@Anonymous wrote:It doesn't have to be late on several accounts at the same time thus creating a good "reason" for it. Say one different account was late every year for 5 years from 2011-2015. I don't think its arguable that this person is far more risky than someone else with the same number of accounts, all clean except one from say 2013. Under FICO scoring models however, at least FICO 08, all other things being equal these two individuals may possess scores that are quite similar as both could be in the same dirty bucket. If you were to survey 100 random people whether they are off the street or loan officers and ask them which of these two people they think would be most likely to be late again, I think you'd get 100 identical answers. Yet, scoring may not reflect this. Seems bogus to me.
Well, what's the intrrinsic measures? I'd suggest they're time since late and severity of the late.
Let's say in your example someone was 30 days late from 10/10 to 10/15, however absurd that is on a single tradeline with no other issue on the file.
Compared to me with a singleton late on 10/15, I don't know if I follow that's way less risky honestly.. I'll reiterate it's not what you or I think as theoretically rational human beings, it's what the data says. We just don't have access to the data... without a lot more information of people with lates watching their scores (which may happen with something like the EX FICO service) and really caring about this, well, that'll probably never be the case, I don't see us making much headway on this.
Assuming I'm around long term (call it the next 5 years) I can track a singleton 30 day late as it moves through my report on TU, but fundamentally it doesn't change anything: lates bad, older lates better, no lates good.
@Anonymous wrote:Then make it a more drastic, more common example. 5 tradelines over 5 years that are littered with lates... 90's, 60's, whatever compared to a person with just a single 90 day late on 1 tradeline. Because of diminishing (really "no", not "diminishing) returns it probably doesn't matter much if there's 5 late payments or 25 total across the 5 accounts as both individuals will reside in the same dirty bucket. Would you really argue that the person with poor payment history (multiple lates across multiple accounts for half a decade) is equally risky as someone with a lone 90 day on 1 account?
In my opinion no, but how can you state that neither FICO nor UW takes that into account? FICO 8 marketing (what they use to sell the scores to lenders) suggested that patterns were now a thing that would be penalized more heavily. This sounds like a pattern certainly by my definition on it.
FICO development in general, the customers are the lenders... if there's something that the customers are doing in UW, FICO's going to try implement that in the algorithm. Discounting paid collections is another example of that, if the lender data suggests there's a difference now, FICO went and looked too and changed it. Or something similar on that, lenders (and competition from Vantage) does drive the algorithm's development.
"Would you really argue that the person with poor payment history (multiple lates across multiple accounts for half a decade) is equally risky as someone with a lone 90 day on 1 account?"
In my opinion no = no I would not.