Appendix C: Testing for Attrition Bias

This appendix presents a variety of additional tests that we conducted to examine the sensitivity of our estimates to potential selective attrition.  We first show the relationship between missing outcomes and observable characteristics for a variety of different outcome measures.  We next examines the sensitivity of the ninth grade reading score results to a variety of sample selection correction methodologies, including generic bounding, worst-case bounding and optimistic bounding.

Supplemental Authors: 
Julie Berry Cullen, Brian A. Jacob, Steven Levitt