September 28 – Financial lending institutions use a variety of measures, such as personal demographics and financial and career history to make decisions on loan applications. New research from Columbia Business School suggest that institutions should add a new measure – the verbiage authored by applicants on loan applications to determine the likelihood of default.
Research by Oded Netzer, professor of business and marketing at Columbia Business School, together with Alain Lemaire, doctoral student at Columbia Business School, and Michal Herzenstein, professor at the University of Delaware found that a borrower’s word usage is predictive of future payment at a similar scale as their financial or demographic information. In fact, the textual information had similar predictive power as the more commonly used credit grade, demographics and other financial measures.
“When it comes to applying for a loan, everyone has a ‘tell,’ ” said Netzer. “Our findings suggest that similar to body language effects, people leave traces of their intentions, their financial and emotional states, as well as their personality traits in what we term the involuntary sweat in the words.”
The paper, When Words Sweat: Identifying Signals in the Text Loan Applications, determines that loan defaulters are more likely to exhibit the writing styles of extroverts and liars. Specifically, a borrower was more likely to default if the following criteria were present on their applications:
- They avoid mentions of themselves;
- They use short-term focused words such as a “day or “week”;
- They mention financial and general hardship (e.g. bankruptcy, child support, divorce);
- They tend to use more explanation words such “explain,” “why,” “because”;
- They adopt appreciative and good-mannered words like “God bless” and “thank you”; or,
- They reference other sources of inspiration such as God, family members, and chance
Conversely, borrowers who use achievement words such as “promotion” or “graduation,” demonstrated through their word usage high degree of financial literacy, or who used long-term focused words such as “one year,” were associated with positive loan repayment. In sum, the findings imply that whether intentional or not, borrowers’ writing disclose their true situation and/or intent.
About the Research
The research uses data from the site Prosper.com, where potential borrowers submit requests for loans for a specific amount (between $1,000 and $25,000 for this study), as well as a specific maximum interest rate they are willing to pay. The authors used text-mining and machine-learning tools to process and analyze over 138,000 loan requests. The results were obtained after controlling for the borrower’s credit grade as well as other loan-specific and borrower-specific financial and demographic, which captures the financial implications of the borrower’s life circumstances, and the interest rate given.
To learn more about the research being conducted at Columbia Business School, please visit www.gsb.columbia.edu.