Parental Gender Identity and Its Relationships to Racial Socialization Self-Efficacy, Beliefs, and Perceptions in Transracial Adoption: Using a Machine Learning Decision Tree Model

Background and Purpose:

A growing number of non-heterosexual parents have adopted a child of a different race. However, there is limited understanding about their needs in racial socialization practices, which are critical in transracial adoptees’ healthy ethnic and racial identity development. This study used a machine learning decision tree model to identify racial socialization self-efficacy, beliefs, and perceptions associated with parental gender identity among transracial adoptive parents.

Methods:

Data were drawn from the Modern Adoptive Families Study, which included 1,616 parents in foster care, private domestic adoption, or international adoption. Parents were mostly Caucasians (90%), mothers (87%), married (71%), and college educated (70%). Approximately 11% were lesbian mothers or gay fathers.

The 29-item Transracial Adoptive Parenting Scale (TAPS) was used to measure racial socialization beliefs and perceptions. Responses were on a 6-point Likert-type scale from 1 (strongly disagree) to 6 (strongly agree), with higher scores including stronger endorsement of racial socialization. The 7-item Racial Socialization Self-Efficacy Scale (RSSES) was used to assess parental feelings of self-efficacy in enacting racial socialization practices. Responses were on a 5-point Likert-type scale from 1 (not at all confident) to 5 (highly confident), with higher scores indicating a higher level of parental confidence in their ability to enact racial socialization practices. The TAPS and RSSE items and married status were examined to identify their associations with parents’ gender identity by a machine learning decision tree model, chi-square automatic interaction detection (CHAID).

Results:

Findings showed the intersectionality of adoptive parental gender identity with adoption type, married status, RSSES items, and TAPS items. The most significant predictor associated with adoptive parents’ gender identity was adoption type, followed by variables named “Helping my child feel a sense of belonging within a community of people from his or her birth culture makes me a better parent,” “I believe that discussions of racial differences with my child may do more harm than good,” and “Talk about my feelings about racism and discrimination with my child.” These predictors emerged as critical variables intersected with adoptive parents' gender identity upon repeated decision tree constructions. The overall accuracy, a percentage of correct predictions for the machine learning model, was 83%.

Conclusion and Implications:

The findings suggested that adoptive lesbian parents were more likely to utilize international adoption; the majority of them were not married but partnered. Among the foster care or private domestic adoptive parents, gay parents were less likely to discuss racial differences with their child. The machine learning approach to identifying adoptive parents’ needs could be a promising way to detect the intersection of gender identity, racial socialization self-efficacy, and racial socialization beliefs and practices. This approach eventually be a basis for developing a supporting model for adoptive parents who are either lesbian mothers or gay fathers. Further research is needed to explore the relationship between the identified intersection and social and emotional support needs of lesbian mothers and gay fathers. Such a relationship study would inform development of an efficient supporting model for children whose adoptive parents are not heterosexual.