Lending Club Whole Loan Program: One Year Later

The following article is a guest post from Bryce Mason, the founder of P2P-Picks, about the whole loan program at Lending Club. He looks at the percentage of loans reserved for this program and whether or not participating investors are getting a random sample of all the loans as was initially promised by Lending Club.

On September 28th, 2012, LendingClub announced that, as opposed to its standard fractional investment business model, it would begin setting aside some loans that could only be purchased in their entirety–or as whole loans. Part of that announcement were promises that whole loans would be chosen randomly from the general pool and that anyone would be able to participate in the whole loan program. With the one-year anniversary of the whole loan program right around the corner, has LendingClub met its promises?

Before we dive into answering these two questions, it might be useful to understand how many loans are part of the whole loan program. Using a data set of all issued loans with a listing date of at least September 27th, 2012 (the first day of the program), we can see that 26% have been partitioned off as whole loans. According to the blog post referenced above, the program came about because some institutional investors felt that owning the entire loan allowed them special legal and accounting treatment that was more consistent with their goals. Thus, the whole loan program was set up with institutional investors in mind, and we might expect its use to scale with them. It appears that the probability of a loan being selected for the whole loan program is a parameter that LendingClub can set depending on their institutional investors’ needs. The table below shows that, as the whole loan program began, fewer than 20% of loans were selected. Subsequent months brought selection up to about 30% on average, with selection slowing down in May 2013 to a more average range.

Month% WholeTotal Loans

Note: Data extract ended on August 7th.

Can Any Investor Join the Whole Loan Program?

The easiest question to address first is whether LendingClub allows any investor to join the whole loan program. It appears to me that the answer to this question is yes, with a caveat. Due to my business of independent P2P loan underwriting, I recently asked for whole loan access from my account representative via my personal account. He merely asked that I reply to his email with a copy of a paragraph disclaimer about the risks associated with investing in whole loans. A few hours after I sent in that text, whole loans started appearing on the platform. While I cannot speak for everyone, it certainly seems that a process is in place to provide retail investors with whole loan access.

Given that whole loans are a large chunk of originations and that most retail investors do not have the account balance to participate with sufficient diversification, the retail investor’s most important consideration should be whether LendingClub has made good on its promise to set aside whole loans randomly. Any other selection method would invariably produce different pools of loans for institutional and retail investors, ultimately yielding returns that would probably be preferential for one group.

A Random Selection of Loans for Large Investors

Thankfully, LendingClub appears to be fulfilling its random selection promise. Using the same data from above, we performed simple t-tests to determine if a number of borrower attributes differed by initial listing status (whole vs. fractional). If LendingClub selected whole loans randomly, then there would be no average difference between the pools.

AttributeFracWholeps.d.dt-test (p<0.05)
FICO (low)6956960.0128.70.04different
Loan Amount ($1k)14.614.
Rate (%)14.514.
60-month (%)22.722.30.250.420.01not different
Employer N/A (%) different
Public Records0.110.110.880.410not different
Income ($1k)72.874.30.01590.03different

As one can see from the table, the averages of a number of attributes are statistically different across initial listing status. However, with almost 90,000 loans in this study, there is so much statistical power that even very small differences might be deemed statistically significant. More important in this case is whether the reader finds the differences practically significant. The mean FICO score, for instance, differed by 1 point between the two groups. This difference (d) represents a little less than four one-hundredths of a standard deviation–a very minor difference–and is the largest of the bunch in those terms. So, while there are statistical differences, overall there are do not appear to be any meaningful differences between whole and fractional loans.

LendingClub should be commended on its transparency, allowing investors the opportunity to hold them accountable for their promises. It is very rare in the investing world to have access to this kind of data.

[Update: The July and August numbers in the above chart have been updated slightly to remove the “Policy Code” of 2, which was a recent addition to the LoanStats download file. These records are void of most borrower attributes and the loans do not adhere to the current underwriting policy of Lending Club – hence the number 2 in the policy code field. According to the data, this program began on July 9th, 2013. Originations stemming from this program accounted for about 5% of total origination dollars since it began. Interestingly, the average funded amount of these loans is roughly half of other LendingClub loans. There will be an upcoming post on these policy code 2 loans explaining this new program in more detail. Thank you to an anonymous reader for pointing out the existence of this second whole loan program.]

Peter Renton is the chairman and co-founder of LendIt Fintech, the world’s first and largest digital media and events company focused on fintech.

LendIt Fintech conducts three conferences a year for the leading fintech markets of the USA, Europe, and Latin America. LendIt also provides cutting-edge content all year long via audio, video, and written channels.

Peter has been writing about fintech since 2010 and he is the author and creator of the Fintech One-on-One Podcast, the first and longest-running fintech interview series.

Peter has been interviewed by the Wall Street Journal, Bloomberg, The New York Times, CNBC, CNN, Fortune, NPR, Fox Business News, the Financial Times, and dozens of other publications.

Notify of
Newest Most Voted
Inline Feedbacks
View all comments
Aug. 29, 2013 10:27 am

I like that you highlighted that statistical significance doesn’t mean it’s meaningful. This is often missed — good post 🙂

Aug. 29, 2013 12:19 pm

Is this analysis completed on listings or fulfilled loans? If its on completed loans I would challenge the notion “If LendingClub selected whole loans randomly, then there would be no average difference between the pools.”. Speaking from my experience on Prosper, the Whole listings fund faster than the fractional listings which has a significant impact on the fulfilliment rate (whole loans listings fund at a 5% higher rate). The listings have identical composition, but the completed loans for the whole loan program skew towards larger loan sizes and lower APRs (groups that take longer to fund on the fractional side). I think your numbers show the same thing (but to a lower degree) in respects to loan amount and rate.

Bryce Mason
Aug. 29, 2013 2:02 pm
Reply to  aaron

Per the article, this is all –issued– loans. There are tons of loans that were listed but later were withdrawn or rejected.

Bryce M.
Aug. 29, 2013 5:15 pm
Reply to  aaron

My purpose here was to see whether the decision to mark a loan for the whole loan program was random. That decision comes before investors decisions whether to buy the loan whole or how fast a fractional loan fills. Thus, the things you mention above don’t have anything to do with my research question. If I’ve missed something here, please help me understand.

Aug. 29, 2013 5:59 pm
Reply to  Bryce M.

It certainly does, because you are analyzing the funded loan pool and NOT listings. As you point out LC’s decision to list an application occurs prior to individual investors decision to fund or not to fund. LC’s promise to investors is to maintain homogenity of the listings in each of the pools of listings. Since each pool exhibits very different funding velocities (which influence ultimate conversion rates) the loans generated will certainly exhibit a selection bias.

My guess is that if you were to analyze listings the $200 differential you see between whole loans and fractional loans would disappear. I believe this is largely a function of high dollar listings taking longer to fund when entering the fractional vs. the whole loan market and thus less high dollar conversions in the fractional pool.

Anyways, great analysis. Even with taking into account this bias, its obvious that LC whole loans and fractional loans aren’t meaningfully different.

B. Mason
Aug. 29, 2013 7:52 pm
Reply to  aaron

It seems quite unlikely that the listings would not be random, and then the process of rejection and borrower withdrawal would somehow correct for it (such that my analysis showed randomness). I would have preferred to look at listings, but LC does not provide that data in an easy to download form. One would have had to archive all BrowseNotes files (and also have had Whole Loan Access) since inception.

Bryce M.
Aug. 29, 2013 12:21 pm

If others think of additional metrics that would be known at the time of loan listing that would be interesting to add to the above table, please let me know. I can include them and Peter can update the table with a few additional rows. I didn’t want to make a gigantic table, but if we think it would be good to expand it some, we can.

Anil @ PeerCube
Aug. 29, 2013 5:38 pm


Good analysis. Couple of questions for you:

1. Do we assume that the data for this analysis came from historical loan data file Lending Club provides?

2. If the answer to Q1 is affirmative, do you know if Lending Club includes whole loans in the historical data file?

I am under the impression that the historical loan data file don’t include the loans that were issued ‘whole’. The data file only includes the loans that were offered as ‘whole’ but were not picked up by ‘whole’ lenders and were later released to ‘fractional’ pool.

Also, did you review the whole loans by loan grade?


Bryce M.
Aug. 29, 2013 5:42 pm

The source is the standard LoanStats file provided to all investors via their website. You can tell whether a loan was initially marked for the whole loan program via the “initial_list_status” variable, which takes on values of “w” and “f.”

I am under the assumption that all loans originated are reflected in the LoanStats file. The origination totals by month match their statistics dashboard perfectly. And, their origination fees in their most recent 10-Q (April, May, June 2013) are about 3.7% of the total originations in the LoanStats file. Seems reasonable to me.

I did not feel compelled to view by grade, as the interest rate is highly correlated.

Bryce M.
Aug. 29, 2013 5:52 pm
Reply to  Bryce M.

Oh, a further reason why I believe all loans are reflected in the LoanStats file. Suppose that the whole loans that were purchased in their entirety were not present in the LoanStats file (i.e., just the leftovers that go to the fractional market are present). Then, given that we see no difference in the average characteristics of the “leftovers” vs. the fractional pool, this implies that institutions who buy whole loans are doing random note selection from the whole loan pool (not even naively targeting the high interest notes). This seems highly unlikely. Therefore, the LoanStats file contains every loan, even those bought by 1 entity. Reductio ad absurdum.

Dan B
Dan B
Aug. 29, 2013 6:09 pm

This is certainly an area of interest which I can see you’ve done a rather competent & dispassionate job in covering. I’m sure it goes some way in reassuring those investors who may have suspected the process to be less than random.
Since I haven’t been shy in my criticism, when I felt it was warranted in the past, I thought it appropriate to commend you on a job well done here. I look forward to reading your future contributions.

B. Mason
Aug. 29, 2013 9:56 pm
Reply to  Dan B

Thanks, Dan. I’m honored to be the recipient of both sides of your brutal honesty.

B. Mason
Aug. 29, 2013 10:22 pm

For the interested reader, the update noted at the end of the article is an interesting development. There are essentially two whole loan programs. It will be interesting to see how much of their originations are a function of these custom underwriting offerings. It’s so new, though, that it’s hard to know what to make of it yet. Surely, however, they are not randomly chosen from all applications. If they were, why wouldn’t LendingClub simply tell those institutional clients just to use the web platform? In any case, their existence is outside of the scope of this article, which covers all of the loans that were issued from the web platform.

B. Mason
Aug. 29, 2013 10:23 pm

Another footnote that Peter said made my article too long ;).

The actual promise regarding randomization was that it would be random within loan grade. However, for simplification, I elected to compare across all grades simultaneously. This is valid, because an environment that is random within grade is still random overall (although it may not be uniformly across grade). Randomizing within grade allows LendingClub to control volume a little better for their institutional clients, in case one month they need more “A” graded notes, and in another month they need more “D” graded notes. A separate analysis of the percentage held back for the whole loan program by grade revealed that LendingClub has rarely used this feature to any large degree.

samuel hu
samuel hu
Aug. 30, 2013 6:30 am


With the whole loan program, do you know if the note that a lender purchases is still an obligation of L/C or does L/C transfer the underlying loan to a SPE and the lender then purchases a note issued by the SPE (akin to the LC Advisers structure).

As larger institutions put more money into LC Notes, do you know how they are addressing the risk that LC itself goes insolvent (with the effect that the underlying loans would be used to satisfy claims of other credits other than the note investors)


Bryce M.
Aug. 30, 2013 7:25 am
Reply to  samuel hu

Sam, you should do more due diligence, but I have to think that is precisely the “special legal and accounting” features mentioned in the article above by LC. Funds own the note, I’m pretty sure. Maybe check LCs 10-Q and see if the obligations are 25% less than you might expect?

Sep. 1, 2013 10:29 pm

I understand that for the big players who have millions of dollars available this makes sense, but who of the small players would commit 30k in one loan (unless you have millions)?

Bryce M.
Sep. 2, 2013 10:46 am
Reply to  Martin

Precisely why it is important for the whole loans to be randomly selected.

Apr. 2, 2016 12:39 pm

I think they may have been random back in 2013, but they don’t seem so random anymore. My filter criteria gets around 10x more whole loans than fractionals.