“[I]f the market is competitive, there are two possibilities. The market may reach a stable equilibrium in which individuals are all rationed in the amount they can borrow, this ration being so severe that no one defaults. Alternatively, the market may oscillate in an unstable fashion, lenders entering and making short-run profits, but in the long run being forced to leave.” Dwight M. Jaffee and Thomas Russell, “Imperfect Information, Uncertainty, and Credit Rationing,” Quarterly Journal of Economics, Nov. 1976.
The two biggest U.S. P2P lenders, Prosper and Lending Club, started operations in 2005 and 2007, respectively. Over the past decade, their business has grown so that they now originate more than $10 billion in loans per year. The public information provided by Lending Club gives us an opportunity to judge how they are doing. Our conclusions are consistent with the opening citation from Jaffee and Russell. At first, P2P lending returns appear remarkably high (adjusted for volatility), but growing evidence of adverse selection highlights how difficult it will be to sustain growth.
When we last wrote about P2P lending, we suggested that profitability might be a consequence of the booming economy (see here and here). Specifically, we noted that the bulk of the loans were for credit card consolidation, and that credit card default rates tend to move one-for-one with the unemployment rate. We concluded that one would need to see performance in a recession before judging P2P’s long-run potential. That is, when you are making consumer loans, it is relatively easy to make money as the unemployment rate falls from 10% to 3.5%. However, profitability over the course of an entire business cycle, including periods when joblessness is rising, is an entirely different story.
Well, maybe there is no need to wait.
Before we get to the details, it is worth observing that the entire consumer lending market is quite large. Total consumer borrowing―including home equity, auto loans and student loans―is currently about $4.5 trillion. Of this, nearly $1 trillion is credit card debt, most of which is held by commercial banks. While we have not found an aggregate measure, given that Lending Club has $9.6 billion outstanding, it seems very unlikely that P2P originations account more than one percent of consumer credit or five percent of credit card debt. So, while P2P has grown, it remains very small.
Turning to some specifics, we collected “adjusted net annualized return” data from Lending Club’s website. As far as we can tell, this is the investor return, net of fees and defaults, at maturity. Focusing on the period since 2010, cumulative originations were $40.4 billion, with charge-offs of $3.3 billion. The average interest rate across all credit categories (labelled A to G) was 13.3%, with a net return of 5.4%. Looking at quarterly data, we are able to compute a risk-adjusted excess return (over Treasuries) for each loan grade. The following chart plots the Sharpe ratio—the average excess return divided by its standard deviation—for each grade, together with those for various benchmark asset classes.
Sharpe Ratios: Lending Club loans (by grade) and various asset benchmarks (1Q 2010-3Q 2019)
Notes: We compute Sharpe ratios based on excess returns. For the Lending Club loans, excess returns are relative to the three-year constant maturity U.S. Treasury rate. For the intermediate bonds, U.S. long-term Treasuries and the S&P 500, excess returns are relative to three-month constant maturity U.S. Treasury rates. With the exception of the Berkshire Hathaway and Renaissance Technology estimates, which we do not compute, all calculations use quarterly data from 1Q 2010 to 3Q 2019.
Sources: Lending Club data are from lendingclub.com; S&P500 data are from Shiller; intermediate bonds (VBMFX) and U.S. long-term Treasuries (VUSTX) are for Vanguard Index portfolios; information on Berkshire Hathaway is from Frazzini, Kabiller, and Pedersen; and the estimates for Renaissance Technologies are for the Medallion Fund from Zuckerman.
We find this figure astonishing. Investments in A and B grade Lending Club loans display Sharpe ratios that exceed estimates for Renaissance Technologies—one of the most successful hedge funds ever—and far surpass Warren Buffett’s Berkshire Hathaway. If this weren’t enough, the P2P returns have very low correlations with equity markets. For grades A to C, the correlations with the S&P 500 range from 0.15 to 0.22. To put it bluntly, these returns and correlations seem too good to be true.
One might conjecture that the data are smoothed, but we strongly doubt that subtle adjustments could lead to such extraordinary results. Instead, we take them at face value, and ask (in line with Jaffee and Russell) whether this business is scalable. It is one thing to obtain fabulous returns when lending one or two billion dollars per year, as Lending Club did between 2010 and 2014. It is quite something else to do it at the current level of $8 to $10 billion, let alone for a sizable proportion of the $1 trillion credit card business.
The real question is whether any P2P lender can overcome a key problem that plagues unsecured consumer lending in a competitive market: adverse selection. This problem, which Jaffee and Russell highlight, is also the basis for the seminal paper by Stiglitz and Weiss. The logic is as follows. As the probability of default rises, a lender will require a higher interest rate to compensate for the risk. But, higher interest rates attract borrowers who are worse risks. Put differently, the pool of willing borrowers at a high interest rate shifts adversely relative to the universe of those wishing to borrow at a rate they expect to pay.
To quote Stiglitz and Weiss, individuals “are willing to borrow at high interest rates because they perceive their probability of repaying the loan to be low.” The stark implication is that as the interest rate on loans rises, lenders’ profit margins will at first increase, but then fall. The following figure reproduced from their paper shows the relationship. Above some interest rate, the riskiness of borrowers will be such that no one will lend because increased lending will diminish margins. Information asymmetries lead to credit rationing.
Theoretical relationship between interest rates charged and expected returns to lenders
Source: Stiglitz and Weiss.
Knowing that higher interest rates attract a riskier pool of borrowers, banks and other traditional lenders have a number of mechanisms to increase profitability. Not only are there rigorous screening processes, but many lenders require collateral. Auto loans, secured by the cars themselves, are a clear example. For consumer credit, at the first sign of trouble, banks can reduce credit lines. For corporate loans, banks use covenants to restrict borrowers’ activities and to require repayment should borrower cash flow slip.
Returning to Lending Club, following Stiglitz and Weiss, we can plot the adjusted net annualized returns against the average interest rate charged on loans for two sample periods. The black dots (and black dashed fitted line) use data from 2010 to 2014, a period when the P2P lender originated $6.6 billion of loans. The red dots and fitted line are for the 2015-2018 interval, when origination surged to $27.3 billion. Data are quarterly and include all six of the credit categories (F and G are combined). Shown on the horizontal axis, average interest rates rise as the loan grade deteriorates from A to F/G, so the riskiest loans are to the right of the figure.
Relationship between interest rates charged and realized returns, 2010-2018
Source: lendingclub.com. We note that, since Lending Club loans are all between 24 and 60 months, much of the data beyond the end of 2016 relies on the firm’s projected, not actual, loss rates.
The figure shows exactly what theory predicts. Looking at the 2010 to 2014 period, the black line turns down, but at an average interest rate above 20%. Adverse selection appears limited in this tiny market. However, as expanding originations attracts more borrowers, the pool becomes markedly worse. Looking at the red line, we see that the realized return starts to turn down at an interest rate that is somewhere between 10% and 15%. Not only that, but with interest rates of 25% or more, realized returns are close to zero or negative. Over the past few years, as far as we can tell, Lending Club severely cut back originations in the lowest credit category. It’s easy to see why.
When we first started to think about P2P lending five years ago, we asked whether these new platforms have any advantage over traditional intermediaries in overcoming information asymmetries—including both adverse selection and moral hazard. Banks and consumer credit companies have been around for a long time. They have substantial expertise in addressing exactly these information challenges, and have used a range of tools to build and sustain a vast market for consumer credit.
It would be nice if P2P firms could find a way to provide unsecured credit cheaply to underserved parts of society. But, unless something changes, the evidence we have now highlights a challenge that is only likely to intensify as they seek to attract more borrowers.
The bottom line: we have to ask whether P2P lenders have already hit the wall—even before the next recession.