Pitfalls of traditional methods for predicting default
By Mary Ellen Biery, Research Specialist, Sageworks
Exposure to credit risk continues to be a focus of regulators and financial institutions years after the financial crisis, but traditional methods of predicting default may have drawbacks. Institutions that use “homegrown” scoring methods or proprietary scoring models should be cognizant of problems related to the reliability and consistency of data that is used to create the model and perform the analysis on a specific borrower. This post examines some of the pitfalls of the traditional methods.
Lack of standardization and objectivity
Currently, it is difficult for banks and corporate credit offices to easily quantify the credit risk associated with private-company commercial lending in an automated and standardized manner. Bankers and other credit officers rely on the borrower’s credit score along with a manual reading of the business’s financial statements for signals of their financial health and creditworthiness. “Homegrown” scorecards may be so informal or easily altered that different users at a bank can employ their own methods of inputting data or analyzing results to achieve their desired outcome. They may also lack external validation, which makes it difficult to defend with bank examiners. Additionally, some vendor models may be built on data that isn’t standardized, such as self-reported survey data or payment histories from unreliable or inconsistent reporting sources.
Unrepresentative data
The data sources used to develop proprietary, vendor scoring models may not reflect the types of businesses or loans that will be analyzed. For example, a model built using public company data may not be the best predictor of private company default, since they don’t have similar requirements for auditing financials and making financial information public. In the same way, a model built using data from global companies may not provide the best method for predicting default by U.S. firms. Further, a model built to assess manufacturing firms isn’t ideal for evaluating retailers, just as a model built using large companies may not be best for predicting default by smaller companies.
Myopic focus vs. global assessment
Most models for scoring business credit risk do not take into account the owner(s)’ personal financial situation and its potential impact on the probability of default. Instead, they rely on business financials only. That’s a big issue when it comes to scoring privately held companies, especially smaller ones.
For more in-depth information on the pitfalls and how a probability of default model can help, download the whitepaper titled: Pitfalls of Traditional Methods for Predicting Default.