Association for Public Policy Analysis & Management

In recent years policymakers and accreditation agencies have strengthened outcomes-based accountability systems for teacher preparation programs (TPPs) and have encouraged programs to use evidence for continuous improvement. For many teacher educators, one response to this data-driven context has been support for the creation and widespread adoption of teacher candidate performance assessments. Relative to traditional licensure exams, these performance assessments (e.g. edTPA, PPAT, NOTE) offer a broader and more authentic measure of teacher candidate performance (Mitchell, Robinson, Plake, & Knowles, 2001). States and TPPs can use these performance assessments to determine candidates’ readiness to teach—linking candidate scores to high-stakes licensure decisions—and as a source of evidence for program improvement (Bastian, Lys, & Pan, 2017; Pecheone & Chung, 2006; Peck, Singer-Gabella, Sloan & Lin, 2014).

Currently, there is little evidence as to whether performance assessment scores go on to predict the effectiveness of early-career teachers (Bastian, Henry, Pan, & Lys, 2016; Goldhaber, Cowan, & Theobald, 2017). This evidence is crucial to the utility of performance assessments: if performance assessments do not predict graduate effectiveness, then states and TPPs should critically examine whether and how they act on performance assessment data. Given concerns that performance assessments may disproportionately prevent minority candidates from entering teaching, this evidence is particularly important to the diversity of the teacher workforce.

Therefore, to inform the accountability and improvement efforts of states and TPPs, I estimate whether performance assessment scores predict the effectiveness of early-career teachers. Specifically, I connect edTPA data from 2013-14 through 2015-16 for three TPPs in North Carolina (a total of 1,980 candidates) to administrative data from the North Carolina Department of Public Instruction (NCDPI) on teacher value-added and evaluation ratings. Following prior work on licensure exams, I perform two main analyses: (1) screening models that assess whether candidates who score above certain edTPA thresholds are more effective and (2) signaling models that assess whether higher edTPA scores predict teacher effectiveness. My preferred analyses control for a rich set of classroom and school characteristics and use a TPP fixed effect to account for differences in program practices and quality.

Results indicate that edTPA scores are predictive of early-career teacher performance. Screening analyses show that candidates meeting hypothetical cut-scores for licensure have higher value-added estimates and evaluation ratings. Signaling analyses indicate that the edTPA total score and Instructional domain (identified through factor analysis) predict higher value-added estimates; the total score and all three edTPA domains predict higher evaluation ratings. Further results show that edTPA scores are a valid screen and signal for the effectiveness of minority candidates and that the association between edTPA scores and teacher performance is strongest for graduates from the TPP that has used edTPA the longest and made scores consequential for licensure.

These analyses advance the nascent evidence-base on performance assessments and their utility for policymakers and teacher educators. Continued work should use data from more locations and years and further test how structural features—e.g. high-stakes consequences, TPP experience with the performance assessment—influence predictive validity results