Peer effects on the United States Supreme Court

Using data on essentially every U.S. Supreme Court decision since 1946, we estimate a model of peer effects on the Court. We estimate the impact of justice ideology and justice votes on the votes of their peers. To identify the peer effects, we use two instruments that generate plausibly exogenous variation in the peer group itself, or in the votes of peers. The first instrument utilizes the fact that the composition of the Court varies from case to case due to recusals or absences for health reasons. The second utilizes the fact that many justices previously sat on Federal Circuit Courts, and justices are generally much less likely to overturn decisions in cases sourced from their former “home” court. We find large peer effects. For example, we can use our model to predict the impact of replacing Justice Ginsburg with Justice Barrett. Under the the assumption that Justice Barrett's ideological position aligns closely with Justice Scalia, for whom she clerked, we predict that her influence on the Court will increase the Conservative vote propensity of the other justices by 4.7 percentage points. That translates into 0.38 extra conservative votes per case on top of the impact of her own vote. In general, we find indirect effects are large relative to the direct mechanical effect of a justice's own vote.

Inference on semiparametric multinomial response models

We explore inference on regression coefficients in semiparametric multinomial response models. We consider cross‐sectional, and both static and dynamic panel settings where we focus throughout on inference under sufficient conditions for point identification. The approach to identification uses a matching insight throughout all three models coupled with variation in regressors: with cross‐section data, we match across individuals while with panel data, we match within individuals over time. Across models, we relax the Indpendence of Irrelevant Alternatives (or IIA assumption, see McFadden (1974)) and allow for arbitrary correlation in the unobservables that determine utility of various alternatives. For the cross‐sectional model, estimation is based on a localized rank objective function, analogous to that used in Abrevaya, Hausman, and Khan (2010), and presents a generalization of existing approaches. In panel data settings, rates of convergence are shown to exhibit a curse of dimensionality in the number of alternatives. The results for the dynamic panel data model generalize the work of Honoré and Kyriazidou (2000) to cover the semiparametric multinomial case. A simulation study establishes adequate finite sample properties of our new procedures. We apply our estimators to a scanner panel data set.

Average crossing time: An alternative characterization of mean aversion and reversion

This study compares and contrasts the multiple characterizations of mean reversion in financial time series as regards the restrictions they imply. This is accomplished by translating them into statements about an alternative measure, the “Average Crossing Time” or ACT. We argue that the ACT measure, per se, provides not only a useful benchmark for the degree of mean reversion/aversion, but also an intuitive, and easily quantified sense of one time series being “more strongly mean‐reverting/averting” than another. We conclude our discussion by deriving the ACT measure for a wide class of stochastic processes and detailing its statistical characteristics. Our analysis is principally undertaken within a class of well‐understood production based asset pricing models.

Making summer matter: The impact of youth employment on academic performance

This paper examines New York City's Summer Youth Employment Program (SYEP). SYEP provides jobs to youth ages 14–24, and due to high demand for summer jobs, allocates slots through a random lottery system. We match student‐level data from the SYEP program with educational records from the NYC Department of Education and use the random lottery to estimate the effects of SYEP participation on a number of academic outcomes, including test taking and performance. We find that SYEP participation has positive impacts on student academic outcomes, and these effects are particularly large for students who participate in SYEP multiple times.

Panel experiments and dynamic causal effects: A finite population perspective

In panel experiments, we randomly assign units to different interventions, measuring their outcomes, and repeating the procedure in several periods. Using the potential outcomes framework, we define finite population dynamic causal effects that capture the relative effectiveness of alternative treatment paths. For a rich class of dynamic causal effects, we provide a nonparametric estimator that is unbiased over the randomization distribution and derive its finite population limiting distribution as either the sample size or the duration of the experiment increases. We develop two methods for inference: a conservative test for weak null hypotheses and an exact randomization test for sharp null hypotheses. We further analyze the finite population probability limit of linear fixed effects estimators. These commonly‐used estimators do not recover a causally interpretable estimand if there are dynamic causal effects and serial correlation in the assignments, highlighting the value of our proposed estimator.

Imposing equilibrium restrictions in the estimation of dynamic discrete games

Imposing equilibrium restrictions provides substantial gains in the estimation of dynamic discrete games. Estimation algorithms imposing these restrictions have different merits and limitations. Algorithms that guarantee local convergence typically require the approximation of high‐dimensional Jacobians. Alternatively, the Nested Pseudo‐Likelihood (NPL) algorithm is a fixed‐point iterative procedure, which avoids the computation of these matrices, but—in games—may fail to converge to the consistent NPL estimator. In order to better capture the effect of iterating the NPL algorithm in finite samples, we study the asymptotic properties of this algorithm for data generating processes that are in a neighborhood of the NPL fixed‐point stability threshold. We find that there are always samples for which the algorithm fails to converge, and this introduces a selection bias. We also propose a spectral algorithm to compute the NPL estimator. This algorithm satisfies local convergence and avoids the approximation of Jacobian matrices. We present simulation evidence and an empirical application illustrating our theoretical results and the good properties of the spectral algorithm.

A notion of prominence for games with natural‐language labels

We study games with natural‐language labels (i.e., strategic problems where options are denoted by words), for which we propose and test a measurable characterization of prominence. We assume that—ceteris paribus—players find particularly prominent those strategies that are denoted by words more frequently used in their everyday language. To operationalize this assumption, we suggest that the prominence of a strategy‐label is correlated with its frequency of occurrence in large text corpora, such as the Google Books corpus (“n‐gram” frequency). In testing for the strategic use of word frequency, we consider experimental games with different incentive structures (such as incentives to and not to coordinate), as well as subjects from different cultural/linguistic backgrounds. Our data show that frequently‐mentioned labels are more (less) likely to be selected when there are incentives to match (mismatch) others. Furthermore, varying one's knowledge of the others' country of residence significantly affects one's reliance on word frequency. Overall, the data show that individuals play strategies that fulfill our characterization of prominence in a (boundedly) rational manner.

Controlling for presentation effects in choice

Experimenters make theoretically irrelevant decisions concerning user interfaces and ordering or labeling of options. Reanalyzing dictator games, I first show that such decisions may drastically affect comparative statics and cause results to appear contradictory across experiments. This obstructs model testing, preference analyses, and policy predictions. I then propose a simple model of choice incorporating both presentation effects and stochastic errors, and test the model by reanalyzing the dictator game experiments. Controlling for presentation effects, preference estimates become consistent across experiments and predictive out‐of‐sample. This highlights both the necessity and the possibility to control for presentation in economic experiments.

Linear regression with many controls of limited explanatory power

We consider inference about a scalar coefficient in a linear regression model. One previously considered approach to dealing with many controls imposes sparsity, that is, it is assumed known that nearly all control coefficients are (very nearly) zero. We instead impose a bound on the quadratic mean of the controls' effect on the dependent variable, which also has an interpretation as an R 2‐type bound on the explanatory power of the controls. We develop a simple inference procedure that exploits this additional information in general heteroskedastic models. We study its asymptotic efficiency properties and compare it to a sparsity‐based approach in a Monte Carlo study. The method is illustrated in three empirical applications.