Research

Working Papers

Insurance Pricing, Distortions, and Moral Hazard: Quasi-Experimental Evidence from Deposit Insurance, Forthcoming, Journal of Financial and Quantitative Analysis

Pricing is integral to insurance design, directly influencing firm behavior and moral hazard, though its effects are insufficiently understood. I study a quasi experiment in which deposit insurance premiums were changed for U.S. banks with staggered timing, generating differentials between banks in both levels and risk-based “steepness” of premiums. I find evidence that differentials in premiums resulted in distortions, including regulatory arbitrage, but also provided strong incentives to curb moral hazard. I find that firms that faced stronger pricing incentives to become (or remain) safer were more likely to subsequently do so than similar firms that faced weaker pricing incentives.

Publications

Economies of Scale in Community Banks, (with Stefan Jacewitz and Troy Kravitz), 2020, FDIC Staff Studies, 2020-06

Using financial and supervisory data from the past 20 years, we show that scale economies in community banks with less than $10 billion in assets emerged during the run-up to the 2008 financial crisis due to declines in interest expenses and provisions for losses on loans and leases at larger banks. The financial crisis temporarily interrupted this trend and costs increased industry-wide, but a generally more cost-efficient industry re-emerged, returning in recent years to pre-crisis trends. We estimate that from 2000 to 2019, the cost-minimizing size of a bank’s loan portfolio rose from approximately $350 million to $3.3 billion. Though descriptive, our results suggest efficiency gains accrue early as a bank grows from $10 million in loans to $3.3 billion, with 90 percent of the potential efficiency gains occurring by $300 million.

Outcome-Robust Mechanisms for Nash Implementation, (View Full Text pdf), 2019, Social Choice and Welfare, 52(3), 497-526

Mechanisms for Nash implementation in the literature are fragile in the sense that they fail if just one or two players do not follow their equilibrium strategy. A mechanism is outcome-robust if its equilibrium outcome is not affected by any deviating minority of players. Is Nash implementation possible with outcome-robust mechanisms? I first show that in the standard environment, it is impossible to Nash-implement any nonconstant social choice rule with outcome-robust mechanisms even if a small number of players are partially honest. If simple transfers are used and if at least one player is partially honest, however, any social choice rule is Nash implementable using an outcome-robust mechanism. The mechanism presented in this paper makes no assumptions about how transfers enter players’ preferences except that transfers are valuable. Moreover, it has: no transfers in equilibrium, arbitrarily small off-equilibrium transfers, and no integer or modulo games.

Criminals' Response to Changing Crime Lucre, 2016, Economic Inquiry, 54(3), 1464-1483

How do criminals respond to changes in the benefit from committing a successful crime? This question is relevant for understanding the effectiveness of crime-fighting policies that reduce demand for illegal goods, disrupt black markets, and otherwise eliminate cheaper avenues to illicit gain. However, the literature has not sufficiently addressed this question, partly because finding a reliable measure of crime lucre is difficult. Using proprietary data on cargo theft, I match historical prices of various goods with their thefts and estimate the price elasticity of theft to be 1.225 over a cumulative 7-month horizon.

Quarterly Fiscal Policy Experiments with a Multiplier-Accelerator Model, (with David Kendrick), 2014, Computational Economics, 44, 269-293

Fiscal decisions are made on an annual basis, but can more frequent decisions improve stabilization? This paper uses a small model in a quadratic linear tracking stochastic control framework to compare a scenario in which fiscal policy is changed quarterly to one in which it is changed once a year. We first report on the use of counterfactual experiments in the 2007-2010 period of a major downturn in the economy. We find in one experiment that quarterly changes in policy stabilize output levels in the economy better than annual changes with a slightly larger increase in debt. In a second experiment we find that when weight changes are used to get roughly equal stabilization results, the increase in the debt level over the counterfactual periods is lower with quarterly than with annual policy changes. In the second part of the paper we repeat the two experiments in a Monte Carlo framework. The Monte Carlo results are consistent with the counterfactual ones, suggesting that a shift from annual to quarterly fiscal policy could provide either better stabilization with a slightly larger increase in the debt level or similar stabilization results with a small increase in the debt level.

Other Papers

Safety in Mechanism Design and Implementation Theory, 2014, Available at SSRN: http://dx.doi.org/10.2139/ssrn.2478655  

A mechanism is unsafe if a small number of agents deviating unexpectedly can make the mechanism deliver an outcome regarded as bad by a large number of other agents. Under Nash behavior, the direct approach of designing mechanisms with only safe Nash equilibria is impossible in many environments. A weaker approach (double implementation in Nash and safe equilibria) making just some Nash equilibria safe is enough to guarantee the safety of players in equilibrium with mild behavioral assumptions. In general, to be safe, a mechanism must restrict the set of outcomes that agents can achieve by deviating, introducing a trade-off between safety and individual autonomy. Also, requiring a mechanism to have safe equilibria implement a social choice rule may require that other equilibria deliver outcomes that are socially undesirable.

A Graph Theoretic Characterization of Implementation Theory Problems

This paper shows that any implementation problem can be formulated as a question about the existence of a graph that solves a (nonstandard) graph coloring problem. This connection to graph theory places implementation theory problems in a general class of Constraint Satisfaction Problems (CSPs) of which graph coloring is a special case. Based on results from the CSP literature, I propose algorithms to find the simplest (in an arbitrary notion of simplicity) mechanisms, if they exist, to solve any implementation problem in an arbitrary solution concept. I present a condition that generalizes Maskin monotonicity to solution concepts other than Nash equilibrium.

Engineering

State-Space Realization for Nonlinear Systems, 2008 (Georgia Tech Thesis)

How can a black box be controlled when what is inside is unknown? Building a model suitable for control theory is the first step. Modern Control Theory is based on a class of models called state-space models, which completely describe the behavior of all the relevant variables of the system. On the other hand, the common way to model a black box is by developing a mathematical relationship between its inputs and outputs (an input-output model) using some kind of regression or curve fit. Input-output models give information about external variables to a system, but do not say much about a system's internal workings. State-space realization theory studies when, and how, an input-output model can be transformed into a state-space model (if such transformation is possible we call the input-output model realizable). My thesis studies the state-space realization problem for nonlinear systems. First, I show that the necessary and sufficient conditions in the literature for realizability of input-output maps of nonlinear systems can be simplified. Second, I introduce a very general form for input-output maps and prove that it is always realizable.