Expert Commentary

“Teacher value-added”: Effects on students’ test scores, college attendance and earnings

2013 Harvard study seeking to quantify the short- and long-term impact of fourth- through eighth-grade teachers by observing them in multiple classrooms.

High-school teacher (iStock)

Intuitively, we believe that good teachers play a positive role in the lives of their students, but the debate over how best to measure teacher performance is a lively and ongoing one. Many researchers have concluded that merely looking at observable characteristics, such as a teacher’s years of experience, level of education, or how their students perform on standardized tests is problematic when it comes to making strong claims. In recent years, one attempt to tackle this measurement challenge has been to consider the concept of “teacher value-added” (VA) — an effort to measure the contributions of teachers to student achievement by holding other relevant variables (income, family, etc.) constant. A teacher’s VA is determined by his or her students’ average score gain on an achievement test at the end of the school year, controlling for other influencing characteristics such as prior academic performance.

Recent research using this VA framework indicates that teachers can indeed have long-term positive or negative effects on their students starting as early as elementary school. Building off of a 2011 study from Harvard and Columbia universities, Harvard University economist Gary Chamberlain further explores this concept as he seeks to quantify the short- and long-term impact of hundreds of fourth- through eighth-grade teachers by observing them in multiple classrooms. In his paper, “Predictive Effects of Teachers and Schools on Test Scores, College Attendance and Earnings,” Chamberlain considers three student outcomes: average score on a math or reading test given near the end of the school year, fraction of the students in a given class attending college at age 20, and average income at age 28. By holding constant additional relevant variables, he is able to estimate predictive effects of teacher quality.

Key findings from the study include:

  • A one standard deviation (SD) increase in the teacher factor implied a predicted increase in earnings at the age of 28 of $186. (However, Chamberlain notes this figure lacks precision.)
  • When controlling for parental characteristics, a one SD increase in the teacher factor (based on test scores) had a predictive effect on college attendance of 0.13 percentage points.
  • About 38% of the students examined in the study went to college. By isolating the variable of teacher quality, the study found students were nearly 1% more likely to go to college if they had a particular teacher of higher quality: “The teacher effect of 0.99 percentage points could reflect skills that are relevant for college attendance but are not measured by the test scores. These skills could be some combination of skills students bring to the class … and skills developed during the class, in part due to the contribution of the teacher.”

Chamberlain notes, as others have, that the value-added model is not flawless. In his study, for instance, although he added several important controls, he lacked data for other variables that could affect student outcomes, such as parental education. This omitted variable bias could confound the causal effects attributed to the VA measurement of teacher quality. Indeed, other researchers have noted that even the most rigorous randomized control trials can include significant errors.

It is also worth noting that even if studies like this determine that teacher quality can have long term effects on students, questions about how to translate this into effective education policy remain. Charter schools or increased funding for private schools? Teach for America — yes or no? Find some current research on these and other topics here.

Keywords: higher education

About The Author