Do teacher expectations matter? In particular, can teacher expectations influence student educational outcomes? Yes, says a new IZA paper authored by researchers at Johns Hopkins University’s Department of Economics and American University’s School of Public Affairs.
Scholars, pundits, educators, and policy makers have long speculated whether, and to what extent, teacher expectations create self-fulfilling prophecies that perpetuate black-white gaps in educational achievement and attainment. However, with the exception of experimental studies of the “Pygmalion Effect” in which teachers’ beliefs about student aptitude were experimentally manipulated, credible evidence on the causal relationship between teacher expectations and long-run student outcomes is lacking for two general reasons:
- Externally valid data on teachers’ subjective expectations are rarely collected.
- Observational analyses of the relationship between teachers’ expectations and student outcomes are plagued by a thorny endogeneity problem. It is often the case that teacher expectations are correlated to students’ outcomes. This could arise for two reasons. On the one hand, teacher expectations may simply be accurate forecasts, in which case teacher expectations do not directly affect student outcomes. On the other hand, incorrect (i.e., biased) teacher expectations could directly affect student outcomes by initiating self-fulfilling prophecies in which investments made in or by students are altered, thereby leading to outcomes that resemble teachers’ initially incorrect beliefs.
In their IZA paper, Nicholas Papageorge, Seth Gershenson, and Kyungmin Kang address these challenges head-on by developing and fitting a structural econometric model of the education and teacher-expectation production functions to unique, nationally representative survey data of the 2002 cohort of U.S. tenth graders (and their teachers).
The empirical strategy exploits a unique feature of these data: two teachers provide their educational expectations for each student. Ensuing disagreements between teachers about particular students together with information on students’ grades and test scores allows for the identification of students’ objective probabilities of completing college.
Comparing these objective probabilities to teachers’ stated expectations identifies the bias inherent in teachers’ expectations, which are allowed to enter the education production function as inputs that can directly affect the likelihood of college completion via self-fulfilling prophecies.
The analysis yields four main results:
1. Teacher expectations matter
Teacher expectations have causal impacts on the probability that students complete a four-year college degree. The elasticity of the likelihood of college completion with respect to teachers’ expectations is about 0.12, which is consistent with the impact of other K-12 educational inputs on college completion (e.g., class size). This result provides insights into the mechanisms through which the long-run effects of K-12 teachers might operate.
2. Over-optimism may help
Assessing which teachers are more accurate is not necessarily the most salient question when investigating the black-white gap in teacher expectations. This is because on average, all teachers are overly optimistic about students’ ability to complete four-year college degrees. However, the degree of over-optimism is significantly larger for white students than for black students, especially when black students are evaluated by white teachers.
This answers the unresolved question in previous research by the authors, which finds that white and black teachers systematically disagree about particular black students’ potential to graduate college, but does not identify “which teachers are wrong.” More generally, this result highlights an important nuance that is frequently overlooked in discussions of biased beliefs: unbiased (i.e., accurate) beliefs can be counterproductive if there are positive returns to optimism or if there are socio-demographic gaps in the degree of over-optimism, both of which are true in these data.
3. Variations in passiveness
Teachers frequently disagree for transitory, arguably unimportant reasons. For example, when a student behaves passively in English class but not in math class, this affects the English teacher’s expectation but not the math teacher’s expectation. This type of within-student, within-semester variation in passiveness represents transitory, arguably random departures from the student’s baseline (steady-state) level of passiveness.
While the baseline level of passiveness likely does affect educational attainment, daily departures from it should not. This insight is central to the identification strategy employed in both the structural and reduced-form analyses.
4. Hiring more black teachers
Finally, the model is used to conduct two counterfactual, but realistic, policy simulations: hiring more black teachers and de-biasing white teachers. The efficacy of these policies varies with students’ capacities for post-secondary educational success upon entering the tenth grade. Specifically, for black students with low objective probabilities of college completion upon entering the tenth grade, hiring more black teachers is more effective than de-biasing white teachers.
However, neither policy has a particularly large effect on college completion, suggesting that investments earlier in the lifecycle are crucial, which is consistent with prior literature. Nonetheless, the study indicates that achievement gaps can be addressed to some degree even when students are older, albeit with varying degrees of effectiveness. For black students with relatively high (above the median) objective probabilities of college completion, both policies are about equally effective.
Thus, the most effective mix of policies implemented for high school students depends on whether college completion is fairly likely at the time students enter the tenth grade. This information, paired with relative costs, could inform decision-makers as to the optimal mixture of policies.