It is confirmed that APRA are seeking to impose a doubling of the minimum capital required for home loans – the banks have all received their letters confirming this over the last couple of days. This will mean two things – a home loan will be both more expensive and difficult to get in Australia.
As background (and without getting too technical) the amount of capital (the safety margin, in a way) that a bank is required to keep against any lending it makes is calculated based on past losses and the current economic situation. The ‘LGD’ is one of the factors in there. Double one of the factors and the safety margin doubles with it. This safety margin costs the bank money – so this is passed on in the form of increased interest charges on the home loan.
APRA have said that the floor on LGD for home loans will be 20%. This is just plain silly. There is already a level of conservatism built into the Basel II Accord, with a 10% LGD floor and to double it, without there being any form of justification other than the regulator’s gut instinct, will just be a dead-weight loss to the economy.
Over the last decade APRA have consistently tried to make home loans more expensive and difficult to get, first requiring a 20% deposit on all home loans to qualify for preferential treatment, then discouraging the low-doc style home loans and now this.
If there were some justification in the loss history I could understand it – but this is not the case. The only justification seems to be the fear that losses may increase at some stage in the future. If this sort of regulatory game playing purely to keep APRA happy becomes common the banks may as well stop lending for housing.
Shall we call this one the skid-mark factor?
19 comments
22 February, 2007 at 19:56
aus
If the ‘doubled’ home loan capital position was still below current levels, then how does increasing it increase the cost of home loans?
In addition, the appropriate LGD is that experienced in a downturn, not current economic conditions.
22 February, 2007 at 20:48
ozrisk
aus,
If the true risk of home loans was at such a level that very little capital needs to be held for the risk that they go bad (the modelled level in Australia is currently around the 4% level for the LGD factor) the actual cost of having that loan on your books is very low. Put in additional capital (as the accord does) up to 10% and the cost of capital to finance that loan also doubles. APRA’s additional increment further doubles the amount of capital.
The actual incremental cost on the loan is not huge (on a back of the envelope calculation on a $400K mortgage the cost is around $320p.a.) it is still an additional $320 out of the family pocket for no real benefit to anyone.
I would agree on the downturn LGD – but 4% is the modelled through the cycle LGD, with some conservatism already built in.
25 February, 2007 at 14:53
aus
Thanks Ozrisk,
I think the point i was trying to make regarding increases in costs to consumers is really a relative one. That is, relative to current capital, the capital held under Basel II will likely decrease (to reflect the true risk of home loans). That is, relative to the current cost of home loans, the cost of home loans under Basel II will not increase. This is true whether the regulator makes the floor 10, 20 or even 30%.
The question that interests me then is that if the capital outcomes reduce the cost of home loans (which i am pretty much speculating that it will), will banks pass on the cost savings to the consumer?
Regarding the downturn LGD – from my reading of paragraph 468, unless a bank can prove that PD LGD correlation doesn’t exist (hard pressed given the extensive literature on the subject), it is then required to calculate a downturn LGD. A TTC LGD (which is essentially a long run average) will not meet the requirement of a downturn LGD as it is biased downwards by data from benign periods. Given the extensive benign Australian experience this bias is likely to be considerable.
I guess what i’m trying to say is that the LGD (whatever it is) is not based upon the loss history as i doubt given tax laws (to keep data for 7 years only) that banks have an extensive enough history to go back to the last downturn (whenever that was). Therefore, as you say the appropriate LGD is most likely based upon the premise that losses will increase in the future and not the data history. Let’s not forget that banks are holding capital for losses that occur in at least the 99.9th percentile of the loss distribution (which I’m sure is a very big number).
25 February, 2007 at 16:16
ozrisk
aus,
Correct on the relative point – the amount of capital to be held for home loans will almost certainly decrease under Basel II, even with a 20% floor. Given the highly competitive environment (and the number of non-APRA regulated entities out there lending) to me it would be a very brave bank that did not pass on the savings, so the comparison to the non-floor capital level is, IMHO, a valid one.
The fact that a lot of the drop in the margins on home lending has been driven by the non-regulated sector serves, to me, to emphasise the point.
On 468, the second sentence is the interesting one – the requirement to estimate the downturn LGD in sentence one is then, effectively, modified in sentence two to be “…[]not be less than the long-run default-weighted average loss rate given default…”. This means it does not have to be more than the long-run (TTC) LGD – but the added conservatism is probably a good idea, which is why the 10% floor is not something I am strongly opposed to – only mildly.
I would also suggest that the last sentence in 468 is important (and this is not something I had thought of before on this)
To me, the imposition of a 20% capital floor by regulatory fiat is not the way to achieve this. The use of a blunt instument is not a good way to encourage development – it will give rise to methodological sloppiness as the whole point will be only to show you are below 20%. If APRA could show that, given the current risk profile, the downturn LGD in Australia is 20% then I would be somewhat mollified.
On the loss history – the banks stopped deleting this data a while ago and the long term history was also available from Moody’s – which most of the banks have utilised in one way or another. Since the last downturn we have also had the 80% LVR imposed, further reducing the risks in this area. I just cannot see how this is justified, given all this.
One other point, and this is a real quirk, the 10% (and now 20%) floor only applies to retail home loans. If a bank grants a loan secured by RRE to a business the floor does not apply.
So, for a cheap home loan, all you need to do is to put it through your business.
28 February, 2007 at 23:22
bruce
Part of the problem, I feel, is at what level the floor is calculated. At the moment the official (UK) guidance on this is light. lgd%s and the floors can be calculated at case level, segment level or whole portfolio level. For many lower risk lenders the lowest capital will be achieved by setting a single LGD at the 20% floor. However, it would be more risk sensitive to group the portfolios into segments with like risk of loss (ie by ltv) and calculate the LGDs subject to the minimum floors. For some of these segments the LGD could be close to zero but would be raised to 20% by the floor and for the higher risk segments the LGD would be beyond the floor. By incouraging a single LGD, there is a loss of sensitivity of the models. If a lender moves to more risky lending the models may not indicate a change in LGD. This lack of sensitivity to risk is contrary to the principles of Basel II.
1 March, 2007 at 10:12
ozrisk
Bruce,
I would agree – even the 10% floor in the accord is contrary to this principle. Doubling it makes it even less valid.
6 March, 2007 at 18:57
Riskopedia
Here is my thought in this 20% LGD, my gut feel of how APRA came to 20% is:
They have done a stress test on scenario where 30% dropped in property price and its related impact to LGD and PD (you can search on google, “APRA-Insight-3rd-Quarter-2003-Stress-testing-housing-loan-portfolios.pdf”). One para there which was interested that “Historical data supplied by the mortgage insurance industry shows that Australian LGDs have a long-run average of approximately 20 per cent.The APRA model assumes that this average LGD applies to loans with an original LVR of 76 to 80 per cent. ….”
what do u guys think?
6 March, 2007 at 19:23
APRA 20% LGD floor « Riskopedia
[…] by riskopedia in LGD, Retail Risk, Basel II. trackback Been reading an interesting article in Ozrisk blog “APRA increases the cost of home loans – again”. With those risk people in the […]
6 March, 2007 at 20:05
ozrisk
It does provide the first reasonable rationale I have seen for the 20% floor – not, I hasten to add, that I agree with it.
For those that are interested, the article is here.
To me, even if the analysis of the Insurance Council of Australia – and the methodological leaps APRA take with it – is correct, imposing a blanket 20% floor is not supportable.
I note from page 10 of the article (for example) that the LGD on loans at 60% LVR or less are estimated to have a 3% LGD – so why the blanket 20%? Even if this modeling is correct loans at an original 60% LGD should not be penalised for the perceived sins of those having an 80% LVR.
I am trying to find a copy of the LMI paper from the Insurance Council which is the source of the 20%, so, if I find it, I will expand on the above.
In the mean time, I cannot see how this meets the para 468 “…appropriate approach…” criteria.
6 March, 2007 at 22:50
Riskopedia
There is another article I have found
“http://www.apra.gov.au/RePEc/RePEcDocs/Archive/discussion_papers/dp0014.pdf“
Again, similar para. on P.8 “Loss-given-default Loss-given-default (LGD) in the proposed MER model is based on the long-run average LGD of 20 per cent reported by LMIs, across all LVR buckets, from 1980 to 2000. Consistent with economic intuition, the proposed model allows LGD to vary with LVR. This increases the model’s sensitivity to risk compared with the current model, which has a flat LGD across all LVR buckets.”
Regardless of what results we have in MER model, I have couple of comments:
– From a general logical pt. (also proven in APRA stress testing results), a bank would expect to get a Higher LGD% for those High LVR Young accounts during downturn. But if a loan got below 70% – 80% LVR (just assume no LMI), even if the property price dropped by 20% (during a downturn). The “LGD” is very unlikely to be 20%, when it pretty much means the price of the property would of dropped by another 10-20%. (i.e. the asset value cannot cover the remaining loan)
– LMI industry normally deals with High risk loan, e.g. LVR > 80% (at least) or high risk customer (from bank’s perspective). As for retail banks, large proportion of the loan is normally
6 March, 2007 at 22:51
Riskopedia
cont..
– LMI industry normally deals with High risk loan, e.g. LVR > 80% (at least) or high risk customer (from bank’s perspective). As for retail banks, large proportion of the loan is normally
6 March, 2007 at 22:57
Riskopedia
– LMI industry normally deals with High risk loan, e.g. LVR > 80% (at least) or high risk customer (from bank’s perspective). As for retail banks, large proportion of the loan is normally less than or equal to 80%. Hence, would not be surprised if retail bank has a low LGD figures (again, forget about the whole beign economic environment). So, I do question the use of LMI results on 20% LGD.
– Basel Accord indicated that a bank needs to segment LGD if can do. So, if one bank decided to have multiple LGDs values (whether based on LVRs, Age of Account etc.), I do not believe to set a 20% floor is appropriate when it kinda contradict what Basel is trying to achieve. Else, why we would have segmented LGD in the first place?
7 March, 2007 at 00:04
ozrisk
I agree – a 20% floor makes the capital less risk sensitive.
If there is anyone form APRA out there who wishes to put their view I would be interested to hear it – ‘coz I can’t work it out.
7 March, 2007 at 20:42
bruce
From a simple view point, A high quality low risk mortgage book could have a PD (for a big chunk of the book) below the floor of 0.03%. As PD is correlated with LDG to an extent. These same loans are likely to have an LGD below the floor.
So the Risk Weight percentage for these loans would be 0.98% of EAD under the 10% LGD floor rule.
An RW% of 1% for mortgages may be a step too far for the regulator (Under the Basel 1 rules all mortgages had a RW% of 50%). Doubling the LGD floor would increase this to just short of 2%.
So it maybe as simple as they dropped their bottle.
7 March, 2007 at 22:32
ozrisk
Bruce,
I think we should organize some Parliamentary hearings on this. I would love to hear that explanation on the formal record.
Just as a side point – I am not sure many consumer loans would get below 0.03% PD (I will have to remember to ask coldies on this) as this would constitute a S&P AAA rating. Technical defaults are likely to be a little more plentiful than this. I do agree it would be very low, though.
So far I have not heard one nice thing said about this idea – and I have spoken and listened widely on it. C’mon – there has to be one of you out there that agrees with this.
7 March, 2007 at 23:34
Riskopedia
Using the QIS 5 workbook from http://www.bis.org/bcbs/qis/qis5.htm.
Lets take the following scenario:
– Let say majority of Non-defaulted home loan has approx 1% TTC PD (using APRA stressing testing results, P.1)
– Downturn LGD is the 10% floor. (for argument sake)
– EAD is $1,000,000,000
–> In Standardised, it would have $500,000,000 as RWA. (50% of EAD)
–> In IRB approach, The RWA would be $125,330,946, and RWA is 12.53% of EAD.
If we now change to 20% floor
–> RWA would now be $250,661,891. I.e. double of what we have before. RWA is now 25.06% of EAD.
Now, what I have heard before is that even after the Basel II implementation next year, the “savings” on capital would be capped at 10% discount for couple of years (again, can correct me on the statement). Using the above as an example, I am supposed to save 75% of my current capital requirement under 10% LGD floor. But I would of only “save” 7.5% using the 10% cap rule, which, it would still be above the 20% LGD floor.
Given all the above, we all know what APRA is trying to do, but I agree with Ozrisk that there needs to be a hearing on APRA points of view (so that at least we know APRA didn’t pick a number out from the air). As a side note, all the above calculation is only on Pillar 1, if you then add on the Pillar 2 stuff, the RWA could easily gone back to 50% of EAD if also applied the 20% floor.
However, I do think that APRA will take it case by case from different banks for discussion sake. (But the likelihood of changing the mindset, may be very unlikely).
8 March, 2007 at 01:02
ozrisk
Riskopedia,
Thanks for that. I would suggest you add in the 1.06 factor to get the numbers post QIS5, but the change is minor. The capital change between 10 and 20% LGD is over $10m on your scenario. Chuck in the value of the CBA’s home loan book (2006 annual report, p142) with the rest of the numbers you have and you are talking $1.5bn in additional, un-needed, capital. Ouch. Realistically it will be more than this, though as the PD of 1% is conservative.
I like the point about adding in the pillar 2 issues – to which you can also add the K for operational risk.
As I note in the post on the CBA the effect of this is not going to be evenly spread – it will penalize the banks overweight in home loans to retail clients.
I would love to see some rationale from APRA on this.
15 March, 2007 at 19:35
aus
For an APRA justification see the speech by Bernie Egan (p7) to the RMA which suggests that the floor is temporary and will be in place until banks address certain ‘deficiencies’…. Any thoughts ozrisk on what he’s talking about?
http://www.apra.gov.au/Speeches/upload/MEETING-THE-CHALLENGES-OF-THE-IMPLEMENTATION-OF-BASEL-II-Bernie-Egan-130307.pdf
15 March, 2007 at 22:08
ozrisk
To be honest, aus – no. Any problems I have seen relate to the corporate portfolios where there can be less data. I must assume he is talking about the long run LGDs being under the estimate APRA has from the ICA (as from riskopedia above) and therefore (in APRA’s understanding) wrong.
I wonder what criteria APRA will use for ascertaining these as “correct” later? That they give a number near 20%?
Thanks also for pointing me at the speech – I must confess I had missed it. I think it will provide grist for a few posts over the next few days.