A Proposed Simulation Technique for Population Stability Testing in Credit Risk Scorecards

Credit risk scorecards are logistic regression models, fitted to large and complex data sets, employed by the financial industry model probability of default potential customers. In order ensure that a scorecard remains representative population, one tests hypothesis population stability; specifying distribution customers’ attributes constant over time. Simulating realistic sets for this purpose is nontrivial, as these multivariate contain intricate dependencies. The simulation practical interest both practitioners researchers; may wish consider effect specified change in properties has on its usefulness from business perspective, while researchers test newly developed technique credit scoring. We propose based specification bad ratios, explained below. Practitioners can generally not be expected provide parameter values scorecard; models simply too many parameters make such viable. However, often confidently specify ratio associated with two different levels specific attribute. That is, comfortable making statements “on average new customer 1.5 times likely an existing similar attributes”. method which used obtain ratios. proposed demonstrated using example, we show simulated adhere closely paper provides link Github project R code generate results.