To my understanding, only retail exposures secured by residential properties will be subjected to a 10% LGD floor by the new Accord (ref. 266). While most other credit exposures may be bound to a risk weight or capital floor imposed by local supervisor, it will be based at their own discretion. However, not many supervisors have came up with the idea of floor yet, therefore none of the banks that we have come across have specifically imposed a floor to their LGD or risk weight calculation.
Note that the LGD floor of 10% of residential mortgage is to be applied at the pool level, not individual facility level.
It is in fact possible to have an unstressed LGD of 0% or even negative LGD value in certain asset classes, such as commercial leasing, where the bank may profit from the sale of an recovered asset (in extreme circumstances). Having said that, this doesn’t occur very often. However once you stress the LGD to an economic downturn LGD, it is rare that LGD will still be 0% particularly if you include intangibles, cost items such as cost of recoveries and the time value of money which cannot be recovered (even if the exposure is secured by 100% collateral value).
As a result LGD cannot be 0% because of the internal cost and economic value, however they can be very close to zero if there is evidence to support this (such as mortgage insured exposures).
However due to paragraph 266, residential mortgage exposures will be subjected to 10% LGD floor for the purpose of capital requirements even if it is mortgage insured.
As for other assets, there is no incentive for banks to imposed a LGD floor for internal assessment, however they are subject to a floor for capital requirements (based on supervisory discretion and therefore outside of the bank’s control).
I recommend that banks impose a floor of 0% on their LGD to:
- cater in extreme scenario where a bank does make a profit in the event of a recovery transaction;
- allow for irregular data error; and
- add a margin of conservatism (given the fact that banks may make money out from a default transaction).
There is no incentive for Banks to impose a LGD floor apart from the Accord requirement.
Any comments?
- Should bank imposed a LGD floor for each of their asset classes?
- Should bank assigned 0% LGD for mortgage-insured or over-collaterised assets?
8 comments
20 August, 2006 at 01:24
Peter
Hi all, thank you for creating this forum for knowledge sharing I find it very useful and informative for the uninitiated.
I have a question regarding the article LGD Floor, just curious to know in further details when can we expect to have negative LGD (besides having bad data, and wild estimates of costs and recoveries) or even having LGD > 100%.
If such problems exist, what are the implications of a negative LGD or LGD greater than 100% for the retail mortgage portfolio.
Would a statistical model circumvent this issues?? If not can we do anything about this problems or just assumed blindly the 10% LGD floor.
Much appreciated for any information that you can share.
20 August, 2006 at 15:44
ozrisk
Peter,
As the post says, these situations will be rare. A negative LGD may occur where the terms of the facility allow for the recovery of more than the amount outstanding at the date of default. This could occur where, as above, the lease arrangements allow for the bank to keep the full value of the collateral, regardless of the actual limit or where full recovery is expected and penalty interest, or other fees higher than the cost of recovery can be imposed.
The problem is the adverse behavior that could come out of this. I remember working with ine bank’s model that indicated that, in the event of default, the correct procedure was to lend more as this would result in higher profits. This would, I would expect, give rise to at least some adverse regulator comment.
As for an LGD ceiling – given the interplay between the EAD and LGD models, I would expect this to come out of the EAD model, rather than the LGD model.
It is quite common for the loss on individual accounts to be higher than the facility limit: credit cards are a good example. More generally, though, once a large portfolio was behaving like this it would be normal to look either at the collections process or the actual product. Unless the PD was very low, it is unlikely a product like this would be economic.
I have asked our resudent credit modeller to have a look at this as well and he might want to add to this.
8 February, 2007 at 20:34
Shyamsundar Baliga
LGD > 100% is quite possible, though it is not clear how to deal with it. How it is possible is as follows.
LGD is defined as LGD% = Loss ÷ EAD at the time of default. Now consider an EMI based retail exposure. Say ‘default’ is defined as a condition where any EMI is more than 90 days past due (DPD).
Now say an account crosses 90 DPD, and EAD of the account at the time of crossing 90 DPD is $1,000. (EAD would include principal outstanding, EMIs overdue, and any other charges receivable from the borrower.)
Say the account is closed 3 months after the default was triggered. In the meantime the EAD has risen in these 3 months to say $1,200, due to accrual of interest and other charges which are unpaid.
If at the time of closure no recovery is possible then the entire EAD of $1,200 would have to be written off. Hence Loss = $,1200.
In such a case LGD = Loss ÷ EAD at the time of default
= $1,200/1,000 = 120%.
Hence it is possible to derive a LGD estimate of more than 100%. Any thoughts on the above or on how to interpret or apply this estimate?
8 February, 2007 at 21:55
ozrisk
Shyamsundar,
The interplay between the EAD and LGD factors is an interesting one, I agree. In the circumstances you have outlined, I would agree that the LGD would need to be adjusted, but whether the failure to receive fees and excess interest given the account has moved into recovery counts as a “loss” is a legitimate question. However, the costs of instructing lawyers, court costs, baliffs etc. are all part of the costs of the recovery process and so should be included in the calculation where this sort of action is used.
Para 336, while not strictly relevant, gives a good idea of the interplay between these factors – showing that, under the advanced methodology CCFs can be incorporated into either LGD or EAD calcs.
In this instance, I would lean towards adding the sorts of costs you have identified into the LGD factors when they relate to the size of the loss and into the EAD where they are a fixed currency amount.
For example, interest charges are a percentage of the exposure – to me, this means they get added into the LGD. Fees would be less dependent on the facility size, so perhaps counting these as part of the EAD would be appropriate – but, again, I am not sure you should be counting fees accrued after the default as being, strictly speaking, a “loss”.
That said, most costs are going to relate more to facility size, and so should be regarded as a percentage – and in the LGD. You are probably not going to engage lawyers to recover a delinquent $1,000, but you are likely to get pretty serious about a missing $1,000,000, costing a lot in legal fees.
8 February, 2007 at 21:58
ozrisk
As I said above, though – if an entire portfolio of loans is acting this way, either the PD would have to be low or the business is unlikely to be very profitable – unless the fees and interest is very high.
8 February, 2007 at 23:47
Shyamsundar Baliga
It may not be correct to ask if a certain item should be included in EAD or LGD. These are not mutually exclusive.
EAD is supposed to include the entire amount receivable from the borrower, including principal outstanding, interest accrued, fees outstanding, and charges recoverable.
Similarly “loss” for LGD computation is supposed to be “economic loss” which should include all the components above.
Hence interest accrued after the account goes into recovery should ideally be included in the loss amount, but would not be included in the EAD amount by definition.
The observation you make about an entire portfolio acting this way is absolutely perfect.. There is one more reason why this could happen. In a part of a certain bank’s retail portfolio that I am observing, the sticky acccounts used to remain under recovery for very long periods of time before being written off, resulting in high amounts of accruals during recovery period. In such cases LGDs are typically in the range of 110% to 180%.
9 February, 2007 at 01:07
ozrisk
I would agree they are not mutually exclusive – it is just important to include them in one – but not both. You are also correct on the EAD – but the problem is to come up with an exact definition – something the Accord does not do. The closest is in para 474 “the expected gross exposure of the facility upon default”.
This is usually taken to mean exposure at regulatory default (90 DPD or other reasons) – but what, exactly, is “exposure”? Does it include all the steps the bank reasonably expects to take to try to recover?
As a quantification of EAD, para 308 is looking at how much the bank’s capital would be reduced if you just wrote it off, so I would interpret this to mean that if you just wrote it off at default and did not pursue it that would mean your EAD is lower (and LGD higher).
In practice, though, it does not really matter. As long as your modeling is consistent you can incorporate recovery costs into either factor and the result will be the same.
Just make sure you document it properly – you don’t want excessive pillar 2 action from the regulator!
.
On the long recovery period – I have seen similar problems. The important thing from a banking point of view is that the fees and interest accrued during this period are not counted as income to the bank only to be written off later. This would tend to overstate current profits and drop future ones from an accounting point of view.
If you are using the accrued amounts in your capital calculation the results would be to increase the amount of capital needed now – not the right result. I would strongly suggest that, once an account is placed in recovery the fees and charges not be accrued. This would then drop the LGD and reduce the EL figure.
The hit would come to current year (accounting) profit.
16 February, 2007 at 14:05
ozrisk
Looks like APRA will be imposing a 20% LGD floor on RRE – comments on this here, please.