How to Make Better Lending Decisions with a Probability of Default Model
As technology and data collection improve, banks and credit unions are finding ways to use this information to improve their loan decision making and thus improve their asset quality in the long term.
The best way to maintain asset quality is to make good loans…and that involves validating that borrowers will be able to service debt, even if economic conditions change in the future. However, qualifying borrowers is always easier said than done, as bankers must:
• Understand what data needs to be collected from borrowers
• Spread the financial statements
• Calculate a global cash flow
• Use ratios to benchmark vs. other businesses in the same category
• Figure out how the proposed loan impacts the borrower’s cash flow
The days of using a gut feel for loan decisioning are over, and Big Data or at least data-driven decisioning is preferred. Banks and credit unions have access to more data than they can use to help them identify potential customers, potential cross sales, understand relationship profitability and riskiness of borrowers – predicting future defaults and distinguishing good customers from bad.
But even if a financial institution doesn’t have a complete data modeling team on staff, a small amount of financial data from the borrower can improve the quality of information used in decisioning.
A probability of default model (PDM) is a system for objectively quantifying future credit risk. It is not a new concept, but it is gaining popularity within banks and credit unions that are seeking defensibility and efficiency in credit management. A PDM looks at multiple characteristics of the borrower to assess the risk of a loss from failure to make loan payments. Often these models are customized by industry classification of the borrower and loan type to capture the risk of particular industries.
Probability of default models will require different information depending on whether it is being applied to a specific loan or an entire portfolio. When analyzing a specific loan, a PDM can help in the following situations:
• Pre-screening borrowers without investing the time and resources required for a full spread
• Substantiating credit analyses with an automated and predictive measure of risk, especially for new relationships
• Reducing the weight that more subjective risk measures carry in risk ratings
• Systematizing and automating, in some cases, the risk rating for loans and delineating between loans that often pool in one or two risk ratings
• Understanding the global probability of default
• Guiding loan pricing to more closely reflect risk
• Identifying loans or prospective loans that may need tighter covenants or administrative controls put in place
• Simplifying annual loan review for applicable loans to save time for the loan officers or administrators
• More quickly identifying loans that may in time default, giving the relationship manager (loan officer, analyst, administration officer) more time to right the situation• Providing an added layer of quantifiable and defensible risk measurement, which could make for easier exam and audit relationships
While developing and implementing a PDM at a financial institution can be a slow process, it can help improve the credit analysis process significantly and help make loans more defensible to examiners. It can also be a source of efficiency, giving institutions a competitive advantage in their response times to borrowers.