Working Papers

We study PAYGo financing for smartphones and other consumer durables in developing countries. Using a dynamic structural model estimated from a fintech pricing experiment, we show that technology-enabled lending generates significant welfare gains equivalent to a substantial income increase, while remaining profitable for lenders. [SSRN] [Slides]

We study a cash loan contract that requires borrowers to make a cash deposit prior to disbursement. The deposit is credited toward the loan principal at disbursement and does not alter repayment incentives. In a randomized controlled trial, we find that the deposit requirement reduces loan take-up but selects borrowers who repay substantially better, raising lender profits. The deposit screens borrowers on both observable and latent characteristics: high-risk borrowers are less likely to take deposit loans and, controlling for all observables, deposit-loan takers repay at a higher rate, particularly among low-risk borrowers. [SSRN]

We study non-price deal terms in merger contests. Using a hand-collected dataset of U.S. mergers, we show that non-price terms are pervasive, differ systematically by bidder type, and often determine the winner. Our structural model reveals that these terms carry substantial economic value for targets. [Draft]

We study the optimal organizational structure for innovation over time, analyzing complementary R&D tasks with moral hazard and privately observed discovery quality. We characterize when a solo innovator is optimal versus a team structure and how the optimal allocation shifts over the project lifecycle. [SSRN] [Draft]

We study how artificial intelligence affects welfare in general equilibrium when production requires both finding and solving problems. Problems are solved in knowledge hierarchies, and AI deployed in solving splits the hierarchy: greater capability affects the layers above it, whereas lower cost affects the layers below. Either form of progress can lower welfare by reallocating solvers into the congested activity of finding problems. A tax on finding corrects this distortion; when it is unavailable, the optimal tax on AI can be positive or negative. [SSRN] [Draft]