Joseph P. Hall

We study how debt moratoria affect loan repayment and banking relationships using a nationwide experiment with consumer loans in India. In the experiment, borrowers receive forbearance offers presented as an initiative of their lender or the result of government regulation. Borrowers who are offered forbearance by their lender are 3.7 percentage points (8.6 percent) less likely to default than borrowers who receive an identical offer presented as a result of government regulation. Borrowers who are offered forbearance by their lender also have causally higher trust in banks and demand for future interactions with the lender in a follow-up experiment.

This paper measures household discount rates using default decisions rather than asset prices or choices in laboratories. When a borrower decides whether to default, the relative sensitivity of that decision to short-run payment burdens versus long-run debt obligations reveals the borrower's rate of time preference. I formalize this insight in an endogenous default model and show that two comparative statics—the effect on default of a change in monthly payments and of a change in total repayment—are sufficient statistics for the discount factor. Applying this method to over six million U.S. auto loans, I estimate an average annual discount factor of β = 0.79, implying a 27% annual discount rate—far steeper than the 4% typically assumed in macroeconomic calibrations. Discount rates vary sharply with income but are nearly flat across credit scores, suggesting that high default rates among low-score borrowers reflect income risk and costs of default rather than impatience. Extending the method across consumer debt contracts at maturities from two to seventeen years reveals a steeply declining term structure of discount rates—including within borrower—evidence of a discount rate schedule rather than a single rate. These findings provide the first large-scale, revealed-preference distributional estimates of time preferences for households who do not participate in asset markets.

Before 1970, most short-term U.S. consumer credit was extended by merchants; today banks extend it through credit cards. Using the 1978 Marquette decision, which unexpectedly ended binding usury regulation, I estimate that the induced 20% decline in merchant lending raised retail net entry—my proxy for merchant profits—by about 4%, most for small firms. New establishment-level data on card acceptance, hand-collected from archival Yellow Pages, combined with a model separating cost reduction from merchant substitutability, show lower costs rather than competition drive the effect: banks' scale and risk-bearing cut accepting merchants' costs by roughly 4%.

We study the extent to which online deliberation aggregates information before collective votes. Our setting is Arbitrum DAO, a decentralized organization controlling a $3.5 billion treasury, where token holders debate proposals on a public forum before each binding vote. Forum activity predicted vote closeness before March 2024 but became uninformative after, coinciding with language models capable of producing indistinguishable governance prose — and the signal degrades at a time when relatively few posts are actually AI-generated. A model of endogenous posting, reading, and voting, derived from primitives, shows this is a general phenomenon: the signal value of a forum collapses at a discrete tipping point at a realized AI share strictly below the level that would mechanically destroy reading value, and the collapse is hysteretic — reducing AI volume afterward does not restore the forum. Calibrating to the data, the forum generated net social benefits equivalent to saving at least 858 words of research effort per delegate per proposal, all destroyed at the tipping point. Delegates who rely on AI hold far less voting power than those who do not, which is inconsistent with power capture by insiders. Our results suggest that the mere availability of indistinguishable AI writing can degrade the governance capabilities of economically important organizations, even when such AI is rarely used in equilibrium.