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

This paper studies factors that determine efficient information processing. I exploit a unique small business lending setting where the entire codified information set that loan officers use in their decision-making is observable (to the researcher). I decompose the loan officers' decisions into a part driven by codified hard information and a part driven by uncodified soft information. I show that a machine learning model substantially outperforms loan officers in processing hard information. Using the machine learning model as a benchmark, I find that limited attention and overreaction to salient information largely explain the loan officers' weakness in processing hard information. However, the loan officers acquire more soft information after seeing salient hard information, suggesting salience has a dual role: it creates bias in hard information processing, but facilitates attention allocation in new information acquisition.