by WAIN Street
A firm’s financial health determines its ability to repay debts. The promise of big data is to incorporate an ever increasing pool of information about a firm’s financial behavior to make better predictions. And by blending in qualitative factors such as social media chatter, big data seeks to make a quantum leap in assessing business credit risk.
For smaller firms, there is another factor that has thus far been largely ignored—the credit risk of peer firms as identified by industry, geography, size, etc. True, economy- and industry-wide trends are routinely analyzed, provide underwriting guidelines, and influence individual credit decisions. But thus far, credit quality assessment of peer firms has never been systematically blended into a firm-level creditworthiness gauge. Perhaps, because the data necessary to draw usable conclusions has not been readily available.
The reason for systematically using information about peer firms in assessing risk is simple: Main Street suffers together. Ask any local economic development expert. It’s the bane of their existence. Businesses suffer together when enveloped by a local major employer. Businesses also suffer together when an industry is disrupted. Remember the local video rental store? Yes, smaller businesses can make a stand. Witness the independent bookstores and local cafés that continue to thrive. And yes, David v. Goliath makes for interesting cocktail conversations. However, that is not the experience of the vast majority of businesses as they run out of cash after being torpedoed by events external to them. For smaller firms, what happens to their peers is foretelling.
Last spring, we introduced the Credit Quality Map – a ranking of business segments based on state and industry sector credit quality. We have tweaked it a little. And tested it a lot. And we find it to be consistent in ranking business segments. About two-thirds of the time, segment credit quality is sticky—it remains the same over the following twelve months. Stated for the benefit of a lender, with about 60% accuracy, the Credit Quality Map can tell a lender today how a specific segment will rank twelve months hence.
The next step is to evaluate how blending peer segment knowledge with firm-specific credit assessment can better predict borrower performance. There is no FICO for business credit risk. There are some nationally recognized business credit scores; some younger players; and every new entrant in the burgeoning alternative lending space has a big data inspired credit scoring model. We would like to evaluate each one of them. But for now, we consider a leading, nationally recognized score provider.
A Simple Test
Our goal is to answer the question: What could we have told a lender last year that would have helped improve performance?
Using a nationally recognized provider, we obtained the August 2013 and August 2014 credit scores for 5,000 randomly selected businesses having fewer than 20 employees. We created an Enhanced 2013 score by blending the August 2013 Credit Quality Map with the August 2013 score. We used the August 2014 score as a proxy for current performance.
The Enhanced 2013 score is stricter. In the table below, the Cutoff Percentile value represents the percentile below which a firm is classified as good using the 2014 score. The Pass Ratio represents the ratio of firms classified as good by the Enhanced 2013 score versus the 2013 score.
Table 1: Failure analysis of 2013 score versus Enhanced 2013 score.
The Enhanced 2013 score has a lower Miss Rate – the percent of bads that were not identified. By using the Enhanced 2013 score, a lender would have rejected more badcredits up-front.
The Enhanced 2013 score offers lenders a second level filter for decisioning. In the table below, the rows and columns represent deciles of the 2013 and Enhanced 2013 scores, respectively. The cell values are firm counts. Three regions have been colored to illustrate the advantage provided by the Enhanced 2013 score in the case of a 40% cutoff – accept firms ranked at or below the 40th percentile for risk. The green region represents agreement between both the scores. In this case about 60% of the firms that were accepted by the 2013 score were also accepted by the Enhanced 2013 score. The orange region represents firms that might benefit from an additional review. The Enhanced 2013 score ranks about 40% of the firms accepted by the 2013 score at a higher risk. The yellow region represents additional opportunity. The more restrictive Enhanced 2013 score expanded the pool of borrowers by about 2.5% by ranking firms rejected by the 2013 score as less risky.
Table 2: Decisioning advantage of Enhanced 2013 score.
Eliminate the Bad and Focus on the Good
By blending a firm’s credit score with an assessment of the firm’s peers, lenders can improve their decisioning process. Lenders can reject more future problem borrowers; identify and review borrowers that merit a deeper dive; and identify creditworthy borrowers that would have been otherwise ignored.
