oxdoc
Posts: 19
Likes: 20
|
Post by oxdoc on Dec 1, 2019 19:16:10 GMT
Hello, I was just checking AC's default rates at www.assetzcapital.co.uk/defaults-and-losses and the difference between the expected and actual losses up to 2016 seems strikingly large - the “expected lifetime default rate at origination” is between 5.0-6.9%, but the "actual lifetime default rate” is much higher, being between 14.1-53.4%. Is it right to interpret this to this mean that AC has a track record of greatly under-predicting default rates? With these numbers, it seems like a large part of the return depends on AC's ability to recover defaults. Default rates since 2018 are not yet large, but I suppose they are too young to be able to project the lifetime default rate from. |
|
sl75
Posts: 2,092 
Likes: 1,245
|
Post by sl75 on Dec 2, 2019 9:16:55 GMT
Hello, I was just checking AC's default rates at www.assetzcapital.co.uk/defaults-and-losses and the difference between the expected and actual losses up to 2016 seems strikingly large - the “expected lifetime default rate at origination” is between 5.0-6.9%, but the "actual lifetime default rate” is much higher, being between 14.1-53.4%. Is it right to interpret this to this mean that AC has a track record of greatly under-predicting default rates? With these numbers, it seems like a large part of the return depends on AC's ability to recover defaults. Default rates since 2018 are not yet large, but I suppose they are too young to be able to project the lifetime default rate from. From the page you linked: "This is prior to actual and expected recoveries (which are taken into account in the bad debt rate figures above)."
Any defaulted loan gets added to that subtotal, even if full recovery has already been made resulting in no eventual loss to investors.
|
|
ashtondav
Member of DD Central
Posts: 1,814
Likes: 1,092
|
Post by ashtondav on Dec 2, 2019 9:18:25 GMT
Yep, I don’t understand because their headline is predicted losses 3.85%, actual 3.75%.
|
|
IFISAcava
Member of DD Central
Posts: 3,692
Likes: 3,018
|
Post by IFISAcava on Dec 2, 2019 9:26:06 GMT
I read it several times and I still didn't understand it.
I'm really not stupid, and financially literate, and I understand other sites stats quite easily - it shouldn't be that hard to work out what is going on.
Clearer communication please Assetz!
|
|
sl75
Posts: 2,092
Likes: 1,245
|
Post by sl75 on Dec 2, 2019 9:31:24 GMT
Yep, I don’t understand because their headline is predicted losses 3.85%, actual 3.75%. Presumably the 3.85%/3.75% figures are annualised, unlike the "Lifetime" figures shown on the linked page.
An annualised rate of 3.85% and a lifetime rate of 4.5% would imply an average loan lifetime of about 14 months.
|
|
ashtondav
Member of DD Central
Posts: 1,814
Likes: 1,092
|
Post by ashtondav on Dec 2, 2019 10:49:54 GMT
Maybe but that doesn’t explain oxdoc’s discrepancy between expected lifetime defaults and actual.
And I would have thought loan length was longer than 14 months
|
|
corto
Member of DD Central
one-syllabistic
Posts: 851
Likes: 356
|
Post by corto on Dec 2, 2019 11:30:31 GMT
The average loan duration is 40 months
14.6% funds are in currently suspended loans (based on original loan values; actual values should make this look better because amortising loans shrink over time; those numbers are not in the loan book). Assuming all suspended loans are in default, which is a worst case assumption, and a recovery rate of 85% as per years 2013-15 (stated on that statistics page) one would predict a 2.2% bad debt rate.
|
|
ashtondav
Member of DD Central
Posts: 1,814
Likes: 1,092
|
Post by ashtondav on Dec 2, 2019 16:42:09 GMT
The average loan duration is 40 months 14.6% funds are in currently suspended loans (based on original loan values; actual values should make this look better because amortising loans shrink over time; those numbers are not in the loan book). Assuming all suspended loans are in default, which is a worst case assumption, and a recovery rate of 85% as per years 2013-15 (stated on that statistics page) one would predict a 2.2% bad debt rate. Thanks for that, Corto. I can understand that logic when you put it that way. But how good good is ACTUALLY at predicting defaults. According to their headline they’re gold star. According to the stat quoted in post 1 they’re a little below bronze.
|
|
corto
Member of DD Central
one-syllabistic
Posts: 851
Likes: 356
|
Post by corto on Dec 2, 2019 17:39:02 GMT
The average loan duration is 40 months 14.6% funds are in currently suspended loans (based on original loan values; actual values should make this look better because amortising loans shrink over time; those numbers are not in the loan book). Assuming all suspended loans are in default, which is a worst case assumption, and a recovery rate of 85% as per years 2013-15 (stated on that statistics page) one would predict a 2.2% bad debt rate. Thanks for that, Corto. I can understand that logic when you put it that way. But how good good is ACTUALLY at predicting defaults. According to their headline they’re gold star. According to the stat quoted in post 1 they’re a little below bronze. I do agree and would like to add that I deliberately left that consideration to the reader.
|
|
ashtondav
Member of DD Central
Posts: 1,814
Likes: 1,092
|
Post by ashtondav on Dec 2, 2019 18:49:07 GMT
Aha, so like the workings of the PF it's a mystery wrapped in an enigma...
|
|
corto
Member of DD Central
one-syllabistic
Posts: 851
Likes: 356
|
Post by corto on Dec 2, 2019 19:28:22 GMT
Sort of, as there are many unknowns. Still there is a message in the 2.2%. You pointed at it: it might well be closer to a lower bound than an upper one. That informs crucially the average level of interest you have to drive the MLA at in order to get higher returns than (eg) in the 90 day account (on the long run, assuming perfect 90d performance at the capped rate, maximally diversified operation in the MLA, and no SM tricks). There will be a critical point at some level, above which one cannot go (without tricks or very good DD) because higher interest loans have a very clear tendency to crunch (just sort the loans by interest rate and count the duds). I suspect the reversion level is around 7%, but there is (of course) not enough data and that number is entirely a rough estimate.
|
|
oxdoc
Posts: 19
Likes: 20
|
Post by oxdoc on Dec 2, 2019 20:48:48 GMT
Hello, I was just checking AC's default rates at www.assetzcapital.co.uk/defaults-and-losses and the difference between the expected and actual losses up to 2016 seems strikingly large - the “expected lifetime default rate at origination” is between 5.0-6.9%, but the "actual lifetime default rate” is much higher, being between 14.1-53.4%. Is it right to interpret this to this mean that AC has a track record of greatly under-predicting default rates? With these numbers, it seems like a large part of the return depends on AC's ability to recover defaults. Default rates since 2018 are not yet large, but I suppose they are too young to be able to project the lifetime default rate from. From the page you linked: "This is prior to actual and expected recoveries (which are taken into account in the bad debt rate figures above)."
Any defaulted loan gets added to that subtotal, even if full recovery has already been made resulting in no eventual loss to investors.
Yes but that applies to both the "expected" and "actual" default rates, so it doesn't look to me like it should account for any of the large difference between the values. The default rate doesn't depend on recoveries anyway. Even if AC are good at predicting losses, being bad at predicting default rates would seem to say something about their ability to assess the risks of loans. The loss rate also includes the recoveries and the 3.85%/3.75% figures look to refer to that - the default rates and losses figures could be consistent if recoveries were somehow much better than predicted, although I doubt that's the case, so I still feel confused...
|
|
ashtondav
Member of DD Central
Posts: 1,814
Likes: 1,092
|
Post by ashtondav on Dec 3, 2019 9:02:03 GMT
From the page you linked: "This is prior to actual and expected recoveries (which are taken into account in the bad debt rate figures above)."
Any defaulted loan gets added to that subtotal, even if full recovery has already been made resulting in no eventual loss to investors.
Yes but that applies to both the "expected" and "actual" default rates, so it doesn't look to me like it should account for any of the large difference between the values. The default rate doesn't depend on recoveries anyway. Even if AC are good at predicting losses, being bad at predicting default rates would seem to say something about their ability to assess the risks of loans. The loss rate also includes the recoveries and the 3.85%/3.75% figures look to refer to that - the default rates and losses figures could be consistent if recoveries were somehow much better than predicted, although I doubt that's the case, so I still feel confused... So am I. Very confused. And the nature and content of this thread’s responses shows it is indeed confusing...
|
|
sl75
Posts: 2,092
Likes: 1,245
|
Post by sl75 on Dec 3, 2019 10:50:04 GMT
From the page you linked: "This is prior to actual and expected recoveries (which are taken into account in the bad debt rate figures above)."
Any defaulted loan gets added to that subtotal, even if full recovery has already been made resulting in no eventual loss to investors.
Yes but that applies to both the "expected" and "actual" default rates, so it doesn't look to me like it should account for any of the large difference between the values. The default rate doesn't depend on recoveries anyway. Even if AC are good at predicting losses, being bad at predicting default rates would seem to say something about their ability to assess the risks of loans. The loss rate also includes the recoveries and the 3.85%/3.75% figures look to refer to that - the default rates and losses figures could be consistent if recoveries were somehow much better than predicted, although I doubt that's the case, so I still feel confused... With more than 85% recoveries on the earliest defaults shown, there's a huge difference between "defaults" and "losses due to defaults" (i.e. the "bad debt" figures).
The latter are much better defined - a default rate BEFORE recoveries is entirely dependent on your internal policy of exactly which stage any given loan is marked as "defaulted", whereas losses due to defaults are pretty unambiguous once enough time has passed for all the defaults and recovery actions to have come to a conclusion.
It seems to me most reasonable to interpret the default projections as the anticipated losses due to defaults.
It's possible of course that the original way those figures were presented was as anticipated losses BEFORE recovery action, and they're just trying to quietly redefine it retrospectively for the older tranches due to the number of loans they subsequently classed as defaults before recovering most or all of the money... but ultimately it's the end result we're interest in!
|
|
registerme
Member of DD Central
Posts: 6,624
Likes: 6,437
|
Post by registerme on Dec 3, 2019 13:23:43 GMT
If this is unclear (I haven't looked at it beyond reading this thread) it would seem reasonable to me to ask AC to clarify things. stuartassetzcapital, can you comment? |
|
