cancel
Showing results for 
Search instead for 
Did you mean: 

Very interesting footnotes about credit models from a data mining textbook

tag
MattH
Senior Contributor

Very interesting footnotes about credit models from a data mining textbook

 

My work, research and development for a pharmaceutical company, consists mostly of data mining.  I began studying credit scoring and hanging around this board primarily out of intellectual curiousity about how similar data mining tools are used in a very different domain.

 

Well, a textbook I'm currently reading for my work, Handbook of Statistical Analysis and Data Mining Applications, makes some very interesting points about credit modeling.  Anybody who is familiar with how academic authors think knows you should read every footnote, because some of the best stuff is buried in those footnotes (and I'm the sort of detail-oriented person who always reads the fine print).  In this book I found two nifty footnotes about credit models:

 

1. After briefly summarizing the current financial disaster as "second only to the Great Depression with accompanying multiples of previously normal volatility," the authors remark, "If a model can't weather such a change, it should at least signal 'don't know'" when it sees input data completely unlike what it was designed to handle.

 

2. "credit profits are nonlinear with risk, and remind us of the triage system established during the Napoleonic Wars...credit companies make the most profit on individuals in the middle category of 'woundedness' -- those who can't pay their balance but keep on trying.  But they lose 5-10 times as much on clients just a little worse off."  Their discussion of triage even includes an aside about a Monty Python character insisting "it's just a flesh wound!" after losing a two arms and a leg in battle.  This footnote ends by observing "recent downturns in the economy have severely punished the stocks of companies that aggressively sought that customer niche -- especially if they did not give obsessive attention to model quality."

 


Message Edited by MattH on 02-02-2010 08:30 AM
TU 791 02/11/2013, EQ 800 1/29/2011 , EX Plus FAKO 812, EX Vantage Score 955 3/19/2010 wife's EQ 9/23/2009 803
EX always was my highest when we could pull all three
Always remember: big print giveth, small print taketh away
If you dunno what tanstaafl means you must Google it
Message 1 of 5
4 REPLIES 4
newstart2010
Blogger

Re: Very interesting footnotes about credit models from a data mining textbook

lol @ Monty Python's "Its just a flesh wound"

 

I remember seeing a documentary on collection agencies, credit card companies, and the state of today's average credit worthiness.  It was amazing to see from the credit companies' point of view how they did not want the A+ credit ratings, but they wanted the "almost bad enough" category because that was where the meat and potatoes of their profit came from.  I sort of always realized that, but it never totally hit me till I saw that documentary.  Hearing about some CEO quietly sitting in on a presentation of some new software that would help the customers stay on top of their finances, and help the company weed out people with less than perfect scores and the CEO asking why he would want to do that.  It was like someone slapped me with the obvious glove.

 

Do I think they are wrong in doing it?  nope.  I am a business woman.  I know where my bread is buttered, and how you always have to look for the biggest bottom line.  BUT!!!  If these companies and banks are willing to take the risk to achieve a larger profit margin they need to take the risk completely and face the downfall when the economy goes in the toilet and those clients who were "just bad enough but will still pay" become "so bad off nothing will get paid".  Just like those who over-leveraged to take advantage of an upturned economy and are now upset that they are in foreclosure and bankruptcy, the banks and credit companies need to feel the sting of this overachieving as well, or nothing is learned.

I live my life like I type, fast and with a lot of mistakes.
Spacebar broken. Watch for finger.

02/04/2015 || TU 08: 728 EX 08: 709 EQ 08: 748

Message 2 of 5
MattH
Senior Contributor

Re: Very interesting footnotes about credit models from a data mining textbook

 


@newstart2010 wrote:

@Anonymous @ Monty Python's "Its just a flesh wound"

 

I remember seeing a documentary on collection agencies, credit card companies, and the state of today's average credit worthiness.  It was amazing to see from the credit companies' point of view how they did not want the A+ credit ratings, but they wanted the "almost bad enough" category because that was where the meat and potatoes of their profit came from.  I sort of always realized that, but it never totally hit me till I saw that documentary.  Hearing about some CEO quietly sitting in on a presentation of some new software that would help the customers stay on top of their finances, and help the company weed out people with less than perfect scores and the CEO asking why he would want to do that.  It was like someone slapped me with the obvious glove.

 

Do I think they are wrong in doing it?  nope.  I am a business woman.  I know where my bread is buttered, and how you always have to look for the biggest bottom line.  BUT!!!  If these companies and banks are willing to take the risk to achieve a larger profit margin they need to take the risk completely and face the downfall when the economy goes in the toilet and those clients who were "just bad enough but will still pay" become "so bad off nothing will get paid".  Just like those who over-leveraged to take advantage of an upturned economy and are now upset that they are in foreclosure and bankruptcy, the banks and credit companies need to feel the sting of this overachieving as well, or nothing is learned.


 

Exactly.  The only interest my wife and I pay is our mortgage, and the only card that charges us annual fees is my AMEX Gold, which I keep mostly because it's my oldest tradeline.  If anybody else began charging fees, we'd take our business elsewhere and from our credit histories they know we could.  All they're gonna get from us is transaction fees.  And if merchants start passing those fees along to the customer then we'll shift more of our purchases to debit cards: we've always got more than enough in the bank to pay cash for everything.  The main reason we use credit cards is because consumer protection laws for credit card transactions are much stronger than for debit cards.  Also, if some crook hits one of our accounts, with a credit card it's only the bank's money at risk.  Some years ago my wife's account got hit with check fraud, and although we did get all the money back it took a while to sort things out; fortunately there was enough in my account to cover all our purchases until hers got straightened out.

 

TU 791 02/11/2013, EQ 800 1/29/2011 , EX Plus FAKO 812, EX Vantage Score 955 3/19/2010 wife's EQ 9/23/2009 803
EX always was my highest when we could pull all three
Always remember: big print giveth, small print taketh away
If you dunno what tanstaafl means you must Google it
Message 3 of 5
Anonymous
Not applicable

Re: Very interesting footnotes about credit models from a data mining textbook

Yeah, I used to be one of those helping to drive up the profits in the financial industry.  Late fees, interest payments, etc, etc.  I can't believe how much money I've wasted that way over the years.  Might as well have piled it up and burned it.

 

 

Message 4 of 5
RobertEG
Legendary Contributor

Re: Very interesting footnotes about credit models from a data mining textbook

You are both right on point!

Late payments allow CCs to impose late fees, jack up interest rates, lower credit limits, etc.

Think that might be THE reason they offer all of those low, or zero, interest promotionall offers and "convenience checks"?

As setup for future profit?  DUH!

Your analysis of the misconception of credit scores is intriguing.

Credit scoring is a risk analysis of the chance of default on debt, primarily within the next two years.

So they can establish your initial approval for credit based on wanting only the highest FICO score, and then set your interest rates, credit limit, etc, based on those scores.

I agree.  Credit scoring may be linear to risk, but profits are not linear to risk.

Would a CCC not fully welcome an occasional late payment, thus allowing them to increase your interest rate, and maybe also reduce your CL and abbrogate all "promition limts" on the account?

An applicant with a 780 FICO score may be a low risk under FICO scoring, but not a potential $$ source for future account mishaps.

Profit.

Good post!

 

 

Message 5 of 5
Advertiser Disclosure: The offers that appear on this site are from third party advertisers from whom FICO receives compensation.