Relative Predictive Power of Factors per Scorecard

[Anonymous]

If the mods believe this should be on the other thread then move away. I just thought it would be nice to have a thread to discuss if our advice has been spot on or this presentation highlights some areas which we can refine our recommendations.

The document that pizza found was excellent.

http://www.jchs.harvard.edu/publications/finance/babc/babc_04-12.pdf

 It has some implications for advice we give on this forum. Some of this we probably already knew, some of it we may not have thought of. Though this is in the NextGen section it is implied for "classic" scores, though not to the same degree.

For those without delinquincies Utility is king, followed by age (if your history is clean), then mix and new credit.

For those with delinquincies - Payment history is king. So advice to dilute bad accounts with good tradelines is excellent advice. Age and new credit categories are much less emphasized and thus will likely be outweighed by better payment history on more tradelines. Mix matters little along with inquiries and new accounts (as negative factors).

Figures 8 b and 8c likely apply to "classic" FICO scores, though they are not as refined as NextGen.

Still in general we note that for young credit - utility, history, then new credit are the important factors, with age and mix less powerful.

The longer you hold credit the more important history becomes (explaining why a 30 day late dings an 800 score so much), and utility less so (though perhaps it applies its pain at a much lower util level), and age, mix, and new credit are much more important factors. This may explain why someone with 12 years of credit history may only get a 25 point drop from going from 90% util down to 50.

[Anonymous]

I wouldnt necessarily say that a 30 day late effects an 800 score more than a someone with, say, a 600 score. The calculations still would have to factor in length of credit history and all the other "negative" things on that report.


This might be old news to the season guys, but might be beneficial to someone new on the boards:


Lets say that you have two individuals with 700 scores. They both just receive their very first and only 30 day late. One has 10 years of history, while the other has 5. When the scoring model takes into account risk the person with the lower history could take a much bigger hit because the odds of the person with 10 years of history to get a 30 or 90 day late in the next two years is going to be less.