The problem with peer-to-peer lending

 

Mark Gimein has an important story today about Prosper, which is required reading for anybody — myself included — who has some hope about peer-to-peer lending helping to disintermediate banks and get credit flowing again to individuals and small businesses.

Prosper was one of the first companies to market in the peer-to-peer space, and in hindsight its arrival at the height of the credit boom was incredibly ill-timed. If you lent money through Prosper back then, when most of its loans were extended, there’s a very high chance that you’ve lost money — in some cases, a lot of money. “Of investors with a portfolio of loans that are an average of at least two years old,” notes Gimein, “folks who have lost money outnumber those who’ve earned 6 percent annual return by more than six to one.”

Prosper was founded in the days before everybody had heard the term “model risk”, back when it made perfect sense that credit risk could be modeled accurately by a computer algorithm using little more than a FICO score.

One of the big problems that Prosper ran into — the massive credit crunch and the ensuing Great Recession — could reasonably be considered to be a one-off event with a low likelihood of happening again. But another is endemic to the model: Prosper borrowers with a given FICO score are inevitably going to be more likely to default on their debts than most other people with the same credit score.

It wasn’t meant to be that way. Peer-to-peer lending was meant to create a personal connection between borrower and lender, and therefore make borrowers more likely to repay their debts than people faced with large obligations to hated, faceless banks. But it seems that adverse selection effects overwhelmed the site’s attempts to be warm and fuzzy. As Gimein explained in an email to me,

I think in this case the adverse selection issues are insurmountable. Folks go to P2P loans almost always because they can’t get money through conventional channels, and often there is a reason. I’ve cut up the Prosper numbers in a bunch of ways, and one thing I’ve noticed is that some of the worst returns come from folks with okay credit who are willing to pay very high interest rates: they’re willing to pay a lot because their finances are in worse shape than they seem.

Also, re: adverse selection, this reminds me of a story I heard years ago from a guy who started a company that marketed credit cards online. What he found was that when you set up a site and have people come to you, you get a really dangerous class of borrowers. This is why credit card companies don’t really make much effort to get people to go to their websites and apply. They would rather *offer* than let people ask. Because ultimately a paradox of lending is that the people who are more likely to repay are those who *don’t need the money*. And Prosper attracts those who do need it.

I’m still hopeful that Lending Club, in particular, can succeed in this space; it certainly doesn’t suffer from the kind of egregiously misleading public communications that Gimein details at Prosper. But insofar as Lending Club can succeed where Prosper seems to have failed, it will have to do so through overcoming the adverse-selection problem with an extremely tough and diligent underwriting program which rejects as much as 90% of the people asking for loans. As such, it’s only likely to make a difference at the margins — and it might find it hard to make a profit itself out of the 1% take it skims off each loan.

Effective underwriting is difficult, labor-intensive, and expensive. And there’s always a worry that at some point any peer-to-peer intermediary will start cutting corners on the underwriting front in an attempt to make more money. Which could be disastrous for lenders.

Update: Gimein points me to the second chart on this page, which shows that Prosper, too, funds less than 10% of the loans that get applied for there. Maybe rejecting lots of loans isn’t, in and of itself, a sign of diligent underwriting.

Update 2: Lending Club CEO Renaud Laplanche emails to explain the difference between his shop and Prosper:

The “10% funding rate” of Prosper and Lending Club are different in nature: Lending Club approves 10% of the loan applications – that’s an underwriting decision. These 10% most creditworthy loans are made available on the platform for investors to invest in, and all loan listings get fully funded. Currently, the platform is “demand constrained”, meaning that we have more investors willing to invest in these loans than loans available. We are increasing our marketing efforts on the borrower side to make sure demand catches up with supply.

Prosper’s 10% is very different in nature: most loan applications received by Prosper get listed on their platform, and only 10% actually get funded, either because of insufficient supply of investors funds, or just because investors don’t want to fund the other 90% of the loans. The question here is whether the 10% that get funded are “the right 10%”? Marketplaces need 2 things to be efficient: sufficient supply and demand, and no information asymmetry between buyers and sellers. I believe Prosper’s marketplace lacks both. So would Lending Club’s by the way, which is why we are making the credit decisions and setting the rates. The consequence is that Lending Club’s 10% are those loans that are the most creditworthy, based on factual information from the credit reports, employment and income verification, anti-fraud measures, etc.

 

An Open Letter to The Big Money: Retract Your Story

01/19/10 posted by Prosper Blog

Today, The Big Money, an outlet we and many others respect, published a disappointingly inaccurate story about Prosper and the peer-to-peer lending industry.  It’s unfortunate that the author’s data analysis and perspective relied almost entirely on a hodgepodge of anonymous sources.  If higher reporting standards had been upheld, the reality that Prosper has shown great promise and performed well on a relative basis over the last three-years would have been self evident.

We’ve requested that the editors at The Big Money retract Mark Gimein’s erroneous perspective on Prosper and the peer-to-peer lending industry.  We look forward to their response.  In the meantime, we’d like to set the record straight.

Mr. Gimein discusses Prosper’s loans in the context of only cumulative unit default rates rather than in terms of the average annual returns lenders have earned.  For example, Mr. Gimein states that 39% of loans that have had a chance to come to maturity (originated prior to 12/31/2006) have defaulted.  What he doesn’t say is that the annual yield on these loans was 16% and the annual loss experienced by lenders was actually 20%, resulting in an annual average return of negative 4%.  Although this return is negative, put in the context of the largest recession in generations, and the performance of other asset classes during the same time period, this paints a very different and more accurate picture of how lenders have fared on Prosper.

Mr. Gimein continues to use his flawed methodology to state that 54% of loans with an interest rate of 18% or greater have defaulted, leaving the impression that lenders on these loans have lost over half of the funds that they lent, and that losses ran roughly three times the interest rate on loans.  Again Mr. Gimein is equivocating annual interest earned with cumulative default rates over a three year period.  Lenders on these loans lost 10% on an annual basis, and while not positive, it’s a far cry from the 54% loss that Mr. Giemein flawed analysis leads the reader to believe.

After quoting these cumulative loss results out of context, Mr. Gimein’s bottom line is “After you take defaults into account, investors have lost money on most of their Prosper lending.”  Mr. Gimein makes this statement without providing any actual returns data, again leaving the false impression that lenders have lost 39% to 54% on their Prosper lending.  The truth is that the median return across all Prosper lenders was negative 3.2%.  In addition, 39% of lenders have made money on their Prosper investment.  While we would have preferred all of our lenders to have made a profit, a low single digit loss for Prosper lenders in the context of the worst recession since the Great Depression shows great promise for peer-to-peer lending as an alternative asset class for investors.  Even a broad index like the S&P 500 saw an annualized loss of 6% in the past three years.  Most individual investors have experienced performance substantially worse than this in their investment portfolios and 401k accounts.  Something Mr. Gimein fails to discuss.

Mr. Gimein also fails to mention that historically there has been a significant amount of social lending through Prosper’s marketplace.  Social loans are deliberately underpriced relative to their stated risk by lenders in order to benefit borrowers with unique or challenging circumstances.  Although we do not have a way to isolate the impact of these loans on performance, there is no doubt that they have had a downward impact on some lender returns.

Mr. Gimein also seems to find fault with the steps Prosper has taken to improve the lending process and provide lenders with additional information to improve their lending decisions.  Prosper has instituted a minimum credit score requirement of 640 to request a loan from Prosper’s lenders as well as a bid floor. Prosper also introduced a new rating system in July 2009 that incorporates the historical performance of over 29,000 Prosper loans into the rating of new borrowers looking for loans.  Although the new rating system uses the same letter grades to rank order risk, the meaning of the letters has changed significantly. This has resulted in a change in the composition of Prosper’s listings, which allows lenders to more accurately assess risk and set prices for prospective borrowers.  This change in rating methodology is well documented on Prosper’s site and lenders can easily compare the impact of Prosper’s new rating system on loans originated under the older system.

The early results from the new rating system are excellent.  Prosper is estimating returns for lenders above 10% for loans originated since our July re-launch and the early data supports these expectations.  Below is a graph comparing the delinquency performance of loans originated by year with the loans originated in the third quarter of 2009.  As you can see, the proportion of loans that are 31 to 120 days past due for the loans originated under the new rating system are dramatically lower.

Prosper was launched to the public in February of 2006 and was about 18 months old when the credit crisis turned our economy upside down.  The crisis that followed saw a dramatic increase in defaults for all classes of consumer loans.  Large banks with years of lending experience saw a dramatic increase in consumer defaults and posted significant losses.  Prosper was clearly not immune from the economic environment, but looked at in the proper context, peer-to-peer lending has weathered the storm relatively well and as a result is well positioned for a bright future.  At a time when financial markets are in upheaval and consumers are facing a dwindling set of credit alternatives, Peer-to-Peer lending deserves better than a flawed, out of context evaluation from a seasoned journalist, and a respected Web site, that should hold themselves to a higher standard.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.