Assessing Human Information Processing in Lending Decisions: A Machine Learning Approach

Effective financial reporting requires efficient information processing. This paper studies factors that determine I exploit a unique small business lending setting where am able to observe the entire codified demographic and accounting set loan officers use make decisions. decompose officers’ decisions into part driven by hard uncodified soft information. show machine learning model substantially outperforms in processing Loan can only process sparse of useful identified focus their attention on salient signals such as large jumps cash flows. However, “red flags” highlight acquire more result suggests is an allocation device: It guides humans allocate limited cognitive resources information, task which have advantage over machines.