Peer to Peer Lending: The Relationship Between Language Features, Trustworthiness, and Persuasion Success

Abstract

This study examined the relationship between language use and persuasion success in the Peer-to-Peer (P2P) lending environment where unaffiliated individuals borrow money directly from each other using a textual description to justify the loan. Over 200,000 loan requests were analyzed with Linguistic Inquiry and Word Count (LIWC) software. The use of extended narratives, concrete descriptions and quantitative words that are likely related to one's financial situation had positive associations with funding success which was considered to be an indicator of trust. Humanizing personal details or justifications for one's current financial situation were negatively associated with funding success. These results offer insights into how individuals can optimize their persuasiveness by monitoring their language use in online environments.

Keywords:

• Persuasion
• Peer-to-Peer Lending
• Online
• Language
• LIWC

Acknowledgements

Many thanks to Jeff Hancock and Amy Gonzales for their helpful comments and suggestions on earlier drafts of this paper.

Notes

1. As exceptional loans can never receive more than 100% of requested funding and particularly poor loans can never receive less than 0% funding, there is a floor and ceiling on the possible range of values for the dependent variable. Therefore, if the data are treated as purely linear as is the case in a standard OLS regression, the large number of results bunched at 0 and 100% funding will bias the results. Thus, for the robustness check, a Tobit regression was used instead which accounts for this censoring of possible observations of the dependent variable. Results using the alternate Tobit specification are available from the author upon request.

2. The Probit regression does not have an equivalent R ² to that found in an OLS regression so the McFadden's Pseudo R ² value is reported instead. This is default Pseudo R ² value provided by Stata for a Probit regression but like most Pseudo R ² measures cannot be directly compared to standard R ² values. Results for alternate goodness-of-fit measures are available upon request from the authors.

3. All values reported in the text are from Table 3, which includes just the financial variables and theoretically supported language variables.

Additional information

Notes on contributors

Laura Larrimore

Laura Larrimore is in the Department of Communications at Ithaca College

Li Jiang

Li Jiang is in the Department of Communications at Cornell University

Jeff Larrimore

Jeff Larrimore is in the Department of Economics at Cornell University

David Markowitz

David Markowitz is in the Department of Communications at Cornell University

Scott Gorski

Scott Gorski is in the Department of Communications at Cornell University