Probability of Default Models
Here, and later in the paper, the default is understood as one of the following signals for its registration:
• A bank’s capital sufficiency level falls below 2 %.
• The value of a bank’s internal resources drops lower than the minimum established at the date of registration.
• A bank fails to reconcile the size of the charter capital and the amount of internal resources.
• A bank is unable to satisfy the creditors’ claims or make compulsory payments.
• A bank is subject to sanitation by the Deposit Insurance Agency or another bank.
We propose a forecast probability of default (PD) model, which is based on the relationship between banks’ default rates and public information. We have constructed a quarterly bank-specific financial database on the basis of Mobile’s information from 1998 to 2011: data in accordance with Russian Financial Reporting Standards, taken from bank Balance sheets and Profit and Loss statements.
During a 14-year period there were 467 defaults in compliance with our definition, as well as 37 bank sanitations. The quarterly database created has a good coverage of default events and the banking sector. We have applied a binary choice logistic model to forecast default probability. The maximum likelihood approach is used to estimate the model. The sample was split into two parts: “1998-2009”—to estimate models, and “2010-2011”—to test the predictive power of the models.
Financial ratios used as explanatory variables were determined from the literature review and common sense. They were tested on their separating power between bankrupt and healthy banks, as well as being divided into blocks according to the CAMELS methodology. We have also employed non-linearities in our model and found the optimal lag on financial ratios.
• Macroeconomic variables are highly correlated, and there were only two variables used in order to account for the effect of the macroeconomic environment on bank performance: quarterly GDP growth rates and the Consumer Price Index. We also controlled for the impact of the following on a bank’s default probabilities:
• Monopoly power of a bank on the market (with the Lerner index).
• Its participation in a Deposit insurance system (with a dummy variable).
• The territorial location of the bank’s operational activity (Moscow or regional)
Our key findings (Karminsky et al. 2012) were that:
• Banks with extremely high and low profitability have higher default rates due to their impact on the default probability of the profit-to-assets ratio (poor and risky banks).
• Banks with a higher proportion of corporate securities in assets carry a higher risk of a price crash on the market.
• Lower turnover on correspondent accounts in comparison with total assets increases the probability of default (a bank’s potential inability to make payments).
• Banks with a considerable number of bad debts are less stable.
Additionally, a growing consumer price index increases a bank’s default probability:
• Inflation reduces the real return on loans.
• Depositors are able to withdraw money and deposit it into the bank again at a higher interest rate or spend it.
We have also found that banks with a higher monopoly power are financially stable. Moscow-based banks have higher PDs on average.
We have found no evidence that a bank’s participation in the Deposit insurance system influences its PD. The explanation is that the set of System participants is too diversified. The out-of-sample prediction performance of the model (for 20102011) is prominent: over 60 % of bank failures were correctly classified with a moderately sized risk group.
At the same time, the developed model underestimated the default probabilities for 2009. This result reveals some unrecorded channels that significantly increased the risks during the period of the recent financial crisis.