Measurement of Core Variables in Banking
Sunday, Jan. 6, 2019 8:00 AM - 10:00 AM
- Chair: Claire Brennecke, U.S. Federal Deposit Insurance Corporation
Banking Operational Footprints: Measuring Geographical Presence
AbstractThis paper proposes a framework for measuring and quantifying a bank's operational footprint in terms of physical geography. Within this framework, we investigate the various data sources available to regulators and the general public in order to construct this measure. We also provide useful information in helping a researcher determine how to adjust their measure in light of the data limitations that might exist. Using branch location as proxies for the location of either depositors or borrowers, researchers typically extend their analysis to the effect of geography on a variety of issues including credit risk, funding risk, real economic impact, M&A or competition, and much more. Currently, publicly available geographic data on deposits is limited to Summary of Deposits (SOD) data provided by the FDIC on an annual basis. On the asset side, there is Community Reinvestment Act (CRA) related data (including HMDA) for banks. We have access to a unique dataset of loan and depositor information taken at the time of bank failure. We cross reference all data sources and provide insight on how representative and accurate SOD data is relative to the true state of geographical dispersion and provide some insight on how the accuracy changes depending on a few observable or controllable factors.
Nonlinear Bank Loan Loss Provisioning
AbstractLoan loss provisions are one of the most critical accruals reported by banking institutions. Charged against net income, loan loss provisions directly affect a bank’s profitability, regulatory capital, and the net carrying value of loans (via a contra-asset valuation allowance account). Understanding how banks provision for future credit losses is an important topic that draws increasing attention from both academics and regulators. An extensive body of literature has used statistical models to analyze the behavior of loan loss provisioning. The routine approach is to estimate a linear regression of loan loss provisions on some nondiscretionary predictor variables that proxy for new information regarding changes in loan portfolio quality (e.g., changes in nonperforming loans), treating the residuals of the regression as the “discretionary” portion of loan loss provisioning. The predominance of this linear estimation strategy is evident in Beatty and Liao’s (2014) survey, where all the nine reviewed models adopt a linear specification. While conceptually intuitive and appealing, the linear specification imposes a specific functional form linking loan loss provisions to changes in expected future cash flows from loans whose validity has been seldom tested. We find evidence that this linear specification poorly fits the observed pattern of loan loss provisioning, casting serious doubt on models that use the residuals from this model. On average, the linear specification will underpredict discretionary provisioning for most banks - those with little change in their portfolio quality, while overpredicting provisioning for banks that do see a significant shift.
Loss Given Default, Loan Seasoning and Market Fragility: Evidence From Commercial Real Estate Loans at Failed Banks
AbstractCommercial real estate (CRE) loan losses are a major contributor to bank failures, yet they are not well understood. We present new evidence of a relationship between loan seasoning (loan age at default) and loss given default (LGD), using a newly available dataset of over 14,000 distressed CRE loans from 295 banks that failed and were resolved by the FDIC during the recent financial crisis. Loan seasoning is negatively related to both the probability and severity of loss, and is distinct from other factors such as loan amortization over time, asset price changes and loan vintage effects. This suggests a new theory for credit cycles in periods of high CRE loan growth – even if the origination quality has not deteriorated – as aggregate losses may be higher due to the changing composition of loan seasoning in the industry. In addition, we find several other unique results related to LGD in times of distress.
- G0 - General
- E3 - Prices, Business Fluctuations, and Cycles