Shrinkage of Value-Added Estimates and Characteristics of Students with Hard-to-Predict Achievement Levels

Working Paper 17
Publisher: Cambridge, MA: Mathematica Policy Research
Apr 12, 2013
Mariesa Herrmann, Elias Walsh, Eric Isenberg, and Alexandra Resch
This working paper investigates how empirical Bayes shrinkage, an approach commonly used in implementing teacher accountability systems, affects the value-added estimates of teachers of students with hard-to-predict achievement levels, such as students who have low prior achievement and receive free lunch. Teachers of these students tend to have less precise value-added estimates than teachers of other types of students. Shrinkage increases their estimates’ precision and reduces the absolute value of their value-added estimates. However, this paper found shrinkage has no statistically significant effect on the relative probability that teachers of hard-to-predict students receive value-added estimates that fall in the extremes of the value-added distribution and, as a result, receive consequences in the accountability system.