Optimal Information Acquisition Strategies: The Case of Online Lending (2410.05539v1)

Abstract: Online lending has garnered significant attention in IS literature, particularly platform lending, but direct (balance sheet) lending is increasingly critical. This paper explores optimal information acquisition strategies for direct online lenders, addressing the broader question: Should a decision-maker rely on multiple lean experiments or opt for a single grand experiment? We first examine a model where an online lender issuing unsecured loans maximizes its expected NPV at an exogenous interest rate, finding that a lean experimentation strategy is optimal. However, when the interest rate is endogenous, the choice between lean and grand experimentation depends on the demand elasticity. If elasticity is increasing or constant, the lender prefers a grand experiment, offering the same loan terms in each period. We also analyze consumer segmentation and demonstrate how higher income variability benefits the lender through more effective experimentation. In addition, we investigate hybrid information architectures that combine dynamic experimentation with traditional static models. Our results show that the hybrid architecture enhances lender profitability, offering a flexible approach that integrates sequential learning with static information. The study contributes to understanding how different information architectures affect lending strategies, experimentation, and profitability in online lending.