Comments on: IFRS retail provisioning methodology

By: Andrew

Andrew — Sat, 23 Jan 2010 10:26:02 +0000

Kim,
What coldies said applies if you are using a roll rate approach to using the IBNR – one that I would not advise unless you have a full understanding of the method and are correctly applying it.
The absense of a full system to calculate this (which can be tacked on to an advanced Basel II compliance system) the way I would typically advise using the IBNR approach needs a bit of work to get right, but it involves taking a sample (say 30 randomly chosen) of the loans that went bad in each category and examine the files to see how long the emergence period for each really was. This then becomes the emergence period to be used for that portfolio.
The percentage loss on the portfolio is a matter of objective fact once the emergence period is known, but this can be adjusted as per the Standard for changing economic conditions.
The only remaining question is what portfolio to apply it to – and this should be the portfolio as it was as at the balance date minus the emergence period. This solves the problem of recognising losses on day one loans as you are not doing so – you are only recognising losses on the portfolio as it was at a point in the past.
Of course, there are better ways to do it once you have a Basel II Advanced system, or a method of tracking roll rates and perhaps progresion through grades, but the method outlined above complies with the Standard and is fairly simple to apply.

By: Kim

Kim — Fri, 22 Jan 2010 07:56:54 +0000

Andrew,

What we are actually doing right now for retail loans is to calculate IBNR for non-default loans since in this case, loss event has yet to be identified (for simplicity we generally use the 91 days past due as the point in time where we identify loss event). When we calculate impairment losses on these IBNR loans/non-default loans we just include in the formula, the emergence period or the loss identification period (LIP) as we call it here.

In order to avoid recognizing day 1 loss on newly written loans, we just exclude those new loans (granted during the past month or two) in the calculation of IBNR losses.

If the emergence period is not applicable to retail loans, as discussed in the article, then we dont have to calculate IBNR losses due to low detection risk (based on my understanding we calculate IBNR losses because of the risk that we have yet to identify loss events on accounts that we still classify as performing/good accounts)because the accounts are updated regularly in the books.

Is my understanding correct. I appreciate your help.

By: Andrew

Andrew — Thu, 21 Jan 2010 14:11:29 +0000

Kim,

This piece is quite old, and consensus has moved on a bit. If using an IBNR approach then the period is probably not immaterial, and the method of finding the emergence period is really up to the bank and the auditor – within the bounds of reality.
The difficulty that coldies is identifying with this piece was that the Standard prohibits the identification of an impairment on newly written assets – which will occur on an IBNR approach if you use the outstandings at the balance sheet date.
This can be fixed by the use of the size of the portfolio at the balance sheet date minus the emergence period.

By: Kim

Kim — Thu, 21 Jan 2010 12:04:11 +0000

I would like to ask one question on this article, if emergence period is generally not applicable in retail banking due to the short time difference between the loss event and the detection, then bank’s should not calculate impairment for IBNR loans and only calculate for default loans. Is my understanding correct? Im a little bit confused on this part.

By: amar

amar — Wed, 14 Feb 2007 02:59:05 +0000

A way to tackle the problem may be through the initial recognition of the asset i.e. if the market expects that a certain % will default, then it would price it in accordingly and therefore the fair value would incorporate the expected loss on zero arrears assets.