Poster Paper: What You See is What You Get? Examining the Relationship Between Teachers' Observation Scores and the Fade-out of Teacher Value-added Measures

Saturday, November 4, 2017
Regency Ballroom (Hyatt Regency Chicago)

*Names in bold indicate Presenter

Olivia L. Chi, Harvard University


The research literature consistently demonstrates that the impact of teacher value-added fades over time. In other words, the fraction of a teacher’s impact on student test scores that persists into future test scores decreases in the following school years. There exist many hypotheses on the mechanisms that cause fade-out. One of these hypotheses is that value-added measures may be capturing some transitory gains, such as those that may come from “teaching to the test.” This hypothesis is troubling, as it raises concerns that value-added measures may overstate the amount of permanent knowledge that teachers impart on their students (Jacob et al. 2010).

This paper attempts to shed light on this hypothesis by investigating the relationship between teachers’ instructional quality (as measured by classroom observation scores), the persistence of teacher value-added, and students’ longer-run test scores. Using data from a small state, this paper addresses the following research questions: 1) Do teachers’ classroom observation scores explain variation in the persistence of teacher value-added? 2) Do teachers’ classroom observation scores explain variation in students’ longer-run test scores (that is not captured by short-run value-added measures)?

Research suggests that teachers’ classroom observation scores capture different aspects of teachers’ performance than do value-added measures. Observation scores could potentially capture aspects of teaching that contribute to persistence of teacher-imparted knowledge, yet are not captured by short-run value-added measures. If so, teachers with higher observation scores may have higher persistence of teacher effects.

Furthermore, observation scores could potentially capture aspects of teaching that are not reflected in short-run test scores or short-run value-added measures, but eventually lead to higher longer-run test scores. If so, teachers’ observation scores may explain variation in students’ longer-run test scores that is not captured by short-run value-added measures.

Understanding how to improve methods that estimate and incorporate teachers’ lasting impacts may improve alignment between teacher instruction and desired outcomes for long-term student learning. Many teacher evaluation systems use multiple measures of teacher performance, including student test scores and classroom observations. Investigating the additional information provided by observation scores and whether they provide insight into teachers’ impact on longer-term knowledge can inform policy decisions regarding the relative weights of these two components within a teacher evaluation system.

In this paper, I attempt to contribute to a larger body of research that seeks to improve measurement of teacher quality by incorporating teachers’ lasting impacts on students’ test scores. The analyses in this paper provide some evidence to suggest that, in math, teachers’ observation scores explain variation in students’ longer-run test scores that is not captured by the persistence of teacher value-added.