and fathers TPR by level of interaction were significant, but more importantly, a key component of the pattern between level of agency-father interaction and TPR was similar to the pattern between interaction and the discharge option of adoption (see figure 6) which was 42.9 percent for unknown fathers, 22.2 percent for identified-only fathers, and 33.1 percent for contacted fathers. TPR and adoptions are lowest when fathers have been identified but not contacted. As TPRs are supposed to be executed as a precursor to adoption, it is reasonable that the patterns would be similar and that the possible reasons for the higher rates among unknown and contacted fathers would be similar as well. As discussed above, having an unknown father might be an indicator of other family dysfunction or father contact may be influenced by the agencys need to pursue TPR because adoption is the childs case goal.
Second, we examined the relationship between father-child involvement and TPR. This analysis involves a smaller set of cases because only fathers who were contacted were categorized by father-child involvement (see page 5). Highly involved fathers were significantly less likely to have their parental rights terminated than not-involved fathers (10.2 percent vs. 29.5 percent, p<.001) and involved fathers (10.2 percent vs. 30.5 percent, p<.001). There were no statistically significant differences in fathers TPR between involved and not-involved fathers. Identical to the results of the fathers, mothers of children with highly involved fathers are significantly less likely to have their parental rights terminated than mothers of children of not-involved fathers (13.3 percent vs. 37.1 percent, p<.001) or involved fathers (13.3 percent vs. 33.1 percent, p<.001). As noted earlier (footnote 25 on page 14), we do not have measures of mothers involvement so we could not examine or control for this aspect of the case.
As in the analysis immediately above, the patterns between father-child involvement and TPR are similar to the pattern between father-child involvement and the discharge outcome of adoption. TPR and adoption are much less likely to occur if there is a highly involved father. No involvement by the father or minimal involvement increases the likelihood of TPR and adoption. The similarity in pattern is likely because a case goal of adoption requires the termination of both parents rights. If the court is moving toward terminating a mothers parental rights, it is likely that termination of a fathers rights will also proceed. Additionally, lack of involvement or minimal involvement on the part of the father may serve as a proxy for family dysfunction, capturing the influence of factors not measured in the analysis.
[ Go to Contents ]
While causality cannot be determined from these data, the association between father involvement and discharge outcomes is clear. Nonresident fathers involvement with their children is associated with a higher likelihood of a reunification outcome and a lower likelihood of an adoption outcome. Children with highly involved nonresident fathers are also discharged from foster care more quickly than those with less or no involvement, in part because of their higher probability of reunification. Reunification discharges typically happen more quickly than adoptions throughout the child welfare system.
Children with reunification outcomes were more likely to experience a subsequent maltreatment allegation than children with adoption outcomes. However, contrary to the expressed fears of some caseworkers and child welfare administrators, nonresident fathers contact with the child welfare agency and involvement with their children is not, in the aggregate, associated with subsequent maltreatment allegations. In fact, among children whose case outcome is reunification, usually with their mothers, higher levels of nonresident father involvement are associated with a substantially lower likelihood of subsequent maltreatment allegations.
That fathers lack of involvement was associated with a higher likelihood of an adoption outcome for children was expected. It is unlikely that the agency would pursue TPR or adoption for a child whose father is providing support to and visiting with his child. The greater likelihood of reunification is more difficult to interpret. Only 4.1 percent of the cases had a permanency goal of reunification with the father. Though the data did not allow us to distinguish between reunification with mothers and reunifications with fathers or other individuals we feel fairly confident that most of the cases discharged to reunification involved reunification with the mother. Perhaps a fathers support makes reunification with the mother more realistic in some way. Or it could be that the fathers participation influences the opinions of caseworkers or judges about the mothers capabilities. The pathway through which father involvement and reunification outcomes operate needs further consideration.
The findings on what type of father involvement matters are very intriguing but must be interpreted with caution. That nonfinancial support had as strong an effect on reunification outcomes as did financial support points to the possibility that the meaning of the support may be as important as the monetary value. Another possibility is that nonmonetary support is likely to be received by the child or the custodial family, whereas cash support may be retained by the government to recoup foster care costs and, therefore, not be identified by the family as a contribution in support of the child. The lack of positive effects of visitation was unexpected and may be an artifact of the data collection limitations.
We cannot determine with the available administrative data the amount or frequency of financial or nonfinancial support provided by the nonresident father or the frequency of visits. While our survey data include some information on frequency of support payments and father-child visits, it is limited by the knowledge of the caseworker and small sample sizes. We also do not have information on the content or tone of the father-child visits, such as whether the father and child interact with one another appropriately or whether the mother, in cases appropriate for reunification, was supportive of the visitation. Such data are very difficult to collect, however. We also do not know the extent of paternal relative involvement.
All of these factors may somehow influence casework and case outcomes. While some qualitative research has focused on casework with nonresident fathers of children in foster care further research is needed if we want to better understand the association between nonresident fathers involvement and a greater likelihood of reunification. Additional exploration of these factors through research would help us to understand whether and how improved efforts to contact and engage fathers prospectively would produce tangible benefits for children in the areas of safety, permanency and well-being. The results presented here were initial and exploratory, but they suggest that engaging fathers has the potential to improve outcomes for children.
Since the original What about the Dads study was published in 2006, the Childrens Bureau within the Administration for Children and Families has funded a Quality Improvement Center on Nonresident Fathers and the Child Welfare System (QIC-NRF). Spurred both by the original studys findings and by results of the first round of Child and Family Services Reviews which showed inconsistent effort to engage fathers, the QIC-NRF was designed to promote additional knowledge development regarding the engagement of nonresident fathers whose children who are involved in the child welfare system. Operated by the American Humane Association and its partners the American Bar Association Center on Children and the Law and National Fatherhood Initiative, that project has two phases. In phase I the team has identified sub-themes, knowledge gaps, service gaps, and research priorities. As this follow up report is being published the QIC-NRF is just embarking on phase II of its activities, in which subgrants will be awarded to several sites for the purposes of experimentation and demonstration focusing specifically on the child welfare system's contact and engagement of nonresident fathers, with attention to the collaboration between courts, community systems, fatherhood program providers, nonresident fathers, and paternal kin. Using a collaborative approach, the QIC-NRF will serve as a laboratory for innovation, application, and learning. More information on the QICs activities and products is available on its web site, http://www.fatherhoodqic.org.
[ Go to Contents ]
 Available online at http://aspe.hhs.gov/hsp/06/CW-involve-dads/index.htm
 See pgs. 128-142 of What About the Dads? Child Welfare Agencies Efforts to Identify, Locate, and Involve Nonresident Fathers (http://aspe.hhs.gov/hsp/06/CW-involve-dads/index.htm) for a comprehensive description of results from the multivariate analyses.
 Identified father cases include cases in which the worker said the fathers name was known at case opening and cases in which the worker answered yes to whether the agency has identified an alleged father, i.e., does anyone at the agency think they know who the father is, at the time of the interview.
 Contact was defined broadly to include in-person, telephone, voicemail or written communication with the caseworker or another staff member.
 Cases with fathers who died prior to case opening were considered ineligible and excluded from the sample. A small number (n=25, 1.7 percent of eligible fathers) died after case opening and are included in descriptive analyses.
 Identified father cases are hereafter referred to as identified only. Identified cases in which the father was also contacted are classified as contacted cases. It is important to note this classification is determined by caseworker responses at the time of the telephone survey. A portion of identified only cases may have moved into the contacted category after the time of the survey.
 Theses figures do not total 1,958 due to removal of 60 cases from the second (identified, not contacted) category for analytic purposes. The 60 cases had identified fathers; however, the caseworker responded dont know to whether or not the father had been contacted. To ensure distinct categories, we eliminated these cases from the agency-father interaction analyses.
 A total of 862 of the 1,071 identified and contacted father cases contained information on financial, non-financial and whether the father had visited the child since case opening. Cases with missing or unknown responses in any of the three categories were excluded from the father involvement analyses.
 For more information, http://www.ers.usda.gov/Briefing/Rurality/UrbanInf/
 Data on race/ethnicity was not included for fathers and mothers due to the correlation to childs race. Fathers education was not included due to large amounts of missing information. Caseworkers were asked whether paternity had been established. If yes, the father was considered a legal father.
 Barriers included unreachable by phone, out of state, incarceration, problem transportation, homeless or housing instability, out of country, and language barriers. For more information see page 83 in What About the Dads? Child Welfare Agencies Efforts to Identify, Locate, and Involve Nonresident Fathers.
 Problems included alcohol/drugs, criminal justice, inadequate housing, unemployment, domestic violence, prior finding of abuse/neglect, mental/physical health, lack of child care. For more information see page 88 in What About the Dads? Child Welfare Agencies Efforts to Identify, Locate, and Involve Nonresident Fathers.
 The vast majority of mothers in the cases were residing in the home from which the child was removed. However, there were 323 mothers who did not reside in the home and these mothers are categorized as nonresident mothers.
 A father is categorized as having a good relationship with the childs mother if the caseworker reported that they are friends, are romantically involved on a steady basis, or involved in an on-again off-again romantic relationship.
 Our earlier study matched caseworker responses to administrative data in the states child support systems. For this report, only information obtained from caseworker responses on fathers support is used.
 Fluke, J.D., Shusterman, G.R., Hollinshead, D., & Yuan, Y.T. Rereporting and Recurrence of Child Maltreatment: Findings from NCANDS. (Washington, DC: U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation, 2005).
 Discharge outcomes were missing for three cases, which have been excluded from this analysis.
 Adoption is comprised of all types of public agency adoptions including adoption by foster families, relatives, and strangers.
 The chart reports weighted percents and unweighted Ns so calculations made by the reader to determine the numbers of cases in specific subgroups will provide the estimated Ns which may differ slightly from the actual number in that subgroup.
 U.S. Department of Health and Human Services, Administration for Children and Families, Administration on Children, Youth and Families, Children's Bureau. 2006. Preliminary Estimates for FY 2005 as of September 2006 (13). Washington, DC: U.S. Department of Health and Human Services. http://www.acf.hhs.gov/programs/cb.
 FY 2005 AFCARS data show that 25 percent of foster children had stays of five months or less.
 Because some survey questions were not asked in cases where the father was not contacted, multivariate analyses could not be conducted on this research question. Very limited data are available on cases with unknown fathers.
 Discharge outcomes were missing for three cases, which have been excluded from this analysis.
 We are unable to measure involvement for mothers in the same way as fathers. Caseworkers were asked whether the mother had visited with the child since the time of case opening (87.5 percent had done so), the questions regarding financial and non-financial support, while asked of the mother, cannot be interpreted in the same way. Mothers of the children in the sample were resident mothers at the time of the childs removal from the home. Thus, payment of any financial and/or non-financial support would only have occurred since the time of removal.
 A total of 862 of the 1,071 contacted father cases contained information on financial, non-financial and whether the father had visited the child since case opening. Cases with missing or unknown responses in any of the three categories were excluded from the father involvement analyses.
 As noted earlier, this variable reflects financial support known to the caseworker and may differ from the administrative records of the child support enforcement system.
 Other consists of almost equal parts guardianship cases, living with relative, and youth who emancipated from foster care. Children living with relatives in a foster care arrangement would not have been considered a discharged case.
 Rob Geen and Karin Malm, 'Casework Practices with Kinship Foster Parents', in Kinship Care: Making the Most of a Valuable Resource, ed. by Rob Geen (Washington, D.C.: Urban Institute Press, 2003), p. 117-120.
 All of these variables were controlled for the multinomial logit models.
 It is important to note that the caseworker was asked if the father had ever visited with the child. However, workers knowledge of visitation may have been limited to agency-sanctioned visits which, depending upon agency practices, may require proof of paternity. Marriage to the mother at the time of the childs birth would have formally established paternity.
 Minton, C., and K. Pasley. 1996. Fathers Parenting Role Identity and Father Involvement: A Comparison of Nondivorced and Divorced, Nonresident Fathers. Journal of Family Issues 17 (1), 26-45.
Seltzer, J. A. 1998. Father by Law: Effects of Joint Legal Custody on Nonresident Fathers Involvement with Children. Demography 35 (2): 135-146.
 The survey question related to visitation asks the caseworker whether the father has ever visited with the child at any point in the case. Thus, the child welfare agency/juvenile court could pursue termination proceedings on a nonresident father who, at one point in the case, had visited the child.
 ODonnell, J.M., W.E. Johnson, Jr., L.E. DAunno, and H.L. Thornton. 2005. Fathers in Child Welfare: Caseworkers Perspectives. Child Welfare.
[ Go to Contents ]
This report was prepared by the Urban Institute under contract to the Office of the Assistant Secretary for Planning and Evaluation of the U.S. Department of Health and Human Services, #HHS-100-01-0014.
The report's authors are Karin Malm, Erica Zielewski, and Henry Chen. Karin Malm and Erica Zielewski are research associates at the Urban Institute. Henry Chen, former research associate at the Urban Institute, is now conducting research at Harvard Business School.
Laura Radel and Linda Mellgren, both of the Office of the Assistant Secretary for Planning and Evaluation, were the federal project officers for the contract.
The authors wish to thank Marla McDaniel and Michael Pergamit for reviewing this report and providing helpful comments and suggestions.
"execsum_37.pdf" (pdf, 62.59Kb)