Author Topic: Who's putting money in this one and why?  (Read 5683 times)

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Who's putting money in this one and why?
« on: July 30, 2014, 01:05:15 PM »
In my "Slow to fund" thread I pointed out a loan that scored 923 (of 999) in my scoring system but was funding much slower than other similar loans I have followed. Anil an lascott pointed out a few items that concerned them about what otherwise looked like a very solid 19.52% loan.

While my scoring systems max score is 999, and although I haven't seen one before, it is possible to get a negative score. One popped up this morning which scores -90 ...

https://www.lendingclub.com/browse/loanDetail.action?loan_id=23293085

Almost everything, item for item, that responders cautioned about regarding loan 22474197 are problems with this loan, TOGETHER WITH NEARLY EVERY MAJOR INDICATOR we have ever discussed here

A) DTI = 34.88%
B) income < $50k
C) PTI > 10% (payment to income)
D) Open credit lines = 21
E) Total credit lines = 45
F) Revolving accounts = 33
G) Accts open last 24m = 6
H) Delinq last 2 yrs = 1
I) mnths since 90days PD = 8
J) Accts ever 120 days PD = 2
K) Bank cards = 17,  Active = 10
L) BLE risk 128.5 (High risk from Peercube - I'm learning guys)
M) FICO = 665
N) Title =    "administartive asst" [SIC]  who can't spell administrative!

I'm exhausted. This is an underpaid over extended secretary living well beyond his/her means by card juggling.
How can LC assign a 13.35% interest rate to this loan (which is funding smoothly ...why?) and nearly 20% to 22474197?

I have to think this is some relative of Renaud Laplanche or someone at LCs secretary with benefits.
Steve

trevor

  • Newbie
  • *
  • Posts: 39
    • View Profile
Re: Who's putting money in this one and why?
« Reply #1 on: July 30, 2014, 01:46:15 PM »
People who believe that "every note is created equal" and don't use any filters. And this is exactly the reason why every loan is NOT created equal...

At least it was a loan for debt consolidation.

Lovinglifestyle

  • Hero Member
  • *****
  • Posts: 897
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #2 on: July 30, 2014, 02:04:35 PM »
I looked at it on PeerCube.  Even worse news there.  I wouldn't have touched it if it were a G.  But then again, LC doesn't seem to consider 3 mortgages and debt (68K) over annual income to be a bad thing. Not to mention the last credit card opened one month ago.

I understand the proof required to be credit-worthy enough to get mortgages and the irregularities of rental income vs. output, but to me it spells expense and trouble. 

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #3 on: July 30, 2014, 02:48:43 PM »
I saw those things too on Peercube and InterestRadar. I just got tired of listing red flags. Like I said in another post, my model differs so widely from the credit risk assigned by LC I am often concerned they have more information than they are sharing with us.
Steve

Fred93

  • Hero Member
  • *****
  • Posts: 2116
    • View Profile
Re: Who's putting money in this one and why?
« Reply #4 on: July 30, 2014, 02:58:07 PM »
I just got tired of listing red flags. Like I said in another post, my model differs so widely from the credit risk assigned by LC I am often concerned they have more information than they are sharing with us.

Have you validated your method of scoring with any mathematics? ... or is it intuitive? ... or what?

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #5 on: July 30, 2014, 03:30:57 PM »
I've done quite a bit of testing mostly with NickelSteamroller. And I don't think I'm doing anything that I hasn't been endorsed elsewhere on the board. The only reason I download csv file and score in Excel is I do move some fuzzy thresholds around based on ratios and other algebraic relationships between variables that I can't do with the 3rd party tools.

The simplest example is PTI (payment to income ratio which I have described elsewhere). So rather than have a single upper limit on DTI, I have an upper limit on DTI + PTI. lascott informed me LendingRobot includes PTI but it still won't permit me to filter on the sum.

I have been using it for 3 months with no problem loans so far. All my problem loans where purchases I made before I started scoring this way. I have more modeling to do (and coding - I intend to use the LC API) in large part because I am very nervous about the present day validity of the historical record, most of which was collected in the depths of the worst recession since the 30s.  I keep the account small while I experiment, learn, and hope the historical record overestimates the risk of default in a growing or stable economy.
Steve

Ran

  • Full Member
  • ***
  • Posts: 145
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #6 on: July 30, 2014, 03:42:30 PM »
In my "Slow to fund" thread I pointed out a loan that scored 923 (of 999) in my scoring system but was funding much slower than other similar loans I have followed. Anil an lascott pointed out a few items that concerned them about what otherwise looked like a very solid 19.52% loan.

While my scoring systems max score is 999, and although I haven't seen one before, it is possible to get a negative score. One popped up this morning which scores -90 ...

https://www.lendingclub.com/browse/loanDetail.action?loan_id=23293085

Almost everything, item for item, that responders cautioned about regarding loan 22474197 are problems with this loan, TOGETHER WITH NEARLY EVERY MAJOR INDICATOR we have ever discussed here

A) DTI = 34.88%
B) income < $50k
C) PTI > 10% (payment to income)
D) Open credit lines = 21
E) Total credit lines = 45
F) Revolving accounts = 33
G) Accts open last 24m = 6
H) Delinq last 2 yrs = 1
I) mnths since 90days PD = 8
J) Accts ever 120 days PD = 2
K) Bank cards = 17,  Active = 10
L) BLE risk 128.5 (High risk from Peercube - I'm learning guys)
M) FICO = 665
N) Title =    "administartive asst" [SIC]  who can't spell administrative!

I'm exhausted. This is an underpaid over extended secretary living well beyond his/her means by card juggling.
How can LC assign a 13.35% interest rate to this loan (which is funding smoothly ...why?) and nearly 20% to 22474197?

I have to think this is some relative of Renaud Laplanche or someone at LCs secretary with benefits.

I will take it if it's E grade loan with >20% interest. C2 with 13.5% is a killer.
But I still want to answer your question why someone will buy this loan. Except that the income is a little low, no pubic record and no 6m inqury, so not a too bad loan. Also my system went south yesterday because I made a change that could not be tested well, and I bought 10 loans that are still waiting for takers today :'(

Ran

  • Full Member
  • ***
  • Posts: 145
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #7 on: July 30, 2014, 05:06:17 PM »
I just got tired of listing red flags. Like I said in another post, my model differs so widely from the credit risk assigned by LC I am often concerned they have more information than they are sharing with us.

Have you validated your method of scoring with any mathematics? ... or is it intuitive? ... or what?
Here is a sample of my portfolio loss rate vs NS loan loss rate in the same period between 8/1/14-3/1/14
Grade Total    Charge Off/Late 31+   My Loss Rate       NS Loss Rate   Yield Advantage
C   $4950   $100                         2.0%                     3.4%              +1.4
D   $8725   $200                         2.3%                     5.19%              +2.9
E   $3750   $175                         4.7%                     7.31%              +2.6
F   $1300   $100                         7.7%                    11.08%              +3.4
My holding size is small because I have gradually changed my focus to Folio IGP/LATE loans since beginning of the year.
But it still speaks a lot about model advantage.
People say you cannot beat stock market, that is right. However, LendingClub issues the loan at fixed rate who is only a market player, not THE market. So you still can beat LendingClub.
« Last Edit: July 30, 2014, 05:10:10 PM by Ran »

turing

  • Full Member
  • ***
  • Posts: 222
    • View Profile
    • PeerToPeerQuant.com
    • Email
Re: Who's putting money in this one and why?
« Reply #8 on: July 30, 2014, 05:23:27 PM »
I'm exhausted. This is an underpaid over extended secretary living well beyond his/her means by card juggling.
How can LC assign a 13.35% interest rate to this loan (which is funding smoothly ...why?) and nearly 20% to 22474197?

This is the exact reason I think there is room for improvement on LC's model.  If notes like this are getting funded, there is room for improvement and therefore money to be made by doing it better than them.

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #9 on: July 30, 2014, 05:47:04 PM »
Here is a sample of my portfolio loss rate vs NS loan loss rate in the same period between 8/1/14-3/1/14
Hey Ran,
I'm having trouble with your stated date range and not guessing well in NS. What date range did you use? I don't think my account is not old enough to see anything meaningful in this type of analysis.
Steve

ksb1234

  • Newbie
  • *
  • Posts: 7
    • View Profile
Re: Who's putting money in this one and why?
« Reply #10 on: July 30, 2014, 06:03:48 PM »
Just to confirm, this note was listed at 6am (Pacific time) today?

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #11 on: July 30, 2014, 06:11:25 PM »
I think that is right. Can't be sure it wasn't 10 pacific time but it was one or the other.
Steve

Ran

  • Full Member
  • ***
  • Posts: 145
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #12 on: July 30, 2014, 06:53:41 PM »
Here is a sample of my portfolio loss rate vs NS loan loss rate in the same period between 8/1/14-3/1/14
Hey Ran,
I'm having trouble with your stated date range and not guessing well in NS. What date range did you use? I don't think my account is not old enough to see anything meaningful in this type of analysis.
I basically pick issue date 8/1/2013 to 3/1/2014 on NS Analytics->Back Testing page, and increase loss estimate to 1 for Late (31-120 days) loans. my portfolio was acquired within that date range, though heavy buying was in Sept-Oct last year, so my loss rate is even lower than NS comparable loan population. Late 31-120 is really when loan heads to charge off, which only happens after 3-6m into loan issuance. So if you are new to LC, you have to wait for 1/2 year before you can see whether your model performs as expected.
As to loan performance, I do not think simple filtering is a good tech. Many loan characteristics are correlated, so you have to develop a regression based model based on selected independent variables, then score the loan. For example, I do not use public record in my model, though it does affect loan performance. The effect of public record is actually better represented by a collection of 2~3 other loan characteristics.
« Last Edit: July 30, 2014, 06:55:40 PM by Ran »

hoggy1

  • Sr. Member
  • ****
  • Posts: 401
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #13 on: July 31, 2014, 03:53:00 PM »
Many loan characteristics are correlated, so you have to develop a regression based model based on selected independent variables, then score the loan. For example, I do not use public record in my model, though it does affect loan performance. The effect of public record is actually better represented by a collection of 2~3 other loan characteristics.
I agree with you and have weeded out many early mistakes double testing items but I chose differently with regards to your example. I do use public records as a catch-all for bankruptcies, tax liens and civil judgements and don't buy loans with any public records (they can't score 900). Why do you think it is better represented by these other items?
Steve

Ran

  • Full Member
  • ***
  • Posts: 145
    • View Profile
    • Email
Re: Who's putting money in this one and why?
« Reply #14 on: July 31, 2014, 05:07:53 PM »
Many loan characteristics are correlated, so you have to develop a regression based model based on selected independent variables, then score the loan. For example, I do not use public record in my model, though it does affect loan performance. The effect of public record is actually better represented by a collection of 2~3 other loan characteristics.
I agree with you and have weeded out many early mistakes double testing items but I chose differently with regards to your example. I do use public records as a catch-all for bankruptcies, tax liens and civil judgements and don't buy loans with any public records (they can't score 900). Why do you think it is better represented by these other items?
People having had public records typically have limited access to credit lines, mortgage, lower FICO, and a few other terms. My model just have better prediction significance when public record is removed and two other parameters added