We study the rise of prescription stimulant use for academic performance using a model with relative-performance externalities. Students compete for rank, where taking a stimulant improves effective performance but entails costs. Because payoffs from academic competition depend on others' performance, students' stimulant usage exhibits strategic complementarities: as more individuals use stimulants, others face stronger incentives to follow in order to avoid falling behind. We identify the model using prescription trends in the U.S. and our own survey evidence on students' incentives, beliefs, and perceived costs. The estimated strategic complementarities are large enough to generate multiple equilibria, including one with low use and another with very high use, in which around 90 percent of students take stimulants. We use the estimated model to evaluate policies that (i) reduce the salience of relative evaluation, (ii) limit the competitive advantage conferred by stimulants, particularly through academic accommodations, (iii) separate academic competition between stimulant users and non-users (i.e., separate curves), and (iv) tighten prescribing standards to reduce non-medical use. Our estimates show that separate curves yield rankings that most closely reflect underlying ability and are the policy that least discourages true ADHD patients from taking prescribed stimulants.

We compare two similar yet distinct early childhood education (ECE) programs while addressing the concerns associated with comparing them. We revisit data from a series of randomized early childhood education interventions to investigate the effects of ECE participation at ages 0 to 3 on a child's cognitive outcome. We document treatment effect heterogeneity in ECE programs by drawing on insights from the causal forest algorithm following Athey and Wager (2018). In particular, children accrue different effects from participation in ECE programs, and the populations differ across programs. Hence, a natural question is what the effect would be if one were randomized into another population. Using a forest built on federal program data and applying state program data to obtain treatment effect estimates for populations resembling each other, we consider treatment heterogeneity and differences between the two program characteristics. The results suggest that when designed and targeted well, ECE programs may be a very effective tool to improve the lives of the disadvantaged population.

We develop a new dynamic continuous-time model of optimal consumption and savings with endogenous liquidity constraints. In addition to exogenously imposed liquidity constraints, we endogenize a liquidity constraint over which individuals can maintain a living standard. We show that the endogenously determined liquidity constraint becomes tighter with a higher living standard. The optimal strategies with endogenous liquidity constraints are derived in closed form. We find a significant discontinuity and dramatic change in the effect of endogenous liquidity constraints on the optimal strategies, which in turn is determined by levels of current borrowing against future income. We show that consumption changes with respect to changes in wealth are greater with higher living standards when the amount of borrowing is large. However, this result can be reversed when the amount of borrowing is small. These findings are particularly important in addressing the interdependence of consumption and liquidity constraints to maintain a living standard in today's inflation crisis.