The unintended consequences of standardized test performance labeling on students' educational investment decisions

State-based standardized testing is used widely both for school accountability and as a graduation requirement. In this thesis, using a quasi-experimental regression-discontinuity design, I draw causal inferences about the unintended consequences of such policies for Massachusetts students. On these tests, Massachusetts divides a continuous measure of student performance (the test score) into several categories (i.e., "Advanced", "Proficient", "Needs Improvement", or "Failing"). I explore whether the classifications that students receive affect their subsequent educational attainments. I examine the effects of labeling in two different testing regimes: (1) the state's high-stakes exit-examination system, which features tests that students must pass in order to graduate from high school, and (2) state tests that are used to hold schools and districts accountable but carry no official consequences for students.

Across the board, I find that these labels matter: earning a better label improves the subsequent educational attainments of students who have essentially equal proficiency near the cutoff. In this thesis, I present evidence about performance labeling in three separate papers. First, I find that earning a more positive label on a low-stakes test affects the college-going decisions of urban, low-income students. Consistent with a Bayesian updating model, these effects are concentrated among students who report -- before taking the test -- that they do not plan to attend a four-year college. Second, I find that barely failing the 10th grade exit examination in mathematics increases the probability that students drop out and pursue the GED by approximately 3 percentage points. Third, under the state's exit-examination policy, students must pass tests in both mathematics and English language arts in order to graduate from high school. I present results from a new regression-discontinuity approach in which scores on multiple tests are used to assign students to different treatments. I find that barely failing both examinations reduces the probability that students on the margin of passing graduate from high school by 7.6 percentage points. These effects are concentrated among students scoring near the joint cut score. Taken together, my research shows that labeling matters and that the effects of performance labeling represent important unintended consequences of state-based testing policies.