Diamond and Rajan (2000, 2001) argue that banks create liquidity by issuing deposits to fund difficult, illiquid firms that otherwise cannot obtain funding. Since deposits may lead to bank runs, this resulting financial fragility is essential for liquidity creation. We revisit the Diamond-Rajan model of financial intermediation and show that a bank with an optimal financing structure is not subject to runs. Our contract rests on three simple notions. First, each bank creditor has the right to demand repayment at every instant. Second, the repayment is given by the value of a pre-specified fraction of the bank’s assets. Third, some creditors are more senior than others: their repayment demands are prioritized. In contrast to Diamond and Rajan, we find that financial fragility is detrimental to liquidity creation.

This paper proposes a theory of large-scale government bond purchases by central banks in an environment with endogenous information acquisition. Information acquisition by private investors lowers sovereign bond yields by reducing uncertainty, and makes prices more responsive to new information. I show that asset purchases by the central bank discourage information acquisition. Using the case of Italian bonds and the start of ECB purchases in 2015, I document through various measures that price informativeness indeed significantly declined with purchases. When the sovereign can be subject to self-fulfilling debt crises, however, this reduction in information acquisition can be beneficial. I show that by impairing price informativeness, asset purchases can avoid the occurrence of roll-over crises, generating large welfare gains. A key property of the model is that substantial purchases may be required, while small interventions have ambiguous welfare consequences. When the sovereign expects the central bank to carry such programs, it leads to excessive indebtedness, forcing the central bank to run an inflated balance sheet to avoid roll-over crises.

Shocks to a bank’s ability to raise liquidity at short notice can trigger depositor panics. Why don’t banks take a more active role in managing these risks? We study contingent risk management (hedging) in a standard global-games model of a bank run. Banks fail to hedge precisely when the exposure to a shock is most severe, just when risk management would have the biggest impact. Higher bank capital and broader deposit-insurance coverage crowd out hedging by banks that already manage risk, yet encourage more banks to establish risk management desks in the first place. The model also yields testable implications for hedging incentives and policy design.