What's new

Loss component under SMA


New Member
Hi @David Harper CFA FRM,

I'm just reading through the Standardized Measurement Approach SMA. I was wondering how exactly the "loss component" part will be calculated in practice.

Let's assume we are in Bucket 3 and we have 10 Losses of EUR 150mm on average every year.

The loss component is 7x the average = 7 * EUR 1.5bn = EUR 11.5bn

Additionally we have to add to the loss component given the size of the losses, but how exactly do we do that?

Will we have to add another 7 * EUR 1.5bn to the loss component for losses > EUR 10mm and another 5 * EUR 1.5bn for losses > 100mm EUR or will we just have to add it once? Is it then the full average or only the part above EUR 10mm respectively above EUR 100mm?

The original source doesn't clarify so you are my last hope.

Thanks so much in advance!

David Harper CFA FRM

David Harper CFA FRM
Staff member
Hi @Roddefeller I don't yet have a finished SMA model (nor do I seem to be able to find one "off the shelf") but since you asked I have started one here at https://www.dropbox.com/s/0v5a5fvvmp5b4vt/sma-loss-multiplier.xlsx?dl=0
I can't warranty the accuracy so far, but at first blush, the basic mechanic appears straightforward to me (I could be missing something). I am using the source Standardised Measurement Approach for operational risk (here at https://www.dropbox.com/s/s1krptbuhppu6xq/R56-Basel-Committee-on-Banking-Supervision-0316.pdf?dl=0)

I have the function in for the BI component (see below), where the BI determines the BI component. To finally get to your question, my (mere) interpretation is that the formula does intend to double- and triple-count the higher loss events. For example, if the BI component is 5.0 billion (corresponding roughly to a BI of about $24.16 billion) then hypothetically something like:
  • $400.0 mm average total annual loss
  • $300.0 mm average total annual loss only including loss events above $10; i.e., I assuming this must be smaller as it is a subset because the phrase "average total annual loss" is exactly the same
  • $200.0 mm average total annual loss only including loss events above $100
Then, in my example, the loss component = 7*400 + 7*300 + 5*200 = $5,900, such that the internal loss multiplier = LN[exp(1) - 1 + (5,900/5,000)] = 1.064. Intuitively, it makes sense that the multiplier "penalizes" via the effect of the the larger losses. That's my preliminary thought (btw, here is some recent critical commentary http://www.garp.org/#!/risk-intelli...3CA7oEAG/basels-new-approach-operational-risk) ... Good luck on the exam!

Last edited: