Abstract: This paper asks: How can we test for credit rating bias? Testing for bias is challenging because researchers do not observe credit rating agencies’ (CRAs’) private information about issuers. I address this challenge by developing a general framework of credit ratings and using it to derive a nonparametric test for rating bias. Ratings are biased if the CRA assigns higher ratings to issuers it deems riskier than others. Using a novel regulatory dataset that indicates whether issuers purchased ratings, I apply the test to Moody's corporate bond ratings and find evidence of bias consistent with conflicts of interest under the issuer-pays model. First, I find that ratings are relatively more biased for issuers that tend to generate more rating revenue. Second, I find evidence of bias for solicited ratings but not for unsolicited ones. The framework is flexible and has several applications beyond credit ratings.

Abstract: In order to study updating rules, we consider the problem of a malevolent principal screening an imperfectly Bayesian agent. We uncover a fundamental dichotomy between underreaction and overreaction to information. If an agent's posterior is farther away from the prior than it should be under Bayes' law, she can always be exploited by the principal to an unfettered degree: the agent's ex ante expected loss can be made arbitrarily large. In stark contrast, an agent who underreacts (whose posterior is closer to the prior than the Bayesian posterior) cannot be exploited at all.

Abstract: We study competition between credit rating agencies (CRAs), who compete by flexibly designing rating schemes and posting prices. A variety of equilibria coexist, distinguished by the number of ratings purchased (in equilibrium) by the asset's issuer. The equilibria can be ranked by the amount of information produced and in terms of welfare; notably, total surplus maximization benefits CRAs at the expense of the issuer. Ratings design allows CRAs to relax competition in the interim (ratings) pricing game to various degrees. Our framework provides novel insights into competition between CRAs and regulation consistent with empirical findings.

Abstract: In late 2013, TransUnion extended the length of repayment histories on its credit reports from one month to thirty months. Shortly after, some credit card lenders stopped reporting actual repayment information, thereby withholding significant information about borrowers' credit risk. Using a dynamic model, I study the effect on borrowers when lenders adjust their disclosure policies in response to changes in the set of feasible disclosure policies. Lenders first choose how long to report borrowers' repayments and then compete for borrowers in each period. Lenders have heterogeneous costs from not reporting repayments, leading to asymmetric equilibrium disclosure strategies. I use comparative statics with respect to the lenders' prior (about the borrowers' risk) and the length of repayment histories to derive a difference equation for equilibrium interest rates. Using mortgage data from Freddie Mac and variation in borrowers' credit scores as a proxy for lenders' prior, I use difference-in-differences to estimate the joint effect of TransUnion's new credit reports and changes in lenders' reporting decisions on mortgage interest rates and borrowers' costs.

Abstract: We study welfare analysis for policy changes when supply behavior is only partially known. We augment the robust-demand approach of Kang and Vasserman (2025) with two supply primitives—intervals of feasible pass-through and conduct (market-power) parameters—applied to two equilibrium snapshots. A simple accounting identity distills the supply-side contribution to welfare to a simple integral expression. From there, we deduce that the bounds are produced by a single-threshold "bang-bang" path. This, plus a modification of Kang and Vasserman (2025)'s demand-side characterization, delivers simple bounds for consumer surplus, producer surplus, tax revenue, total surplus, and deadweight loss. We also study an ad valorem extension.

Description: We develop a general framework in which a decision-maker (DM) takes a binary action after observing informative signals. We provide two tests for taste-based discrimination, assuming the econometrician observes part of the DM’s information, the DM’s decisions, and ex-post outcomes. The tests are robust to the inframarginality problem and omitted variable bias. We test for taste-based discrimination in police stop decisions using the NYPD Stop, Question, and Frisk data.

Description: Executive boards and executives make high-stakes choices under uncertainty and ambiguity. Standard governance frameworks typically assume common and known risk preferences and beliefs. We study robust decision-making when risk preferences and beliefs are heterogeneous or unknown. In this setting, we examine how collective decision-making shapes board oversight, executive pay, and responses to shareholder activism. We aim to inform the design of governance protocols that robustly improve decision-making.

Description: While wage inequality has risen across developed countries, we document a substantial 33% decline in Portuguese wage inequality from 2000 to 2021, driven primarily by reduced between-firm wage variation. During this period, Portugal's statutory minimum wage increased 109% nominally, and collective bargaining agreements (CBAs) covering nearly all workers set sector-specific wage floors. This paper explores how minimum wage policy and CBAs affect firms' wage-setting mechanisms—including rent-sharing and bargaining power—potentially contributing to the compression of between-firm wage variation. To facilitate this analysis, we digitize the first comprehensive CBA dataset using large language models to extract information from unstructured official documents.