Default or profit scoring credit systems? Evidence from European and US peer-to-peer lending markets

Abstract For the emerging peer-to-peer (P2P) lending markets to survive, they need employ credit-risk management practices such that an investor base is profitable in long run. Traditionally, relies on credit scoring predicts loans’ probability of default. In this paper, we use a profit approach based modeling annualized adjusted internal rate returns loans. To validate our models with traditional models, data from European P2P market, Bondora, and also random sample loans Lending Club market. We compare out-of-sample accuracy profitability within several classes statistical machine learning including following: logistic linear regression, lasso, ridge, elastic net, forest, neural networks. found outperforms standard for Bondora More specifically, as opposed across all are 24.0% (Bondora) 15.5% (Lending Club) higher, whereas 6.7% 3.1% higher proposed models. Moreover, results not driven by manual selection suggest investing more Finally, even if consider sampling bias, set superior consists almost exclusively Thus, contribute literature suggesting paradigm shift market prefer