In the first chapter, I use a fuzzy regression discontinuity design to study the impact of auction-based emergency liquidity at the onset of the 2007-09 crisis. My empirical design uses the presence of binding auction capacity constraints to isolate variation in short-term borrowing from the Federal Reserve. I find that the growth in marginal winners' outstanding loan commitments is about 15% higher than marginal losers. This effect stems through more lending: marginal winners' loan growth was about 8% higher than that of marginal losers. These effects arise despite the ability to replicate this auction-based liquidity's funding structure through the discount window. They highlight that banks would have relied considerably less on the Federal Reserve had only the discount window been in place during the crisis. I discuss the role of discount window stigma in partly explaining these results. In the second chapter, I study the relationship between confidentiality and contract enforcement. I specifically examine whether publicly exposing past-due borrowers increases their lenders' propensity to seek loan recovery through the legal system. I explore this question in the context of a large rise in past-due loans within India's government-owned banks, which led a bank employees' association to disclose the largest defaulters' identities and loan amounts. Overall, this group increased the probability of legal enforcement by about 10%, with most of this effect arising from their initial threat to expose these defaulters. Lenders targeted past-due borrowers that could likely have avoided default, which is consistent with a desire to reduce the perception of corruption that might follow from disclosure.