PhD candidate publishes empirical analysis of Prosper

Sanjeev Kumar, a Doctoral Candidate at the University of Michigan, has written an interesting paper about Prosper called Bank of One: Empirical Analysis of Peer to Peer Financial Marketplace. He just presented the paper at Americas' Conference on Information Systems and has made a copy of the PowerPoint presentation here. It's a fascinating academic study on Prosper and may be the first of it's kind. Here is the paper's abstract:

Peer to peer financial marketplaces provide a platform for individual lenders and borrowers to interact and transact. These marketplaces disintermediate the traditional financial services business models. In this exploratory paper we study the operation and effectiveness of one such marketplace: Prosper.com. We analyze six months of lender, borrower and loan repayment data to answer preliminary research questions about lender behavior, market effectiveness and antecedents of loan default. We show that lenders mostly behave rationally and charge appropriate risk premiums for antecedents of loan default. We also show that there are mismatches between risk premiums charged and relative importance of factors that drive loan default. We then explore the dynamic process of lenders adjusting their lending strategies to reduce these mismatches. We analyze the effectiveness of the group reputation used in the marketplace and show that it is not effective in promoting good borrower behavior. Our analysis provides a base for future research in this exciting and evolving context. Our results provide directions for practice applications as well as future research in design of financial marketplaces, investing and risk mitigation strategies and improving the effectiveness of peer-to-peer financial marketplaces.

Kumar attempts to answer the following research questions:

• Do lenders follow rational lending practices while lending on the marketplace? That is, what factors affect their lending strategy and whether they are in line with rational expectations?
• Do lenders follow efficient lending practices while lending on the marketplace? That is, what are the antecedents of loan default and whether lenders charge appropriate risk premiums for factors that drive loan defaults?
• What is the impact of group reputation systems on borrower and lender behavior?
• Is the marketplace evolving to achieve higher lending efficiency? That is, are lenders adjusting their lending strategies and charging appropriate risk premiums as more information becomes available on antecedents of loan defaults and the relative importance of these factors that drive loan defaults?

Here's a look at a portion of his results:

The results are certainly interesting. One thing that catches my eye is the homeowner determination. Based on Matt's analysis, homeowners are actually a greater risk of default than non-homeowners. Yet, according to the dataset Kumar used, being a homeowner is not a significant variable but Prosper lenders bid down homeowner loans. A careful look at Kumar's dataset could help increase returns for lenders.

Later this week I intend to provide a deeper look at Kumar's paper but for now check out his PowerPoint.