Lending Club High vs. Low Risk Loan Experiment Complete

Bank Deal: Earn 1.00% APY on an FDIC-insured savings account at Barclays.

Lending Club High vs. Low Risk Loan Experiment Complete

A little over three years ago, I initiated an experiment with Lending Club. I took $1000 and split it into two auto-selected loan portfolios. With the first $500, I cranked the risk slider all the way down to the lowest level. With the other $500, I cranked it all the way up to the highest level.

The end results was two portfolios of 20 x $25 notes each. The average rate on the low risk portfolio was 9.85% while the average rate of the high risk portfolio was 15.05%. At the time, these were all 36 month notes so the portfolios have (more or less) run their course.

The low risk portfolio finished with a balance of $518.58 — that’s a 3.7% total gain over three years.

The high risk portfolio finished with a balance of $502.49 (plus $0.52 due on a loan that is running late). Assuming that last little bit comes in, that’s a 0.6% total return over three years.

It’s important to keep in mind, however, that I wasn’t reinvesting the proceeds within these portfolios. Rather, I was using the money to pick additional loans in my hand-selected portfolio. Thus, these numbers aren’t all that reflective of my real returns.

Some interesting tidbits…

Both portfolios suffered the same number of defaults (five each), though the defaults occurred much earlier — and thus caused greater losses of principal — in the high risk portfolio.

In the low risk portfolio, there were three A and two B loans charged off.

In the high risk portfolio, there were four D and one E loans charged off.

My single best loan generated $32.89 in payments ($7.89 profit over the $25 invested) whereas the worst loan didn’t receive a single payment, resulting in a complete loss of $25. Both of these were in the high risk portfolio.

So there you have it… Based on a small sample of loans that were auto-selected a little over three years ago, it appears that lower risk loans provide slightly higher returns than higher risk loans. Apply for your own account

Note that these were completely static portfolios, without any selling of troubled notes.

15 Responses to “Lending Club High vs. Low Risk Loan Experiment Complete”

  1. Anonymous

    I agree with many of the other posters. While your attempt was commendable, the experiment was flawed from the beginning due to the sample size being too small. With only 20 notes, one good note or bad note, could easily sway the numbers too much in either direction. If you mine the statistic from LC’s site and other places like Lendstats, etc. you’ll find that returns are actually greater on the riskier notes with the exception of the “G” rates notes (the riskiest they issue). Again, I applaud your effort.

  2. Anonymous

    I invested 2500 over period to lending club. At this moment total 107 loans are (six loans of $50 and remaining each $25) current. None in grace period. two loans once went in to grace period but eventually came back and now current. Didn’t go to grace period next time. 2 are in funding and 2 are fully paid in just 4 months. As soon as interest exceeds 25$, I reinvest. All notes are hand picked by me. Lowest interest rate is 6.03% on 5 notes and highest interest rate is 23.33% on 1 note.

    From Nov 2011 till date, total interest is 119$. Cash balance is 5.81$.

    Nov: New fund 500
    Jan 2012: New fund 1000
    Apr 2012: New fund 1000

    Current return rate is 11.32%. I hand pick loans with total amount less then 12000 and only for business (that risk I am willing to take) or credit card debt consolidation (People pay 20-23% rate on credit card debt so I guess they will pay off 10-15% lending club loan easily.)

  3. Anonymous

    This was an interesting experiment. I think the key takeaway is that Nickel invested in the LC auto-selected portfolios.

    I have been skeptical in the past what parameters LC uses to put these together.

    I have had success with LC due to the time I take reading each note application. Yes, it takes time, but it’s money – my money. I prefer to lend to people who have steady employment, no delinquencies and a good FICO score.

    I think Nickel’s experiment doesn’t discount P2P lending or LC, but LC’s auto-selected portfolios have some issues.

  4. Anonymous

    I think the novelty of the idea three years ago caused lenders to be over eager. Loans were misplaced and returns sucked. I made 1.9% in my experiment. Now that everybody is slamming the model, I wonder if loans are misplaced the other way now, and the next 3 years could be great.

  5. Anonymous


    While we wait for some of these commenters to take the sticks out of of their ….., I’d like to thank you for following up and posting the results! I actually remember when you first began this experiment, and I was excited to see how it would all play out. If anything, I completely agree with you that P2P lending has changed a lot within the last 3 years.

  6. Anonymous

    I’m sorry but though I normally enjoy your posts, I have to say that the only thing that is interesting is that you didn’t realize prior to starting this experiment the simple fact that whatever results you’d get was bound to be inconclusive at best, & downright misleading at worst. 20 notes as a sample size? Really? Are you kidding?

  7. Nickel

    Wesley: The numbers above reflect return of principal + interest and late fees. It’s a fair apples-to-apples comparison between the high and low risk portfolios but it’s not a valid measure of the true return.

    While I was willing to underwrite the experiment (to the tune of $1k in essentially “blind” investments) I wasn’t willing to carry it on indefinitely. I thus decided to pull the money out of these portfolios as it came in. The end result was an ever-decreasing balance in both portfolios, which means you can’t simply scale the net profit against $500 to calculate the overall rate of return.

    Here are the actual numbers…

    “High Risk”:
    Return of principal: $408.30 (almost $92 lost to defaults)
    Interest received: $94.03
    Late fees received: $0.15
    Outstanding balance: $0.52 (one loan is running late)

    “Low Risk”:
    Return of principal: $452.63 (not quite $48 lost to defaults)
    Interest received: $65.89
    Late fees received: $0.16
    Outstanding balance: zero

    Unfortunately, I’ve tracked all of my Lending Club investments as a single account in Quicken (originally; more recently in MoneyDance) so it’s not a simple thing to estimate the true rate of return and Lending Club doesn’t do it for you.

    Yes, LC provides a net annualized return, which is similar to what you’re looking for, but they do this for the account as a whole and don’t offer that sort of information on a per-portfolio basis.

    I hope this helps.

  8. Anonymous

    I don’t understand your logic. If this didn’t include reinvestment, how can you even measure the return?

    The High Risk portfolio returned more in interest, but all of that (as well as all of the Low Risk interest) went to fund other portfolios of loans.

    Moreover, if all the returns were reinvested, how did the two portfolios end up with balances at all?

    One of us is missing something here.

  9. Nickel

    Ray: At the time I set up the portfolio, this is what the rate landscape looked like. In fact, there weren’t any loans paying 24% or higher. Recall that, at the time, they only had 36 month loans.

    As I stated in the article, these portfolios were constructed via automatic selection of loans using their risk slider. When cranked to the lowest level, the average rate was 9%ish, and at the highest level the average rate was 15%ish. To ask for a greater spread is silly since it didn’t exist at the time.

    Keep in mind that Lending Club is not Prosper. Lending Club is typically more selective in who they’ll accept as a borrower which contributes to this lower interest rate spread between the lower and higher risk borrowers.

    As for the sample size, my regular portfolio has had upwards of 500 loans, but this is a less variable collection of loans as I selected them to fit specific criteria. The experimental portfolios had 20 loans apiece. You are correct that this is a relatively small sample size but it is what it is. If you’d like to bankroll a larger experiment, be my guest. 🙂

    For what it’s worth, I actually saw a lot of consistency within the portfolios, which would suggest that the my results were quite representative. For example, of the five loans that defaulted in each portfolio, all five happened early in the high risk portfolio and all five happened late in the low risk portfolio.

    Anyway, a better criticism would be that I only ran this experiment once, starting three years ago. Could an investor starting out today expect similar results? I have no idea, but it’s not unreasonable to think that the landscape has changed. For example, Lending Club may have gotten better at pricing risk. After all, if Lending Club (or Prosper or whoever) is doing their job right, you should get similar returns at all levels of risk.

    I would also be reticent to extend the results from one lending platform to another (e.g., applying lessons learned from Lending Club to Prosper or vice versa) since they are entirely different beasts.

  10. Nickel

    Chris: As I noted, these numbers aren’t reflective of real returns because I didn’t reinvest the proceeds into the same portfolio. Rather, as the money came in on the experimental portfolios I re-deployed it in portfolios where I was actively selecting loans.

    In other words, the portfolio size dwindled over time and the true returns are thus considerably higher than the numbers reported above. I could do a bunch of fancy math to figure it out, but that’s not the point. The point was to compare the relative returns between strategies and for that these numbers work. The low risk outperformed the high risk.

  11. Anonymous

    I really enjoy reading your blog, but I have to see that your sample size is so small and the difference between 9.85% while the average rate of the high risk portfolio was 15.05% is so minor that I can’t call that low risk and high risk.

    Low risk would better be defined as a rate below 10% assuming the loan is evaluated correctly and high would be 24% or higher.

    Without having a larger sample size of say 250-500 loans each, you really can’t draw a correct conclusion either way.

    If you look filter loans correctly you can easily obtain 10% over time. Right now I’m earning 16%+ and my avg loan is 28% on Prosper with almost 600 total loans. On lending club where I have a much smaller sample but I’m earning 18%+. While I don’t expect these returns to continue, I do expect to earn close to 12% over the course of my loans by using proper criteria with filtering on loans.

    Your article could have easily of had a 9% return for the “low” risk loans and a 15% return for the “high” risk loans, so while I do appreciate the article I find it suspect at best due to the small sample size.

    Once loans get to around the 10 month mark you normally have an idea of what they will do over the course of the entire loan.

    P2P is here to stay, just takes some initial work setting up your criteria and saving the filters for future investing.

  12. Anonymous

    Wait a sec..these returns are crap. 3.7% over three years in an extremely illiquid account? I suspect that you also exposed yourself to much more risk than you anticipated. How does the risk/reward ratio of Propser really compare to the S&P for example?

  13. Anonymous

    These are very interesting results. I have the same feeling that higher quality A and B loans most likely will outperform lower quality loans.

    BTW, will you be interested in sharing the loan IDs in these two portfolios, date of last payment received and date of default/charged-off? If borrower entered in payment plan, I would like that information too, for example, when, what was new payment.

    I am currently analyzing the lending club historical loan data file on my blog. As these files are snapshot in time, I am trying to collect more information on how these loans were performing over time, when they entered late, grade period and other status, how much principal was charged off, etc.


Leave a Reply