Working papers results

This paper proposes a test of racial bias in capital sentencing based upon patterns of judicial errors in lower courts. We model the behavior of the trial court as minimizing a weighted sum of the probability of sentencing an innocent and that of letting a guilty defendant free. We define racial bias as a situation where the relative weight on the two types of errors is a function of defendant and/or victim race. The key prediction of the model is that if the court is unbiased, ex post the error rate should be independent of the combination of defendant and victim race. We test this prediction using an original dataset that contains the the race of the defendant and of the victim(s) for all capital appeals that became final between 1973 and 1995. We find robust evidence of bias against minority defendants who killed white victims: in Direct Appeal and Habeas Corpus the probability of error in these cases is 3 and 9 percentage points higher, respectively, than for minority defendants who killed minority victims.

-

We introduce and study finitely well-positioned sets, a class of asymptotically "narrow" sets that generalize the well-positioned sets recently investigated by Adly, Ernst and Thera in [1] and [3], as well as the plastering property of Krasnoselskii.

-

There is a large literature on social interactions and still little is known about the economic mechanisms leading to the high level of clustering in behavior that is so commonly observed in the data. In this paper we present a model in which agents are allowed to interact according to three distinct mechanisms, and we derive testable implications on the mean and the variance of the outcomes within and across groups. The empirical tests allow us to distinguish which mechanism(s) generates the observed patterns in the data. In our application we study the performance of undergraduate students and we find that social interactions take the form of mutual insurance. Such a result bears crucial policy implications for all those situations in which social interactions are important, from teamwork to class formation in education and co-authorship in academic research.

This paper contrasts measures of teacher effectiveness with the students' evaluations for the same teachers using administrative data from Bocconi University (Italy). The effectiveness measures are estimated by comparing the subsequent performance in follow-on coursework of students who are randomly assigned to teachers in each of their compulsory courses. We find that, even in a setting where the syllabuses are fixed, teachers still matter substantially. The average difference in subsequent performance between students who were assigned to the best and worst teachers (on the effectiveness scale) is approximately 43% of a standard deviation in the distribution of exam grades, corresponding to about 5.6% of the average grade. Additionally, we find that our measure of teacher effectiveness is negatively correlated with the students' evaluations of professors: in other words, teachers who are associated with better subsequent performance receive worst evaluations from their students. We rationalize these results with a simple model where teachers can either engage in real teaching or in teaching-to-the-test, the former requiring higher students' effort than the latter. Teaching-to-the-test guarantees high grades in the current course but does not improve future outcomes. Hence, if students are myopic and evaluate better teachers from which they derive higher utility in a static framework, the model is capable of predicting our empirical finding that good teachers receive bad evaluations, especially when teaching-to-the-test is very effective. Consistently with the predictions of the model, we also find that classes in which high skill students are over-represented produce evaluations that are less at odds with estimated teacher effectiveness.

This paper uses a structural, large dimensional factor model to evaluate the role of 'news' shocks (shocks with a delayed effect on productivity) in generating the business cycle. We find that (i) existing small-scale VECM models are affected by 'non-fundamentalness' and therefore fail to recover the correct shock and impulse response functions; (ii) news shocks have a limited role in explaining the business cycle; (iii) their effects are in line with what predicted by standard neoclassical theory; (iv) the bulk of business cycle fluctuations are explained by shocks unrelated to technology.

Starting with the seminal paper of Gilboa and Schmeidler (1989) an analogy between the maxmin approach of Decision Theory under Ambiguity and the minimax approach of Robust Statistics -- e.g. Huber and Strassen (1973) -- has been hinted at. The present paper formally clarifies this relation by showing the conditions under which the two approaches are actually equivalent.

-

Many separable dynamic incentive problems have primal recursive formulations in which utility promises serve as state variables. We associate families of dual recursive problems with these by selectively dualizing constraints. We make transparent the connections between recursive primal and dual approaches, relate value iteration under each and give conditions for it to be convergent to the true value function.

-

In real-life elections, vote-counting is often imperfect. We analyze the consequences of such imperfections in plurality and runoff rule voting games. We call a strategy profile a robust equilibrium if it is an equilibrium if the probability of a miscount is positive but small.

All robust equilibria of plurality voting games satisfy Duverger's Law: In any robust equilibrium, exactly two candidates receive a positive number of votes. Moreover, robust- ness (only) rules out a victory of the Condorcet loser.

All robust equilibria under runoff rule satisfy Duverger's Hypothesis: First round votes vare (almost always) dispersed over more than two alternatives. Robustness has strong implications for equilibrium outcomes under runoff rule: For large parts of the parameter space, the robust equilibrium outcome is unique.

This is a survey of some of the recent decision-theoretic literature involving beliefs that cannot be quantified by a Bayesian prior. We discuss historical, philosophical, and axiomatic foundations of the Bayesian model, as well as of several alternative models recently proposed. The definition and comparison of ambiguity aversion and the updating of non-Bayesian beliefs are briefly discussed. Finally, several applications are mentioned to illustrate the way that ambiguity (or "Knightian uncertainty") can change the way we think about economic problems.

-

The available empirical evidence suggests that the distribution of income and its composition play an important role in explaining tax noncompliance. We address the issue from a macroeconomic point of view, building a dynamic general equilibrium Bewley- Huggett-Aiyagari model that jointly endogenizes tax evasion and income heterogeneity. Our results showthat the model can successfully replicate the salient qualitative and quantitative features of U.S. data. In particular, the model replicates fairly well the shape of the cross-sectional distribution of misreporting rates over true income levels. Furthermore, we show that a switch from progressive to proportional taxation has important quantitative effects on noncompliance rates and tax revenues.