One of the problems with comparing default rates at Lending Club and Prosper is that you are really comparing apples with oranges. What do I mean by that?
Each company has a completely different underwriting model and they grade loans differently. Does a AA-grade loan on Prosper match up with an A-grade loan on Lending Club? It depends. If the interest rate, loan terms and the credit data of the borrowers are all similar then I would say yes. But I have heard from many borrowers who have received very different interest rates from both companies when applying for a loan with identical terms.
P2PXML Brings Standardization to P2P Lending
Earlier this year I wrote about the release of P2PXML, an attempt to bring an open standard to analyzing p2p loans. P2PXML is an open source project by Michael from Nickel Steamroller and it is very useful in doing a legitimate comparison between Lending Club and Prosper. The problem has been these companies store data in very different ways so even though you can download the entire loan history from each company it has been difficult to make realistic comparisons.
P2PXML stores the data from both companies in exactly the same format. It is not perfect because not all the same data is available from both companies but it does provide enough information to make the first real apples to apples analysis of default rates.
Interest Rate Groups Are the Key
When you look at a Grade A loan on Lending Club it could carry an interest rate of anywhere between 5.42% and 9.63%. On Prosper the ranges are even larger. So, instead of looking at loan grades we should really be looking at interest rates.
P2PXML assigns a Rate Group to every loan based on the interest rate; assigning each loan to a bucket with a range of 2%. So, Rate Group 1 has a range of 0 – 1.99%, Rate Group 2 is 2 – 3.99%, and so on. Now, there has never been a loan issued below 4% so for our analysis we started with Rate Group 3 (R-03) – the bucket for interest rates between 4 – 5.99%).
To make a true apples to apples comparison, in my analysis I only included 36-month loans issued by both companies after July 18, 2009 (the end of Prosper’s quiet period). Then I tallied up the total number of loans in each rate group and what percentage of those loans had defaulted. You can see the results of this analysis in the chart below.
There is No Clear Winner on Default Rates
It is interesting that there is no clear winner here when it comes to default rates. Lending Club does a little better than Prosper for interest rates below 12% and then Prosper is better between 12% and 18%. It is a bit bizarre that Prosper’s best performing Rate Group is actually R-09 (16 – 17.99%) where there have been 1089 loans issued and just 16 defaults.
I cutoff the comparison at an interest rate of 22%, even though this meant eliminating over half of the loans issued on Prosper, because I was focused on interest rates where both companies issued many loans. Lending Club issues very few 36-month loans above 22% – just 26 have been issued since July 2009.
You can take a look at the table below which has the data used to create the chart. You can see that Lending Club has the bulk of their loans between 6 and 16%. Prosper, on the other hand, has a pretty even spread between 6% and 22%. What is not shown is that Prosper had 10,296 loans with a rate of 22% or more out of their total number of 18,448 three-year loans. So around 56% of Prosper’s loans are not included in this analysis.
[table id=29 /]
[Update: It was pointed out in the comments that this table would be more useful with loan age included. I have added that now in the table above and you can see it gives a more complete picture. Prosper has a more mature portfolio in every rate group except for R-09 which explains in part why they are doing so well in that rate group. So, from this we could say that Prosper is doing a better job in their underwriting than originally thought.]
Underwriting is a Moving Target
While this is the closest we have come to a true apples to apples comparison on default rates it certainly isn’t perfect. For one thing the underwriting rules and interest rates are constantly changing so a loan that may have been 9.5% last month could be 10.5% this month.
Also it should be noted that until December 2010 Prosper had an auction-based model where interest rates were “decided” upon by investors. So that certainly skews results as well and may account for default rates that are somewhat inconsistent with risk.
I am interested to hear from others on this. Did you expect any differences here? Is this analysis even worthwhile? Please let me know in the comments.