Authentic, teacher report measures are a popular approach to assessing young children, but prior research has shown that teacher over- and under-estimates of early elementary children's skills can be associated with child characteristics like race and gender and are often associated with achievement gains over time. The current study extended this work by examining teacher over- and under-estimates of preschool students’ academic skills with the widely-used measure, Teaching Strategies GOLD. 1045 children (ages 46–61 months, mean = 54.4 months, SD = 3.7) from 89 publicly-funded preschool classrooms were rated by teachers using Teaching Strategies GOLD in the fall and were independently assessed on language, literacy, and math by trained assessors in the fall and spring. Results of multilevel path models indicated that the greatest discrepancies between teacher ratings and direct assessments were for children whose direct assessment scores were farther from the classroom mean; discrepancies were not significantly associated with children's race/ethnicity or gender. Discrepancies were associated with achievement gains from fall to spring, with children whose skills were overestimated making greater gains than achievement scores alone would have predicted. Indirect pathways indicated that teacher discrepancies partially transmitted the effects of higher and lower fall skills to children's spring outcomes. Results are discussed as they relate to the use of teacher report measures as formative assessments, and the potential sources of systematic error in these ratings. (author abstract)