Meaningful Differences in Discrepancy Patterns between Self-Report and Observed Parenting in a Sample of Homeless Families

Introduction: Previous studies utilizing multi-method measurement of parenting demonstrate that different reporters and measurement types yield less shared variance than expected (Dishion, et al., 1998). In particular, self-reported parenting practices are often only weakly correlated with observed parenting. While easily dismissed as error variance, these discrepancies between measurement sources may in fact be meaningful (Reyes, et al., 2011). The current study employed a person-centered approach to explore such discrepancies between measures of self-reported and observed parenting, with the aim of 1) identifying common profiles of discrepancies in a high-risk sample of parents and 2) examining if these discrepancy subgroups are associated with parent and child characteristics.

Methods: The study relied upon baseline data from a family-focused preventive intervention for youth conduct problems (Gewirtz, 2007). Participants were recently homeless families (138 mothers) with child between 6-12 years old (223 children, 51.1% boys). Latent class analysis (LCA) was conducted to find patterns in different measures of parenting: self-report parenting (6 Parent Relationship Questionnaire scales) and observed parenting (5 observational coded scales from a Family Interaction Task). Then multilevel SEM (MSEM) was used to examine parent predictors (support, depression, anxiety) and child behavior outcomes (composite score of observed behavior, teacher and parent report) associated with class membership.

Results: LCA results revealed a best-fitting model with four subgroups. The High/High (27.7% of sample) was high in both self-report and observed parenting; HighSR/LowO (32.9%) was high in self-report but low in observed parenting; LowSR/HighO (26.8%) was low in self-report but high in observed parenting; and Low/Low (12.7%) was low in self-report and low in observed parenting. MSEM results at the child level indicated children were more likely to have behavior problems in Low/Low than in High/High. At the parent level, Low/Low had less social support and more anxiety compared to High/High. HighSR/LowO reported higher social support and LowSR/HighO reported lower social support than High/High.

Conclusions: Results of this study support the existence of distinct subgroups characterized by patterns of discrepancies in two types of parenting measures. These subgroups where characterized by distinct levels of social support, anxiety, and child behavior problems. Our results encourage researchers to consider the methods for measuring parenting carefully due to the inconsistent information that differing strategies yield. Furthermore, ignoring such discrepancies (e.g., through the creation of composite scores across sources) may suppress meaningful differences among parents.