Making Sense of Soft Information: Interpretation Bias and Ex-Post Lending Outcomes

Abstract

We explore whether behavioral biases impede the effective processing and interpretation of soft information in private lending. Taking advantage of the internal reporting system of a large federal credit union, we delineate three important biases likely to affect the lending process: (1) limited attention (or distraction), (2) task-specific human capital, and (3) common identity. Specifically, we find that using soft information in lending decisions leads to worse loan quality when loan officers are busy or before weekends and around national holidays, when loan officers have earlier sales experience, and when both the officers and borrowers are men. Overall, we provide novel evidence of non-agency-related costs in the use of soft information in lending decisions.