Association for Public Policy Analysis & Management

Research demonstrates that highly effective teachers substantially increase student achievement during the year in which they teach (e.g., Nye, Konstantopoulos, & Hedges, 2004; Rivkin, Hanushek, & Kain, 2005; Rockoff, 2004). More recent evidence establishes that these highly effective teachers also positively affect adult outcomes such as students’ lifetime earnings (Chetty, Friedman, & Rockoff, 2014). Put simply, teachers impact student achievement in the short-term, which translates to impacts on adult outcomes in the long-term. What is less known, however, is the connection between the short- and long-term effects—the extent to which teacher effects persist and contribute to student learning in the years following direct instruction. We call these medium-term effects.

In the value-added literature, the persistence of teacher effects is characterized as a process by which teachers’ short-term effects continue to impact the learning gains of students in the years following direct instruction (McCaffrey, Lockwood, Koretz, Louis, & Hamilton, 2004). While it is often assumed that these effects diminish over time, a testable assumption, there is an often-untested assumption that teachers who produce short-term learning gains are also highly effective for producing medium-term gains (Rothstein, 2010).

In this paper, we estimate short-term and medium-term teacher effects on standardized test scores to provide insight into the following questions: (Q1) What is the magnitude of teacher effects on test scores in the medium term and does the magnitude vary by grade level?; (Q2) To what extent are teachers who improve students’ short-term learning the same teachers who improve students’ medium-term learning?; (Q3) Do characteristics such as years of experience or principal observation ratings predict the variability in medium-term effects?

To answer these questions, we combine student and teacher-level administrative files from the Tennessee Department of Education to follow students and their teachers over time. Our data include student test scores in grades 3 to 8 from 2006-07 to the present that can be matched directly to teachers. We place students into cohorts and follow them for six years.

Our preliminary results suggest there is meaningful variation in medium-term teacher effects, between 0.13 and 0.20 standard deviations (Q1). We also find that teachers most effective in the short term are not necessarily the most effective for medium-term learning (Q2); correlations between short- and medium-term effects range from 0.4 to 0.5, consistent with Rothstein (2010). We continue to identify which teacher characteristics explain variability in medium-term effects using our state-level administrative data from Tennessee (Q3).

Overall, students are likely to retain or remember information from one year to the next because of the important role that teachers have in imparting knowledge in a classroom. Our results will enable school leaders to gain novel insights into the ways teachers contribute to a cumulative learning process, one which encompasses short-term and longer-term knowledge. While Chetty et al. (2014) find that highly effective short-term teachers improve adult outcomes, policymakers do not know if medium-term effects are the primary driver of their results. The results from this study serve as a first step to inform this discussion.