Taking a Close Look at Lending Club’s Underwriting Changes

I was in San Francisco last week visiting Lending Club and Prosper. So I took the opportunity to sit down with the risk management team at Lending Club to delve into the underwriting changes they introduced late last year.

As Anil from Random Thoughts pointed out recently at first glance it seems that Lending Club has moved to a riskier credit profile. Given that Lending Club now allows borrowers with a bankruptcy or a current delinquency it is easy to see how one would get that impression.

But Lending Club said the changes have resulted in a better ranking of risk that should lead to a more conservative credit profile overall. How is that possible given the supposedly more lax underwriting criteria? They went through a whole presentation with me to explain these changes in detail.

A Brief Summary of the Underwriting Changes

Before I get into the detailed explanation let’s first look at some of the changes that were made. Many of the changes were detailed in the Random Thoughts post and you can always read the latest underwriting requirements in the Lending Club prospectus. Here are the five main changes:

  1. Maximum number of credit inquiries is now 6 for all borrowers (before it was maximum of 3 inquiries for FICO scores of less than 740 and up to 8 inquiries for scores of 740 or more).
  2. Current delinquency now allowed.
  3. Revolving credit balance maximum of $150,000 restriction removed.
  4. Major derogatory record (meaning a bankruptcy) now allowed.
  5. Maximum credit utilization of 98% restriction removed.

Swapping in and Swapping Out

According to Lending Club the latest underwriting changes did two things. It removed the highest risk borrowers that were previously being approved and it added back in the best borrowers from previously declined populations. This was obviously not a change that was taken lightly – a huge amount of analysis has gone into this decision.

The problem with the old underwriting model is that if a borrower failed just one of the major credit criteria then it didn’t matter how pristine the rest of their credit report was they would be declined. The example was given of a senior member of Lending Club’s executive team who was declined for a loan. This person has a high credit score, a good six-figure salary, no delinquencies and a long history of excellent credit. But he had a HELOC on his house of more than $150,000 and that number is included in the revolving credit amount. So he was over the $150,000 revolving credit balance limit and was therefore rejected for a loan. Under the new model his loan is approved and rated an A1.

Now if there is one strike against a borrower other factors are taken into consideration. This is the key point here. Rather than just deny everyone based on one data point they are now considering hundreds of factors in the underwriting decision. This includes all the credit data they pull from TransUnion, the information gained during the verification process as well as the self-reported information from the borrower loan application.

The point they stressed with all this is that if someone does have, say, a $250,000 revolving credit balance or they have five credit inquiries in the last six months the rest of their credit report has to be excellent. The bar is set higher for these people, which is how Lending Club is able to add these looser criteria while at the same time decreasing overall risk.

How Can Lending Club Be So Confident With Their New Model?

This is a question I posed to them. My thinking was that sometimes these red flags should provide a legitimate concern. How do they know that their model will perform better?

It turns out they had a very credible answer for this. Lending Club went back through and analyzed their declined borrower population. As or this writing Lending Club has declined over 822,000 loan applications representing over $10.7 billion in potential loans. That is quite a decent sized data set.

So, what they did was pull the credit data from many of these declined borrowers and they back tested their new underwriting model. In particular they looked for credit lines that were opened at around the same date as the original borrower application and looked for a successful payment history.

For borrowers that passed the new model they could see their credit score history as well as what kind of payment performance they had experienced with their various open credit lines. They could see who had trouble keeping their obligations and who paid on time every month. From this they could determine who they really should have approved. This back testing gave Lending Club a high degree of confidence in their new model.

But What About Bankruptcies?

Sure, I can understand some of this credit data providing a false indication of creditworthiness but what about a bankruptcy? That is a serious issue and one that infuriates many investors.

They made two points around bankruptcies. One, for someone to have a bankruptcy on their credit report and get back to a FICO score of 660 is not an easy thing to do. This means they most likely have had an excellent payment record since their bankruptcy. Two, declaring bankruptcy a second time is a much more difficult thing to do and so these people were more likely to be consistent payers. The analysis they did of the declined borrower history confirmed this.

So there you have it. The new underwriting model that should provide even better returns for investors going forward. While Lending Club tweaks their underwriting every quarter this seems to be the biggest change they have made in several years, probably since 2008. I came away satisfied that these latest changes were made in the best interests of investors. But if you have your doubts you can easily just filter out these changes and stick primarily with loans that meet the old underwriting criteria.

What do you think? As always I am interested in your comments.


  1. Roy S. says

    When investors create their filters, what they are basically doing is back testing various scenarios. I think this is positive news for LC, and this makes me hopeful for the new (yet-to-be-disclosed) changes at Prosper, too.

    • says

      Roy, I know Prosper are also tweaking their underwriting model and including many more factors than they have historically. I think the whole p2p space is getting more sophisticated with risk management going forward.

  2. Henry Miller says

    First ‘get the data'; much of what happens in this world is counterintuitive. The data available for backtesting seems convincing. Second, set the interest rate in accordance with the data. I am willing to buy D, E, F and G notes partly because the higher interest rates compensates me for the greater risk. I don’t expect pristine metrics from a 20% interest rate loan; just reasonableness. All in all I’m satisfied with the new underwriting standards.

    • says

      Henry, One of the things I didn’t mention in this article but is worth mentioning in response to your comment is this. LC wants to price risk much more accurately than before. This means that most loans within each grade will perform in a similar fashion. They want to eliminate or at least minimize the advantages that note pickers have had by analyzing the loan history. I think that will be tough given the millions of combinations of paramaters but that is the direction they are moving.

  3. says

    Fascinating example of the LC executive! I agree with Roy and Henry, this is generally positive news. I am always eager for there to be a wider selection of borrowers to choose, and am happy LC has found a way to include these people they previously denied.

  4. Bryce M. says

    I’m enamored with the backtesting method they used. Seeing who among the 800,000 rejected people had loans taken out elsewhere around the same time and monitoring those payments as an outcome is brilliant.

  5. says

    Great article, Peter. I am glad you were able get more information and reasoning behind underwriting changes. The switch from a redline model to a holistic one is a nice change which I am sure will payoff for LC and lenders in few years.

    But I expect a year or two will be rough while LC fine-tunes the new model. Uncertainty brings more opportunities for data crunchers like us, so no complains from me. :-)

    Did LC mention anything about providing more details about new underwriting process with the lender or through prospectus?

    • says

      Hi Anil, I agree it might take a while before the new underwriting model is settled. In fact it will always be a moving model I expect so there should always be opportunities.

      As for the underwriting process they do not want to share more details other than what they shared with me. They do not want borrowers to be able to game their system not to mention give away their secrets to their p2p competitors…

  6. Daniel G. says

    Thanks for this information. It is great to hear more of the inner workings of LC.

    So if the problem is with LC pricing risk better in the future and creating a lack of “speculation” and expert picking algorithms, at lease the flip side is that it can become a more successful passive investment (i.e. like an index fund) for many who do not want to actively manage the fund. Of course, things will always be in flux, so there will always be opportunities to analyze trends for both LC and the investor.

    • says

      Hi Daniel, I think that is their goal. They would like an investment in LC to be as passive as possible just like an index fund. But as I have stated above, there will likely be opportunities for savvy investors to beat the averages for many years to come, it just might become a little more difficult.

        • says

          Fair point. And I think that is what LC are getting at. Ideally, they would like a PRIME account to provide the same returns as someone who is actively picking notes given the same weighted average interest rate.

          • Roy S. says

            I agree with you Peter that they are trying to make the entire investment more passive. I would be curious to know what percentage of people already treat it as such, but I’m not sure how one would go about figuring that out. I know a lot of people who read your blog are trying to get 15%, 20% or more in returns and like to try and exploit the weaknesses in the current UW models. But if they can make it more of a passive investment while keeping returns in the double digit range, then I’m fine with that. A 10% ROI by any means is still a great investment return. Personally, I would prefer something closer to 12% before making it purely a passive investment for myself.

          • Dan B says

            Yeah well, there are always going to be some people actively picking notes who outperform the averages & outperform Prime. The other thing that is rarely discussed is that there are always going to be some other people who actively manage, yet under perform the averages & under perform Prime. The above are realities. One way to dramatically change this reality is to eliminate note picking. I hope it will never happen………….as I’m out the door, if it does. Why? Because I know what my returns are & I KNOW what Prime’s returns are (don’t we Peter?) The gap is enormous & is the main reason why I’m here.

            Now, as far as I’m concerned all this is interesting “cocktail party” type conversation………………….but it does miss the point as to why these changes were made……………& that was to reject fewer borrowers & drive more business through the door. The volume & especially the loan “approval” percentages have gone up substantially in the past year.

            I have no problem with any of the above. But I do resist any attempt to spin this as an investor friendly move or any suggestion that this was motivated by a desire to achieve a higher quality overall borrower (& therefore, loans that will perform better)

          • Roy S. says

            “But I do resist any attempt to spin this as an investor friendly move or any suggestion that this was motivated by a desire to achieve a higher quality overall borrower”

            I would have to disagree with you, Dan. Obviously, this is not a savvy investor friendly move. By savvy, I mean those who spend the time doing their due diligence to exploit the current UW model to achieve higher returns than average. Obviously, it is to the savvy investor’s best interests to have flaws in the UW model that can be exploited to achieve higher returns than the average. So, for those of us on this board, it may not be the most investor friendly move. HOWEVER, for the average investor this is a friendly move.

            Every time LC or Prosper adjusts their UW model it is an attempt to better price risk. Yes, there will be fewer B borrowers given a C or D grade lowering the returns of those who have worked to figure out which borrowers are more likely to be rated a higher risk than they really are. But to the average investor, it means that a borrower given a B rating can be assumed to be more accurately rated and priced.

            Yes, the result does also achieve fewer borrowers being immediately disqualified and thus drives more business through the door. But this, too, is in my opinion a good thing. More loans on the platform means more loans from which to choose. I find it hard to say that more choice = a bad thing for lenders. In fact, I would say the opposite is ture…assuming, of course, the UW models are accurate.

            At this point, I lean towards the changes being a good thing for the industry. It may suck for you and those like you who wish to achieve 20%+ returns, but I think overall it is the best move for the industry. I would like to see them keep returns as high as possible for investors, but obviously this is also supposed to be beneficial for the borrowers, too. I don’t agree with sacrificing the benefit of borrowers for the benefit of lenders. Both should be able to benefit…and that means better pricing of risk for the borrowers (i.e. interest rates commensurate with risk). …and driving more business through the door is also better for the industry. LC and Prosper have to find the balance between pleasing the borrowers and lenders as well as their own bottom line. If they are in the red in perpetuity, then that is definitely a bad thing for investors. It is the same when it comes to balancing retail investors and institutional investors.

            This a nascent industry, and it will continue to evolve. Those here who are investing in it today may not be doing so tomorrow (e.g. you stated that if they eliminate Note picking, you will be gone). Remember when they used the auction model? Yeah, well…things change. I think moving away from the auction was a good move (I waited until after they ditched the auction model before I started investing), but not every move will end up being a good one. You may be right that this may not be an investor friendly move, and it may in fact prove detrimental to investor returns. Not all change is good – in fact, the worst change occurs when people try to change things for the sake of change (I can’t tell you how many times I have had to listen to some idiot state how change is always a good thing and we should continually implement changes; it isn’t always a good thing). But good or bad, this is at the very least a learning experience for LC.

          • Dan B says

            Roy S…………….Why would these changes “suck” for me? In fact these changes have minimal effect on me, or on anyone else, who uses their own filters & sets their own borrower requirements.

            And I didn’t say that these changes were bad for the industry or LC. What I did say is to call these changes for what they are actually about & not for what their marketing department like us to believe they are about.

          • Roy S. says

            I think everyone knows that ultimately all changes are about trying to improve the bottom line for LC. I don’t believe anyone here disputes that. I think the question on hand is whether investors believe that it is also a positive change for investors. I don’t factor in what LC’s (or Prosper’s) marketing departments have to say. Obviously, they want to promote everything they are doing in as positive a light as possible. I generally evaluate what they are doing, and with the limited information I have in my possession, I come to my own conclusions. Here, I believe the changes LC is implementing are positive.

  7. Danny S says

    I can agree with almost all the changes made…

    Not declining borrowers because of failing to meet a single test… sure.
    Past BK, but a CS now of 660+… sure
    Removing some of the restrictions (ie revolving credit balance)…sure.

    But, CURRENT delinquency? no way. I would never ever look at such a loan for investment.

    • says

      Hi Danny, Everyone has to invest in criteria that feels right to them. I like no delinquencies as well but if you do analysis on the loan history you will see that it makes little or no difference to returns in many cases.

  8. Jacob says

    So to clarify… LC pulled which credit report on the 822,000 rejected? Did they use the one at time of borrower application or did they re-pull a report after the borrower had been declined? Confused?

  9. Hrant says

    So, I am trying to see as to what other “serious” investors are truly achieving as their returns..over a few years? Can you elucidate me on the long term 20% returns some are inferring to? I am putting in a lot of time picking, and am averaging around 10.5% over the past several years. Even tried to invest thru LC funds- not attractive due to returns, lock-ups, huge minimums, counterparty risks, etc. If anyone can guide me to a more automatic & consistent returns over long term, would be very interested in investing thru them or their methodology.
    Please reach out to me.
    The “changes” at LC are all good to increase loans, and bottom line for LC, as a business, and all should be mindful of this. P2P is an evolving model, on the way up.

    • says

      The 20% numbers you see bandied around are all from relatively new investors. Once a portfolio has an average age of more than a year those numbers are virtually impossible to sustain.

      Let’s look at Prosper where investor returns are public. According to Prosper Stats there are 15 investors who are earning 15% or more with at least 500 loans invested and an average portfolio age of at least 12 months. If you take that average age up to 18 months then that number reduces to 8 investors with the best investor earning 16.75%. See that list here. Now, Prosper has an average interest rate much higher than Lending Club, so the top investors at Prosper will likely do doing a little better than at LC.

      • Roy S. says

        Peter, where is your screen name among the list? According to this post (http://www.lendacademy.com/my-quarterly-p2p-lending-results-q4-2012/) your returns are above 15% for both of your Prosper accounts. I know one of your accounts is SLN-10, but it does not appear on the list. The problem I have with the list you link to is that you exclude portfolios that are heavily weighted with younger Notes. It is possible that there are portfolios where the average Note age for the portfolio is younger than 12 months, BUT the ROI of all Notes aged past 12 months still has an ROI over 15%. So your post is not entirely accurate. I just wanted to point that out. I’m not sure it really detracts from your argument, but it does beg the question. It would be interesting to see what the distribution of seasoned return numbers are from Prosper. I know that Prosper uses 10 months instead of the 12 or 18 you use here, but it should demonstrate a similar point. And Prosper’s seasoned return numbers only exclude Notes younger than 10 months rather than entire portfolios where the average Note age is less than 12 or 18 months.

        • says

          Roy My returns are indeed above 15% for my Prosper accounts according to my XIRR() calculations but at Prosper Stats I am around 13% for my main SLN-10 account. Also, my average loan age is just 10 months so I am excluded from that list. There is no easy way to find the seasoned return numbers on Prosper Stats which is why I focused on the average loan age for the entire account.

  10. Hrant says

    Thank you for the clarification. So, have a Prosper act, am trying to close it due to the reason that once a loan becomes late (even by one day), one cannot sell the loan!
    LC adding funds to steadily, currently have approx 1000 loans soon to double to 2000, in it for approx 4-5 years…manual investing,,,a lot of time spent.
    Again, does anyone know if can invest 10-50-100 G in my name act, and someone else picks loans automatically that has any sort of good track record for a fee that justifies it?
    Where can any of the same stats in LC as above can be obtained? As far as top, average, and returns of more than 3/4/5 years?
    Thank you in advance, and keep up the great work. Much appreciate it:)

    • says

      There are options as to your investing question but nothing I can share here. I will send you an email.

      There is no data on LC investors returns – the information is not publicly available.

  11. Hrant says

    Just out of curiosity…just tried to invest…only 420 loans???
    At one time we were over a thousand loans, even a lot more. Any ideas as to why the reduction drastically?
    Also, as to the late/delinquent borrowers, why doesn’t LC get collection agencies to go after them, as not worthwhile for 25$ investment, yet well worth it for the $1,000-$35,000. investments made collectively.
    Thanks in advance for the input.

  12. Kate D. says

    Peter, it seems to me that as long as we are allowed to individually select loans, those of us who are knowledgeable, selective and patient will have an edge. My primary reason for saying this is that Lending Club has to follow highly regulated banking rules and I don’t. I know Lending Club is not allowed to discriminate according to residency–yet results show certain states have higher defaults, certain states have higher unemployment, certain industries and job categories are in greater demand and less vulnerable to economic variables. I doubt if Lending Club is allowed to discriminate purely on higher income (unrelated to other angles, such as relative to amount being borrowed, etc.) yet I keep reading research that shows higher incomes have lower default rates. They certainly are not allowed to discriminate according to age–and yet, the various age categories behave differently when back-tested. I am less likely to give loans to people whose first credit was earlier than 1980 or later than 2003, for example. The amazing thing is that most people pay back their loans. Kudos to Lending Club borrowers!

    My account is less than a year and one-half old (and I was a fool in some of my earlier selections,) so it’s hard to tell yet if my selection is making a meaningful difference. The only measure so far is that I am slightly above the maximum LC gives for loans of my rating and time length. I’m learning all I can about selection.


Leave a Reply

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.