Working Paper: NBER ID: w17960
Authors: Timothy N. Bond; Kevin Lang
Abstract: Although both economists and psychometricians typically treat them as interval scales, test scores are reported using ordinal scales. Using the Early Childhood Longitudinal Study and the Children of the National Longitudinal Survey, we examine the effect of order-preserving scale transformations on the evolution of the black-white reading test score gap from kindergarten entry through third grade. Plausible transformations reverse the growth of the gap in the CNLSY and greatly mitigate it in the ECLS-K during early school years. All growth from entry through first grade and a nontrivial proportion from first to third grade probably reflects scaling decisions.
Keywords: Black-White test score gap; scaling decisions; early childhood education; test scores
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 |
|---|---|
| scaling decisions (D79) | Black-White test score gap (I24) |
| certain transformations (C69) | gap appears to shrink from kindergarten to third grade (I24) |
| other transformations (C69) | significant increase in the gap (I24) |
| scaling decisions (D79) | no growth in the gap (O40) |
| scaling decisions (D79) | doubling of estimated increase in the gap (F62) |
| scaling decisions (D79) | substantial gap of nearly half a standard deviation (I24) |