Lending Club Proprietary Model Ranking

[faeriering]

So I was just reading Anil's site and he was saying that LC no longer publishes how they are setting interest rates . . . I couldn't believe they would strip this transparancy from their site.  So I had to double check : http://www.lendingclub.com/public/how-we-set-interest-rates.action

And sure enough, they are only showing their risk modifiers for Loan amount and loan term.

I feel a little betrayed by this, since I used to try to focus on moving nobs in my filters that weren't covered by their risk profile.

[yojoakak]

They used to list 9 steps. http://web.archive.org/web/20121031160219/http://www.lendingclub.com/public/how-we-set-interest-rates.action

Now they list just 4.

Is it possible they reworked their assessment to use just 4?

What did Anil say about it?

[Zach]

So I was just reading Anil's site and he was saying that LC no longer publishes how they are setting interest rates . . . I couldn't believe they would strip this transparancy from their site.  So I had to double check : http://www.lendingclub.com/public/how-we-set-interest-rates.action

And sure enough, they are only showing their risk modifiers for Loan amount and loan term.

I feel a little betrayed by this, since I used to try to focus on moving nobs in my filters that weren't covered by their risk profile.

While transparency is appreciated by investors, you have to consider the situation from LC's perspective - if they divulge the intellectual property to Prosper or another emerging competitor they'd be giving away a distinct advantage without having to do much work for it.

[Peter]

So I was just reading Anil's site and he was saying that LC no longer publishes how they are setting interest rates . . . I couldn't believe they would strip this transparancy from their site.  So I had to double check : http://www.lendingclub.com/public/how-we-set-interest-rates.action

And sure enough, they are only showing their risk modifiers for Loan amount and loan term.

I feel a little betrayed by this, since I used to try to focus on moving nobs in my filters that weren't covered by their risk profile.

I am going to do some investigation and will try and get the head of risk management at LC to answer my questions. Will hopefully have some answers in a post some time soon.

In the meantime everyone should read Anil's post from this morning:
http://andirog.blogspot.com/2013/01/recent-changes-in-minimum-credit.html

[faeriering]

sorry for the late reply.  For completeness sake, here's the link to the article I was reading:

http://andirog.blogspot.com/2013/01/lending-club-loans-impact-of-recent.html

[Simon]

I'm interested in the reasoning behind this decision. Did they do it so as to stop people from gaming the system when they applied for loans? Or are their reasons more about not wanting to do the work and communicate?

[Zach]

I'm interested in the reasoning behind this decision. Did they do it so as to stop people from gaming the system when they applied for loans? Or are their reasons more about not wanting to do the work and communicate?

I don't think its about "gaming the system" - borrowers who are savvy can still do just that.
Instead of borrowing $35k, they could borrow $34,975 and save 1 or 2 risk modifier dings.

It all comes down to competition and potential competition in my opinion.

[faeriering]

I think it's because they don't want lenders to know how many "less than prime" borrowers they are making loans available to now.  I am a bit of a pessimist though.  I don't see what advantage/disadvantage they would have in the market place for publishing their criteria besides making their lenders more informed investors.  Prosper doesn't copy LC rate setting model.

[AnilG]

I also don't see the benefits of not describing how LC assigns credit grade. It is going to reduce the reliability and trustworthiness of credit grade as loan quality proxy.

The more concerning for me is the presence of borrowers with current delinquencies, previous bankruptcies, tax liens, collections and Lending Club credit grade model not accounting for these factors in assigning credit grade. I understand expanding borrower base but the loan quality model should account for such negative factors.

Right now, too many such loans show up in grade A to C. I believe no lender will rate borrowers with previous bankruptcy who runs up credit card balance again as same risk as the borrower who has no previous bankruptcy and does the same.

I think it's because they don't want lenders to know how many "less than prime" borrowers they are making loans available to now.  I am a bit of a pessimist though.  I don't see what advantage/disadvantage they would have in the market place for publishing their criteria besides making their lenders more informed investors.  Prosper doesn't copy LC rate setting model.

[breitenm]

I find it slightly worrying that they changed their methodology. Without knowing what the new method is, it's hard to compare loan quality across time. It's harder to do any accurate statistics from historical data if the meaning of a D rating changes somewhere in the middle even slightly. On the other hand IIRC Fico also adjusts their scoring from time to time, but they put in a lot of effort to preserve the meaning of the scores. I hope Lending Club will explain in a bit more detail what they changed and if the meaning of the levels is still roughly the same.

[Zach]

I also don't see the benefits of not describing how LC assigns credit grade. It is going to reduce the reliability and trustworthiness of credit grade as loan quality proxy.

The more concerning for me is the presence of borrowers with current delinquencies, previous bankruptcies, tax liens, collections and Lending Club credit grade model not accounting for these factors in assigning credit grade. I understand expanding borrower base but the loan quality model should account for such negative factors.

Right now, too many such loans show up in grade A to C. I believe no lender will rate borrowers with previous bankruptcy who runs up credit card balance again as same risk as the borrower who has no previous bankruptcy and does the same.

It shouldn't really make much of a difference so long as the same credit/listing data is both available and consistent throughout the loan database. Services like Interest Radar or P2P Analytics are using risk factors that don't specifically count on the credit grade assigned by LC, rather a large combination of the factors that correlate most with defaults.

I am confident LC knows what they're doing and I'm sure much statistical analysis went into expanding their borrower customer base, while not compromising investor returns - only time will tell for sure.