The Practice of Credit Risk Modeling for Alternative Lending (303113)

Keywords: Alternative Lending, Credit Risk Scorecards, Logistic Regression, Model Monitoring, Banking, Financial Regulators, Web-Based Platform for Lending

In recent years, data scientists in the credit risk profession have experienced less freedom to deviate from industry-accepted practices because the biggest users of credit risk models (banks and large lenders) are the very institutions that face increased regulatory pressure since the 2008 crisis. Regulators have thus been less anxious to bless practices that deviate from the “scorecard and cutoff score” paradigm. But things are changing. The lending industry is demanding innovation and creativity from its data scientists more than ever. The rise of alternative lending, driven mainly by Marketplace lenders (aka Pier-to-Pier lenders), is built on the notion that web-based platforms can integrate customer-facing technology and Big Data to acquire, fund, and underwrite loans in a manner that is more targeted and efficient than banks currently do today. Big Data techniques have increased in popularity but logistic regression performed on structured data is as good an option for quantifying credit risk as ever. In the “new world” of lending it is not the model fitting techniques that need to change, as much as it is the treatment of samples, potential predictors, and model refits.