Decoding risk assessment: Exploring the efficacy of credit scoring vs. profit scoring in P2P lending

The conventional practice in the financial industry involves the assessment of creditworthiness through default scoring models, primarily focused on predicting the likelihood of default. However, this research paper challenges the status quo by introducing benefits of the concept of profit scoring – a novel approach that shifts the focus from mere default prediction to forecasting potential returns on a debtor when extending credit to clients. Using publicly available peer-to-peer (P2P) lending data, this study meticulously examines the efficacy of profit scoring compared to traditional default scoring methods. Through the implementation of machine learning (ML) algorithms, our analysis demonstrates that profit scoring outperforms default scoring in average return on portfolio of out-of-sample credit applications. In order to achieve comparable results, the paper uses the same family of algorithms for both tasks: profit scoring (predicting return on a loan) and default scoring (classifying defaults and non-defaults). Furthermore, the paper reveals that the factors influencing loan profitability exhibit disparities when compared to those driving default probabilities. The results underscore the transformative potential of profit scoring, offering lenders a more holistic perspective on risk assessment. By focusing on potential internal rate of return rather than default probability alone, financial institutions can make more informed decisions, leading to enhanced portfolio performance and profitability.