A central tenet of economics is that agents respond to incentives, yet the real-world effects of performance pay remain a puzzle. Simple models predict a direct increase in effort, while critics warn of a narrow targeted effort. This paper provides new evidence on this debate by studying a statewide bonus program that rewards Advanced Placement (AP) teachers a piece-rate for each student in their class passing the standardized test. We use marginal analysis to show that the bonus induces targeted effort and to explain the mechanisms driving variation in incentive strength, which we test empirically. Using student-course level data linked to teachers and a difference-in-differences design, we first document a significant average effect: the program increased student pass rates by 2.4 percentage points. We then decompose this average effect and report two key insights. First, contrary to the "teaching to the bubble" hypothesis, we find robust evidence of gains concentrated at the top of the AP score distribution. Second, we document heterogeneity across multiple dimensions, suggesting that the success of performance pay is jointly determined by the design of the incentive, the nature of the task, and the agent's baseline performance and workload.

Do policies that broaden educational access also foster success? We study this question by analyzing North Carolina’s universal Advanced Placement (AP) exam fee waiver policy. Using student-course level administrative data, we exploit within-student variation on a sample of students who took multiple AP courses to estimate the policy's effect on exam participation (access) and pass rates (success). We find that the fee waivers significantly increased exam participation but had no overall effect on the pass rate for these enrollees. This, however, masks a robust 3 percentage point increase in the pass rates among low-SES students. We also find imprecise but suggestive evidence of gains among underrepresented minorities (non-Asian and non-White). A complementary analysis, leveraging the full sample of AP courses, shows that fee waivers had the greatest impact in courses where predicted financial barriers to exam participation were highest, and that the policy's benefits far exceed its cost. Finally, our results help reconcile the seemingly disparate findings from prior work on AP exam funding.

Does public health insurance mitigate the functional consequences of adverse health shocks? I investigate this question by exploiting variation induced by Tennessee's 2005 Medicaid disenrollment, which abruptly terminated coverage for approximately 170,000 childless adults. Using Behavioral Risk Factor Surveillance System (BRFSS) data from 1997 to 2010 and a difference-in-differences design, I estimate Intent-to-Treat effects. I first document that the reform significantly increased the uninsured rate and led to a deterioration in observable health capital, marked by a 15% rise in severe obesity prevalence and the share of respondents reporting 'poor' health status. I then document a puzzle: while reported days of functional incapacitation increased by 27%, the underlying frequency of physical and mental symptom days remained statistically unchanged. I reconcile this divergence by showing that the "incapacitation ratio"—the probability that a symptom day leads to incapacitation—increased by 17%. These findings suggest that public health insurance functions as a technological buffer, enhancing resilience and preventing adverse health symptoms from disrupting daily functioning.

Does cutting Medicaid eligibility unlock employment? I revisit the influential study of Garthwaite, Gross, and Notowidigdo (2014) examining the 2005 TennCare disenrollment, which terminated public health insurance coverage for approximately 170,000 childless adults. Using the same dataset (CPS ASEC), I show that their "employment lock" evidence is mechanically driven by a steep employment decline in the control group, consistent with the demand-side fiscal contraction of the reform. I find no statistically detectable increase in new hiring or transition to employer-sponsored insurance among the disenrolled. Instead, the labor supply response reflects a "labor market scramble": gains are concentrated predominantly in informal self-employment within sales occupations. External validation using data from the Behavioral Risk Factor Surveillance System confirms a simultaneous rise in disability-related labor force exits. These findings suggest the policy did not spur formal labor market integration, but rather bifurcated the population into survival-driven gig work and labor force withdrawal.