Many educators who succeed at raising test scores also fail at keeping students fulfilled, new research suggests.

students in class
Is a good teacher one who makes students enjoy class the most or one who is strict and has high standards? And are those two types even at odds?

A new study that tries to quantify this phenomenon finds that on average, teachers who are good at raising test scores are worse at making kids happy in class.

"Teachers who are skilled at improving students' math achievement may do so in ways that make students less happy or less engaged in class," writes University of Maryland's David Blazar in the study, published in the peer-reviewed journal Education Finance and Policy.

The analysis doesn't suggest that test scores are a poor measure of teacher quality, but does highlight the different ways teachers may be effective.

The research uses data from four school districts across three states between 2010 and 2013; in one year, students were randomly assigned to fourth- and fifth-grade teachers, allowing researchers to study what effect different teachers had on students. Those students were also surveyed about their behavior, self-efficacy, and happiness in class.

A large body of past research has found that teachers have a meaningful impact on student test scores, and a number of more recent studies have found that teachers also impact other measures-sometimes called non-cognitive outcomes-such as behavior and attendance.

The latest study asks a few big questions.

First: Do teachers have an impact on students' attitudes and behavior, as measured by student surveys? Here, the answer is convincingly yes, consistent with the emerging research.

Second: Are the statistical estimates-often called value-added measures-of teacher impacts on test scores and non-cognitive skills accurate? The study examined this by comparing the statistical estimates to the results from from random assignment, and it found that the answer varies. Value-added measures are quite accurate for predicting test scores-an important finding in light of the charged debate on whether to judge teachers by these metrics. But it concludes that the statistical models are often biased for measuring impact on student attitudes, suggesting that attempting to evaluate teachers in this regard may be misguided.

Finally: Is a teacher's performance, measured by test scores, similar to performance according to other measures? This question is especially important because it's key for understanding how to think about teacher quality and how to evaluate it.

The study concludes there was only a weak relationship between test score performance and student behavior and feeling of efficacy in math. But when it came to student happiness, there was a moderate negative association-on average, greater test score gains meant less happy students.