Detecting Financially Distressed Companies at the Earliest Possible Stage

BRAINKRUPTCY applies artificial intelligence models to estimate, as accurately as possible, the measure of bankruptcy risk that indicates the level of a company’s financial distress. BRAINKRUPTCY identifies companies which exhibit severe financial distress and are at imminent risk of filing for bankruptcy, based on these estimated measures of bankruptcy risk.

The empirically estimated measures of bankruptcy risk incorporate relevant information about a company. This information is processed through a bankruptcy risk model. BRAINKRUPTCY uses a two-step process:

In the first step, the model applies historical data of the previous ten years, which includes:

• Solvency status
• Comprehensive accounting and market data
• Shareholder structure
  
• Insider activities
• Macroeconomic data

In the second step, the most recent information about a company is applied to the model.

The applied bankruptcy risk model is updated at the beginning of each year and it is applied for the first time at the beginning of the second quarter. Changes in the estimated measures of bankruptcy risk are further updated with periodically disclosed information:

• Accounting data is disclosed quarterly in 10-Q filings
• Shareholder structure information is disclosed quarterly in 13F filings
• Macroeconomic data is released in a weekly or monthly cycle

For example, if a specific company files a 10-Q filing which provides new accounting information, the measure of bankruptcy risk for that specific company is updated to incorporate that most recent filing.

The applied bankruptcy risk model also incorporates daily market data on a company’s stock price and market value.

For the first trading day of each week, BRAINKRUPTCY provides updates to estimate the measure of bankruptcy risk for all listed US companies.

BRAINKRUPTCY empirically estimates the measures of bankruptcy risk for listed US companies. The measure of bankruptcy risk indicates the level of a company’s financial distress. The measure of bankruptcy risk is a metric measure, and ranges between zero and 1. Higher values indicate greater financial distress.

BRAINKRUPTCY displays the distribution of the measures of bankruptcy risk for all listed US companies. The distribution is skewed to the right – the vast majority of listed US companies are healthy. There are still a considerable number of companies that are either financially distressed or on the verge of bankruptcy.

An integral part of the applied bankruptcy risk model is the cut-off-rate, which represents the threshold between listed US companies that are financially healthy, and listed US companies that exhibit severe financial distress. The latter group of financially distressed companies specifically includes listed US companies which are at imminent risk of filing for bankruptcy within the next 12 months.

If the measure of bankruptcy risk is above the cut-off-rate, this indicates that the company is in severe financial distress and is at imminent risk of filing for bankruptcy.

An enhanced analysis of a company’s financial distress looks at the current measure of bankruptcy risk, and compares it to the previous measure of bankruptcy risk. Ernest Hemingway observed that bankruptcy occurs first gradually and then suddenly. The interrelation between the current measure of bankruptcy risk and the previous measure of bankruptcy risk provides further insights, such as indicating if the company is accelerating toward the sudden phase of potential bankruptcy.

The discriminatory power of the applied bankruptcy risk model can be assessed by analyzing the out-of-time Receiver Operating Characteristic (ROC) curve and the associated Area Under Curve (AUC).

If the ROC curve coincides with the dashed gray line, the AUC value is equal to 0.5. That means a bankruptcy risk model that distinguishes between solvent and bankrupt companies as accurately as tossing a coin. If the ROC curve coincides with the solid gray line, the AUC value is equal to 1.0, and the bankruptcy risk model offers a perfect bankruptcy prediction.

The BRAINKRUPTCY model of bankruptcy risk provides a very accurate bankruptcy prediction, because the historic AUC value since 2006 is in the range 0.86 to 0.95. The historic median value since 2006 is 0.92. The small gap to a perfect bankruptcy prediction is due to the variance of the measure of bankruptcy risk.

The cute-off-rate of the bankruptcy risk model applied by BRAINKRUPTCY distinguishes between companies that are financially healthy, and companies that exhibit severe financial distress. The latter group of financially distressed companies specifically includes listed US companies that will actually file for bankruptcy within the next 12 months.

Application of the cut-off-rate leads to two errors with respect to bankruptcy:

First, the alpha error indicates companies which are actually bankrupt, but which are mistakenly classified as financially healthy companies. The historic median value since 2006 is about 17.3 percent, which means that about 82.7 percent of bankrupt companies are correctly identified at an early stage.

Second, the beta error indicates companies which are actually solvent, but which are mistakenly classified as being at imminent risk of filing for bankruptcy. The historic median value since 2006 is about 13.6 percent. These beta-error companies exhibit severe financial distress, although they remain solvent within the next 12 months.

Avoiding investments in these financially distressed companies means that about four out of five actual bankruptcies will not have an impact on an investor’s performance.