American Indian Fragile Families and the Marriage Initiative: A Replication Study

Beginning in the mid-1990s, the federal government, supported by both Republican and Democratic administrations, has allocated roughly $1.5 billion to promote “healthy marriage initiatives.” A major target of these initiatives have been unmarried parents, or what researchers call fragile families. Over the past two decades, studies have examined this issue within the general population. This study applied three areas of the marriage initiative used by McLanahan (2006) to American Indian people: potential participation in marriage promotion programs, potential impact of marriage programs, and likelihood of marriage. Data for 3,152 women were examined from the Fragile Families and Child Wellbeing Study, including 154 who self-identified as American Indian. This study showed that American Indians exhibited a high willingness to participate in marriage promotion programs. American Indians were less likely than other racial/ethnic groups to see marriage as better for children. This study underscores the need to understand American Indian families and their unique approaches to developing healthy marriage and family structures. For marriage promotion programs to work, they should reflect the cultural practices of the individual American Indian communities.

marriage, relationship skills, and budgeting; (c) marriage education, focusing on skills for non-married, pregnant women, and non-married fathers; (d) pre-marital education and marriage skills training for engaged couples; (e) marriage mentoring and marriage skills training programs for married couples; (f) divorce reduction programs; and (g) programs to reduce the disincentives to marriage. The policy also provided funding for research, technical assistance, and demonstration projects (ACF, 2004). McLanahan (2006) examined three perspectives of the initiative using data gathered from the Fragile Families and Child Wellbeing study (Reichman, Teitler, Garfinkel, & McLanahan, 2001). The potential for participation in marriage promotion programs, the potential impact of marriage programs, and the likelihood of marriage were examined. Further, McLanahan examined evidence to indicate whether the HMI was founded on reliable data and if it was likely to be successful. Using data gathered from the Fragile Families and Child Wellbeing study, McLanahan concluded that the three initial assumptions of the HMI were empirically supported, and the programs funded by the initiative would generally have a positive influence on relationship stability and marriage. In addition, McLanahan (2006) suggested that the HMI should "guardedly move forward" (p. 5).
As part of the HMI, American Indian families provide unique insight into issues relating to fragile families. In this study, it should be noted that participants self-identified as American Indian with no restrictions placed on residential status or need to officially belong to a federally recognized American Indian tribe. American Indians encounter many challenges as they try to form and sustain healthy marriages. The proportion of American Indians who live below the poverty line is almost twice the national average (U.S. Census Bureau, 2016). American Indian children are twice as likely to live in poverty as are their non-Indigenous counterparts (National Healthy Marriage Resource Center, 2012; U.S Census Bureau, 2010; White, Godfrey, & Moccasin, 2006). Further, one of the greatest challenges of American Indian communities under this policy is the high percentage of unmarried families (Administration for Children and Families Healthy Marriage Initiative [ACFHMI], n.d.).
Although previous studies using the Fragile Families data have examined African Americans and Mexican Americans, little research has been done in areas that address the strengths, needs, and issues of American Indians (Reichman et al., 2001). In this study, the term American Indian matches the language used in McLanahan's study (2006) and the original Fragile Families and Child Wellbeing research (see also Reichman et al., 2001). Therefore, the current study applied McLanahan's research on the HMI to American Indian people. As such, the same three areas were examined to better understand American Indian families that are defined by McLanahan (2006) as fragile: (1) potential participation in marriage promotion programs and activities, (2) potential impact of marriage programs, and (3) likelihood of marriage. Examination of these areas increases the knowledge-base of a Western perspective on American Indians and marriage and provides valuable information for funding future programs.

Potential Participation in Marriage Promotion Programs
The first area that McLanahan (2006) examined was whether unmarried parents were willing to participate in marriage promotion programs. Supporters of marriage promotion programs point to research that shows that many unmarried parents desire to marry, suggesting that they are open to participating in marriage promotion programs (Brown, 2000;Bumpass, Sweet, & Cherlin, 1991;Hawkins, Blanchard, Baldwin, & Fawcett, 2008). However, another group of researchers claim that most parents will not participate because their relationships are casual, they have been involved in traumatic or violent relationships, they prefer cohabitation to marriage, or, they simply do not have time to participate (e.g., Edin & Reed, 2005;Roberts & Morris, 1998).
There is little research about American Indian families and marriage promotion programs. There are, however, federal programs set up and designed specifically for marriage promotion among American Indians. For example, as part of the ACF implementation of the HMI in 2002, the Native American Healthy Marriage Initiative (NAHMI) was launched. NAHMI was designed to strengthen relationships and families in American Indian communities by promoting healthy marriages, responsible fatherhood, and child well-being (ACFHMI, n.d.). Then in 2008, President Obama signed into law the Healthy Marriage and Responsible Fatherhood initiative the provided demonstration grants in this area (ACF-HHS, 2014). NAHMI was established to utilize culturally competent strategies to strengthen American Indian families and communities. Similarly, the Administration for Native Americans (ANA) supports projects to improve child well-being by forming and sustaining healthy marriages and strengthening families in Indigenous communities. Unfortunately, data about these programs is difficult to obtain from tribal communities due to limited access and measure variability, making it hard to determine if NAHMI has been successful or not. Further, funding opportunities around NAHMI have been sparse since 2012, while funding for healthy marriage and responsible fatherhood initiatives continues (ACF-HHS, 2018).

Potential Impact of Marriage Programs and Likelihood of Marriage
McLanahan (2006) also examined whether relationship enhancement programs increased marriage rates. Supporters and critics appeared on both sides. Research on a number of relationship enhancement approaches (including marriage education and relationship skills building activities) used with the general population found that relationship enhancement programs increased marital satisfaction and stability (Carroll & Doherty, 2003;Cowan, Powell, & Cowan, 1998;Stanley, Blumberg, & Markman, 1999). However, critics pointed out that the data used were predominately conducted on married, middle-class couples and that these results may not be generalized to other couples (Dion, 2005;Jarchow, 2003). Further, critics questioned whether the emphasis on relationship skills was warranted. They argued that low wages, high unemployment, trauma, domestic abuse, and other socio-demographic elements could pose much stronger barriers to actual marriage (e.g., Legal Momentum, 2006;Lerner, 2004;Onwuachi-Willig, 2005).
McLanahan (2006) also examined whether marriage promotion policies actually increased marriage and whether marriage was thought to be best for children. McLanahan reported that there is widespread disagreement among researchers about whether marriage would make children better off financially and physically. A number of studies (e.g., McLanahan & Sandefur, 1994;Waite, 1995;Waite & Gallagher, 2000) found that marriage is beneficial and that children who grew up with both biological parents were more successful across a broad range of outcomes than children who grew up with a single parent. However, several researchers believe the benefits of marriage have been overstated and it is not so much because the parents were married as the characteristics of parents who marry and stay married versus characteristics of parents who divorce or never marry (Axinn & Thornton, 1992;Booth & Johnson, 1988;Cherlin, 2010).
There are few studies about American Indian people that specifically examine relationship enhancement programs. For example, American Indian children who live with their own married parents have better physical health than children in other family forms (Radel, Bramlett, Chow, & Waters, 2016). American Indian children whose parents divorce have higher rates of psychological distress and mental illness (Kenney & Singh, 2016). Marriage is associated with better health and lower rates of injury, illness, and disability for both American Indian men and women (Administration for Children and Families Healthy Marriage Initiative, Health and Human Services, n.d.; Goins et al., 2018). Additionally, marriage increases the likelihood that American Indian fathers have good relationships with their children and married women appear to have a lower risk of experiencing domestic violence than do cohabiting or dating women (Sapra, Jubinski, Tanaka, & Gershon, 2014).
Several factors may differentiate American Indian families from other American families. According to the Annie E. Casey Foundation (2012), 39% of American Indian adults are married (compared to 60% of Whites) and American Indians have higher divorce rates than African Americans, Whites, and Hispanics. Single-parent families account for 52% of American Indian households, compared to 24% for Whites, 66% for African Americans, and 41% for Hispanics or Latinos. American Indian families include extended family members. Selected community members are often included in their definition of family, reducing the expectations on nuclear family members (Day, 2014a;Red Horse, Martinez, & Day, 2001). Due to multi-generational trauma and lost parenting skills from residential schools and poverty (Day, 2014b), American Indians also have the highest rate of reported child abuse or neglect at 21.7 per 1000 children compared to 10.7 per 1000 for White children (Child Trends, 2015).
Additional research is needed due to the lack empirical studies on this critical topic with this understudied population. McLanahan's (2006) research has set the stage for some crucial discussions about the marriage agenda and the impact of marriage on couples and families from a Western legal, policy, and practice perspective. As a result, the purpose of this study was to examine these three areas with the American Indian sample of the Fragile Families and Child Wellbeing study (see Reichman et al., 2001) and to compare findings with those of McLanahan.

Method Data
Data from the Fragile Families and Child Wellbeing Study were used (Reichman et al., 2001). The data set sampled mothers with children born between 1998 and 2000 in 20 cities with populations greater than 200,000. Stratified sampling was used to identify 4,898 children and mothers (with an oversample of unmarried parents) who were interviewed within 48 hours of their child's hospital birth. Additional waves of data were collected at ages 1, 3, 5, 9, and 15. Because of the study's explicit interest in programs that promote marriage among the unmarried and the explicit focus of these programs on women (Manning, Trella, Lyons, & Du Toit, 2010), the sample was limited to women who were not married at baseline. The response rate at birth was 87% for unmarried mothers. An additional 11% of the sample was lost at age 1 and 3% more were lost at age 3. Thus, 73% of the sample remained by age 3. The outcomes for the current study were assessed at birth, age 1, and age 3.
Multiple imputation techniques were used to impute values on six of the dependent variables. Although imputation on dependent variables is not without controversy, it has become an increasingly accepted practice in recent years ( Van Ginkel, Linting, Rippe, & van der Voort, 2019;Von Hippel, 2007). The variable with the highest number of imputed values had approximately 25% missing. Some respondents did not answer any of the dependent variables. As a result, the values were not imputed for these individuals because their data may not have been missing at random. A final analytic sample included 3,152 women, including 154 women who self-identified as American Indian but not Hispanic or of multiple racial identities (Bendheim-Thoman Center for Research on Child Wellbeing, 2018).

Dependent Measures
Potential for participation measures. Following McLanahan's (2006) lead, various measures were used to tap the potential participation in marriage promotion programs of the 154 women who self-identified as American Indian. Specifically, eight dichotomous measures were indicated if program participation is likely. Willingness to participate in a marriage promotion program was dichotomized to indicate if a respondent was very likely to do so. Participants who responded that the chance of marrying the baby's father was greater than 50% were coded as having a high chance of marriage. Distrust of men was originally dichotomized in the data and respondents indicated whether they did or did not trust individuals of the opposite sex. Individuals who agreed or strongly agreed that marriage was the best relationship type for children were coded as believing marriage was the best relationship type.
Potential impact of marriage promotion programs. Based on the work of McLanahan (2006), the possible effects of marriage promotion programs by racial/ethnic group was addressed. The two explicit goals of these policies are to improve child wellbeing and reduce poverty. As such, four outcomes were examined to determine improvement after marriage (married by the time their child reached age 3, or wave 3 of the data). Marriage was examined through the lens of race/ethnicity to assess whether American Indians might fare better or worse on these outcomes after marriage. Three outcomes were about parenting from the mother's perspective: whether the mother engaged in cognitive stimulation (play games, read stories, and tell stories) during the week, whether the mother engaged in warm parenting (sang songs, hugged the child, and played with child) during the week, and whether the mother used corporal punishment (spanking). All three were dichotomous measures. A fourth dichotomous measure indicated if the family was below the poverty line when the child was age 3.

Likelihood of marriage.
Perhaps the simplest way to assess the potential impact of marriage promotion programs is to see if marriage is more likely due to being a parent. Thus, a dichotomous dependent variable indicating whether the respondent married the child's father by the time the child was age 3 was added. Again, differences were assessed by race/ethnicity on this variable.
Key independent variables. One focus was on racial and ethnic differences in the potential participation and impact of marriage promotion programs. More specifically, differences between American Indians and other racial/ethnic groups in the data were examined by the race/ethnicity the mother identified via a set of dichotomous variables. At the time of the birth of their children, mothers were asked to identify what "best describes" their race and/or ethnicity. The racial and ethnic groups included on the questionnaire were American Indian, non-Hispanic Black, non-Hispanic White, Latina, and other race/ethnicity (including multiracial).
Potential program participation characteristics are likely to correlate with one another and are important to consider as well. As a result, this study included willingness to participate in a program, high marriage chance, high gender distrust, marriage is best for kids, and father involvement in our models. Each variable was measured in the same manner as the dependent variables described above.

Control variables.
Various socio-demographic variables which may be associated with marital attitudes, the likelihood of marriage, gender distrust, father involvement, and parenting behaviors were controlled for. This study also controlled for father's involvement with a continuous measure formed from a series of dichotomous variables indicating whether the baby will have the father's last name, whether the father's name is on the birth certificate, whether the father helped to buy things for the baby, provided transportation, whether the father visited the baby in the hospital, and whether the mother wants the father involved in raising the child. These items were added together to form a scale (α= .80) which ranged from 0 to 6, with higher values indicating greater father involvement. McLanahan (2006) noted that how involved a father was with a child can affect one's willingness to marry as well as parenting styles.
Socioeconomic characteristics such as living below the poverty line, using public assistance, and employment in the prior year were measured with dichotomous variables. Relationship status at baseline was measured with dichotomous variables for cohabitation, visiting relationship, and other relationship types. Finally, educational attainment data were also gathered; response options included: less than a high school diploma, high school graduate, some college, and at least a college graduate.

Analytic Strategy
In order to compare our data with McLanahan's (2006) findings, we used the same analytic strategy as they employed, with the American Indian sample. Logistic regression models were used for the outcomes except in the case of father support, which was analyzed with Ordinary Least Squares (OLS) regression. Several variables were imputed, and the results were averaged over five imputations. The first set of models focused on potential participation in marriage promotion programs and on racial/ethnic differences. The second set of models addressed the potential impact of marriage promotion programs by considering differences by marital status at age three and racial and ethnic differences in this effect. As a result, marriage of the mother by the time the child reached age three and race/ethnicity were incorporated in these models. Finally, the likelihood of marriage by potential participation in marriage promotion programs and race/ethnicity was examined. The data assessed whether the respondent was married by the time her child was age 3, by willingness to participate in marriage promotion programs, high marriage chance, and high gender distrust. Thus, this examination focused on the possibility that a desire to participate will affect marriage entry and if this varies by race/ethnicity.

Descriptive Results
When examining mother's demographic data, approximately 5% of the sample was American Indian, 55% non-Hispanic Black, 14% White, 27% Latina, and 14% of another race. While racial and ethnic minorities are disproportionately represented in the sample, because the focus of research and the data set are on fragile families, this discrepancy makes sense. Approximately 43% of respondents were in poverty and used public assistance, though the correlation between these two variables was not very high (r= 0.31). Nearly 68% of respondents had worked in the prior year. Fifty-two percent of women had cohabited with their child's father, 35% were in a visiting relationship, and the remaining 13% were in other arrangements. The sample was disproportionately less educated than the general population with 39% not completing high school and an additional 34% having graduated high school. Only 3% had graduated from college. Table 1 provides proportional differences between American Indians and non-American Indians on a number of dependent measures. American Indians had a significantly higher proportion than non-American Indian counterparts who were willing to participate in marriage promotion programs, high distrust in men, lived below the poverty line, and married by age 3 of their child. A significantly lower proportion of American Indians were less likely to use corporal punishment (spanking) on children.  (Reichman et al., 2001); Note: *chi square test indicates difference between American Indians and non-American Indians at p<.05

Logistic Regression Models
The first set of models examined racial/ethnic differences on potential participation in marriage promotion programs. Regarding willingness to participate, the study found that only participants who self-reported white were significantly less likely to report a willingness to participate in marriage promotion programs than were American Indians (see Table 2). Interestingly, all racial/ethnic groups appeared to have a relatively high willingness to participate in marriage promotion programs. However, when examining high marriage chances, Whites reported they were more likely to marry the child's father as compared to American Indians. No other differences were observed in this area. When we looked at women's distrust of men, Whites were significantly less likely to distrust men than were American Indians. However, women from other racial/ethnic groups were significantly higher on distrust than were American Indian women. Finally, we examined potential participation in marriage promotion programs and its impact on children. White people, Latinas, and other race/ethnicities were all significantly more likely to see marriage as better for children than did American Indians (see Table 2).
The second set of models examined the potential impact of marriage programs and whether American Indians might fare better or worse on these outcomes after marriage. There were no differences in the effect of marriage across stimulation, warmth, or corporal punishment. However, Blacks and Latinas were less likely than were American Indians to engage in cognitive stimulation and Blacks were more likely to use corporal punishment. Results also showed that American Indians were much more likely to be in poverty after marriage than were any of the other groups (main effect vs. interactive effects). Thus, it appears that marriage does little to alleviate this challenge for American Indians.  (Reichman et al., 2001); * p<.05, ** p<.01, ***p <.001 (two-tailed tests) The third set of models addressed the likelihood of marriage by potential participation in marriage promotion programs and race/ethnicity. Respondents were assessed based on whether they were married by the time their child was age three, by willingness to participate, high marriage chance, and high gender distrust. Interestingly, results showed that American Indian women were more likely to marry than most other groups. Similarly, willingness to participate in marriage promotion programs was strongest for American Indians and somewhat weaker for other racial/ethnic groups. When looking at high marriage chances, this was the only significant difference for American Indian women. Despite higher odds for Latinas than American Indian women, distrust had little effect in this model.  (Reichman et al., 2001); 1 indicates American Indian as reference group; * p<.05, ** p<.01, ***p <.001 (two-tailed tests)

Discussion
Given that very little research has been done about American Indian fragile families and the impact of HMI, the purpose of this study was to add critical information to the profession's knowledge-base. This study examined three areas of marriage promotion among American Indians. The analysis of this data identified several components from each area in understanding the impact of marriage on American Indian families.
In examining data pertaining to the potential participation in marriage promotion programs, American Indians exhibited a high willingness to participate. American Indians were lower than other racial/ethnic groups in whether they saw marriage as beneficial for children. This goes against what others have found in the general population (McLanahan & Sandefur, 1994;Waite, 1995;Waite & Gallagher, 2000) and may suggest cultural differences in the definition of family and child-rearing practices. Prior research by White et al. (2006) suggests that American Indians have a more communal philosophy when it comes to raising children. This might suggest that, in addition to the mother and father, some American Indian people feel it is important for the well-being of children to involve their extended family, friends, and community. For example, according to Purzycki (2004), in the Lakota culture a child's mother's sister would be considered their second mother figure and his/her father's brother would be a second father figure. Therefore, American Indian women may not feel compelled to marry for the sake of the children.
In this study American Indian mothers considered their chances for marriage to be good, yet this enthusiasm for marriage did not come to fruition as often as hoped for, as indicated by a smaller percentage of couples actually getting married within 3 years of the baseline survey. This could indicate that while American Indian women see marriage as important, there are other factors that need to be considered if a marriage promotion program is to be successful in American Indian communities. Foremost among those factors is likely the impact poverty has on potential marriage relationships in some cases. Perhaps another way to measure this is child well-being, rather than just marriage, as an outcome measure.
This study found that American Indians were much more likely to be in poverty before and after marriage than any other group. While some researchers found that marriage has a positive impact on poverty in the general population (Lerman, 2002;Sawhill, 2014;Young, 1990), it appears from this study that marriage does little to alleviate these circumstances for American Indian families. Finally, American Indians were less likely to achieve higher levels of educational attainment, which could explain some of these results.
During the analysis, several limitations were encountered. The relatively small sample of American Indian participants limited the generalizability of findings. American Indians make up roughly 1.5% of the U. S. population and were almost 5% of this study. Because of this small sample size, it was difficult to detect significant differences with other racial/ethnic groups. Second, this study used data from large urban areas. These findings may have been different if respondents included rural and/or reservation areas. Taking the location and sample size factors into account, findings regarding American Indian fragile families and the marriage initiative variables may not be generalizable for all American Indian people. In addition, cultural backgrounds, individual tribal differences, and historical issues were not considered. Although marriage promotion programs may have attempted to provide culturally appropriate services, participants came from many different Indigenous cultures. It should also be noted that in the descriptive results, nine chi square tests were conducted using a .05 alpha, thus increasing the chance of a type I error. Finally, while the fragile families project is ongoing, data for this study occurred almost 18 years ago. Thus that age of the data needs to be taken into account when looking at current policy and practice implications.
Despite these limitations, this study offers several useful implications for practice with American Indian families. First, in this study, American Indian fragile families do not greatly differ from the general population in their willingness and desire for marriage and commitment to those relationships. Also, marriage promotion programs in American Indian communities appear to be needed and wanted. Here, cultural experts suggest that for marriage promotion programs to work, they must reflect the cultural practices of the individual American Indian communities, which is difficult in large urban areas (Nicotera, Walls, & Lucero, 2010). Therefore, they believe that successful programs are those in which tribal communities have input and are engaged in every aspect of the program (National Healthy Marriage Resource Center, 2012).
While fragile families of all races and ethnic groups are at risk, this study found that American Indian fragile families are particularly at risk because of the high incidence of poverty and its deleterious effects among American Indian children. Additionally, while poverty can impact mental health, culturally competent practitioners can have greater influence in helping American Indian fragile families by addressing poverty and the economic issues these families experience and by including extended family in assessment and interventions. Because the Fragile Families (Reichman et al., 2001) study only included American Indians living in cities with a population over 200,000, future studies in this field should address the rural American Indian fragile families and the direct impact of poverty in these areas as well (Saasa & Limb, 2017). Perhaps more meaningful data could be gathered when examining the relationship between marriage and child well-being.
Further, additional, continued, and focused funding in these areas is needed. As noted earlier, while specific funding for NAHMI seems to have not taken place recently, ACF continues to fund initiatives and grants for healthy marriage and fatherhood initiatives. It is recommended that this funding source, similar to Temporary Assistance to Needy Families (TANF), include line items and funding streams that target American Indians.

Conclusion
This study aimed to understand American Indian families and their unique needs for developing healthy marriage and family structures. By learning about the need for culturally competent, evidence-based marriage promotion programs and poverty programs addressing these needs, fragile families can be strengthened. With increased understanding, this study intended to promote additional research on American Indians that might help in providing specific steps for improving and strengthening these families. As the definition of marriage and families has been redefined by the US Supreme Court ruling of same-sex marriage, marriage promotion programs must also be redefined. American Indians broad definition of marriage among Two-Spirit, lesbian, bi-sexual, transgender, and gay partners need to be incorporated in these programs to strengthen families and support children.