@RobertEG wrote:
IMHO, the issue it NOT whether any FICO score is better or worse than any non-FICO score.
Fair Isaac, of course argues, that FICO scores are more predictive, and the CRAs argue that that their non-FICO algorithms are more predictive.
Who knows? Who cares?
What matters is what the lendor/creditor will use, and the vast majority of credtors base their decsions on FICO scores.
That may be due primarily, as the non-FICO score vendors argue, only because FICO has been around longer, and thus have captured the mjority of the scoring business, but as consumers,. it is all academic. It all boils down to what the lendor is using. Most use FICO, so who cares about a score that your lendor is not using?
Exactly so. A well-designed scoring algorithm will do one purpose very well and other purposes not as well. A poorly-designed scoring model might not even accomplish its primary purpose very well, but for the sake of this discussion let's consider an algorithm that does its purpose very well for a little thought experiment.
Assume the FOOKO company decides to invent their own credit scoring model. They obtain large volumes of data, comparable to the data behind the FICO score, and pay lots of statisticians lots of money over an extended period of time analyzing the data to develop a completely independent scoring model that is nearly as accurate for the purpose of predicting whether a person will get into credit difficulty as are the FICO models. In fact it turns out where a "perfect" model would be 90% accurate in predicting default (no model based on the past history can be 100% accurate because sometimes stuff happens to people that just can't be predicted so in the absence of a time machine no predictor can be perfect), some version of a FICO model is 85% accurate and the FOOKO model is 80% accurate in predicting default. But since it is an independent model, for predicting this version of a FICO score the FOOKO model turns out to be, say, 75% accurate. This can easily happen if the models are independent: two models that have nearly the same accuracy for predicting default can be much less accurate for using the output of one model to predict the output from the other model.
Now here is the key point: if you are a lender then you do not need to care very much about how well some other scoring algorithm predicts a customer's FICO score, you care mostly about how well it uses a customer's credit past to predict that customer's credit future. Indeed, if the FOOKO algorithm really does almost as well as FICO at predicting credit issues but does not correlate especially well with FICO scores that might be to your advantage as a clever lender because that gives you a major commercial opportunity to establish a profitable niche: identify people whose FOOKO scores are much higher than their FICO scores and offer them credit on terms more favorable than other lenders offer them. If the FOOKO score really is (1) almost as good as the FICO score for predicting default but (2) not very closely correlated with the FICO score, then in that difference lies a profitable market niche for a clever lender. However, as recent history has demonstrated many lenders are not terribly clever: instead of doing genuine statistical arbitrage (which is the technical term for using a different model of comparable accuracy to identify people who do better than the standard model would predict), lenders merely lowered their standards! The smart way to be a subprime lender would have been to do high-quality research and identify underserved populations who actually are better credit risks than conventional lenders realize. A few small nonprofits who actually did just that are doing OK even in the current recession. But in the recent bubble lots of companies made buckets of money by simply lowering standards without giving a whole lot of thought to the risks they were taking. Now in the aftermath, just about every lender is terrified to do anything that might be construed as reckless lending, so I doubt any big lender will dare move away from FICO scores.
Of course I have no idea whether any of the competitors to FICO scores are in fact like my hypothetical FOOKO scores. Maybe some other model really is almost as good as FICO for a lender's purposes. Maybe some other model is even better than FICO for a lender's purposes. I have no way to know. What we do know as consumers is that whatever their possible merits for predicting what a lender wants to know, none of the various FAKO scores is very good for predicting our FICO scores. In particular, none of the FAKO scores does better than this free estimator does at predicting my FICO score, and therefore none of them is worth my money.
And as consumers, we don't really care how well some scoring model performs as a predictor of what the lender should know. The market reality at this time is that the vast majority of lenders and others who use credit scoring models to make decisions affecting us are using some version of a FICO scoring model. So for our purposes as consumers, unless and until some competitor to FICO captures a significant share of the market among those making decisions that affect us, we need to know our FICO scores and none of the competitors have much value for us.
Message Edited by MattH on 09-25-2009 07:42 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
Starting Score: TU: 821/EQ: 800