While many education policies target test scores as a contemporaneous measure of student learning, a common concern is that these policies may generate higher test scores in a way that fails to translate to more important student outcomes in the long run. I use administrative data from North Carolina and two regression discontinuity designs to estimate the impact of school accountability pressure under No Child Left Behind on elementary students' test scores and their long-run outcomes at the end of high school. I find modest positive effects on elementary test scores and a significant increase in SAT scores years later. There is some evidence for a small increase in high school GPA, mixed evidence for an increase in students intending to attend a 4-year instead of a 2-year college, and no effect on high school graduation or intention to attend any college. Further evidence suggests the effect on SAT scores may be explained by persistent test-score effects in years after accountability exposure. Altogether, these results lend support to a mixed story for No Child Left Behind: while accountability pressure led to a long-run increase in skills captured by tests, these learning gains were not strong or broad enough to yield meaningful improvements in other long-run outcomes like educational attainment.

Does lower teacher quality contribute to short-term and long-term outcomes for disadvantaged students? We leverage transfers of elementary teachers across schools in North Carolina to measure differences in teachers' effects on contemporaneous and future test scores according to students' socio-economic characteristics. We quantify the importance of these differences to account for the observed test score gaps between disadvantaged and advantaged students. Variation in teacher quality accounts for 3 percent of the total variation in contemporaneous test scores. We also find that teacher quality accounts for similar proportions when we consider variability in test scores taken two and three years after. Our estimates are robust to bias-correction methods that account for limited mobility bias.

This paper evaluates the test score effects of individual teacher performance pay schemes implemented in a number of high-need schools in North Carolina. Since performance bonuses were paid to teachers with value-added above a threshold toward the top of the district-wide distribution, I evaluate whether this policy generates larger incentives for teachers with higher probability of attaining the bonus. I find evidence for the opposite: those expected to be further away from the threshold increased their value-added, though the performance incentives did not have a significant overall effect. I show that the single-year value-added estimates are quite noisy, which likely was a reason for this. I also find that almost all of the teachers in these high-need schools were predicted to have value-added below the performance threshold. Both of these factors could explain the lack of overall incentive effectiveness. One possibility is that lower value-added teachers have more scope to improve effort.