This paper studies bank learning through repeated interactions with borrowers from a new perspective. To understand learning by lending, we adapt a methodology from labor economics to analyze how loan contract terms evolve as banks acquire new information about borrowers. We construct "proxy" variables for this information using data from borrowers' out-of-sample, future credit performance. Due to the timing of their construction, banks could not have used these variables directly to price loans. We nonetheless find that these proxies increasingly predict loan prices as relationships progress, even after controlling for possible omitted variable bias. Our methodology provides strong evidence that: (a) bank learning affects loan prices, and (b) relationship benefits are heterogeneous. In particular, higher quality borrowers face differentially lower spreads as their relationship with lenders develop -and banks learn about their quality -while lower quality borrowers see loan prices increase and their loan amounts fall. We further find suggestive evidence that banks incorporate CEO-specific information into loan prices.

for excellent research assistance. The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peer-reviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications.

This paper explores the tension between asset quality and market liquidity. I model an originator who screens assets whose cash flows are later sold in secondary markets. Screening improves asset quality but gives rise to asymmetric information, hindering trade of the asset cash flows. In the optimal mechanism (second‐best), costly retention of cash flows is essential to implement asset screening. Market allocations can feature too much or too little screening relative to second‐best, where too much screening generates inefficiently illiquid markets. Furthermore, the economy is prone to multiple equilibria. The optimal mechanism is decentralized with two tools: retention rules and transfers.