Expected vs Unexpected loss

[valima]

I just started reading the FRM material and I am confused about expected and unexpected loss.
If risk analysis and measurement can help us quantify unexpected losses, then wont they then
turn into expected losses if the company takes measures to cover for those risk for e.g. by allocating
extra capital to cover to these unexpected losses?

 

[Arka Bose]

Hi valima!
Its actually a nice question, well, we can say that the expected loss that we factor into while giving loans comprises of the firm specific risk (and a bit of estimation of the industry related risk)
However, unexpected losses are related to the performance of the economy as a whole. If the economy performs bad, not only the firm may default due to their specific risk but also due to other factors as loan defaults are generally correlated.
This is why we try to cover losses by 'economic capital'

 

[S666]

My reasoning is as follows:

When you say "If risk analysis and measurement can help us quantify unexpected losses, then wont they then turn into expected losses" - I think that would be generally true but unexpected losses by definition can not be quantified exactly - they can only be predicted to a degree of uncertainty - there will always be variation of actual losses around the level of "expected loss". It is the variance of this actual loss around the expected loss that is termed as "unexpected loss".

If there is no "unexpected loss" and actual loss turns out to be exactly as predicted by the "expected loss", that doesn't mean that there still wasn't the RISK of unexpected loss throughout that period.

Of course as we know from the theory of continuous distributions, the probability that actual loss is EXACTLY equal to the expected loss is in effect, zero. So there will always be variation around expected loss....and therefore there will always be some degree of unexpected loss.

 

[Arka Bose]

Hi @S666 , your last line about continuous probability distribution was very very helpful. Thanks!

 

[S666]

@arkabose no problem happy to help!