@Thomas_Thumb wrote:Thanks for the update.
I remain an advocate of a 70% threshold for certain types of installment loans (including auto but not share secured). Again, I wonder if length of payment history affects the scoring impact associated with crossing a high end threshold. [extended payment history => increases points allocated to high end threshold/reduces points associated with low end threshold with total available points being a constant].
What evidence do we have that there's any difference between a share secured loan and any other type of installment loan? Did I miss something in the bickering that I walked away from? Certainly possible.
My own anecdotal data which has been beaten to death suggests there's no difference between a share secured loan vs. a mortage on the classic FICO 8 models (or FICO 04 for that matter) why would an auto loan behave differently? There certainly seems to be no disparity on the best 10% line.
@Thomas_Thumb wrote:
Revelate wrote:What evidence do we have that there's any difference between a share secured loan and any other type of installment loan? Did I miss something in the bickering that I walked away from? Certainly possible.
My own anecdotal data which has been beaten to death suggests there's no difference between a share secured loan vs. a mortage on the classic FICO 8 models (or FICO 04 for that matter) why would an auto loan behave differently? There certainly seems to be no disparity on the best 10% line.
Data from testing presented in the last 6 to 12 months shows no evidence that a threshold exists for share secured loans other than a "low end" 10% threshold. In contrast multiple posters have provided data on an "upper end" threshold for other types of installment loans (mortgages, car loans...) on an individual basis or in aggregate.
In summary information I have read indicates only one threshold for a share secured loan - 10%. Data relating to some other installment loan types indicates an upper end threshold which is not evident with share secured loans - perhaps because they are universally paid down before the loan has any significant age.
So, the difference in behavior was not targeted to the auto loan but to an upper threshold for share secured loans.
I have seen no data indicating a 10% threshold exists for mortgage loans. That would be a hard thing to nail down but, I'd be surprised to see any step change in score due to dropping below 10% on an "aged" mortgage.
Note: There are a number of data points where profiles have scored Fico 08 850s with mortgages or aggregated installment loans in the 40% to 70% on balance to loan ratio. Even accounting for a 20 point potential buffer for some 850s, a 25 to 30 point score boost (as reported witth SS loans dropping under 10%) extrapolates to a score below 850 if mortgage B/L is above 10% ... assuming no high end B/L threshold.
So I reject the null hypothesis that there is no difference among installment loans relative to B/L thresholds.
Perhaps time since a mortgage loan was opened/payment history influences points allocated to B/L ratio?
Has anyone actually tested 70% for a single small SSL?
@Thomas_Thumb wrote:
Revelate wrote:What evidence do we have that there's any difference between a share secured loan and any other type of installment loan? Did I miss something in the bickering that I walked away from? Certainly possible.
My own anecdotal data which has been beaten to death suggests there's no difference between a share secured loan vs. a mortage on the classic FICO 8 models (or FICO 04 for that matter) why would an auto loan behave differently? There certainly seems to be no disparity on the best 10% line.
Data from testing presented in the last 6 to 12 months shows no evidence that a threshold exists for share secured loans other than a "low end" 10% threshold. In contrast multiple posters have provided data on an "upper end" threshold for other types of installment loans (mortgages, car loans...) on an individual basis or in aggregate.
In summary information I have read indicates only one threshold for a share secured loan - 10%. Data relating to some other installment loan types indicates an upper end threshold which is not evident with share secured loans - perhaps because they are universally paid down before the loan has any significant age.
So, the difference in behavior was not targeted to the auto loan but to an upper threshold for share secured loans.
I have seen no data indicating a 10% threshold exists for mortgage loans. That would be a hard thing to nail down but, I'd be surprised to see any step change in score due to dropping below 10% on an "aged" mortgage.
Note: There are a number of data points where profiles have scored Fico 08 850s with mortgages or aggregated installment loans in the 40% to 70% on balance to loan ratio. Even accounting for a 20 point potential buffer for some 850s, a 25 to 30 point score boost (as reported witth SS loans dropping under 10%) extrapolates to a score below 850 if mortgage B/L is above 10% ... assuming no high end B/L threshold.
So I reject the null hypothesis that there is no difference among installment loans relative to B/L thresholds.
Perhaps time since a mortgage loan was opened/payment history influences points allocated to B/L ratio?
Interesting.
I absolutely did get a change when I was playing with my Alliant Share Secure Loan and my USAA Secured CD loan; and it was kinda gold standard level data namely 3B and 1B pulls very close together during my mortgage process where I held my file rigidly fixed and was looking at dates on all my crap for aging (and didn't find any obvious year anniversaries) and seeing an explicit movement on precisely the expected scores that have held over time during everyone's testing.
I'm also not certain that we can extrapolate anything from 850 credit scores, I don't mean to discount your godlike FICO but it makes it difficult for concrete testing when talking degress of pain on beautiful files. That said your file is pretty fixed which makes it good for testing purposes unlike so many others over time contributing data.
Busy reports make for difficult testing and analysis, like mine currently, busier than typical and I got a very, very suspect datapoint on EX (which admittedly isn't the best monitoring solution of the 3B product but here goes since it might be installment related will post it here) and maybe others.
Alert dates, all the credit cards were minimally utilized.
- 7/27 random report on a balance decrease, 721 baseline.
- 8/3 my HELOC reports to Experian, +7 points to 728. $100 out of $27,416. Not sure what to make of this, revolving wouldn't make a difference on this and the terms are very clearly revolving when I look at the base EX report.
- I pay off pretty much all of my credit cards, 8/4, 8/11, 8/13, 8/16 all sitting at the 728
- I pay off my "last" credit card 8/26 and see a drop across all 3 bureaus: -11 points to 727 (-12 on EQ 08 TU 08 for reference). There's still a balance on my HELOC though.
- I rack up a $10 charge on my next to report card (DCU CC at end of month), it reports 9/5: +23 to 740 which is a new record for me, as 739 was my last best score before I got the mortgage.
It doesn't look like the HELOC is being counted as a credit card, and I could make a sloppy analysis and say that +7 is awfully close to what I saw previously when installment testing where I still do think I explicitly got a high breakpoint in there somewhere.
The problem is of course my report isn't fixed, I had an inquiry from Penfed (which some folks have said isn't counted, not sure I believe that) fall off 7/31 so whether it's day of or end of the month, it's gone 8/1 and maybe 7 points. 5->4 countable inquiries (or maybe 4->4 as some have suggested).
I had another inquiry fade 8/23 from Cap 1 and I think everyone's confident that one does count. so 4->3 countable inquiries.
So what to pull from this?
1) My 27K HELOC isn't counting as revolving or at least the number of revolving tradelines with balances as we understand it on any bureau. That's pretty concrete and I don't think that's a sloppy analysis as the no revolving tradelines reporting a balance is well characterized and I've never seen that magnitude of a change on a small balance update ever on my file so it's with a high degree of confidence I assert that's what the -11 and -12 was across 3 bureas. If nothing else in this post is worth a damn, that one fact is very, very useful potentially.
2) Inquiries causing all the bonus, or maybe the HELOC is factoring in somewhere on the installment side (I ain't so sure on this one), but +12 points for an inquiry coming off on my dirty file? Color me skeptical, I've never lost double-digits before and my file hasn't markedly changed (ok one less collection than when I last was really monitoring but the tax lien and lates are still there, I didn't markedly improve my file's beauty as we understand it).
For giggles:
Prior to HELOC report my installment utilzation (aggregate, mortgage, share secure loan, USAA secured CD loan) was either 94.20% or possibly 93.76% depending if my Chase 8/1 payment was reflected already... and if the HELOC is counting as installment as has been suggested it goes to 84.90% and that's certainly there on the most recent 9/5 datapoint.
I'll stop beating the dead horse, as you can see there's some flaws with this analysis and it's awfully hard to pin down what precisely went on here, namely that neither EQ nor TU have shown similar shifts and it really hangs on what had been inserted into the consumer side database at EX at the time any of these datapoints were made... and I've posted exactly nothing to this point on how my file has had some things age (I had a 60 day late from August 2010 age past 6 years for example with that September datapoint).
I should go pull a 1B EX report and see what EX v2 v3 did actually as that might rule out the late or the installment utilization as we understand it. Anyway though that's the level of sophistication required to get a good datapoint, it's not easy if you have a busy file.
ETA: Tme of age, maybe? I had both my secured loans open a year before I started playing with them, and my mortgage is a year old now too though my HELOC is brand spanking new and we obviously see the 10% line with new SSL's so I wouldn't think there'd be much of an age component? For refererence, we know absolutely explicitly that revolving utilization is a point in time metric and that that newly established CL's absolutely count regardless of account open date, so it's reasonable to assume that the same holds on the installment utilization side and age is a different factor altogether.
Also when my HELOC posts next at $5250ish which takes me back above any theoretical installment utilization line of 85%, that'll be another useful datapoint to try to rule out other changes.
So for the 1B current data on EX:
9/5:
FICO 8: 740 (new record)
FICO 04: 687 (not a record but +10 over the last pull)
FICO 98: 732 (also new record as I think I capped out at 731 on a mortgage pull have to double-check that)
FICO 9: 778, or pretty much in line with my EQ FICO 9 score which has always been absurdly (for me) high.
7/1:
FICO 8: 721
FICO 04: 677 (FICO 3 for reference in the report, I still use the model version mea culpa)
FICO 98: 723
FICO 9: 740
Sadly for whatever reason I can't find the full FICO reason codes for FICO 8 which bloody sucks on a 1B report; however, there are some salient changes. For FICO 8 bankcard and auto recently seeking credit was #2, it's now gone off the reason code list, completely: apparently 3 inquiries is less damage than the rest of my file issues, go figure.
FICO 9 had recently seeking credit in the #2 slot on 7/1, it is now at #4 (everything else in same order just this line item moved down two slots, as mentioned 3 inquiries latest June 2016, so looks could be aggregate inquiry count unlike my former now debunked theory of time since inquiry... still a seemingly big shift from FICO 8).
No change on the FICO 04 reason codes between it.
FICO 98 the recently seeking credit line item moved down one.
So inquiries played a non-trivial factor on several if not all the models. Possible it was that as my old installment test where I originally found something started at 82% and down there, and that line didn't get crossed even if under the somewhat sketchy idea that my HELOC is counting as an installment loan... or your and oilcan's point about mortgages being counted differently holds water though I'm struggling to reconcile that with my point gain from the now well characterized 10% breakpoint and then subsequent loss when I got my mortgage.
Either way my HELOC will be reporting non-trivially higher soon on it's way to capping, and then whenever I graduate and go back to my former salary I'm going to be moving with a vengance to just pay off the current mortgage most likely... though 5 years is a long time to be waiting for a datapoint, going to be 6 for my tax lien, plus another 4ish from that point for the mortgage to get under 10%, meh doubt I can do that 
.
@Thomas_Thumb wrote:Oilcan12 changed his name - can't recall what it is at the moment.
In any case we both have postulated that aggregate B/L thresholds may be set to the "highest ranking" installment loan. In other words if SS and mortgage loans do have different thresholds and both types open, the mortgage B/L thresholds would apply to the aggregate.
I remain of the opinion that loan age has some influence on score and paydown of new loans to under 10% is a quick hit that might be realized at a higher B/L threshold for aged loans. I have a hard time rationalizing that 30 or 15 year loans require paydown to 10% to realize full benefit. Heck, Inverse popped to 850 when his liens were removed and his aggregate B/L was around 65%. He certainly did not appear to be in buffer territory as he was at 845 on one CRA and then went to 850 when an inquiry dropped off.
As we know some have reported substantial (25 to 35) point increases when dropping below 70% Ag UT on installment loans but these loans had some age. Can't recall names but one was the poster who purchased jet skis. There was also the data from CAPTOOL which suggests a 70% threshold that was reviewed and referenced in a few prior threads. In contrast SJ reported a couple rather small score increases by dropping his auto loan below 70% but, his loan is only a few months old.
Side Note: Typical 4 to 6 year auto loans, under standard payment terms, would not drop below 70% B/L until after a year's time. Even a 3 year auto loan could be above 70% until one year given initial payments are skewed toward interest vs principal paydown.
Not sure where you are at with your mortgage but some have said 90% could be a milestone.
Ignoring the potential HELOC for the moment I would be just under 95% on my mortgage; if the HELOC is factored (which again I have my doubts) it'd be under 85%. In a month I should be able to tell if the HELOC is factored or not into my installment calculation on EX and as well if 85% is some sort of breakpoint. It is over a year aged now for what it's worth (the mortgage that is).
I'm a little skeptical of a couple of things and admittedly I haven't paid close attention to the forums so my information might be out of date.
1) Obtaining a large number of points for going under the call it 70% installment - I had this discussion with oiilcan but look at the very first page of my absurdly lengthy thread... we're talking 6-7 points on my file, and the odds that it's 3-4x that for a clean file is a little beyond belief. Could be wrong but I have yet to see anything else scale like that in the algorithm.
2) Datapoints are of varying degrees of quality, again I would suggest my first post on the installment utilization research thread is gold standard level data, many other datapoints aren't as I tried to illustrate in my long post up above. The overwhelming majority of files on this forum are nowhere near as static as mine was during that time, who really micromanages their reports to try to report identical revolving utilization for several months in a row? Incredibly folks I would submit.
3) How did idea of the highest quality installment tradeline was the one that was scored? I can't even think of a way to really test that frankly other than to have a mortgage just beyond some breakpoint and then going out and buying a car or similar. I will admit that's a neat way to explain my dataset but I'm not certain that conclusion can be drawn and I would suggest Occam's Razor does apply. Or an auto loan similarly and then taking out a secured loan, that would be more easily testable if someone had financial capability to pay the car ahead through a known breakpoint and then take out a few thousand dollar secured loan.
Heh my memory sucks these days; thanks for the reminders! I need to re-read the whole thing again.
Yeah 10-15 is way more reasonable thank you for that correction.
10-5-16
50% & 40% Thresholds tested.
Effect on FICO 8 scores of paying down car loan, reducing overall installment loan utilization from 58.4% to 39%:
EQ (not available) TU 779 +-0 EX 768 +2
Not much of a bump.
History:
Auto loan reporting with overall installment utilization 87.7%: EQ -30 TU -41 EX -20
Crossing 80% (from 87.7% to 79.1%) : +2 in each: EQ, TU, and EX
Crossing 70% (from 79.1% to 68.8%) : +3 in EQ, +8 in TU, no change in EX
Crossing 60% (from 68.8% to 58.4%): no change
So by paying the installment loans down from 87.7% to 39%, I got back
(a) 4 of the 20 points I'd initially lost in EX
(b) 10 of the 41 points I'd lost in TU, and
(c) [probably 7] of the 30 points I'd lost in EQ.
Next step is crossing 30%; I'm wondering if 30% has significance in the installment
loan calculation, the way it clearly does in revolving credit.
And I'm also considering skipping 30% and 20% altogether, and going to the promised land of 9%.
I don't like having my scores depressed.