Author Topic: What's This "Best Note" Selection Business Anyway?  (Read 30997 times)

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
What's This "Best Note" Selection Business Anyway?
« on: January 12, 2014, 12:57:45 PM »
Been thinking of writing this a while and have an hour or so before football. Folio users may ignore this thread. That's an entirely different animal.

Before we see any loan LC has reviewed it in sufficient depth to its own satisfaction to assign an interest rate sub-grade. Not only do they have all data that is eventually available to us in the browsenotes but, given direct access to the borrower, they clearly have much more. They have a team of seasoned professionals that lead the underwriting process and they have been doing this a long time. They have developed a proprietary loan scoring model and every bit of historic loan data available to us is also available to them. They have access to expensive historic data from the credit bureaus likely not economic for the rest of us to obtain. Finally, history has shown LC has done a pretty good job of achieving its goal to "... provide higher risk-adjusted returns for each loan grade increment from A1 to G5." (actually not so well with D, F and G).

LC policy makers select an interest rate structure that they feel will maximize their fees at any point in time, and loans are uniformly underwritten to conform to that structure. All loans graded XX on any day are to the very best of LC's ability the same and no loans are improperly graded XX when they should be YY.

So why is it that we believe we can use a subset of the data available to LC, filter for zero inquiries, no business loans, income >$3k/mo., etc. and get the "best" notes? Typically when we filter we filter on grade, not sub-grade. That's pretty coarse risk bins. I guess it's possible to create a model better than LC's given the data we have or can get, but I don't see large obvious factors persisting over time.

Nonetheless filtering is widely practiced and by all accounts appears to work. What am I missing here? (Please avoid using the word heretic; also dummy if possible)


Randawl

  • Sr. Member
  • ****
  • Posts: 469
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #1 on: January 12, 2014, 01:50:14 PM »
Been thinking of writing this a while and have an hour or so before football. Folio users may ignore this thread. That's an entirely different animal.

Before we see any loan LC has reviewed it in sufficient depth to its own satisfaction to assign an interest rate sub-grade. Not only do they have all data that is eventually available to us in the browsenotes but, given direct access to the borrower, they clearly have much more. They have a team of seasoned professionals that lead the underwriting process and they have been doing this a long time. They have developed a proprietary loan scoring model and every bit of historic loan data available to us is also available to them. They have access to expensive historic data from the credit bureaus likely not economic for the rest of us to obtain. Finally, history has shown LC has done a pretty good job of achieving its goal to "... provide higher risk-adjusted returns for each loan grade increment from A1 to G5." (actually not so well with D, F and G).

LC policy makers select an interest rate structure that they feel will maximize their fees at any point in time, and loans are uniformly underwritten to conform to that structure. All loans graded XX on any day are to the very best of LC's ability the same and no loans are improperly graded XX when they should be YY.

So why is it that we believe we can use a subset of the data available to LC, filter for zero inquiries, no business loans, income >$3k/mo., etc. and get the "best" notes? Typically when we filter we filter on grade, not sub-grade. That's pretty coarse risk bins. I guess it's possible to create a model better than LC's given the data we have or can get, but I don't see large obvious factors persisting over time.

Nonetheless filtering is widely practiced and by all accounts appears to work. What am I missing here? (Please avoid using the word heretic; also dummy if possible)

I want to first address the above bolded phrase.  While it may be their theoretical intention to do so (and they have recently gotten better at it), it is not the case in our current reality.  That's how you make more money than the other person.  Take a look at the average account return curves.  The vast majority of people fall between 5%-10%.  What separates them?  Do you purport people's accounts above that 10% line are there 100% due to luck and no other factors?  Taking advantage of small inefficiencies and information asymmetries (even as they themselves change) is what keeps me multiple standard deviations above the crowd.

Lovinglifestyle

  • Hero Member
  • *****
  • Posts: 896
    • View Profile
    • Email
Re: What's This "Best Note" Selection Business Anyway?
« Reply #2 on: January 12, 2014, 02:03:40 PM »
I hope somebody better qualified than I (who am not at all qualified) responds regarding the origination fee motivation factor discussed in other threads. 

If LC comes out ahead on loans that go bad, but investors come out behind, that would be grounds for suspicion regarding whose best interests are in mind when grading loans.  I believe in creating the highest good for all concerned (win-win), but the corporation has to look after its own bottom line first for anybody to win.  To what extent it does that may be where filtering has an edge.

Sorry for stating some of the obvious here.  Don't mean to waste your time.

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #3 on: January 12, 2014, 03:04:18 PM »
Taking advantage of small inefficiencies and information asymmetries (even as they themselves change) is what keeps me multiple standard deviations above the crowd.

I tried to make it clear at the beginning that Folio trading is a different animal completely. I'm talking about the purchase of fresh notes only. If Folio trading is part of your strategy my long winded post doesn't apply to you.

Only yesterday on another thread you mentioned Folio trading is part of your strategy and I certainly agree skilled traders have the edge there. Apparently you are one.

core

  • Hero Member
  • *****
  • Posts: 1784
  • Your loss is my gain
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #4 on: January 12, 2014, 03:08:55 PM »
I tried to make it clear at the beginning that Folio trading is a different animal completely. I'm talking about the purchase of fresh notes only. If Folio trading is part of your strategy my long winded post doesn't apply to you.

I don't know... he did say "multiple standard deviations above the crowd".  He couldn't have been talking about Folio because LC doesn't include Folio users on their pretty scatter plots.  So how would he know if he was multiple SDs above the crowd or not?  My conclusion:  He must have been talking about something else, on topic.

Ran

  • Full Member
  • ***
  • Posts: 145
    • View Profile
    • Email
Re: What's This "Best Note" Selection Business Anyway?
« Reply #5 on: January 12, 2014, 03:34:45 PM »
Been thinking of writing this a while and have an hour or so before football. Folio users may ignore this thread. That's an entirely different animal.

Before we see any loan LC has reviewed it in sufficient depth to its own satisfaction to assign an interest rate sub-grade. Not only do they have all data that is eventually available to us in the browsenotes but, given direct access to the borrower, they clearly have much more. They have a team of seasoned professionals that lead the underwriting process and they have been doing this a long time. They have developed a proprietary loan scoring model and every bit of historic loan data available to us is also available to them. They have access to expensive historic data from the credit bureaus likely not economic for the rest of us to obtain. Finally, history has shown LC has done a pretty good job of achieving its goal to "... provide higher risk-adjusted returns for each loan grade increment from A1 to G5." (actually not so well with D, F and G).

LC policy makers select an interest rate structure that they feel will maximize their fees at any point in time, and loans are uniformly underwritten to conform to that structure. All loans graded XX on any day are to the very best of LC's ability the same and no loans are improperly graded XX when they should be YY.

So why is it that we believe we can use a subset of the data available to LC, filter for zero inquiries, no business loans, income >$3k/mo., etc. and get the "best" notes? Typically when we filter we filter on grade, not sub-grade. That's pretty coarse risk bins. I guess it's possible to create a model better than LC's given the data we have or can get, but I don't see large obvious factors persisting over time.

Nonetheless filtering is widely practiced and by all accounts appears to work. What am I missing here? (Please avoid using the word heretic; also dummy if possible)

My take is that mass market always beat single authority. Individual market participant may not do a better job overall compared to LC scoring system, however, for a specific sector, for example, D notes, 60-month notes, some specialized market participant can develop a better model. And if that specialized market participant uses his own model for a sub-set of loans, while using LC for other notes, he will gain advantage. We have seen how LC have assigned higher rates for C/D but lower rates for 60m notes in the past years. Also LC as a company, even they noticed some deficiencies in their system, it takes time for them to react. If a few specialized scoring system providers like IR can cooperate to integrate their system, they can be even stronger.
« Last Edit: January 12, 2014, 03:36:45 PM by Ran »

bobeubanks

  • Sr. Member
  • ****
  • Posts: 273
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #6 on: January 12, 2014, 03:48:07 PM »
I don't see how LC (or Prosper) has much motivation to truly assign interest rates in anything other than being more or less right. They get the origination fee up front, and then get about the same amount of extra income after that regardless of interest rate.

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #7 on: January 12, 2014, 03:49:33 PM »
I hope somebody better qualified than I (who am not at all qualified) responds regarding the origination fee motivation factor discussed in other threads.

Well I'll admit I'm pretty clueless about LC's motivations and policies too. However my understanding is that they do set the rates for all the loans, not just those for us small retail folks. Along with the rate they set comes an expected default rate. Institutional lenders are going to hold LC accountable to that default rate prediction and judge LC in no small measure on its accuracy as their portfolios mature. Seems to me that gives LC motivation to get it right, but that's just my uninformed working theory.

Randawl

  • Sr. Member
  • ****
  • Posts: 469
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #8 on: January 12, 2014, 03:56:59 PM »
Taking advantage of small inefficiencies and information asymmetries (even as they themselves change) is what keeps me multiple standard deviations above the crowd.
I tried to make it clear at the beginning that Folio trading is a different animal completely. I'm talking about the purchase of fresh notes only. If Folio trading is part of your strategy my long winded post doesn't apply to you.

That was clear indeed.  I, too, am talking about portfolio performance of a pure "filter and hold" strategy.  The inefficiencies to which I refer (in part) are regarding LC's imperfect underwriting model.  Not to imply that it should be perfect nor ever could be.  I say show me one that is perfect and I'll start the world's first infallible banking system.

Quote
Only yesterday on another thread you mentioned Folio trading is part of your strategy and I certainly agree skilled traders have the edge there. Apparently you are one.

Yes, I do use Folio, and I also know that when I separately analyze my trading and non-trading portfolios, my aforementioned statements remain true. 
« Last Edit: January 12, 2014, 03:59:27 PM by Randawl »

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #9 on: January 12, 2014, 04:16:17 PM »
We have seen how LC have assigned higher rates for C/D but lower rates for 60m notes in the past years. Also LC as a company, even they noticed some deficiencies in their system, it takes time for them to react.

Sure, that's why I said it was hard to see large obvious factors persisting over time. The LC model gets better, macro economic factors change, etc. If you believe personal filtering is essentially front-running the LC model by including relevant factors more quickly I guess that's a possibility.

Fred

  • Hero Member
  • *****
  • Posts: 1421
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #10 on: January 13, 2014, 01:19:39 AM »
What am I missing here?

For one, if we take LC's grades vs. FICO scores, LC's definition of "best note" is very different than that of FICO's.

Borrowers with FICO 700 have been assigned grades A to G by LC.  On the other hand, LC's grade A borrowers have shown FICO ranging from 660 to 845.

As an investor, I definitely make a distinction between grade A borrower with FICO 660 vs. grade A borrower with FICO 845.

You can see some data in this thread (http://www.lendacademy.com/forum/index.php?topic=1887.msg15772#msg15772).
« Last Edit: January 13, 2014, 01:23:02 AM by Fred »

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #11 on: January 13, 2014, 10:09:45 AM »
For one, if we take LC's grades vs. FICO scores, LC's definition of "best note" is very different than that of FICO's.
Most certainly true. In their 2012 Goldman Conference briefing LC published a Gains chart claiming the KS scoring of their proprietary model was superior to FICO08. Believe it or not over time I don't know, but it certainly must have been true for the set of loans tested (Q2-Q4 2010).

I wonder if LC's model incorporates FICO as a parameter? My guess is that they do not, but they do incorporate the same borrower information used to create the FICO score. Even more I wonder why the two models diverge so greatly as shown in the numbers you referred to. What FICO 705 borrower's information makes the LC model assign a loan grade F or G? Do we have access to that same information in browsenotes, or more importantly do we have access in browsenotes to all the information LC uses in their model? I've never seen this question asked or answered, but as you may suspect my guess is no.

https://s3.amazonaws.com/lcmedia2.lendingclub.com/2012-Goldman-Conference.pdf

Rob L

  • Hero Member
  • *****
  • Posts: 2014
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #12 on: January 13, 2014, 11:27:54 AM »
LC policy makers select an interest rate structure that they feel will maximize their fees at any point in time, and loans are uniformly underwritten to conform to that structure. All loans graded XX on any day are to the very best of LC's ability the same and no loans are improperly graded XX when they should be YY.

Sorry: to correct myself, "all loans graded XX on any day are the same". Not true.
The LC model scores of sub-grade XX loans will vary and some will be better than others. However they all fall within the XX "bin" so to speak, but no loans are assigned XX when they should be YY.
For example, E3 loans have various "best-ness" but no E3 loan is improperly graded as E2 or higher, nor E4 or lower (to the accuracy of the LC model).
We do have the opportunity to select the best E3's (or whatever) loans.

Emmanuel

  • Full Member
  • ***
  • Posts: 157
    • View Profile
    • LendingRobot
Re: What's This "Best Note" Selection Business Anyway?
« Reply #13 on: January 13, 2014, 12:12:31 PM »
The loan issuer may want to optimize its model for consistency, not absolute returns. If LC goal is to lower the 'risk' (risk being defined as discrepancies in default rates), they may cluster together loans with different expected returns. For instance, if loans A, B, and C have the following properties: default risk 5% +/- 2%, 7% +/- 5%, 13% +/- 2%, Lending Club could cluster B and C together, and then investors focusing on returns only will consider B a bargain, because it has the same interest rate than C which a much lower (average) default risk.

rawraw

  • Hero Member
  • *****
  • Posts: 2756
    • View Profile
Re: What's This "Best Note" Selection Business Anyway?
« Reply #14 on: January 13, 2014, 03:04:51 PM »
For one, if we take LC's grades vs. FICO scores, LC's definition of "best note" is very different than that of FICO's.
Most certainly true. In their 2012 Goldman Conference briefing LC published a Gains chart claiming the KS scoring of their proprietary model was superior to FICO08. Believe it or not over time I don't know, but it certainly must have been true for the set of loans tested (Q2-Q4 2010).


No they use FICO.  But they have info FICO doesn't have (income related).  So you'd expect it to be better.