Early Warning Models of Bank Failures: Understanding Bank Bankruptcies during the Late 2000s Great Recession

Summary

This project analyzed the significance of predictors of bank failures using financial performance and macroeconomic variables. The analysis involved both failed and surviving commercial banks that operated shortly before and during the late 2000s Great Recession. Early warning models were developed for several 6-month periods going backwards to around 4 years before actual bank failures occurred. Among the significant predictors identified were more costly funding arrangements, higher interest rate risks, increasing real estate, consumer and industrial credit concentration, declining profits, and deteriorating liquidity conditions. The most compelling result was the insignificance of agricultural lending variables that refute the contention that agriculture is a riskier borrowing sector.

Situation

The global economy experienced a general slowdown in economic activity in the late 2000s that economists and business analysts consider as the worst economic crises experienced since World War II and the longest downturn since the 1930s Great Depression. Dubbed as the Great Recession (Wessel, 2010), the worsening global economic conditions began in December 2007 as declared by the National Bureau of Economic Research (NBER) that took cues from the deteriorating conditions in the labor market (Isidore, 2008). The United States economy was not spared from such global crises. In the local economy, the period of the late 2000s was marked by trends of high unemployment, declining real estate values, bankruptcies and foreclosures, among many other indicators (Rutenber and Thee-Brenan, 2011). A widely accepted theory of the real culprit that significantly launched the onset of the economic crises in the U.S. was the breakdown of the real estate industry (Isidore, 2008). The housing downturn started in 2006 when housing process dropped significantly after reaching peak levels in the early 2000s. This resulted in an abrupt increase in loan defaults and mortgage foreclosures that led to widespread crises in the banking industry. The late-2000s financial crisis led to a surge of bank failures in the United States at an overwhelming rate not observed in many years. The cycle of seizures started in 2007, and by the end of 2010, a total of 325 banks have failed. In contrast, only 24 banks had failed in the seven-year period prior to 2007. California, Florida, Georgia, and Illinois were among the states hardest hit by bank collapses, with 34, 45, 52, 38 failed banks, respectively, since 2007.

Response

Research was conducted to discern the farm sector's role or influence in the ensuing economic crises. Specifically, the research addressed the issue of whether or not agriculture, which is more vulnerable to a wider range of risk and uncertainty factors, has significantly influenced the incidence or probability of banking failures during the current recessionary period. The study adopted a more comprehensive approach in understanding bank failures by a retrospective approach to bank failure prediction involving several time period models. Technical efficiency analysis was also employed to allow the comparative evaluation of internal and external factors that could affect a bank's probability of failure. A bank failure regression model was developed using several cross-sectional datasets for commercial banks from the Banking Call Report database available from the Chicago Federal Reserve Board's website. The banking datasets were also supplemented by macroeconomic indicator variables, compiled at the state level, to determine whether the banks' internal decisions were enhanced or adversely affected by prevailing industry or economic operating environments. The model was applied to different time periods starting from an initial period of six months from actual bank failure, and re-tracking backwards in six-month period increments until 4 years (48 months) prior to the bankruptcy declaration, defined the . These models determined how far back in time were the early warning signals of bank failure detected. Moreover, a technical efficiency model based on the stochastic cost frontier framework was also developed. The calculated technical efficiency scores were endogenously determined by an array of instrumental variables that include the bank performance factors considered in the bank failure prediction models. The relative performance of the TE variable, which was used as a collective measure of overall bank financial performance, vis-à-vis variables that capture the prevailing macroeconomic conditions, was analyzed in a system of equations.

Impact

The results of this research will be very useful in both understanding the ensuing bank crises as well as for developing cautious, more prudent strategic plans to maintain the viability of existing banking institutions. This study's bank failure prediction models produced results that identified important early warning signals that could be detected as far back as 3 to 4 years prior to a bank's declaration of insolvency or bankruptcy (herein referred to as failure). As early as that time (3 to 4 years from bank failure), the ones that are showing early signs of trouble (or exhibiting trends leading to eventual bank failure) are banks that had to resort to more costly funding arrangements (relying more heavily on more costly sources in the national funds markets), experiencing higher interest rate risk (through the uneven mix of short-term assets and liabilities), registering a downward trend in business profits (if not incurring losses yet), and facing pressure to sell less risky assets to improve liquidity conditions. As the time period approaches the eventual bank failure, increasing trends of delinquencies among industrial/commercial and consumer loans become important early warning signals, in addition to the factors already identified as problem areas in the earlier time period models. The most compelling result in the analyses of early warning signals is the notable insignificance of any measure related to the banks' agricultural loan portfolios. Even agricultural real and non-real estate loan delinquencies have not been established to significantly influence the likelihood of bank failure across all time period models. These results confirm our contention that exposure to a seemingly riskier and more uncertain agribusiness operations does not necessarily enhance a banks' tendency to fail. The fact that agricultural loans' delinquency rates are consistently below the banks' overall loan delinquency rates also suggests that either agricultural lenders are generally more cautious in making credit decisions or that agricultural borrowers are actually more prudent in the borrowing decisions especially during recessionary times. The TE analysis also allows the validation of the relative financial strength of agricultural banks vis-à-vis their non-agricultural counterparts. Results of this analysis confirm that successful (or surviving) agricultural banks have been shown to be operating more efficiently than surviving non-agricultural banks. This result only helps refute the contention about the relative higher level of riskiness of loans extended to farm borrowers. The agricultural banks' average TE scores also have been dominant in comparisons between agricultural and non-agricultural failed banks. This study has provided an important clarification on the economic impact of the agricultural sector in the lending industry.

State Issue

Agricultural Profitability and Sustainability

Details

• Year: 2011
• Geographic Scope: National
• County: Clarke
• Program Areas:
  • Agriculture & Natural Resources