9y AAoA threshold on EX8 but not EX2

I don’t know how novel this info is, but I woke up to a very clean aging DP today. My EX8 dropped 3 points and EX2 didn’t budge, when the only thing that changed was an old, closed loan dropped off my report. According to the EX CRM, my AAoA went from 9y to 8y10m.

I read this portion of the Primer a few days ago, but there’s a lot of detail there. Is this the same threshold that KiB confidently reported?

I’ll take a closer look later to see if there’s more to the story.

Good find!

Just a couple of other things I'd look at are before/after loan:revolver ratio as well as AAoIA, as presumably that would have dropped as well.

BrutalBodyShots wrote: ↑Wed Jan 19, 2022 3:37 pm Good find! Just a couple of other things I'd look at are before/after loan:revolver ratio as well as AAoIA, as presumably that would have dropped as well.

Good points. I also wanted to check that AWB% isn’t at issue. I think I’m in the clear there. I went from 6/26 (23%) to 6/25 (24%). I don’t recall anyone suggesting there’s a threshold at 24%. I must admit that loan:revolver ratio is not something I pay attention to. I went from 12:14 to 11:14. Hard for me to imagine that’s at a threshold. I don’t know what my official AAoIA is (or, more importantly, AAoA). I can never get my manually calculated aging numbers to match up with what the CMS systems say. I know the standard advice is to calculate our own and ignore the CMSes, but I’m not sure that’s wise. If they all agree with each other, it seems somewhat likely they also agree with FICO. I trust my math, but I think I must be using inputs that differ from theirs. Is it possible that a card opened around, say March 31 - April 1 could show up to me as April, but be viewed by FICO (and EX.com and MF) as March? Ironically, one of my theories for what was causing the inconsistent aging calcs just got shot down. The account that vanished today was an old student loan. It had been transferred to a different lender years ago, and both versions of it appeared on my EX report. I had speculated that the algorithms might have a way of disregarding accounts that were duplicated by transfer. Nope! The disappearance of the duplicate caused a score change, so it was being (double) counted.

Justaguy wrote: ↑Wed Jan 19, 2022 11:01 pm

BrutalBodyShots wrote: ↑Wed Jan 19, 2022 3:37 pm Good find! Just a couple of other things I'd look at are before/after loan:revolver ratio as well as AAoIA, as presumably that would have dropped as well.

Good points. I also wanted to check that AWB% isn’t at issue. I think I’m in the clear there. I went from 6/26 (23%) to 6/25 (24%). I don’t recall anyone suggesting there’s a threshold at 24%. I must admit that loan:revolver ratio is not something I pay attention to. I went from 12:14 to 11:14. Hard for me to imagine that’s at a threshold. I don’t know what my official AAoIA is (or, more importantly, AAoA). I can never get my manually calculated aging numbers to match up with what the CMS systems say. I know the standard advice is to calculate our own and ignore the CMSes, but I’m not sure that’s wise. If they all agree with each other, it seems somewhat likely they also agree with FICO. I trust my math, but I think I must be using inputs that differ from theirs. Is it possible that a card opened around, say March 31 - April 1 could show up to me as April, but be viewed by FICO (and EX.com and MF) as March? Ironically, one of my theories for what was causing the inconsistent aging calcs just got shot down. The account that vanished today was an old student loan. It had been transferred to a different lender years ago, and both versions of it appeared on my EX report. I had speculated that the algorithms might have a way of disregarding accounts that were duplicated by transfer. Nope! The disappearance of the duplicate caused a score change, so it was being (double) counted.

Definitely check AAoIA (average age of installment accounts) as well. AAoA is supposed to max out at 90 months. So this could be new. I thought KIB had 7 years? I agree it is not AWB or loan ratio. How many total accounts does that bring you to?

Alright, I have triple-checked the open dates for all accounts on my EX report. I still get different AAoA when I manually calculate it than EX shows on their site, and of course they don't display AAoIA anywhere.

Here's what I get before->after removal of the old loan that vanished this month:
AAoA: 8y11m->8y9m
AAoIA: 8y11m->8y6m

According to EX:
AAoA: 9y0m->8y10m
AAoIA: ???

I think I might have just figured out the source of the discrepancy. When calculating the average, should I use standard rounding or round down? For example, when I have 26 accounts with a cumulative age of 2,805 months, the average age without any rounding (2805/26) is 107.8846... Should I treat that like 107 or 108? I have been rounding down, on the logic that ages are always calculated in that way, but perhaps I have taken that rule too literally.

That's a great question. I too have often wondered how to round when considering monthly aging metrics.

I have modified the formula in my spreadsheet to use standard rounding, instead of rounding down. My aging numbers now match perfectly with those shown on EX's site. There have been four instances in recent months when my AAoA crossed over or under 9y. In every instance, I gained or lost exactly 3 points on EX8.

AAoIA does not explain these changes. Its values at those same transition points were as follows:
8y7m->8y8m
8y10m->8y11m
8y11m->9y
9y->8y6m

I think this is pretty strong evidence that AAoA has a threshold at 9y on EX8.

Oh, and EQ appears to have an even higher number as the top-end threshold for AAoA. I'm getting "Short account history" as a negative reason statement on EQ8. The explanatory text says "People with longer credit histories who infrequently open new accounts generally pose less risk to lenders. In your case, the age of your oldest account and/or the average age of your accounts is relatively low." My oldest account is over 27y, so I can't imagine that's too young. EQ has my AAoA at 9y7m.

Justaguy wrote: ↑Sun Jan 23, 2022 11:37 pm Oh, and EQ appears to have an even higher number as the top-end threshold for AAoA. I'm getting "Short account history" as a negative reason statement on EQ8. The explanatory text says "People with longer credit histories who infrequently open new accounts generally pose less risk to lenders. In your case, the age of your oldest account and/or the average age of your accounts is relatively low." My oldest account is over 27y, so I can't imagine that's too young. EQ has my AAoA at 9y7m.

I would tend to agree that AAoA is the factor triggering the reason statement you referenced, not AoOA in your case. It would be cool to continue to monitor that and see when it drops off, at that time citing your age of accounts factors.

Justaguy wrote: ↑Sun Jan 23, 2022 11:37 pm Oh, and EQ appears to have an even higher number as the top-end threshold for AAoA. I'm getting "Short account history" as a negative reason statement on EQ8. The explanatory text says "People with longer credit histories who infrequently open new accounts generally pose less risk to lenders. In your case, the age of your oldest account and/or the average age of your accounts is relatively low." My oldest account is over 27y, so I can't imagine that's too young. EQ has my AAoA at 9y7m.

@Justaguy yes standard rounding. That’s where the false utilization nines came from as well, the rounding down incorrectly. when we learned there is standard grounding that’s when we learned 9.5%, 29.5%, 49.5, etc. Edit: aging may round up. Very very interesting. We previously thought AAoA maxed out at 90 months. If you crossed it four times and got three points on EX at 108 months each time, yes I would say that’s pretty strong evidence! AAoRA & AoORA max out at/< 108 & 240 months (9 & 20 years). BBS gave us those data points when he lost the AAoRA/AoORA code. I don’t know the exact thresholds, but it has to be under those or at those.

Birdman wrote: ↑Mon Jan 24, 2022 5:10 am @Justaguy yes standard rounding. That’s where the false utilization nines came from as well, the rounding down incorrectly. when we learned there is standard grounding that’s when we learned 9.5%, 29.5%, 49.5, etc.

I knew the AMA taught us that standard rounding is used for Util calculations. I didn’t know the same was true for aging. I had reasoned that, as with people’s birthdays, being even 99% of the way from one age to the next isn’t enough for us to round up. Presumably that is true with respect to the age of any particular tradeline, but not when calculating averages. I’m so relieved to finally be able to calculate my aging figures with confidence!

Birdman wrote: ↑Mon Jan 24, 2022 5:10 am AAoRA & AoORA max out at/< 108 & 240 months (9 & 20 years). BBS gave us those data points when he lost the AAoRA/AoORA code. I don’t know the exact thresholds, but it has to be under those or at those.

Hold on. We might be talking past each other. First, I’m seeing indications that the aging thresholds on the upper end are different between bureaus. On EX8, I now believe there’s an AAoA threshold at 9y. On EQ8, I'm getting the “short account history” code I mentioned above. The explanatory text mentions AoOA and AAoA, so I believe there must be an AAoA threshold higher than 9y7m there. On TU8, I’m getting “short revolving history.” My AoORA there is 27y4m. AAoRA there is 8y. Finally, I think we might need to be careful drawing conclusions about things — including but not limited to top-end thresholds for aging — based on the disappearance of reason statements. If, as you argue in the other thread about 70% Util, there are point caps by FICO category (e.g. Length of Credit History), then presumably once a person has all the available points in that category, the algorithms won’t display any negative reasons related to factors in that category. That doesn’t mean they have reached all top-end thresholds, it means there are no more point available for some reason. That reason might be that they are already at the category cap.

I think another angle to look at when considering this discussion is top end buffer.

We all know that top end buffer exists on 8 and we reference it once the needle is pinned at 850. If the buffer exists with the total score, I don't think it would then be out of the question to think that it could also exist in each criteria category.

Here we're of course discussing Length of Credit History. If we say that this category is worth X points, considering the buffer angle couldn't it be possible that a negative reason statement related to that category could still be displayed even if X points are already achieved?

Perhaps on the upper end one of these aging thresholds is only "worth" 2-3 points on a certain scorecard. If one already has a score of X + 5 (buffer) but has that negative reason statement present, their score would still be X+2/3 points and for scoring purposes they'd still have that category maxed out.

I'm not sure if I'm explaining that the best possible way, so apologies in advance if I'm not being clear. I think this would be an interesting topic to discuss though... sort of the breaking down of the top end buffer that we all know is there.

BrutalBodyShots wrote: ↑Mon Jan 24, 2022 8:33 pm I think another angle to look at when considering this discussion is top end buffer. We all know that top end buffer exists on 8 and we reference it once the needle is pinned at 850. If the buffer exists with the total score, I don't think it would then be out of the question to think that it could also exist in each criteria category. Here we're of course discussing Length of Credit History. If we say that this category is worth X points, considering the buffer angle couldn't it be possible that a negative reason statement related to that category could still be displayed even if X points are already achieved? Perhaps on the upper end one of these aging thresholds is only "worth" 2-3 points on a certain scorecard. If one already has a score of X + 5 (buffer) but has that negative reason statement present, their score would still be X+2/3 points and for scoring purposes they'd still have that category maxed out. I'm not sure if I'm explaining that the best possible way, so apologies in advance if I'm not being clear. I think this would be an interesting topic to discuss though... sort of the breaking down of the top end buffer that we all know is there.

I agree, it's a fascinating topic! I've got to tip my cap to @Birdman for being the first person I've seen mention it, over on the other site. My answer to the bolded question above is: absolutely it's possible! It seems to me we are just barely starting to explore this aspect of FICO scoring, so there are probably many nuances that we don't know about. But my hunch is that reaching a category cap would preclude the algorithm from throwing negative reasons there. If there are no points to be gained in that category, then by definition no factors in that category are costing any points. If I'm following your hypothetical, the person would get exactly X points before and after the change in question, right? I agree that buffers can be a useful concept for explaining why maxed-out scores don't change. But that doesn't mean, or even imply, that the software literally uses buffers. Personally, I think of them as "disregarded excess points." I find it hard to imagine the software designers would intentionally recommend accruing more of those. But we're layers deep in speculation here, and I could easily be wrong in many different ways. I should give a little more context to my comments. It has always puzzled me that some scores that aren't particularly close to 850 show fewer than four negative reasons. According to standard theory, that should not happen. But it does. To my knowledge, no one has ever provided anything better than a shrug emoji to explain it, but... I think we might have just found the explanation! As a corollary, I think it's highly likely there are different category caps on different scorecards.

Number of negative reason codes seems to depend on the source, and the source for whatever reason limits negative reason codes to X number at Y score. I've never quite understood that, outside of the generalization "...if the score is top-tier already, no need to focus on the negative reason codes..." or something similar. True for the average person? Probably. True for a Credit Rebel? Absolutely not. I think it's important to remember that many of us even with top notch scores still have many negative reason code factors adversely impacting those scores. The thing is, the majority of them may only be impacting score a point or two or three, basically nothing significant and easily able to go unnoticed with A - buffer or B - omitted reason codes beyond 2, 4, whatever. I guess what I'm starting to think about when it comes to negative reason codes, buffers, and potential caps per category is this. Say in Category A one is capped out in terms of points, but has some factor that's adversely impacting their score 3 points for the sake of discussion. Those 3 points go unnoticed due to buffer. This could be something like an average age of account factor of (say) 9 years verses 11 years. My guess is that this negative reason factor would still be present, but likely either A - farther down the list than other factors so it isn't visible or B - on a very strong profile may be omitted completely by most CMS. Just because it's omitted of course doesn't mean it isn't there. Circling back to your question about the hypothetical and the person having X points both before and after the change in question... I'm not really sure. Maybe the term X "visible" points is better. There could still be a 3 point impact (using the example above) in place, it's just not visible. If there were a different profile change that impacted that sector of the Fico pie adversely by (say) 9 points, perhaps the net visible change would then be only 6 points... meaning that the unseen 3 points are indeed "counting" even if in a capped out state they aren't seen or able to be quantified. This is definitely an exciting topic to ponder!

Indeed!

Just to clarify one thing: when I mentioned instances where fewer than four negative reasons are shown, I meant places like MF that show up to four if they are present. I wasn't talking about intentional suppression in the interface, such as many banks that only show the top two reasons, or the many places that suppress all reasons on scores of 800 and above. Those are annoying practices, but not relevant to what I had in mind.

There was one case I recall from several months ago on MF where someone bona fide -- maybe Rev? -- was talking about a person with a score in like the low 700s, on a scorecard like maybe clean/thin/young/???, and they said something like "only three reasons were present so <the factor under discussion> must have been optimized." I'm doing an awful job of remembering and describing it. I'll try to dig up the post, but it's a needle in a haystack so I'm not very optimistic.

I'm having a great time trying to think of ways to test all this. I love this hobby! So glad to have some like-minded people to geek out with and learn from. I missed y'all.