We use a labor-search model to explain why the worst employment slumps often follow expansions of household debt. We find that households protected by limited liability suffer from a household-debt-overhang problem that leads them to require high wages to work. Firms respond by posting high wages but few vacancies. This vacancy-posting effect implies that high household debt leads to high unemployment. Even though households borrow from banks via bilaterally optimal contracts, the equilibrium level of household debt is inefficiently high due to a household-debt externality. We analyze the role that a financial regulator can play in mitigating this externality.
with Giorgia Piacentino (in JET 173)
We present a model in which asset managers design their contracts to attract flows of investor capital. We find that they make their contracts depend on public information, e.g. credit ratings or benchmark indices, as a way to attract flows, rather than as a way to mitigate incentive problems, as has been emphasized in the literature. Unfortunately, asset managers’ competition for flows triggers a race to the bottom: asset managers use public information in their contracts even though it is socially inefficient. This inefficiency arises because contracting on public information prevents risk sharing.
We develop a theory of banking that explains why banks started out as commodities warehouses. We show that warehouses become banks because their superior storage technology allows them to enforce the repayment of loans most effectively. Further, interbank markets emerge endogenously to support this enforcement mechanism. Even though warehouses store deposits of real goods, they make loans by writing new “fake” warehouse receipts, rather than by taking deposits out of storage. Our theory helps to explain how modern banks create funding liquidity and why they combine warehousing (custody and deposit-taking), lending, and private money creation within the same institutions. It also casts light on a number of contemporary regulatory policies.
with Eva Micheler (in the JFE 127)
Many debt claims, such as bonds, are resaleable, whereas others, such as repos, are not. There was a fivefold increase in repo borrowing before the 2008 crisis. Why? Did banks’ dependence on non-resaleable debt precipitate the crisis? In this paper, we develop a model of bank lending with credit frictions. The key feature of the model is that debt claims are heterogenous in their resaleability. We find that decreasing credit market frictions leads to an increase in borrowing via non-resaleable debt. Borrowing via non-resaleable debt has a dark side: it causes credit chains to form, since if a bank makes a loan via non-resaleable debt and needs liquidity, it cannot sell the loan but must borrow via a new contract. These credit chains are a source of systemic risk, since one bank’s default harms not only its creditors but also its creditors’ creditors. Overall, our model suggests that reducing credit market frictions may have an adverse effect on the financial system and may even lead to the failures of financial institutions.
We develop a model in which collateral serves to protect creditors from the claims of other creditors. We find that, paradoxically, borrowers rely most on collateral when pledgeability is high, because this is when it is easy to take on new debt, which dilutes existing creditors. Creditors thus require collateral for protection against dilution—there is a ‘‘collateral rat race.’’ But collateralized borrowing has a cost: it encumbers assets, constraining future borrowing and investment—there is a “collateral overhang.” Our results suggest that policies aimed at increasing the supply of collateral may backfire, triggering an inefficient collateral rat race. Likewise, upholding the absolute priority of secured debt can exacerbate the rat race.
We explain the endogenous emergence of a variety of lending intermediaries in a model based only on differences in their funding costs. Banks have a lower cost of capital than non-banks due to government safety nets. However, with only bank finance, entrepreneurs make inefficient project choices, forgoing innovative projects for traditional projects. Non-banks emerge to mitigate this inefficiency, using their high cost of capital as a commitment device not to fund traditional projects, thus inducing entrepreneurs to innovate efficiently. Despite earning high returns, non-banks never take over the entire market, but coexist with banks in general equilibrium.
We develop a dynamic model of borrowing and lending in the interbank market, in which banks fund investments through short-term collateralized debt, like repos. This debt is not a perfect substitute for cash: lending banks may not be able to convert their loans to cash to fund their own investments. Hence, lending comes with an opportunity cost that generates positive spreads even absent any credit risk. The opportunity cost channel gives a new explanation for why credit is tight in crises and loose beforehand. Through banks’ collateral constraints, the opportunity cost also determines asset demand and prices, creating a feedback loop that can result in multiple, welfare-ranked equilibria. High-leverage equilibria are inefficient in booms; hence, countercyclical capital regulation can improve welfare.
with Giorgia Piacentino
We present a banking model in which bank debt circulates as “money” in decentralized secondary markets, like banknotes did in the nineteenth century and repos do today. We find that bank debt is susceptible to runs because secondary-market liquidity is subject to sudden, self-fulfilling dry-ups. When debt fails to circulate it is redeemed on demand in a “money run.” Even though demandable debt exposes banks to costly runs, banks still want to issue it. Indeed, to facilitate creating demandable money, banks in our model pool investments and transform maturity/liquidity—they endogenously do something that resembles real-world banking. Thus, the model suggests a raison d’etre for banking based on money creation alone.
We develop a dynamic model of board decision making. We show that a board may be unable to make a decision even if all directors agree on what the best decision is. Thus, the board may be unable to fire a CEO all directors agree is bad—a deadlocked board leads to an entrenched CEO. We explore how to compose the board and appoint directors to mitigate the costs of deadlock. We find that board diversity and long director tenure can exacerbate deadlock. Moreover, we rationalize why CEOs and incumbent directors have the power to appoint new directors: it helps to avoid deadlock. Our model also gives a new perspective on staggered boards and short-termism.
I explore how the cyclicality of firms’ output affects their debt capacity. I point out that, in contrast to received theory, procyclical firms can have an advantage in the funding market. Because they have more assets in booms, when asset prices are high ex post, they have looser collateral constraints ex ante—assets are useful as collateral only when they are valuable. This advantage is present across diverse models of borrowing constraints, and can be a first-order determinant of debt capacity. In dynamic general equilibrium, debt capacity goes up in booms, improving capital allocation and generating fluctuations in firm-level and aggregate variables consistent with stylized facts, even absent technology shocks. Further, “procyclical assets” are priced at a premium.
with Giorgia Piacentino
We present a model to explain why banks hold off-setting debts without netting them out. We find that off-setting debts help a bank to raise liquidity with new debt from a third party, since diluting old debt subsidizes the new debt. Even though a diluted bank is worse off ex post, a network of gross debts is stable ex ante. This is because it provides banks with valuable liquidity co-insurance, since each bank exercises its option to dilute when it needs liquidity most. However, the network harbors systemic risk: since one bank’s liabilities are other banks’ assets, a liquidity shock can transmit through the network in a default cascade.