Bank Behavior & Financial Regulation

The research begins by acknowledging the existing debate surrounding the relationship between bank leverage and risk-taking. Two opposing theoretical viewpoints are presented. The first, based on limited liability, suggests that higher leverage incentivizes banks to take on more risk because equity holders benefit disproportionately from upside gains while being protected from downside losses. The second perspective highlights the role of callable demand deposits, a significant portion of bank debt. A substantial percentage of these deposits are uninsured, which gives depositors a strong incentive to monitor banks and punish excessive risk-taking. This creates a market discipline counteracting the limited liability effect. The paper aims to resolve this ambiguity by providing a rigorously identified causal effect of leverage on risk.

A core challenge in studying the leverage-risk relationship is endogeneity. The study identifies two primary sources: 1) Reverse causality: increased leverage may lead to riskier behavior, but a bank's risk profile might also affect its ability to attract deposits (thus influencing its leverage). 2) Omitted variable bias: unobserved shocks affecting both a bank's assets and liabilities create bias. To address these, the study introduces a novel instrumental variable (IV) approach. The chosen instrument uses data on local bank office deposits and local unemployment rates. This geographical detail allows the researchers to isolate variations in leverage that are exogenous to the bank's risk-taking decisions. The risk measure itself is carefully constructed using new mortgage loan issuances rather than existing portfolio data, which helps mitigate the impact of omitted shocks.

The methodology section delves into the details of the instrumental variable strategy and the chosen risk measure. The IV for leverage is constructed using granular data from the FDIC on deposits at the office level, enabling calculation of each bank's exposure to local unemployment fluctuations. This approach is designed to capture variations in leverage driven by external factors rather than the bank's internal risk choices. However, even this measure could still be endogenous if it's tied to the existing portfolio. To circumvent this, the researchers define their measure of risk-taking based on newly issued mortgage loans. The reasoning is that the riskiness of these new loans directly reflects management's choices, uncontaminated by pre-existing portfolio composition or exogenous shocks that might simultaneously influence both deposits and the existing portfolio. The use of data from the Loan Application Registries (LAR), mandated by the Home Mortgage Disclosure Act (HMDA), provides a comprehensive dataset of mortgage loan applications for the analysis (1999-2016).

The empirical findings using Two-Stage Least Squares (2SLS) and instrumental variable regression demonstrate a positive and significant relationship between bank leverage and risk-taking. This supports the theoretical prediction that limited liability encourages riskier investment strategies. Furthermore, the analysis explored potential non-linear effects by interacting the predicted endogenous variable with leverage deciles but found no significant pattern suggesting the relationship is consistent across the leverage distribution. The results carry important policy implications, indicating that curbing leverage can indirectly reduce bank risk-taking, thereby strengthening the overall resilience of the banking system. The study emphasizes the effectiveness of its methodology in providing a clean identification of the causal effect of leverage on bank risk taking compared to previous studies hampered by endogeneity concerns and noisy portfolio-based risk measures.

This section lays out the theoretical framework surrounding deposit insurance and its potential impact on bank risk-taking. It acknowledges the dual nature of deposit insurance: while it prevents bank runs by eliminating self-fulfilling prophecies of deposit withdrawals (Diamond and Dybvig, 1983), it simultaneously introduces the risk of moral hazard (Cooper and Ross, 2002). Moral hazard arises because insured depositors are less incentivized to monitor banks, as they bear less risk of loss. This reduced monitoring can embolden banks to engage in riskier investments, as the cost of potential losses is shifted to the insurance system. The existing literature, however, presents inconsistent findings on the magnitude and even the existence of this effect, creating a need for more robust empirical analysis. The limitations of previous studies are also highlighted, including the reliance on cross-country comparisons that may conflate deposit insurance policies with other institutional differences and the common use of portfolio-based risk measures that are themselves subject to external shocks.

To overcome the limitations of previous research, this study employs a natural experiment using data from the US banking sector. This strategy is designed to address endogeneity concerns by creating exogenous variation in deposit insurance coverage. The chosen event is a policy change that increased the deposit insurance coverage limit from $100,000 to $250,000 for most banks but left state-chartered banks in Massachusetts with unlimited coverage. This difference creates a treatment group (banks affected by the increase) and a control group (Massachusetts banks). The analysis leverages the fact that all banks in the sample are subject to similar regulations and supervisory requirements. This setup allows for a clearer identification of the causal relationship between the increased deposit insurance coverage and bank risk-taking behavior. This methodology addresses previous endogeneity issues related to the correlation between deposit insurance policy and bank behavior. The study also employs a more precise measure of bank risk, focusing on the riskiness of newly issued mortgage loans, rather than relying on the overall riskiness of the bank’s portfolio.

The study uses a novel measure of bank risk-taking based on newly originated mortgage loans. This is a more direct measure than traditional portfolio-based measures used in previous studies and mitigates endogeneity issues arising from past choices and current shocks impacting the entire portfolio. The data for this analysis comes from the Loan Application Registries (LAR) under the Home Mortgage Disclosure Act (HMDA). The analysis considers various factors influencing loan origination, creating a model that isolates banks' sensitivity to loan-to-income ratios. This sensitivity serves as the primary risk-taking measure. The study also includes a control group of banks in Massachusetts that had unlimited deposit insurance coverage, allowing for a difference-in-differences approach. The researchers acknowledge and address potential issues such as the indirect effect of deposit supply shifts and the impact of the 2008 financial crisis on the results. Careful consideration is given to the timing of the policy change relative to the financial crisis to separate the effects of the crisis itself from the effects of the policy change.

The empirical findings of this study challenge the conventional wisdom regarding the impact of deposit insurance. Contrary to the prediction that increased deposit insurance leads to increased bank risk-taking, the analysis reveals no significant effect of the US policy change on bank risk-taking behavior. Banks did not exhibit a greater propensity to grant loans based on loan-to-income ratios after the increase in the coverage limit. The lack of effect on deposit rates further supports this conclusion by demonstrating that market discipline was not significantly affected. This surprising finding has crucial implications for policymakers as it suggests that the moral hazard channel associated with deposit insurance may not be as powerful as previously thought. The study considers potential confounding factors like the timing of the policy relative to the financial crisis and addresses these concerns in the robustness checks. The results imply a need to re-evaluate policy priorities in a situation where the trade-off is between preventing bank runs and potentially reducing the risk-taking incentives due to deposit insurance.

This section details the data and methodology used to analyze the impact of monetary policy on local economies, focusing on the role of local bank characteristics. County-level data from the County Business Patterns (CBP) are used to measure local economic outcomes, specifically employment and payroll. The choice to focus on county-level data is informed by the observation that while mortgage lending is less localized, corporate lending tends to be concentrated in the vicinity of bank offices. Data from the Federal Deposit Insurance Corporation (FDIC) provides information on bank deposits at the office level, enabling the calculation of each bank's presence within each county. The study uses the federal funds rate as a proxy for monetary policy, recognizing that the peak effect of monetary policy shocks on employment lags by 7-9 quarters (Neville et al., 2012). The analysis employs a regression framework that incorporates the lagged federal funds rate, local bank characteristics, and control variables to account for county-specific effects and other confounding factors. The interaction term between the federal funds rate and local bank characteristics is crucial for assessing the modifying effects of bank attributes on the transmission mechanism.

This section explores how various local bank characteristics influence the transmission of monetary policy to local economic outcomes. The study examines the impact of a change in the federal funds rate on employment and payroll while controlling for factors such as bank size, capital structure, and risk (measured by realised credit risk and probability of default). Contrary to some prior literature (Kashyap and Stein, 1995, 2000), the research finds that bank size does not significantly influence the effect of monetary policy. However, improvements in the capital structure of local banks do dampen the impact of monetary policy changes. This is consistent with the idea that well-capitalized banks have easier access to alternative funding sources when deposits decline due to monetary tightening (Carlino and Defina, 1998). Critically, the analysis reveals that higher credit risk and a higher probability of bank failure strengthen the impact of monetary policy on employment and payroll. This suggests that riskier banks, with less access to alternative funding, react more strongly to monetary policy changes by reducing their loan supply, which amplifies the effect on local economic activity.

The influence of local banking market concentration on monetary policy transmission is investigated using the Herfindahl-Hirschman Index (HHI). The results indicate that higher market concentration and higher average market shares of local banks dampen the effect of monetary policy on employment and total annual payroll. This finding complements the work of Adams and Amel (2011), who found a similar dampening effect on local lending at the metropolitan area level, extending it to the county level. However, the effect on average annual payroll remains largely unaffected by the HHI. To address concerns about endogeneity, the researchers re-estimate their regression model including both the HHI and the interaction between the HHI and the federal funds rate. The results remain consistent, showing that the effect of the federal funds rate on employment and payroll is negative across the whole distribution of the HHI, and that market concentration dampens the effects of monetary policy. The study also considers the roles of bank capitalisation, credit risk, and the probability of default, finding that higher capitalisation lowers the impact of monetary policy, whereas higher realized credit risk and probability of default amplify it.

The study concludes that monetary policy affects local economic outcomes (employment and payroll) as predicted by theory, regardless of the specific composition of local bank characteristics. The effect of a one-percentage-point increase in the federal funds rate results in a decrease in employment ranging from -0.15% to -0.17%, with a similar negative impact on total annual payroll. Importantly, the findings demonstrate a robust relationship between the health of the local banking system and the effectiveness of monetary policy. Well-capitalized banks, operating in less concentrated markets, experience a muted response to monetary tightening, likely due to easier access to alternative funding sources. Conversely, riskier banks, with less access to alternative funding, react more strongly, leading to a more pronounced impact on local economies. The study strongly suggests a need for coordination between monetary and macroprudential policies to account for the interaction between bank health and monetary policy effectiveness.