Working Paper: NBER ID: w19243
Authors: Timothy N. Bond; Kevin Lang
Abstract: We address the ordinality of test scores by rescaling them by the average eventual educational attainment of students with a given test score in a given grade. We show that measurement error in test scores causes this approach to underestimate the black-white test score gap and use an instrumental variables procedure to adjust the gap. While the unadjusted gap grows rapidly in the early school years, particularly in reading, after correction for measurement error, the education-scaled gap is large, exceeds the actual black-white education gap and is roughly constant. Strikingly, the gap in all grades is largely explained by a small number of measures of socioeconomic background. We discuss the interpretation of scales tied to adult outcomes.
Keywords: black-white test score gap; measurement error; educational attainment; instrumental variables
JEL Codes: C18; I24; J15
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Measurement error in test scores (C83) | Underestimation of the black-white test score gap (J79) |
| Adjustment for measurement error using IV (C26) | Larger and consistent education-scaled black-white test score gap (I24) |
| Socioeconomic factors (P23) | Observed gaps in test scores (I24) |
| Measurement error (C20) | Apparent growth of the black-white test score gap from kindergarten through third grade (I24) |
| Adjusted estimates (C51) | No evidence of growth in the test gap over time (O40) |