Date of Award

Spring 2021

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Geanakoplos, John


This collection of essays examines financial intermediation and collateral requirements in economies where limited enforcement of repayment necessitates loans to be backed by eligible collateral. Collateralized lending was a key feature of the 2008 financial crisis, with significant policy interventions in major economies directed towards restoring leverage and financial intermediation. Similarly, given the unprecedented shock of the COVID-19 pandemic, firms that face binding collateral constraints and find it difficult to secure loans on affordable terms may struggle to weather the crisis. Governments and central banks worldwide have instigated large scale policy responses to ease credit conditions. Theoretical models of financial frictions that account for endogenous variations in leverage across the business cycle are thus important for guiding policy design and for analyzing the propagation and amplification of shocks through the financial sector. In Chapter 1, I examine how central banks should intervene to improve credit conditions during a downturn. Specifically, I analyze the setting of collateral requirements in central bank lending facilities. Traditionally, central bank lending during times of crisis followed Bagehot's rule. That is to lend freely to solvent institutions, against collateral that is good in normal times, and at “high interest rates”. The rule is designed so that the central bank can improve credit conditions without taking on any credit risk. Lately, central banks began to deviate from this approach, conducting more direct lending to firms, against a broader class of collateral, and reducing the haircuts imposed on the collateral posted. Which is the more appropriate response? Should central banks take on greater credit risk in order to provide a larger stimulus? To answer the question, I develop a model of central bank intervention in collateralized credit markets. I find that when the downturn is severe it is optimal for the central bank to take on greater credit risk. The analysis suggests that credit facilities set up by the Federal Reserve in response to COVID-19, such as the Main Street Lending Program, can achieve greater participation and effectiveness by easing their terms of lending. In Chapter 2, I present joint work with Agostino Capponi and Stefano Giglio on the determinants of collateral requirements in central clearinghouses for credit default swaps (CDSs). The empirical results in Capponi et al. (2020) demonstrate that extreme tail risk measures have higher explanatory power for observed collateral requirements than the standard Value-at-Risk rule. To provide a theoretical foundation for these findings, we develop a model of endogenous collateral requirements in the CDS market, where counterparties trade state-contingent promises backed by cash as collateral. Trading occurs due to differences in market participants' beliefs about the uncertain states of the world. We show that it is the nature – rather than the degree – of these belief differences that determines collateral requirements in equilibrium. For instance, the equilibrium level of collateral increases both when the optimist becomes more pessimistic (which reduces the extent of the disagreement), and when the pessimist becomes more concerned about tail events (which increases the extent of the disagreement). We can thus point to the clearinghouse’s concerns about extreme tail events as an explanation for the highly conservative levels of collateral observed in practice. In Chapter 3, I propose a generalization of the Binomial No-Default Theorem of Fostel and Geanakoplos (2015). The Binomial No-Default Theorem states that “in binomial economies with financial assets serving as collateral, any equilibrium is equivalent in real allocations and prices to another equilibrium in which there is no default”. I extend this theorem to economies with more than two states of nature when debt can be ordered by seniority. For instance, with three states of nature, borrowers can issue both senior secured debt and junior unsecured debt. The senior secured debt is explicitly backed by the risky financial asset held by the firm, whereas the junior unsecured debt is implicitly backed by the residual value of the firm after the senior creditors satisfy their claims. The interest rates and credit risks associated with each creditor tier are endogenously determined in equilibrium. The Multinomial Max-Min theorem I prove states that any equilibrium is equivalent to another equilibrium where the senior tranche never defaults, and the junior tranche only defaults in the worst state of the world. The expanded theorem allows for the application of the endogenous leverage framework to a richer set of models with more than two states and where some loans are not contractually secured. Finally, in a joint paper with David Miles presented in Chapter 4, I examine the interaction between the real interest rate and capital requirements in the risk taking behavior of banks. In this model, banks can undertake costly screening to discover private information about the probability of success on their potential lending projects. Once this probability is known, each bank sets a cut-off threshold for the likelihood of success on a project that determines whether or not to lend. When banks are highly leveraged, a combination of asymmetric information and limited liability means that banks screen too little (insufficient “participation”) and accept projects that are too risky (insufficient “prudence”) relative to the first-best benchmark. We show that the real interest rate and capital requirements function as imperfect substitutes. Qualitatively, raising either the real interest rate or the capital requirement on banks can increase “prudence” at the cost of decreased “participation”. But the two policy instruments work through very different mechanisms. An increase in capital requirements forces banks to hold “more skin in the game” and is targeted at the subset of poorly capitalized banks in the population. In contrast, an increase in the real interest rate increases the opportunity cost of lending for all banks, and is thus a much blunter instrument. The interaction between these two policy levers means that the optimal capital requirement on banks rises as the interest rate falls, suggesting tougher macro-prudential capital standards in a “lower-for-longer” interest rate environment.