Overqualified borrower problem

[yojoakak]

And it's a bigger deal that Lending Club is encouraging this behavior, enticing borrowers, to line their own pockets with origination fees at your expense.

If the first step is admitting that there is a problem...

...then it looks like we're still on step zero.

[hoggy1]

I was wondering what you and Rob were getting at.  I was curious how you guys were going to magically match these up.

B. It turns out member_id's are NEVER REPEATED. Every unique loan has a unique member_id. Why? What purpose does member_id serve?

This was reported 14 months ago in this thread.  That thread is about this same topic, by the way.

Thanks for this link Core. Before my time here, but good to know that at least a year ago memberid was useless and hasn't been recently altered to make this type of analysis hard.

C. So I use a proxy for member_id, earliest_cr_line opened which is REPORTED to the second.

As trade lines age they fall off the report if they were closed 10 years ago.  Much sooner in some cases.  This will often cause the earliest line reported to change.

Not sure this is true. earliest_cr_line goes back to 1944 in the loans file.

I had no idea that times were reported to the CRAs for each trade line.

I don't believe they are. I thought (still think) this is an unaltered property of the borrower when their credit file is first established.

It seems some of you disbelieve that refinances actually happen at all, so just a handful would be enough.  But there is no question that refinances happen.  The question is did the borrower do it on their own, or did Lending Club send the borrower an offer in order to get another origination fee at our expense. 

I am not a disbeliever. I am as certain as you that refinance happens. Why I couldn't find lots of them in the DATA using this method is what surprised me. The method is not completely unsuccessful though. I'll tell you what I DID find in the next post.

I now remember why this scam used to be common knowledge around here, even though borrowers do not tend to post on this forum:  The Q&A section, and borrower-supplied additional information.  ... Now, unfortunately, LC has taken down these free-form text sections.  Probably to prevent investors from seeing what's really going on.

I don't understand this. What has LC removed?

[hoggy1]

WHAT I DID FIND

This process flagged 422 loans by borrowers with an earlier loan. Certainly not enough to account for refinancing so what are they?

A. When you find one, there is no question the borrowers are the same - every field of credit data is identical
B. Very surprisingly to me, almost all are multiple loans (usually 2) taken out within some few days (often 0-2) after one another. Often the interest rates are identical but not always. This leads me to believe that a larger loan was purposely broken into two parts. This is quite a problem from my perspective. 1) I don't get to see the full amount the borrower is borrowing and 2) I might purchase both not realizing I'm lending to the same borrower. (see these loan pairs in notes file: 807138-833456)
C. The timing of these split loans is suspicious. A large majority of these issue in the last few days of December (2010 most prominent) continuing into the first few days of January. (see these loan pairs in notes file:  621124-642871)
D. Finally a really worrying collection:
 1. Search for a borrower whose earliest_cr_line is "9/4/1995 7:44".
 2. There are 8 loans issued to this borrower all within 5 days of each other (472364,470562,470912,471412,471762,470024,472062,472214)
 3. Those loans total $179,700
 4. The borrower lists 1 Denver , 4 Boston, 2 NY, and one other Mass address (surely LC can tell they have multiple loan requests from the same borrower SS#)
 5. The loan purposes are all different
 6. Three of the loans show "Verified Income". Unfortunately they are all different!

OK Core, explain this one to we simpletons.

[core]

As trade lines age they fall off the report if they were closed 10 years ago.  Much sooner in some cases.  This will often cause the earliest line reported to change.

Not sure this is true. earliest_cr_line goes back to 1944 in the loans file.

These facts are not in conflict.  I said trade lines fall off after they have been closed for 10 years.  It is possible a tradeline has been open since 1944.  I do not know what "earliest_cr_line" is for sure in LC's case.  It could very well be TransUnion's file creation date for you, which is not necessarily the same thing as earliest tradeline.

I'm looking at my official TU paper report right now.  There is no "earliest ever line" date.  All it says is "You have been on our files since 10/01/1993".  A hard INQ probably would have created the record even without an open line, so this is not the same thing.  Otherwise a new credit user could go on a massive app spree with impunity.  I would think.  And there is no time listed. 

I had no idea that times were reported to the CRAs for each trade line.

I don't believe they are. I thought (still think) this is an unaltered property of the borrower when their credit file is first established.

So you're saying back in 1944, some guy manually wrote down the exact time the very first line was opened?  Things took time back then, I imagine days; time of day when someone got round to logging it would have little meaning.  Unaltered property or not, that does not change my point.  It had to have gotten there somehow.  I would simply like to know what the TIME is (not the date nor if it remains unaltered), and whether this time is meaningful at all.  Might be a good way to weed out some people, if creditors do report the exact time of the request.  If I see 2am maybe it's a boozer and I should steer clear.

I don't understand this. What has LC removed?

LC has removed the ability for us to ask questions and have the borrower type in whatever they want, giving their sob story or business plan for their sheep brothel or whatever.  It was a very good way to get the "rest of the story".  Even without the Q&A the unsolicited borrower comments were nice.  Now it's gone.  LC says it's to protect the borrowers.  The 12-stepper cheerleaders here would likely say the vast majority never read those comments anyway so quit yer whining.  Sure.

[core]

OK Core, explain this one to we simpletons.

Alright, I'll start by giving a grammar lesson to you simpletons.  The correct usage would be "explain this one to us simpletons" since it is a direct object.

(see these loan pairs in notes file: 807138-833456)

I was all excited when I read your post, hoping you had stumbled upon yet another LendingClub conspiracy to defraud investors.  In this case though, after reviewing those two loan IDs, it seems very innocent.  And yeah it's definitely the same guy.  2 loans, 22 days apart.  From what I've read here you can have up to two loans active at any given time.

Loan #1: Amount requested $24k, amount funded only $18.5k.  So he walked away 5.5k short and accepted the loan anyway.
Loan #2: Amount requested $5.4k.  (Gee, where have we seen that number?)

Loan #1 Q&A:

Q: What are your monthly expenses related to housing (rent, mortgage(s), home equity loan and / or line of credit, utilities, insurance, taxes, etc)
A: (07/12/2011-13:23) - monthly expenses 1887.00

Loan #2 Q&A:

Q: What are your current debt balances, interest rates, and monthly payments by type (credit cards, student loans, mortgages, lines of credit, etc)?
A: (08/10/2011-04:24) - mortgage 1495, loan 444, credit card 194, credit card 134
This adds up to 2267, for you simpletons

Difference in claimed monthly expenses between loan #1 and loan #2 was $380.  Close enough to that new "loan $444" item.  Especially taking into account reduced interest.  And wouldn't you know it, loan #1's monthly payment was exactly $444.15/mo.

Conclusion:  This guy was perfectly honest about everything and played by the rules.  I don't see anything fun to dig up here.  He didn't get the full amount the first time so it seems legit enough to go back to the well for the difference.  He was up front about the new monthly expense.

2. There are 8 loans issued to this borrower all within 5 days of each other (472364,470562,470912,471412,471762,470024,472062,472214)
 3. Those loans total $179,700
 4. The borrower lists 1 Denver , 4 Boston, 2 NY, and one other Mass address (surely LC can tell they have multiple loan requests from the same borrower SS#)
 5. The loan purposes are all different
 6. Three of the loans show "Verified Income". Unfortunately they are all different!

Good find.  This one is not so simple.  Occam's razor may be of some help here.  There are at least two possibilities:

1) Some guy took out 8 different loans under the same SSN within mere days.  Using several different addresses, different incomes, and got all of his different incomes verified.  No doubt he disguised his voice 8 different times during the phone interviews which LC always did back then.  And after getting away with all this, the guy paid them all back when he could have skipped town.
-- or --
2) LendingClub's programmers screwed up (again)

Which one requires the fewest number of absurd assumptions?

Couple of things I noticed while looking over some of the loans in question:

• Notice the revolving credit balance is $255.  Or 0xFF, the maximum value of an unsigned byte.  Seems suspicious to me.  Unintentional byte shifting could cause this.  Struct alignment issue.  Or 255 would be a nice "round number" for a programmer to use if he was typing in bogus data.
• This "guy" owns a home with no mortgage payment but yet in at least 2 loans "he" lists "Home Buying" as the loan purpose.  Doubtful.
• DTI is the same out to 2 decimal places even when income is different. 
• "Member_xxx's Credit History (as reported by credit bureau on 1/1/06)
  The credit report fetch date is way off here.  Unless Stephanie's many powers include ESP, Lending Club is not going to pull a credit report 3 years before the guy even applies. Again, seems to be bogus data to me.
  

I don't know for sure what happened here, but I'm convinced that it's not the same person and something screwy happened with LC's data.  As usual.  I did not go as far as to look at the payment histories for each loan though.  There might be more clues there.

These could be test cases that found their way into the file.  Or these could be some of those "fake" notes that LC issued to meet excess demand. 

[hoggy1]

Core,

I think we should write a book together. I'll work on the title this afternoon.

[rawraw]

I'd be curious what LC has to say about the stuff above.  Esp if some of their data is screwed up

[rawraw]

I take an early payoff as a win and move on.  With the automated tools, I hardly notice, and it maybe costs me a week or two interest on $25, no big deal.

It costs you far more than a week or two interest, because you're paying that 1% servicing fee to Lending Club on the entire lump sum amount.  Can you make 1% on your principal in a "week or two"?  No.

Rawraw says they have "fixed" it so you can't go negative because of this.  Fine, but you could still have up to 2 months of interest wiped right out.  That seems like a big deal to me.  And it's a bigger deal that Lending Club is encouraging this behavior, enticing borrowers, to line their own pockets with origination fees at your expense.

The other thing is that the borrowers you are losing are good borrowers, so it's adverse selection as well.  It'd be a little different if current loan holders had an option to fund the new loan with the proceeds received from the old loan.  It would suck, but seem a little more fair. 

[hoggy1]

I'd be curious what LC has to say about the stuff above.  Esp if some of their data is screwed up

IF? Every time I set out to study some phenomenon in LC data and get looking closely at it, it is SO SCREWED UP I can't face trying to make the corrections necessary to carry on with the work. I have corresponded privately with a number of members on the subject. I can tell you the historical payment data is a disaster.

[core]

I think we should write a book together.  I'll work on the title this afternoon.

I'm one step ahead of you...

"The Joy of Getting Screwed".  Illustrated.
"Fifty Shades of Incompetence"
"The Call of Greed"
"To Kill a Return"
"The Shafting"
"Hippie Potsmoker to IT Manager in 21 Days"
"A Brief History of Slime"
"The Little Company That Couldn't"
"The Stevenson Street Swindle"
"How to Host a Financial Site on a Console Gaming System"

That should be enough to get you started.  Please try to have the book finished before the IPO. 

[rawraw]

I'd be curious what LC has to say about the stuff above.  Esp if some of their data is screwed up

IF? Every time I set out to study some phenomenon in LC data and get looking closely at it, it is SO SCREWED UP I can't face trying to make the corrections necessary to carry on with the work. I have corresponded privately with a number of members on the subject. I can tell you the historical payment data is a disaster.

Have you spoke to LC?

[hoggy1]

I have sent them various examples in the past and exchanged emails with a "customer service" rep. I got my first real phone call from them   just last week, so at least they are showing a bit of interest. It gives me a bit of hope going forward but I still haven't spoken to a technical person yet.

[rawraw]

I have sent them various examples in the past and exchanged emails with a "customer service" rep. I got my first real phone call from them   just last week, so at least they are showing a bit of interest. It gives me a bit of hope going forward but I still haven't spoken to a technical person yet.

I think there is a hierarchy in customer reps, to weed out noise.  You may want to see if Fred93 has a contact for you, since he spoke with several technical people.

[muflafler]

As a possibly irrelevant analogy, before the financial crisis, mortgage securitizations were generally structured so that the senior securities were protected mainly against prepayment--the biggest risk was seen as being paid back too soon, and the holders of the most senior securities in the securitization would accept a lower return in order to be protected against the risk of prepayment (the more junior securities would be paid back before the more senior securties).

But since the financial crisis, mortgage securitizations are no longer structured that way.  The senior securities are now protected against default--once again, the primary risk is viewed as not getting paid back at all; getting paid back too soon is not viewed as quite so bad anymore after seeing the types of losses that can be incurred when defaults spike.

So you might not like to see a high prepayment rate, but it could be worse.  I personally believe that the loan performance data that everyone here is working off of is skewed by limited sample size, i.e., the vast majority of performance data comes from a time period where the economy was generally improving and does not include a recessionary time period.  Loan performance during a recession is sure to be significantly different (worse), and attitudes about risk among P2P investors may change in similar ways as it did for mortgage securitization investors.

[core]

I don't follow, muflafler.  Yes, a default is worse than a prepayment.  And losing my house due to bankruptcy would be worse than having to deal with a leaky roof.  I don't see what the two have to do with each other.

It's not like you get to choose ahead of time.  If recession hits and defaults follow, prepayments have nothing whatsoever to do with anything at that point.  Unless you're saying 0% is somehow OK because it's better than -100%.  Well I can stay home and make 0%, with a lot less risk.

So find a homeless guy with no credit, get a LC loan in his name, build his credit, and then do a re-fi. Really core, sometimes I think you just don't think these things through.

You just try getting a bum to remain conscious throughout a 15 minute Lending Club phone interview, after he's consumed most of a bottle of Thunderbird or Night Train!  LC already knows what I sound like.  We are getting to be real good pals after the past few years.