Risk Contribution

[liewpw05]

Hi David,

In the Ong reading on Portfolio Effects, the author said "risk contribution is the single most important measure in credit portfolio management". The reasons given are

1. It enables the risk profile of the portfolio to be modified by changing the risk characteristics of an asset
2. better management of return on asset

Although Ong has not elaborated the two reasons, i infer in regards to #1 that by choosing assets that have lower default correlations with each other, this reduces the risk contributions of assets and hence the overall portfolio unexpected loss is reduced. Diversification, according to Ong, is a more effective way to reduce portfolio default risk compared to the use of credit derivatives and securitisation which shifts the credit risk rather than hedging it away totally.

In regards to #2, my thoughts are risk contributions allows economic capital to be allocated to different business units and hence enable the risk-return evaluation of projects and performance measurements of business units to be evaluated by the use of risk measures like RAROC vis-a-vis the return on equity required by shareholders.

What are your views on this?

Thanks

Regards,
Peggy

 

[David Harper CFA FRM]

Hi Peggy,

"by choosing assets that have lower default correlations with each other, this reduces the risk contributions of assets and hence the overall portfolio unexpected loss is reduce" - Yes, could not better say it myself! His (6.5b) is instructive: the RC is a function of the sum of the asset's pairwise (default) correlation with other portfolio assets. The "free lunch" of diversification is to alter the asset in such a way that those pairwise correlations (correlation with asset #1, correlation with asset #2, etc) are reduced. That is cheap method to reducing the RC and to your 2nd point, by extension, the amount of incremental EC needed for the new asset.

On (2), yes, per our other thread, this will allow him to use RAPM like RAROC to compute the obligor-specific return: incremental return given by additional obligor divided by incremental capital required (for the additional obligation). Where incremental capital is some multiple of Ong's RC (recall to other thread, his RC is based on a UL which is only a single std deviation, which under normal distribution purely for illustration sake since we know credits aren't normal, but 1 SD = NORMSDIST(-1) = 15%, so cannot put RC directly into RAROC denominator. It is not enough confidence, must put (multiple of)(RC). And that can go into RAROC as you suggest. Same point here as my newsletter chart: Ong UL < Internal UL and Basel UL because Ong UL is only 1 s.d.)

In short, RC is the risk that's doesn't blend in and disappear due to diversification. So, extra EC (RC*multiple) must be allocated it.

David