Keeping the Little Guy Down: A Debt Trap for Lending with Limited Pledgeability with Ernest Liu (Job Market Paper)
Abstract: Microcredit has so far failed to catalyze business growth among small-scale entrepreneurs in the developing world, despite their high return to capital. This prompts a re-examination of the special features of informal credit markets that cause them to operate inefficiently. We present a theory of informal lending that highlights two of these features. First, borrowers and lenders bargain not only over division of surplus but also over contractual flexibility (the ease with which the borrower can invest to grow her business). Second, when the borrower's business becomes sufficiently large she exits the informal lending relationship and enters the formal sector – an undesirable event for her informal lender. We show that in Markov Perfect Equilibrium these two features lead to a poverty trap and study its properties. The theory facilitates reinterpretation of a number of empirical facts about microcredit: business growth resulting from microfinance is low on average but high for businesses that are already relatively large, and microlenders have experienced low demand for credit. The theory features nuanced comparative statics which provide a testable prediction and for which we establish novel empirical support. Using the Townsend Thai data and plausibly exogenous variation to the level of competition Thai money lenders face, we show that as predicted by our theory, money lenders in high competition environments impose fewer contractual restrictions on their borrowers. We discuss robustness and policy implications.
Targeting High Ability Entrepreneurs Using Community Information: Mechanism Design in the Field with Natalia Rigol and Reshmaan Hussam (New Draft Coming Soon)
Abstract: One of the most difficult problems in development is that the cost of assessing credit risk for small businesses often makes giving loans unprofitable. In a field experiment in Maharashtra, India we asked 1,345 entrepreneurs to rank their peers on various metrics of business profitability and growth and entrepreneur characteristics. To assess the validity of these predictions we then randomly distributed cash grants of about USD100 to a third of these entrepreneurs. We find that information provided by community members is predictive of the marginal return to capital. We horserace the community rankings against a machine learning prediction and find that while the machine learning exercise is able to predict high-return entrepreneurs, community information continues to correctly predict 23% returns per month for the top entrepreneurs even after controlling the machine learning prediction. We use experimental variation in the design of surveys to demonstrate agency problems in obtaining community information when community members know the purpose of the information gathering exercise. We conclude by demonstrating how mechanism design can be used to address these agency problems; monetary incentives for accuracy, eliciting reports in public, and cross-reporting techniques motivated by implementation theory all significantly improve the accuracy of reports.
Making it Safe to Use Centralized Markets: Dominant Individual Rationality and Applications to Market Design with Ran Shorrer (New Draft Coming Soon)
Abstract: A critical, yet underappreciated feature of market design is that centralized markets operate within a broader context; often market designers cannot force participants to join a centralized market. Well-designed centralized markets must induce participants to join voluntarily, in spite of pre-existing decentralized institutions they may already be using. We take the view that centralizing a market is akin to designing a mechanism to which people may voluntarily sign away their decision rights. We study the ways in which market designers can provide robust incentives that guarantee agents will participate in a centralized market. Our first result is negative and derives from adverse selection concerns. Near any game with at least one pure strategy equilibrium, we prove there is another game in which no mechanism can eliminate the equilibrium of the original game.
In light of this result we offer a new desideratum for mechanism and market design, which we term epsilon-dominant individual rationality. After noting its robustness, we establish two positive results about centralizing large markets. The first offers a novel justification for stable matching mechanisms and an insight to guide their design to achieve epsilon-dominant individual rationality. Our second result demonstrates that in large games, any mechanism with the property that every player wants to use it conditional on sufficiently many others using it as well can be modified to satisfy epsilon-dominant individual rationality while preserving its behavior conditional on sufficient participation. The modification relies on a class of mechanisms we refer to as random threshold mechanisms and resembles insights from the differential privacy literature.
Rank-Based Elicitation Schemes for Relative Likelihoods of Events with Vivek Bhattacharya
Abstract: This paper addresses the relative scarcity of incentive compatible elicitation schemes that are simple enough for agents with low numeracy. Specifically we study situations under which rank-based scoring rules, in which agents are asked to order events by their likelihoods, can be used to elicit truthful rankings. Such elicitation mechanisms are commonly used in the field since they only require agents to compare likelihoods of similar events to one another rather than to numerical probabilities, which may be difficult for agents with low numeracy. We first show that commonly used rank-based scoring rules need not elicit truthful rankings from agents who are not risk-neutral. We then provide a variation of a rank-based scoring rule that incentivizes agents to rank arbitrary sets of events truthfully, regardless of risk preferences. This scheme can be used to elicit approximate numerical probabilities as well. We provide a simple, indirect implementation of this scheme using a series of questions that ask agents about the relative likelihoods of pairs of events. Finally, we conduct numerical experiments that show that this scheme elicits probabilities to high precision using relatively few questions.
Research in Progress
Paying for the Truth: The Efficacy of Peer Prediction in the Field with Natalia Rigol
Abstract: There is increasing consensus among development economists that community information is valuable for targeting, and that incentives for accuracy may be a vital part of elicitation. We make an empirical case for peer prediction – a class of mechanisms that overcome many practical difficulties traditional monetary incentives face. Peer prediction allows payments to be contemporaneous with the initial report, reducing surveying costs and eliminating the possibility that respondents don’t trust surveyors to return with remuneration. The primary tradeoff is that peer prediction is complicated to explain in practice, and incentive compatibility relies on assumptions that may not hold empirically.
We report results from a lab-in-the-field experiment in Maharshtra, India in which we compare peer prediction to a simple payment rule relying on ex-post accuracy. Farmers were asked questions about their neighbors and were told that their reports would be used to determine cash prizes. Both payment rules result in comparable improvement in accuracy. Importantly, by imposing structure on the data we also find evidence that one peer prediction mechanism is incentive compatible given empirically estimated subjective beliefs; respondents maximize their subjective expected utility by reporting truthfully. This bodes well for situations that require the repeated use of monetary incentives to promote accurate responses – the message that respondents can do no better than to tell the truth will be reinforced with repeated play. Given the broad applicability and the ease of implementation of peer prediction, we hope that this experiment will serve as a catalyst to verify its usefulness in other contexts.
Does Mistrust Prevent Cooperation: The Case Of Fruit Vendors in Delhi with Matt Lowe
Abstract: Duplicative efforts are commonplace among small-scale firms in the developing world. In ongoing work we study fruit vendors in Delhi who procure their produce from the same central marketplace. Despite clear returns to scale in procurement, vendors typically make several trips to the central marketplace each week for their own produce, occupying substantial time they could otherwise spend selling their goods. We hypothesize that this market inefficiency is driven by potential partners’ inability to trust one another. Motivated by repeated game theory we intend to explore a variety of experimental interventions to build trust. Specifically we will offer random vendors assistance in monitoring and rewarding their partners for good behavior. We also explore behavioral explanations such as overconfidence in one’s own ability to procure produce as potential explanation for vendors’ extreme reluctance to cooperate.