New Study on Prosper Returns and Dynamics

06/10/08 posted by Chris Larsen

One of our primary objectives in making Prosper market data fully transparent and freely available via an API is to allow and encourage anyone to study the Prosper market and consumer credit markets in general. We deeply appreciate the level of diligence and analysis so many have contributed using Prosper’s marketplace data. We are also very encouraged that an increasing number of economists and professional credit analysts have taken the time to conduct deep studies of the market and share their conclusions publicly.

One such study was released yesterday by Economists Ginger Zhe Jin and Seth Freedman of the University of Maryland. The study looks at Prosper’s market since inception to determine average returns and other interesting conclusions about the market. Among their findings:

The average Prosper returns since inception were estimated at 6.05% with a 5.72% standard deviation and were trending up as credit quality continues to improve (see table 9 and figure 8.3).
The highest returns were in Near Prime loans (grades B-D) at an average of 8.27%, followed by Prime loans (grades AA & A) at 6.78% and sub-prime loans at 1.73%.
The probability of default of a Prosper loan peaks at month ten and then edges down (see Figure 9). This is a major reason why Prosper’s performance tool shows more conservative returns. Prosper’s performance tool does not adjust for this later default moderation, which is significant given that the average age of the Prosper portfolio is currently 9.7 months (i.e. at the peak of the default curve).
The study showed that the highest returns occurred for loans priced up to 25%. Loans funded at more than 25% actually showed lower returns because defaults increased disproportionately. This implies that borrowers willing to pay very high rates show higher adverse selection, which is logical (see Figure 8.1).
The study found that lenders learn quickly - adjusting their bids and strategies as performance is revealed. New lenders also learn from their observations of earlier lenders in adjusting to marketplace dynamics.

—

Chris Larsen is the CEO and Co-founder of Prosper.

Posted in Academics, Prosper, Prosper News

5 Responses

Joey Lawrance | June 10th, 2008 at 9:16 pm

I concur with the assessment. Incidentally, this is one of the reasons I don’t use portfolio plans: although I can specify a minimum interest rate, I cannot specify a maximum interest rate. Personally, I’m only comfortable bidding from 10% to 25%. Anything lower isn’t worthwhile, anything higher is too risky. Is there any work on screening listings based on the maximum interest rate?

Martin | June 11th, 2008 at 1:21 am

Joey, my thoughts exactly! I can only specify the minimum rate, not the maximum. Did anyone go through the study and read the assumptions made? Are they agreeable/realistic?

Yankeefan | June 11th, 2008 at 7:59 am

If we take figure 9 at face value as applying to all loans, it would indicate that many loans come current after 12 months or so.

More likely, the longer durations shown reflect a different loan mix than the more recent ones, with different risk characteristics.

Would that the authors had developed “roll rates” or rates of going late by loan age, rather than just an average percentage already late as shown in Firgure 9.

That said, I plan to read the report more carefully- and thank the authors for their work.