The intrepid EdWeek bloggers Alyson Klein and Stephen Sawchuk caught a passage in President Obama’s latest budget proposal that would require states to, “develop a definition of ‘effective teacher’ that is based in significant part on student learning, and to put in place a system that links the academic achievement and growth of students to their teachers and school leaders.” This would complete the gradual progression from requiring a student-teacher data link in any education database that receives federal funding (the 2007 America COMPETES Act), the the State Fiscal Stabilization Fund’s assurance around teacher evaluation metrics and results, and the Race to the Top Grant precluding any state with a teacher-student firewall and insistence that teacher evaluations include student growth as a “significant” component.
This should not be seen as an entirely new policy, but one that’s been developing for at least the last three years.
The research behind using growth scores (the academic gains between two years of standardized tests) as a teacher evaluation metric is young but improving rapidly. In December, Dan Goldhaber and Michael Hansen released a paper for the Center for Reinventing Public Education looked at what would happen if teacher tenure decisions were based entirely off value-added scores. That is, after teachers had a couple years experience and were up for lifetime tenure, could a district accurately cut off the bottom 25 percent of teachers and improve their workforce?
The answer is a qualified yes.
First, the qualifications. The authors focused their paper by looking at value-added measures for teachers who could be accurately linked to students with at least one year of prior math and reading test scores. That limited the pool to almost 20,000 4th and 5th grade teachers observed for up to 11 years for a total of 63,000 observations. *For part of the analysis, they limited the pool even further, to teachers with at least five years of experience in the district, in order to observe the impact of two years of value-added data three years later (the first year after they’ve earned tenure). To enforce tenure rules according to these value-added limitations would obviously only affect a small sub-set of teachers: it would not affect any teacher teaching a grade lower than 4th or higher than 8th and would also exclude any teacher without obvious responsibilities for reading and math results.
These are important qualifications, but, if you can get past them, there are some interesting results. For one, the authors have some nifty graphs showing the growth of teacher effectiveness. As in prior studies, teachers make dramatic gains in effectiveness their first few years, but make only very modest improvements the rest of their career.
Two, differences in teacher effectiveness are wide and persistent across time. This is important, and it means that teachers with 3 years of experience have the same variation as those with 25. You are just as likely to find a great teacher with 3 years of experience as you are a terrible one with 25 years on the job, and vice versa. Other than the first few years, experience does not matter one iota.
Three, and this is the most important one, a teacher’s prior year value-added score is a better predictor of this year’s student’s success than any other teacher factor. That includes experience, whether the teacher holds a Master’s degree, their licensure score, the college they went to, or whether they were fully licensed or not. The value-added model was better than any of these other measures that are commonly used in selection and salary decisions.
Four, rejecting tenure for teachers in the lowest quartile of teacher effectiveness scores after three years on the job would improve a district’s overall teacher quality. In other words, this would not be a random exercise, but one that had some meaningful impact of student lives.
The paper comes with a handful of caveats–the actual improvements could be relatively small statistically, that value-added measures still have questions before being ready for high-stakes personnel decisions, and that any change would affect teacher behavior in ways we cannot predict–but it concludes by reminding readers that, “the results presented here indicate teacher effect estimates are far superior to observable teacher variables as predictors of student achievement, suggesting that these estimates are a reasonable metric to use as a factor in making substantive personnel decisions.”
*Update for clarity