Since 2013, we have observed an increasing number of failed Russian banks with negative capital and falsified financial reporting. We use previously unavailable data for the period 2010 -1H2015 to develop a logit model predicting the probability of bank failure with negative capital. In order to do so, we suggest solutions for the class imbalance and variable selection problems. The models chosen are confirmed to be robust and have longer forecasting horizons compared to previous research. Also, we implement a novel probability-based approach to the out-of-sample forecasting evaluation which confirms a good fit of the selected models to data. The model predicts bank failures in three quarters and finds 33% of actual failures among 5% of banks with the highest predicted probability to fail (out-of-sample). In addition, we make available previously unpublished banking data for Russia.

No abstract

We propose a new approach based on a generalization of the logit model to improve prediction accuracy in U.S. bank failures. Mixed-data sampling (MIDAS) is introduced in the context of a logistic regression. We also mitigate the class-imbalance problem in data and adjust the classification accuracy evaluation. In applying the suggested model to the period from 2004 to 2016, we show that it correctly classifies significantly more bank failure cases than the classic logit model, in particular for long-term forecasting horizons. Some of the largest recent bank failures in the United States that had been previously misclassified are now correctly predicted.

This paper investigates the phenomenon of hidden negative capital (HNC) associated with bank failures and introduces a product mismatch hypothesis to explain the formation of HNC. Given that troubled banks tend to hide negative capital in financial statements from regulators to keep their licenses, we attempt to capture this gambling behavior by evaluating product mismatches reflecting disproportions between the allocation of bank assets and the sources of funding. We manually collect unique data on HNC and test our hypothesis using U.S. and Russian banking statistics for the 2004-2017 period (external validity argument). To manage the sample selection concerns, we apply the Heckman selection approach. Our results clearly indicate that product mismatch matters and works similarly in both U.S. and Russian banking systems. Specifically, an increase in mismatch has two effects: it leads to a higher probability that a bank's capital is negative and raises the conditional size of the bank's HNC. Further, we demonstrate that the mismatch effect is heterogeneous with respect to bank size being at least partially consistent with the informational asymmetry view. Our results may facilitate improvements in the prudential regulation of banking activities in other countries that share similar features with either the U.S. or Russian banking systems.