Naive observation

[brian.field]

This is naive of me, I know, but it is interesting to note that attention given to diversification. It is a pillar of modern portfolio theory, yet a primary area of focus in risk management is the degree to which correlations approach 1 in extremely stressful times....

Isn't it interesting to note that market participants are asked to rely on the benefits of diversification, which are dependent on negative correlations? In order to benefit from diversification, investors must go long assets that are negatively correlated with one another.....and then, once long this portfolio, they are exposed to the risk that correlations approach 1?!!!

After having studied Risk Management for that last year or two via my FRM prep, I still to not see a clear answer for this question. Everyone mentions that correlations increase but no one really offers a solution. Basel requires stress testing and scenario analyses..... So what? Investors are still pushed to diversify! The only solution I see is to identify assets that are always negatively correlated, even in extreme environments....

I haven't given much deep thought to that above - it just occurred to me as I was reading. I am sure I could come up with a better rationale or response.

 

[QuantMan2318]

Hi @brian.field ,

The best way is the Buffett way! or the Original version by Ben Graham, whichever you prefer! Value investing would protect us

 

[Dr. Jayanthi Sankaran]

Hi Brian,

Some thoughts on your "naive observations". It so occurred to me while I was reading Reading 1 of the 2016 GARP FRM Part I, that one could look at "diversification" and "correlation" in the following manner (correct me if I am wrong!)

As you rightly point out, a "long" portfolio of negatively or imperfectly correlated assets will diversify the "idiosyncratic risk" according to portfolio theory.

In risk management we are more concerned about "correlation risk" especially under stress. Lets take a complex portfolio of interest-rate-sensitive assets with different maturities - some "longs" and some "shorts". If interest rates go up, the value of the "longs" go down, and the value of the "shorts" go up. This is a "natural" hedge. In fact, it is possible to effectively hedge this portfolio against a "parallel shift" of the yield curve. Notice that till now, we have not talked about "correlation risk".

Although, the above portfolio is hedged against a parallel shift of the yield curve, it is not hedged against "yield curve" risk. Meanwhile, even when offsetting positions have the same maturity, "basis risk" can arise if the rates of the "longs" and "shorts" are imperfectly correlated. For example, although a three-month Eurodollar deposit and a three-month US Treasury bill pay three month rates, they are not perfectly correlated. As a result, spreads between their yields vary over time.

Hence, a three-month US Treasury bill funded by a three-month Eurodollar deposit represents an "imperfect hedge" or "basis risk". This causes a divergence in their spreads under periods of stress - which is exactly what happened to the LTCM speculative trades during the Russian default crisis.

Hope I am making some sense

Jayanthi

 

[brian.field]

Great points Jayanthi! Thanks.

As I mentioned in my original post, I probably could have thought about it a bit more - had I done so, I might have been able to present a more cogent hypothesis.

My real point is the following:

We are studying "risk management". In doing so, we have been asked again and again to understand, to know, to be able to compute, to employ, various risk metrics, the most common of which is VaR.

It is amazing to me that after all of this effort, we learn that not only is the VaR a weak risk measure - it is actually much worse.

There is a difference between an inaccurate metric and a metric that actually makes things worse!

If the VaR is fundamentally prone to procyicality, then it would be better to do noting rather than to risk manage (assuming reliance on VaR)!

Let me say that again...

If the VaR is fundamentally prone to procyicality, then it would be better to do noting rather than to risk manage (assuming reliance on VaR)!

This is remarkable.

 

[ami44]

I always thought that "the correlation increase in a crisis" is a very fancy expression for the fact that a crisis means, that everything goes down at the same time.

I'm thinking along the lines of Jayanthi, what past crisis has shown us is, that the only hedges that are effective under stress are being short your long position and vice versa. If you do not have any position in any risk factor you are save. Of course if you do not take any risk, you are limiting your potential profit too.

To the topic of procyclicality: i think the problem is not the risk measure, its the connection with regulatory capital, that creates the procyclicality. Any measure of risk would cause that.

Does that make sense?

 

[Dr. Jayanthi Sankaran]

Hi Brian,

Since VAR can be used only under normal market conditions, it is a useless risk measure under periods of "market stress". Further, since it does not tell us anything about the extent of loss beyond the 95% or 99% confidence level, we use the EVT?? And that is where "scenario analysis" and "stress testing" come into the picture...

Hope that makes sense
Thanks!
Jayanthi

 

[brian.field]

Great points @jsankaran and @ami44 !

It makes for interesting discussion!

Brian