Is Default Risk Better than Credit Scores?

04/17/08 posted by Prosper Blog

The following is a guest post by a Prosper member: CreditKarma… You can use the Prosper Performance tool to estimate default risk for Prosper loans.

When I was a marketing and risk analyst with a credit card issuer, the customer’s credit score never really mattered per se. My concern was always with the default risk for any given score. For example, it was more important to know that a 720 FICO™ Score customer had a 1 in 237 probability of default. The score was meaningless to me (the lender) without the corresponding risk probability.

As such, I always wondered why Fair Isaac and credit bureaus used arbitrary scores when default risk is so much more informative to all parties. If I had to speculate, I would guess the following reasons:

Intellectual Property Protection - It is probably much easier to patent and copyright some arbitrary score and score algorithm than it is to patent logistic regression (the math term used in developing scores) and an arbitrary score range.
Consumers Don’t Want to Be Labeled - Consumers may object to such an outright classification as default risk. To say one person is a 450 FICO™ is probably not as offensive as saying that person has a 1 in 3 chance of defaulting on a loan.

Today, the reason is moot as it has become ingrained in consumers. As I read the comments in various blogs and discussion forums, I am still amazed at the level of mis-information accepted as truth. People claiming there certain scores are FAKO’s* or that doing X will yield Y result in your score. I personally think the confusion stems from the number of data sources, models, default variables, and uses of scores. It took me a degree in mathematics, economics, and 10 years of industry experience just to get my arms around credit scores and credit uses.

All this brings me back to the question: why doesn’t the industry just share your default risk? Isn’t that just an easier thing to understand than some arbitrary score range. Some people may object to a probability label but as they say, “a rose by any other name…”

Editor’s Note: *FAKO or Fake-O has been around for years, but the term was popularized by Suze Orman fairly recently. Orman is, admittedly, on Fair Isaac’s payroll.

Posted in Misc, Personal Finance Education, Prosper Lenders

6 Responses

bignosebob | April 17th, 2008 at 11:59 am

Do you know of any web site or other resources that lists an estimated probability of default corresponding to credit scores?

Smarter than you CK, once again | April 17th, 2008 at 8:22 pm

Oh boy, another know it all.

1. The FICO scores (or any other scores used by lenders for that matter) aren’t “arbitrary.” Do you really think that the folks in Marin are sitting there throwing darts at a board filled with characteristics, variables and weights when they build and rebuild their models? If you honestly believe that then stop reading b/c you won’t understand #2.

2. Default risk or “odds to score” as it’s better known is not something that’s the same across the board and it changes over time (something that mortgage lenders learned the hard way). All lenders experience different odds to scores relationships. And, it varies by lender, geography, marketing message, loan products, time, etc, etc.

What your superiors did when you worked at the bank was to perform (or pay someone to perform) retro validations to learn what the default risk was by score range. Then they told you that “at FICO 650-659 we’ll see 500:1 good/bad odds.” And then you made your decision based on that knowledge.

Would you care to make anything else up?

Eliot | April 18th, 2008 at 9:15 am

I somewhat agree with the writer. Why 350-850 why not 0-100? The current range is uncommon. I don’t think anyone is suggesting it is meaningless but perhaps it could be more useful.

Liquid | April 18th, 2008 at 9:43 am

@ “Smarter” - I can tell this is a subject you are quite passionate about, yet I fail to see how your points contradict anything that was in the original post.

Regarding #1, the actual score itself (i.e. 750) is what is being referred to as arbitrary by CK, not the algorithm used to determine the default odds.

Regarding #2, your point is correct but does not contradict the original post. That is why many lenders build custom risk scores and use them instead of FICO or in conjunction with FICO. This information, while interesting, does not at all contradict anything in the OP. As an aside, it is usually a junior analyst who does the default odds “analysis” (it is really just reporting), not a senior analyst or superior as you point out.

It appears what we have here is that you misunderstood what “arbitrary” was referring to which then caused you to furiously pound out a post where you say the OP is less intelligent than you and making things up, when in fact you fail to refute anything in the original post and make yourself look like “another know it all”. How ironic.

Peer Lend | April 19th, 2008 at 8:11 pm

Yes - The reason that the score ranges might -seem- arbitrary (though they are calculated precisely, not by an algorithm, but by the shrinks in marketing) has to do with consumer psychology.

This is related to the reason that highschool kids keep getting more and more points on the SAT just for signing their names, and kindergarten students get a super-dooper-check-mark just for -trying- to make a macaroni necklace, glue-eater or otherwise.

It’s also probably (and forgive me for speculating), somehow related to how Prosper assigns AA-A-B-C-D-E and… HR credit grades. Not to pick on Prosper, everyone else in just about any industry does a similar thing - I mean, who wants to give a customer an “F”? - but I did want to point out that consumers aren’t the only ones who this kind of thinking is “ingrained” in - consciously or otherwise.

It’s -also- probably somehow related to how there are no “13th floors” in most buildings in countries where the general population tends to be superstitious, or how, as George Carlin puts it, no one can any longer be a “loser” - they are, instead, said cheerily, “the last winner”.

Chrisfs | April 23rd, 2008 at 9:57 am

I think a probability would be too fatalistic for many people (I know it would come across that way to me). Saying “Well Mr Jones, you have a 1 in 4 chance of defaulting on this”, seems to take Mr Jones’s actions out of the picture. Implying that sort of ‘fate’ based thinking, is not something you want to do when you are making loans.

I know it is meant to mean “1 out of 4 people with similar credit features will default” but I don’t think it will come out that way. People who have HR scores are unhappy enough as it is. To give them a probability (a relatively high one at that) is in no one’s interest