In part 1 of this topic, I discussed different ways to apply historical default rates to small value P2P loan portfolios. To make more sense out of how historical default rates might affect loan performance, I ran a monte carlo analysis on a select set of hypothetical loans. Here are the simulation results.
To get a better handle on the effect of historical default rates on P2P loan portfolios, I performed a monte carlo simulation on portfolios lending $100, $500, $1000, $2000, and $5000. Since I assumed that the money was spread across investments of $25 each, those dollar amounts correlate to 4, 20, 40, 80, and 200 loans respectively. For each test case, I simulated investing in all A1, C1, E1, or G1 loans (using Lending Club grading terminology) as well as an equal mix of those 4 grades, which I called Div for diversified. I used the following historical yearly default and corresponding loan interest rates (again from Lending Club’s site):
| Grade | Historical Yearly Default Rate | Loan Interest Rate |
| A1 | 0.16% | 7.37% |
| C1 | 1.74% | 11.28% |
| E1 | 3.32% | 14.43% |
| G1 | 4.90% | 17.59% |
Note that these historical default rates are not for actual Lending Club loans, but rather the industry accepted default rate historically found for people with the specified credit rating.
Since loans on Lending Club have 36 month terms, I calculated payments across each of these months, accounting for the appropriate probability of default. Using 36 months is more accurate that defaults/year since borrowers can default at any time and although month 1 and month 12 are both in year 1, the difference of those two default scenarios is significant. The overall results wouldn’t have meant much for one instance of each test case, so I instead ran each case 1000 times. The results show the total value of all of the payments made.
Let me explain the $100 results as an example and then present the others for you to analyze.
$100 Investment
For this small investment amount, you can see that most cases of each portfolio had the same result, which is why you see prominent bands around the max expected returns for $112 $118 $121 $124 and $129. We would expect the number of cases at the maximum to decrease as more money is invested, though the minimum result would also be expected to rise. To read this graph, assume that you had $100 to invest. If you invested in all A1 loans, using the historical default rates and corresponding loan interest rate, the total payments received would be one of the dark blue dots. Had you invested in C1 loans, you’d be a pink dot, and so on. The other dots of the same color (other than the one that is “you”) can be thought of as other people who also invested in the same number and grades of loans. Remember that this simulation gives hypothetical returns using historical default rates and the payments you receive could be worse. It’s just a matter of probability. Now that we’ve looked at the $100 investment results, you can probably guess how the rest will look and analyze them yourself. Here they are:
$500 Investment
$1000 Investment
$2000 Investment
$5000 Investment
General Conclusions;
As more money is invested, the likelihood of getting the full expected return declines, as the occurrence of default becomes more prevalent. Investing $1000 or more always (at least in these simulations) resulted in a profit, whereas investing smaller amount sometimes resulted in a loss. The average return of the riskiest (All G1) portfolio was always the highest.
The results of these simulations indicate two things to me:
1) Investing more money seems to be safer
2) If you do invest more money (in as many loans as possible) the higher returns of riskier grade loans seem worth it. This is likely due to the fact that while the individual loans do carry more risk, the fact that you’re diversifying by investing in so many of them tends to moderate the risk.
If you enjoyed this post, subscribe to my feed via RSS or email.
You can support Richer by the Day by visiting our advertisers and sponsors. A thumbs up from any StumbleUpon users would also be greatly appreciated.
Related Posts
P2P Lending Default Considerations (Part 1)P2P Lending Blog Carnival #4
Lending Club Quiet Period Has Ended
Get Better Rates with Lending Club
Free Lending Club Webinar This Thursday
November 11th, 2008 at 11:27 am
>The average return of the riskiest (All G1) portfolio was always the highest.
This is true for the current data. But remember all loans are still running, so that might change over the loan term.
And do remember that those who invested in the highest rates/risks at Prosper in general did worse than those investing in AA or A loans there.(I am NOT suggesting that Lending Club and Prosper are comparable; just saying that it might be too early for conclusions like that.
Wiseclerk
http://www.p2p-banking.com
November 11th, 2008 at 11:33 am
@Wiseclerk: I meant that the average return of the All G1 return was always the highest in my simulations. Though I used the rates and historical defaults from Lending Club’s site, these simulation results are in no way correlated the the actual loans or default rates you might get there. Availability of loans may also prevent implementing these methods, i.e. you might not be able to find sufficient loans at the desired grade to invest in at any given time.
November 12th, 2008 at 5:02 am
@Mike: I know what you meant. But it is still based on performance so far, since you based it on the past default and interest rates of Lending Club: “I used the following historical yearly default and corresponding loan interest rates (again from Lending Club’s site):”
I appreciate the work you did, my point was that it was a bit early for conclusions. Do this again in 2 years when a part of the loans have completed their loan term.
Wiseclerk
http://www.p2p-banking.com
November 12th, 2008 at 7:10 am
@Wiseclerk: Now I see the disconnect between how we were interpreting what I said. The historical default rates on the Lending Club site are not from its own history, but rather rates established by the banking industry from years (maybe decades) of performance on more traditional loans. If I were to have used actual performance data, my results surely would have the “too early for conclusions” bias you suggest. G Grade loans for the first year (June 1, 2007-May 31,2008) had zero defaults, for example. Sorry for the confusion.
March 31st, 2009 at 9:08 am
I wasn’t sure exactly what was being plotted in your graphs. What are the labels on the axes? Also, in terms of returns on the different loan portfolios by risk, it would be a lot different if you accounted for the lending club “processing fee” which varies from 0.75% for low risk to 3.5% for high risk. Also, I agree with the previous comment regarding applying previous lending industry data, since I’m sure it’s much easier to get a Lending club loan while you’re dying in a hospice program, than in a traditional bank where you have to go in face to face.
Ken
March 31st, 2009 at 9:24 am
@Ken: The x-axis is the number of the trial run. Since I ran each case 1000 times, each x-axis value represents one of those 1000 cases. The y-axis is the amount repaid. So a dot at (35, 105) would mean that in the 35th trial of that scenario, a $100 loan repaid $105. I’ll look into the processing fee variation in more detail. You’re right that using historical default numbers isn’t ideal, but it is about the best we can do until Lending Club has sufficient default history from its own performance. Thanks for your input.
April 11th, 2009 at 12:14 am
I believe that the externally obtained ‘historical default’ rates are wildly inappropriate for experience with Lendingclub (and probably, other microlending vendors). I have over $7K loaned out through Lending club. I have been mildly agressive in the the class of loan chosen, but a review of late payments and defaults suggests that the assigned risk (and hence interest rate) of the loan does not have much predictive power for the loan going into default. Were it not for the signup bonus, I would be in the red, net, and as the loans play out, I full expect to lose money over all. Of the 240 loans I have across 40 portfolios, 7 have been charged off, 15 are in default, 12 are 1 to 10 months late (heading for default), 8 are 16-30 days late, and 22 have been fully paid. The weighted average return is 13.4% (in actuality, this is 0% with the above losses taken into account).
I funded many borrowers seeking credit consolidation. These types of loans seemed much more straightforward than business loans, and relatively safe when put into a risk-diversifying portfolio. My adverse experience here suggests that banks will fare very poorly in the months ahead in on credit card defaults.
I have not yet done an analysis to see if there are systematic differences in my portfolios vs. Lendingclub total loans, but my outcomes differ markedly from what they report.
I think Lending Club’s portfolio of loans with $25 exposure is a brilliant business model, but I have seriously underestimated the default risk involved, and part of that is to blame on Lending Club’s published statistics. I am not complaining too much. 0% is a better ROI than most investments during the last 12 months.
–Carl
April 11th, 2009 at 9:18 am
Carl, I agree with your assessment of Lending Club’s default statistics. The way they present them gives potential lenders a very optimistic picture. My situation is a bit brighter, but my loans are relatively new and none are late…yet.
Unfortunately, these types of forums are the only way for lenders to communicate in a meaningful way. I doubt LC will develop a forum on their site anytime soon. I hope LC management monitors these types of discussions and takes concerns seriously.
April 13th, 2009 at 3:32 pm
@Carl: Thanks for your detailed data…that is certainly interesting. On your last point, I suspect defaults are higher as a result of the credit crisis. So returns are taking a hit, but are still better than many alternative investments, which is a typically goal. My Lending Club portfolio is my best performer among my investments over the past 12 months.
April 13th, 2009 at 3:39 pm
@veggivet: Glad you were able to voice some of your concerns here. You may also find the Wiseclerk P2P lending forum useful.
August 13th, 2009 at 1:44 pm
@Carl: I’m new to Lending Club (LC). What is a portfolio? Do you organize your loans into portfolios? Looking at your numbers, your average loan is $29 (7000/240), total bad loans are 34 (7+15+12), total of bad loans is $991 (34*29/7000) and the number of bad loans as a percentage is 14% (34/240). Is that about right? If so, it differs markedly from what LC suggests on their website.
If one lends more money, does it make sense to make each loan amount larger (e.g. $100 or $200)? Or should one invest slowly and take longer, investing in $25 multiples?
The Monte Carlo simulation is very useful. I had a question as to how it works. Take a look at the A1 credit grade:
A1 0.16% 7.37%
Given that the A1 grade has a 0.16% default rate, do you then randomly select a rate of return which is between 7.37% and 7.37%- 0.16%? Something like return = 1*(100%+ random number between (7.37% and (7.37%-0.16%)))?
Thanks!
Mahesh
September 30th, 2009 at 5:49 pm
Great post. I wonder, have you seen any analysis based on Lending Club returns vs borrower credit scores? I noticed how there were a few posts with concerns about the amount of historical data avaialable on this new platform. Using credit scores would put those concerns to rest.