(with J Granja, E Jiang, G Matvos and T Piskorski), 2024

In the face of rising interest rates in 2022, banks mitigated interest rate exposure of the accounting value of their assets but left the vast majority of their long-duration assets exposed to interest rate risk. Data from call reports and SEC filings shows that only 6% of U.S. banking assets used derivatives to hedge their interest rate risk, and even heavy users of derivatives left most assets unhedged. The banks most vulnerable to asset declines and solvency runs decreased existing hedges, focusing on short-term gains but risking further losses if rates rose. Instead of hedging the market value risk of bank asset declines, banks used accounting reclassification to diminish the impact of interest rate increases on book capital. Banks reclassified $1 trillion in securities as held-to-maturity (HTM) which insulated these assets book values from interest rate fluctuations. More vulnerable banks were more likely to reclassify. Extending Jiang et al.’s (2023) solvency bank run model, we show that capital regulation could address run risk by encouraging capital raising, but its effectiveness depends on the regulatory capital definitions and can by eroded by the use of HTM accounting. Including deposit franchise value in regulatory capital calculations without considering run risk could weaken capital regulation’s ability to prevent runs. Our findings have implications for regulatory capital accounting and risk management practices in the banking sector.

(with S Agarwal, B Morais and K Shue), 2024

While reliance on human discretion is a pervasive feature of institutional design, human discretion can also introduce costly noise (Kahneman, Sibony, and Sunstein 2021). We evaluate the consequences, determinants, and trade-offs associated with discretion in high-stake decisions assessing bank safety and soundness. Using detailed data on the supervisory ratings of US banks, we find that professional bank examiners exercise significant personal discretion—their decisions deviate substantially from algorithmic benchmarks and can be predicted by examiner identities, holding bank fundamentals constant. Examiner discretion has a large and persistent causal impact on future bank capitalization and supply of credit, leading to volatility and uncertainty in bank outcomes, and a conservative anticipatory response by banks. We identify a novel source of noise: weights assigned to specific issues. Disagreement in ratings across examiners can be attributed to high average weight (50%) assigned to subjective assessment of banks’ management quality, as well as heterogeneity in weights attached to more objective issues such as capital adequacy. Replacing human discretion with a simple algorithm leads to worse predictions of bank health, while moderate limits on discretion can translate to more informative and less noisy predictions.

(with E Jiang, G Matvos and T Piskorski), 2023

Building on the work of Jiang et al. (2023) we develop a framework to analyze the effects of credit risk on the solvency of U.S. banks in the rising interest rate environment. We focus on commercial real estate (CRE) loans that account for about quarter of assets for an average bank and about $2.7 trillion of bank assets in the aggregate. Using loan-level data we find that after recent declines in property values following higher interest rates and adoption of hybrid working patterns about 14% of all loans and 44% of office loans appear to be in a “negative equity” where their current property values are less than the outstanding loan balances. Additionally, around one-third of all loans and the majority of office loans may encounter substantial cash flow problems and refinancing challenges. A 10% (20%) default rate on CRE loans – a range close to what one saw in the Great Recession on the lower end — would result in about $80 ($160) billion of additional bank losses. If CRE loan distress would manifest itself early in 2022 when interest rates were low, not a single bank would fail, even under our most pessimistic scenario. However, after more than $2 trillion decline in banks’ asset values following the monetary tightening of 2022, additional 231 (482) banks with aggregate assets of $1 trillion ($1.4 trillion) would have their marked to market value of assets below the face value of all their non-equity liabilities. To assess the risk of solvency bank runs induced by higher rates and credit losses, we expand the Uninsured Depositors Run Risk (UDRR) financial stability measure developed by Jiang et al. (2023) where we incorporate the impact of credit losses into the market-to-market asset calculation, along with the effects of higher interest rates. Our analysis, reflecting market conditions up to 2023:Q3, reveals that CRE distress can induce anywhere from dozens to over 300 mainly smaller regional banks joining the ranks of banks at risk of solvency runs. These findings carry significant implications for financial regulation, risk supervision, and the transmission of monetary policy.

(with G Buchak, G Matvos and T Piskorski), 2022 (R&R, Journal of Political Economy)

We study the frictions in dealer-intermediation in residential real estate through the lens of “iBuyers,” technology entrants, who purchase and sell residential real estate through online platforms. iBuyers supply liquidity to households by allowing them to avoid a lengthy sale process. They sell houses quickly and earn a 5% spread. Their prices are well explained by a simple hedonic model, consistent with their use of algorithmic pricing. iBuyers choose to intermediate in markets that are liquid and in which automated valuation models have low pricing error. These facts suggest that iBuyers’ speedy offers come at the cost of information loss concerning house attributes that are difficult to capture in an algorithm, resulting in adverse selection. We calibrate a dynamic structural search model with adverse selection to understand the economic forces underlying the tradeoffs of dealer intermediation in this market. The model reveals the central tradeoff to intermediating in residential real estate. To provide valuable liquidity service, transactions must be closed quickly. Yet, the intermediary must also be able to price houses precisely to avoid adverse selection, which is difficult to accomplish quickly. Low underlying liquidity exacerbates adverse selection. Our analysis suggests that iBuyers’ technology provides a middle ground: they can transact quickly limiting information loss. Even with this technology, intermediation is only profitable in the most liquid and easy to value houses. Therefore, iBuyers’ technology allows them to supply liquidity, but only in pockets where it is least valuable. We also find limited scope for dealer intermediation even with improved pricing technology, suggesting that underlying liquidity will be an impediment for intermediation in the future.

(with N Artavanis, D Paravisini, C Robles-Garcia and M Tsoutsoura), 2022 

We develop a new approach to identify different categories of depositors during periods of uncertainty and quantify their compensation to remain in the bank. We isolate withdrawals due to liquidity needs, deterioration of fundamentals, and expectation about withdrawal behavior of other depositors. We exploit variation in the cost of withdrawal induced by the maturity expiration of time deposits around unexpected uncertainty events and high-frequency microdata from a large Greek bank. Deposit withdrawals quadrupled in response to a policy uncertainty shock that doubled the short-run credit default swap (CDS) price of Greek sovereign bonds. About two-thirds of this increase is driven by direct exposure to deteriorating fundamentals, and the remainder due to strategic complementarities. We find that depositors need to be offered annualized returns exceeding 50% to remain in the bank during episodes of high uncertainty. Our findings provide new insights into the design of interventions that prevent runs by targeting depositors with the largest propensity to withdraw.

(with E Jiang, G Matvos and T Piskorski), 2022

Is bank capital structure designed to extract deposit subsidies? We address this question by studying capital structure decisions of shadow banks: intermediaries that provide banking services but are not funded by deposits. We assemble, for the first time, call report data for shadow banks which originate one quarter of all US household debt. We document five facts.

(1) Shadow banks use twice as much equity capital as equivalent banks, but are substantially more leveraged than non-financial firms.

(2) Leverage across shadow banks is substantially more dispersed than leverage across banks.

(3) Like banks, shadow banks finance themselves primarily with short-term debt and originate long-term loans. However, shadow bank debt is provided primarily by informed and concentrated lenders.

(4) Shadow bank leverage increases substantially with size, and the capitalization of the largest shadow banks is similar to banks of comparable size.

(5) Uninsured leverage, defined as uninsured debt funding to assets, increases with size and average interest rates on uninsured debt decline with size for both banks and shadow banks.

Modern shadow bank capital structure choices resemble those of pre-deposit-insurance banks both in the U.S. and Germany, suggesting that the differences in capital structure with modern banks are likely due to banks’ ability to access insured deposits. Our results suggest that banks’ level of capitalization is pinned down by deposit subsidies and capital regulation at the margin, with small banks likely to be largest recipients of deposit subsidies. Models of financial intermediary capital structure then have to simultaneously explain high (uninsured) leverage, which increases with the size of the intermediary, and allow for substantial heterogeneity across capital structures of firms engaged in similar activities. Such models also need to explain high reliance on short-term debt of financial intermediaries.

(with J Liberti and V Vig), 2017. (R&R Journal of Financial Economics)

This paper investigates the effect of a change in informational environment of borrowers on the organizational design of bank lending. We use micro-data from a large multinational bank and exploit the sudden introduction of a credit registry, an information-sharing mechanism across banks, for a subset of borrowers. Using within borrower and loan officer variation in a difference-in-difference empirical design, we show that expansion of credit registry led to an improvement in allocation of credit to affected borrowers. There was a concurrent change in the organizational structure of the bank that involved a dramatic increase in delegation of lending decisions of affected borrowers to loan officers. We also find a significant expansion in scope of activities of loan officers who deal primarily with affected borrowers, as well as of their superiors. There is suggestive evidence that larger banks in the economy were better able to implement similar changes as our bank. We argue that these patterns can be understood within the framework of incentive-based and information cost processing theories. Our findings could help rationalize why improvements in the information environment of borrowers may be altering the landscape of lending by moving decisions outside the boundaries of financial intermediaries.

(with S Agarwal, E Benmelech and N Bergman), 2012. (R&R Journal of Political Economy)

Yes, it did. We use exogenous variation in banks’ incentives to conform to the standards of the Community Reinvestment Act (CRA) around regulatory exam dates to trace out the effect of the CRA on lending activity. Our empirical strategy compares lending behavior of banks undergoing CRA exams within a given census tract in a given month to the behavior of banks operating in the same census tract-month that do not face these exams. We find that adherence to the act led to riskier lending by banks: in the six quarters surrounding the CRA exams lending is elevated on average by about 5 percent every quarter and loans in these quarters default by about 15 percent more often. These patterns are accentuated in CRA-eligible census tracts and are concentrated among large banks. The effects are strongest during the time period when the market for private securitization was booming.