How to evaluate credit scores
I have been looking at how CCC’s assess their risk in issuing new cards. I am a patent attorney, but my FICO friends, please do not hold that against me! LOL!
Well, being a FICO-holic, I did a patent search on credit card company evaluation and scoring of applicants for new credit cards, and found an extremely interesting patent that contains surprising input on how credit card companies evaluate applicants. While most scoring models used to create both FICO scores themselves and the related models used by banks to assess credit worthiness for issuance of new cards are kept as closely-guarded trade secrets, First USA Bank applied for a patent in 1998, which issued in 2001 (U.S. Patent No. 6,202,053, which can be viewed to the Patent Office website. This put their scoring methodology in the public domain.
I wont bore you with details of their process, but it is basically this. The bottom line for them was to estimate, when issuing a new card, the statistical probability that the account would become “bad” within the next two years. “Bad” simply means and account that will go 60+ days delinquent in the next two years.
So, if you are still with me, this is what they did. They statistically broke down the general population into 7 categories, listed below, that are recognized credit risk categories…. They then looked at actual credit reports for those in each category, and developed statistics on the percent of total population that would fall into each category, and then developed stats for each category, disclosed in the patent, of those in each category that later had a delinquency of 60 days or more. They used that to project the probability, when issuing a card, of its risk to them.
Here are the results, and my reading of them.
First, definitions of the categories they used:
A “clean” credit file was one that had no record of delinquencies that exceeded 30 days.
A “dirty” credit file is one that had at least on prior delinquency of 60 days or more.
A “thin” credit history is one that does not show at least 3 bankcards.
A “thick” credit history is one that shows 3 or more bank cards.
Percent util in each category is a term know to all of us FICO-holics as simply the current revolving debt to credit limit ratio. They set cutoffs, depending upon the categories listed below, at 25%, 40%, and 70% depending upon the category.
Well, with all that verbosity, here is the bottom line from their patent.
The “bad” rates for each individual category are listed below, which represents the potential that if they issue a card to one person within that category, the chances that they will go “bad” within two years.
But the more interesting thing that can be gleaned from the patent, which is not specifically discussed, since they
don’t disclose in the patent their final decision making process, is that if you multiply the bad rate for each category by the percent of population within each category, you arrive at a simple number which shows the total potential liability to the CCC if they issue a card to any person, as weighed against all cards issued. That, weighed against all cards issued, is their methodology for establishing overall potential liability of an account going over 60 days. The results, tabulated below, are kinda surprising. Those with dirty files that have more than 10 years of credit history or have current &Util of less than 40% are better overall risks to the CCC than those with clean files but with &Util above 40% and extending up to 70%. In fact, the most surprising thing from their data is that if they issue a new card to one who has a thick, clean file with medium percent utlilzation between 25% and 70%, their overall risk might be the highest, even though the category risk is the second lowest, because of the much higher number of customers in that category. You can be dirty, in their minds, if you have a long credit history, and more importantly, if you keep your %Util down. These are their summarized risks… their data, not mine):
Category %Pop %BadRate (%Pop)x(%Bad)
Thick, clean, %Util<25% 29.9% 1.4% 0.4%
Thick, dirty, %Util>40%,
history >10yrs 4.6% 9.5% 0.4%
Thin (3 or less cards) 6.5% 8.7% 0.6%
Thick, dirty, %Util<40%,
history <10 yrs 13.4% 5.1% 0.7%
Thick, dirty, %Util >40%,
Thick, dirty guys and gals are the second lowest risk to them if they have a credit history of 10 years. Hmmmmmm. And, if you are thick and clean, getting a percent utilization of under 25% can take you, at least in First USA’s mind, from the a very high risk to a dersirable customer on their potential risk list. In each category, dirty or clean, the driving factor is long prior credit history and low %Util. They use 25% if you are clean, but they then cut your risk slack to 40% Util, if you be dirty, based on the higher number of applicants in that dirty pool category.even though it puts you in a higher risk group. Sly devils. And we all know that the bottom line for them is to issue you a card, reel you in, and then, if you have even a minor 30 day delinquency (not yet even dirty), they will jack your rate up and soak you! They excuse 30 day lates from their risk factor algorithm when issuing you a card, but use is as the hammer to increase you interest rate after you get it. The fact that they don’t use 30 day lates in their algorithm shows me that they like people going 30 late so that they can change their prior interest rate, and soak you.Source: ficoforums.myfico.com