Effects of Soft Information in Peer-to-Peer (P2P) Lending

Grant Winners

Deans

Abstract

Peer-to-peer (P2P) lending is the debt-based subset of crowdfunding, which is an open call through the Internet for the provision of financial resources by multiple investors to fund products, projects and business ventures. P2P lending success is the result of technological advances creating the efficient transfer of funds between capital-seekers and capital providers. Prior to 2015, the leading sites like Prosper and Lending Club allowed borrowers to voluntarily provide personal information such as marital status, number of children and photos, and respond to peer lenders' comments. The literature has found that these unstructured soft facts help to establish trust and positively influence the likelihood of funding, interest rates and loan defaults.

Recently, in pursuit of fair pricing and openness, dramatic changes occurred in January 2015 to remove such soft information across the leading P2P lending platforms. Questions related to the impact of removing soft information need to be answered. Currently, P2P lending research under the new disclosure policies and fixed rate structure has not been published. A comparison study of funding probability, time and amount, default probability, and investor returns on P2P lending platforms is needed to close this gap in the field. This study has significant practical implications too. As millions of borrowers, mainly startup business owners, seek funding opportunities and more number of individual investors seek financial returns from P2P lending sites, our study will present deeper understanding of the determinants of funding amount and time, and eventual loan defaults. Borrowers and lenders will benefit from our study results as well as policy makers like the U.S. Securities and Exchange Commission and Small Business Administration.

Our research is data intensive. For our research questions, we plan to collect the following data for all loans from Prosper and Lending Club: Funded loan amounts, loan description launch date and time, funding conclusion time, number of lenders, loan expiration date, borrower credit scores, loan default, etc. Monthly measures of Consumer Sentiment Index will be collected from the University of Michigan. We will spend significant time to collect borrower identity information from millions of loan description pages, using big data algorithms and numerous iterations. After analyzing descriptive statistics and correlations of the sample variables, we plan to perform parametric and non-parametric tests of mean difference, comparing the periods before and after the policy change. Then we will use generalized maximum likelihood regressions for the research questions related to funding amount and funding time. For default-related hypotheses, we plan to use binary logistic regressions. Additional robustness tests are expected.