Assessment of pediatric behavior problems often requires rating scales from multiple reporters in different settings (eg, home and school); however, concordance between reporters may be low. Pediatricians must reconcile differences to inform treatment. We sought to examine characteristics predicting parent-teacher concordance on ratings of preschoolers' behavior problems. Methods: Data from 562 preschoolers were used from the Growing Healthy study, an obesity prevention trial in Head Start programs (2011–2015). Parents and teachers completed the Eyberg Child Behavior Inventory (ECBI)/Student Behavior Inventory (SBI) and the Social Competence and Behavior-Evaluation (SCBE). Outcome variables were: parent-teacher concordance (teacher minus parent score on each subscale of ECBI/SBI and SCBE); teacher reports problem behavior, parent does not (children rated in the top quintile of challenging behavior by teacher but not parent); and parent reports problem behavior, teacher does not (children rated in the top quintile of challenging behavior by parent but not teacher). Multiple linear and logistic regression models were created for each subscale outcome, including the following covariates: child sex, child race/ethnicity, parent age, parent education, family structure, parent depressive symptoms, and parenting self-efficacy, and time of school year. Results: Lower concordance was associated with child female sex, and child black or Hispanic race/ethnicity; parent older age, lower education, more depressive symptoms, and greater self-efficacy; and beginning of school year. Conclusions: Low parent-teacher concordance may reflect different perceptions of child behavior. Pediatricians could consider parent depressive symptoms, culture, and implicit bias when interpreting differences in behavior ratings by parents and teachers. (author abstract)