Measurement Errors in Models of the Racial and Ethnic Test Scores Gaps

Hongyun Han, University of Wisconsin at Madison

The measurement errors in children’s test scores have not been well understood in the studies of the racial test score gap. This paper aims to address the cumulative racial gap in test scores by taking measurement errors into account. Using a newly available dataset (the Early Childhood Longitudinal Study), we use structural equation models to estimate the effect of family background on the racial gap in test scores among children from kindergarten to fifth grade. We find that the model with measurement error correction and causal chain fits the data best, suggesting that family background and test scores are latent variables, and minority children face cumulative disadvantages during elementary school. Furthermore, our findings show that latent family background has a stronger influence on achievement among African American fifth graders than their White counterparts, after correcting measurement errors in test scores and family background.