Speakers
WiET 2022

This paper studies which misspecified models are likely to persist when an agent compares her subjective model with competing models. I present a framework where the agent learns about an action-dependent data-generating process and makes decisions repeatedly. Aware of potential model misspecification, she uses a threshold rule to switch between models according to how well they fit the data. The main result provides a characterization of persistent models based on the underlying form of misspecification, the characteristics of the learning environment such as the switching threshold, and the set of competing models that may arise. Misspecified models can be robust against a wide range of competing models---including the true data-generating process, despite the agent having an infinite amount of data. Moreover, simple misspecified models with concentrated priors may have even better robustness properties than correctly specified models. I use these results to provide learning foundations for the persistence of systemic biases in two canonical applications.

This paper studies a dynamic information acquisition model with payoff externalities. Two players can acquire costly information about an unknown state before taking a safe or risky action. Both information and the action taken are private. The first player to take the risky action has an advantage but whether the risky action is profitable depends on the state. The players face the tradeoff between being first and being right. In equilibrium, for different priors, there exist three kinds of randomisation: when the players are pessimistic, they enter the competition randomly; when the players are less pessimistic, they acquire information and then randomly stop; when the players are relatively optimistic, they randomly take an action without acquiring information.

Cities are often divided into local governments, which invest in commuting infrastructure within their jurisdictions. In this paper, I study the implications of this metropolitan fragmentation on the provision of commuting infrastructure and the distribution of economic activity. I develop a quantitative equilibrium model of a city where local governments compete for residents and workers by investing in commuting infrastructure to maximize their land value. In this framework, fragmentation leads to the underprovision of commuting infrastructure and employment decentralization. I calibrate and fit the model to the city of Santiago, Chile, which is divided into 33 municipalities.

Policymakers have expressed concern that a dominant online platform that acts as both a marketplace and an active seller might disadvantage its rival sellers and thereby harm consumers. I examine whether platform competition might be promoted to protect consumers. Perhaps surprisingly, I find that increased platform competition can reduce seller competition, and thereby harms consumers.

I study transient matching with overlapping generations. Workers are born with incomplete preference information and must strategically propose matches. To learn their preferences, workers must temporarily match with firms. Workers freely choose a firm to apply to each period, and firms hire their top applicants, up to a capacity constraint. I build upon techniques from the bandit literature to characterize the unique equilibrium, wherein firms are evaluated as endogenous bandits. Even when time frictions are vanishing, path dependency remains a feature of equilibrium. Providing information to workers benefits those with lower quality matches through eliminating costly search.

Sequential voting is used in a wide variety of settings; legal, political, expert committees, and social decision making. This paper is the first to explain why late voters sometimes follow the first voter, commonly known as herding, and sometimes get to cast the deciding vote within the same framework. I extend the standard sequential voting model with a cost of information and show how the information environment uniquely determines the voting and information aggregation outcome. Equilibrium is characterized by the existence of ’Devil’s advocates,’ players who support an opposite argument to make people think seriously, in information environments where the cost of information and/or the signal precision is high. On the other hand, bandwagons occur where the cost of information is low and/or the signal precision is low. I also show that this mechanism performs better than a simultaneous voting mechanism in all information environments by improving information aggregation in high-cost-high-quality information environments and lowering the cost of information aggregation in low-cost-low-quality information environments. Further, it often performs as well as can a social planner that is not constrained by individual incentive compatibility. Finally, it provides recommendations on welfare-maximizing seniority rules in each information environment.

We study how long-lived, rational, exponentially discounting agents learn in a social network. In every period, each agent observes the past actions of his neighbors, receives a private signal, and chooses an action with the objective of matching the state. Since agents behave strategically, and since their actions depend on higher order beliefs, it is difficult to characterize equilibrium behavior. Nevertheless, we show that regardless of the size and shape of the network, and the patience of the agents, the speed of learning in any equilibrium is bounded from above by a constant that only depends on the private signal distribution.

Beginning with Diamond (1998), an important strand of the literature on optimal taxation has used first order conditions to characterize the behavior of taxpayers. This paper explores conditions under which this first order approach (FOA) is valid. The main theoretical result is that FOA is valid if, roughly speaking, the distribution of taxpayer abilities is not too weighted toward the bottom and the government’s objective function is not too redistributionist. I show by example that if these conditions do not hold, then FOA can lead to an “optimal” income tax schedule that, because it fails to capture incentive compatibility constraints, is very far from the true optimum.

In a world where the majority and the minority group have equal distributions of talent, where candidates are objectively evaluated and no discrimination occurs, underrepresentation is nonetheless highly sticky. The reasons are intuitive. If two samples of different sizes are drawn from the same distribution, the highest realization is likely to belong to the larger sample. If the sizes of the future samples respond to the realized selection in the expected direction—increasing if the selection came from the sample, decreasing or increasing less if it did not—the hysteresis can be very strong. We capture this process with a well-known statistical model: the Polya urn. The richness of existing results and the streamlined model allow us to study and compare different policy interventions. A simple app (https://caron.shinyapps.io/Women-Men-Polya-Urns/) allows readers to run their own experiments. A robust result is that temporary affirmative actions interventions have long-term equalizing effects. At the same time any decline in the quality of selected candidates is self-correcting, even while the intervention lasts.

This paper provides a portable, systematic, and rigorous framework for analyzing how the balance-of-power in a group of people evolves over time. N non-overlapping generations of players compete for control over the group’s wealth by accumulating power – modeled as capital that enhances one’s ability to forcefully redistribute wealth – in a dynamic contest. The group’s initial balance-of-power yields a unique equilibrium trajectory towards one of three steady-states: inclusive steady states – where all players are equally powerful – are reached when the initial distribution of power is roughly uniform; otherwise, the group heads towards dictatorship or oligarchy, where all power is held by one or a few players, respectively. In addition to providing a rigorous explanation for the observed tendency of large organizations to devolve into oligarchies, this paper also makes a sharp prediction about the effect of group size on the strength and feasibility of dictatorships.

I study the relation between limited commitment and learning in auctions. In each period, the seller sets the terms for an auction selling an indivisible good among multiple buyers; but if the item fails to sell, he cannot pre-commit to the terms of future offerings. I find that, in interdependent value settings, the seller's equilibrium revenues are greater than immediately running an efficient, Vickrey auction. In contrast with private value settings, this result persists regardless of how often the seller may interact with buyers. This is because learning among buyers both limits how many times a good can be gainfully re-offered and the information rents that each buyer can demand the seller. Intuitively, buyers lower their valuations in response to their peer's lack of interest. This progressively lowers the trade surplus and compresses the support of valuations. As the dispersion in valuations falls, the seller further extracts an increasing share of the remaining trade surplus.

We study a dynamic Principal-Agent problem where they jointly choose a project to implement where the set of available projects is the private information of Agent. The preferences of Principal and Agent over the possible projects are in perfect conflict. In each period, Agent can propose an available project or stay silent, and Principal can accept or reject a proposal but he cannot commit to his future responses. We first establish a benchmark in which Principal has commitment power and we compare it to the equilibria of the dynamic game. When there are two possible projects, there is always an equilibrium in which Principal attains his commitment payoff when the players are sufficiently patient. The main reason why Coasian dynamics are not always present here is that Principal lacks proposal power and can provide incentives through off-path beliefs. When there are more than two possible projects, we identify a condition on the parameters under which Principal can still attain his commitment payoff.

Availability of information to market participants is often seen as essential to the functioning of markets. We consider a market-design based alternative to improve access to information. We examine designs allowing traders to condition their demands on simultaneously-determined prices in venues in which they do not participate. Facilitating access of information is neutral in competitive markets. Designs allowing cross-venue conditioning can be superior in markets with large traders who have price impact. Our results highlight the role of trader heterogeneity for design and regulation.

This paper builds a macroeconomic framework with heterogeneous firms and heterogeneous innovations. The goal is to quantitatively understand the allocation of innovations across firms and the implication of efficiency in the secondary market on the firm-level growth rates. The model characterizes the entire process of innovating, including both the primary and the secondary market. On the primary market, an inventor decides which firm to work for; on the secondary market, a firm can buy and sell innovations when there is an efficiency difference in using the innovations. The model is calibrated to the US. The result implies that inventors with more effort-sensitive ideas work for smaller firms. Counterfactuals on the secondary market show that it affects not only the firm-level innovation allocation but also the aggregate growth rate. For example, if there were no secondary markets, the allocation shifts to larger firms, and large firms contribute more to the growth than before. The share of innovations created in startups and firms with fewer than 500 employees would decrease from 8.0% to 1.3%. For firms with more than 100,000 employees, this share would increase by more than 10 percentage points. The overall aggregate growth rate in the economy would decrease by 0.16 percentage points.

We model a decision maker who observes available alternatives according to a list and stochastically forgets some alternatives. Each time the decision maker observes an item in the list, she recalls previous alternatives with some probability, conditional on those alternatives being recalled until this point. We provide conditions on observable choice probabilities that characterize the model. We show that the preference ordering, the list ordering the decision maker follows and the memory parameters are uniquely identified up to the ranking of the two least preferred alternatives. We apply our model to study the pricing problem of a monopolist who faces consumers with limited memory. We show that when forgetting is severe, the monopolist is better off charging a lower price than the optimal price in the perfect memory case.

I present a model of a competitive labor market in which agents of different skill levels decide whether to enter the market as a manager or a worker. Their decisions induce a two-sided matching market, and a cooperative assignment game occurs. I show existence and uniqueness of a rational expectations equilibrium under both one-to-one and many-to-one matching, and that instead of strict supermodularity, a stronger condition that I call role supermodularity determines if the equilibrium matching pattern is positive assortative. I use the one-to-one framework to provide a perspective on observed changes in U.S. wage inequality over time. Empirical trends are most consistent with the hypothesis that the increase in productivity to the highest-skilled matches has significantly outpaced productivity gains of other match types, and that has caused highly-skilled workers to cluster together.