Working papers
How should firms dynamically manage a supplier portfolio to improve contracting outcomes? This paper analyzes a large Indian buyer's policy of promoting high-performing suppliers to higher tiers that receive increased order volume. I estimate a structural dynamic principal-agent model to quantify the policy's value and disentangle its two mechanisms: retaining better suppliers by sorting them based on a persistent type, and mitigating moral hazard by using relational value and induced competition to incentivize effort. A novel two-stage estimation strategy first recovers the buyer's policy using an iterative algorithm to handle unobserved supplier tiers, then estimates a dynamic effort choice model of supplier behavior conditional on this policy. The policy improves performance by 12% over random allocation, with selection effects dominating incentive effects (66% vs. 34% of the total value). The economic value of this relational approach is significant—achieving the same performance improvement in a spot-market benchmark would require 18% higher payments to suppliers. Counterfactual analysis shows that a policy optimized for the estimated environment could improve performance by 28%, but the firm's current policy is more robust to facing a less capable supplier pool.
Supply chain resilience via partial integration
How do firms adapt to ensure supply chain resilience? The typical answer, vertical integration, is limited in practice by its high costs and inflexibility. This paper considers partial integration, defined as targeted buyer interventions across firm boundaries, as an alternative. Using novel administrative data from Indian supply chains for fabricated steel products, we show that supplier underinvestment in inputs, stemming from working capital constraints and non-contractible input use, is the primary driver of disruptions in these supply chains. To reduce the incidence of disruptions, the buyer exerts control over supplier processes—through in-person monitoring, contingent contracts, and direct sourcing of raw materials—rather than only using cash advances. This buyer involvement escalates as disruption risk increases: an unanticipated input cost shock leads to direct buyer control of inputs for the most constrained suppliers. We develop a three-stage model that rationalizes these strategies. The model clarifies that buyers control input decisions to prevent resource diversion due to non-contractibility, while allowing suppliers to retain control of production in order to preserve output market incentives. It also predicts that relational buyers with low monitoring and sourcing costs enjoy a comparative advantage in fostering resilient trade with poor regions.
(with Vishan Nigam)
Interest caps, competition, and strategic borrowing: Evidence from Kenya
We study Kenya’s 2016 interest-rate regulation, which capped bank lending but left one digital platform, called M-Shwari, exempt on the lending side while imposing a deposit-rate floor across all lenders. Using borrower-level administrative data, survey data and an RD around the implementation date, we show that lending on the exempt platform rose, with safer borrowers substituting toward cheaper capped credit and riskier borrowers increasing savings to rebuild limits. We build and estimate a simple model of screening and credit limit-setting to interpret these reallocations and compute welfare. The observed carve-out for M-Shwari preserves access for high-risk borrowers but yields a modest aggregate welfare decline relative to pre-policy; a uniform cap counterfactual generates substantially larger losses by eliminating credit for low-score borrowers.
(with Tavneet Suri and Prashant Bharadwaj)
Publications
Risk preferences of learning algorithms
Many economic decision-makers today rely on learning algorithms for important decisions. This paper shows that a widely used learning algorithm—ε-Greedy—exhibits emergent risk aversion, favoring actions with lower payoff variance. When presented with actions of the same expectated payoff, under a wide range of conditions, ε-Greedy chooses the lower-variance action with probability approaching one. This emergent preference can have wide-ranging consequences, from inequity to homogenization, and holds transiently even when the higher-variance action has a strictly higher expected payoff. We discuss two methods to restore risk neutrality. The first method reweights data as a function of how likely an action is chosen. The second method employs optimistic payoff estimates for actions that have not been taken often.
(with Andreas Haupt)
Games and Economic Behavior (2024) | Link
Ex-post implementation with interdependent values
We characterize ex-post implementable allocation rules for single object auctions under quasi-linear preferences with convex interdependent value functions. We show that requiring ex-post implementability is equivalent to requiring that the allocation rule must satisfy a condition that we call eventual monotonicity (EM), which is a weakening of monotonicity, a familiar condition used to characterize dominant strategy implementation.
(with Saurav Goyal)
Games and Economic Behavior (2023) | Link
Single peaked domains with designer uncertainty
This paper studies single-peaked domains where the designer is uncertain about the underlying alignment according to which the domain is single-peaked. The underlying alignment is common knowledge amongst agents, but preferences are private knowledge. Thus, the state of the world has both a public and private element, with the designer uninformed of both. I first posit a relevant solution concept called implementation in mixed information equilibria, which requires Nash implementation in the public information and dominant strategy implementation in the private information given the public information. I then identify necessary and sufficient conditions for social rules to be implementable. The characterization is used to identify unanimous and anonymous implementable social rules for various belief structures of the designer, which basically boils down to picking the right rules from the large class of median rules identified by Moulin (1980), and hence this result can be seen as identifying which median rules are robust to designer uncertainty.
Social Choice and Welfare (2023) | Link
Coverage analysis in millimeter wave cellular networks with reflections
The coverage probability of a user in a mmwave system depends on the availability of line-of- sight paths or reflected paths from any base station. Many prior works modelled blockages using random shape theory and analyzed the SIR distribution with and without interference. While it is intuitive that the reflected paths do not significantly contribute to the coverage (because of longer path lengths), there are no works which provide a model and study the coverage with reflections. In this paper, we model and analyze the impact of reflectors using stochastic geometry. We observe that the reflectors have very little impact on the coverage probability.
IEEE Global Communications Conference (2017) | Link
Works in progress
Disaggregating organizations: The effect of CEOs on firm markups
Do different CEOs within the same firm systematically set different markups, or are markups determined solely by firm-level optimization? To answer this question, we estimate a Two-Way Fixed Effects (TWFE) model of firm markups on CEO and firm dummies. We use the De Loecker et al. (2020) framework to estimate firm-year level markups, and use CEO movements between firms to identify CEO effects on markups. We address limited mobility bias using the leave-out estimator of Kline et al. (2020). To enable meaningful comparisons across different connected sets of firms and CEOs, we apply the normalization procedure of Best et al. (2023). After applying these corrections, we estimate that CEO effects explain 10-15% of the overall variance in markups.
(with Kartik Vira)
Competition and information sharing
In many markets, data providers allow exchange of confidential commercial information between firms, with ambiguous effects on competition. Most of the conduct testing literature restricts the information sets of firms to be complete, or at least known. We consider a case where membership in a data aggregator’s subscription service is unobserved to the analyst, but the distribution of outcomes and some components of costs are. We develop a sequential method to identify firms' latent information structures and test between alternative conduct models. First, the information structure is identified using a firm’s response to rivals’ private cost shocks. Then, firm conduct is identified using standard exclusion restrictions conditional on the information structure. We discuss an application of this method to the poultry processing industry, where an aggregator (AgriStats) shares members’ private cost information for a subscription fee.
(with Lia Petrose)