Screen Twice, Cut Once: Assessing the Predictive Validity of Teacher Selection Tools

Evidence suggests that teacher hiring in public schools is ad hoc and often fails to result in good selection among applicants. Some districts use structured selection instruments in the hiring process, but we know little about the efficacy of such tools. In this paper, we evaluate the ability of applicant selection tools used by the Spokane Public Schools to predict three outcomes: measures of teachers' value-added contributions to student learning, teacher absence behavior, and attrition rates. We observe all applicants to the district and are therefore able to estimate sample selection-corrected models, using random tally errors in selection instruments and differences in the quality of competition across job postings. These two factors influence the probability of being hired by Spokane Public Schools but are unrelated to measures of teacher performance. We find that the screening instruments predict teacher value added in student achievement and teacher attrition but not teacher absences. A onestandard- deviation increase in screening scores is associated with an increase of between 0.03 and 0.07 standard deviations in student achievement and a decrease in teacher attrition of 2.5 percentage points.