Knowledge that Transforms

Abstract We show that teachers’ value added (VA) depends on the VA of the teachers who preceded them. To do so, we use administrative data from North Carolina and find that a one-standard deviation increase in last grade’s mean teacher VA causes a 0.08σ decrease in teacher VA. Controlling for prior teacher assignment eliminates this bias. Under a benchmark policy that releases teachers in the bottom 5% of the VA distribution, 32% of teachers are incorrectly released using traditional techniques. Our results highlight the importance of incorporating dynamic features of education production into the estimation of teacher quality.

Abstract Close elections are important for many reasons, including that consequent election disputes can weaken democratic legitimacy and risk political violence. We quantify the probability of close outcomes in U.S. presidential races with novel applications of empirical election models from several sources. We show that razor-thin margins are very likely under the Electoral College (EC). And we establish that the EC causes this closeness: It would not occur under any plausibly comparable popular vote system. The tendency of the EC to generate close elections is found today and throughout U.S. presidential voting history.

Abstract This paper studies imperfectly perceived private information in insurance markets when contracts endogenously respond. Equilibrium contracts, pooling, and welfare depend on the joint distribution of risk and misperception. In the Health and Retirement Study (HRS), I show that misperceptions typically covary with (medical, long-term care, disability, and mortality) risk type: high types underperceive their risk and low types overperceive. I develop a general model and algorithm to estimate the equilibrium contracts, pooling, and welfare impact of misperceptions that is applicable in many settings. I offer suggestive evidence from U.S. annuity markets that contracts are distorted due to misperceptions, with welfare likely increasing.

Abstract This paper develops a semiparametric sieve approach to estimate impulse response functions of nonlinear time series models within a broad class of structural autoregressive specifications. A two-step procedure flexibly captures nonlinearities without imposing fixed parametric forms. We establish uniform estimation guarantees and propose an iterative algorithm that makes impulse response computation straightforward. Simulation results show robustness to misspecification with only modest efficiency losses. In an application to U.S. monetary policy, we find larger GDP responses to interest rate hikes than in linear models. We also examine oil supply news shocks of varying magnitudes to assess limitations when analyzing large shocks.

Abstract How do trade shocks affect welfare and inequality when human capital is endogenous? Using India’s IT boom and internal migration data, I document that IT employment and engineering enrollment increased with exports, especially in regions with larger college-age populations. I develop a spatial model featuring higher education choice and differential migration costs for college versus work. The IT boom increased welfare by 2.39%, but without educational mobility, gains would be 25% lower and regional inequality 1.5 times larger. Removing endogenous education further reduces gains by over one-third. Education policies like national scholarships could substantially reduce regional inequality from trade shocks.

Abstract We propose a restrictiveness measure for economic models based on how well they fit predefined synthetic data. This measure, together with a measure for how well the model fits real data, outlines a Pareto frontier, where models that rule out more regularities, yet capture the regularities that are present in real data, are preferred. To illustrate our approach, we evaluate the restrictiveness of models in two laboratory settings—certainty equivalents and initial play—and one field setting—takeup of microfinance in Indian villages. The restrictiveness measure reveals insights about each, including that some economic models with only a few parameters are very flexible.

Abstract Why do some individuals evade taxes while others do not? We study this question using administrative tax records from Uruguay linked to a tailored survey of taxpayers. Using third-party reports, we measure individual income under-reporting as an indicator of evasion. We then examine how three factors predict who evades: social preferences (e.g., honesty measured through incentivized laboratory games), peers (e.g., the behavior of current and former coworkers), and economic factors (e.g., the marginal tax rate). We find that social preferences have little power to predict evasion, while economic factors matter more and peer behavior is the strongest predictor.

Abstract We empirically assess whether hindsight bias affects citizens’ evaluation of their political actors. Using an incentivized elicitation technique, we demonstrate that people systematically misremember their past policy preferences regarding how to best fight the COVID-19 pandemic. At the peak of the first wave in the United States, the average respondent mistakenly believed that they supported significantly stricter restrictions at the onset of the first wave than they actually did. Exogenous variation in the extent of hindsight bias, induced through a randomized survey experiment, indicates that hindsight bias has a negative causal impact on the change in trust in government.

Abstract This paper theoretically and empirically examines the role of information in pharmaceutical detailing (promotional interactions between drug representatives and physicians). We start with a theoretical framework in which pharmaceutical firms target detailing visits to physicians who learn about drug quality and prescribe accordingly. We derive several predictions about the role of information in these visits, which we then test empirically using prescriptions and pharmaceutical detailing visit data. We find limited empirical support of learning as the dominant mechanism, though cannot rule it out completely. We conclude with discussing alternative models that may be more consistent with the observed empirical patterns.

Abstract Standard estimators in information provision experiments place more weight on individuals who update their beliefs more in response to new information. This paper shows that, in practice, these individuals who update the most have the weakest causal e!ects of beliefs on outcomes. Standard estimators therefore understate these causal e!ects. I propose an alternative local least squares (LLS) estimator that recovers a representative unweighted average e!ect in a broad class of learning rate models that generalize Bayesian updating. In five of six recent studies, estimates of the e!ects of beliefs on outcomes increase. In two, they more than double.