VaR Perspectives: Driving While Looking In The Rearview Mirror

VAR: Driving while looking in the Rearview Mirror Submitted by Merrill Lynch's RateLab

Comments

[Anonymous]

Great read from Harley, as usual.

[Anonymous]

Starin at the world through my rearview
Go on baby scream to God, he can't hear you
I can feel your heart beatin fast cause it's time
to die
Gettin high, watchin time fly

-2pac

[McGriffen]

As a mildly-informed follower of VAR, seems too squirrelly to reference such a short-dated series for any historical returns (excess, volatility, etc).  10 yrs of daily data is better than 1 yr...I wouldn't invest with mutual funds on a "good" 12-month track record if their 5-yr track sucked eggs.

[ZeroPower]

Looking back >1year on a VaR calculation makes it even more arbitrary than it already is.. As mentioned in the paper, soon the Lehman bankrupcty will not be included in the 1yr moving average so those PMs wanting to take more risk will easily be able to. The only thing a longer timeperiod chart would bring in the case of VaR is more noise. VaR itself is pretty much an unnecessary statistics, only used so you can show your DM or backoffice accounting hacks that 'hey look we're within the respectable guidelines for our positions'.

[McGriffen]

Agree to disagree...it's a valid question.  Pensions and endowments evaluate asset managers on 3-yr or 5-yr (or 10-yr if able) against benchmark & peers.  How is that risk-management could incorporate such a very limited (as measured in months, years) return series in deciding when to lever up/down.  Just seems easily distorted, in particular in volatile periods.

I want every data point going back to LTCM in 1998, and I'd want a manager of let's say, agency MBS, to show me every data point he has for his return series.

Might be talking apples / oranges, but in practical terms measuring a return series should go longer than 12 months.

[Anonymous]

Volatility as studied in statistical analysis seems to be rooted only in all fear or all greed at the extremes of standard deviations.

The applecart of data is upturned because of the brevity of last October's events. In the course of historical data it looks like an overnight event. The overnight great depression. If this is true, the "cure" for all such future catastrophic financial events has been discovered such that blind risk in the pursuit of of a defined objective, much like the VAR itself, can actually be encouraged without limit. Believe me I want to live in that world and I want all of humanity to enjoy it.

The wide gap and thus the gist of the report is to urge positioning now as the real recovery will be much sharper and intense than what we have seen since the "event". Who can argue this if you look at the economic data? If employment goes positive, home and auto sales go positive, credit et al, who that is still left will not then run to the greed side of the ledge with post haste?

Certainly no one that has read this far into my post!

But while the data mimics that part of common sense, it fails to consider the economic data just described. It assumes it will because it only measures greed and fear.

If a corporation can be said to have the rights of a legal entity, then an economy can, in a sense, be said to be a living entity. It needs to breath, to rise and sleep, to follow cycles.

The diagnostic tests of a heart attack patient may show clearance for a marathon, but is this in the long term interest of the patient?

This report tells me that the FED needs to throw some water on the fire now, a 1/4 point, because if they do so when implied dips to reach realized, the shock needed, say a 1% rise in the rate, may not faze the rush to greed need to close that gap and then no amount of tightening will be able to have an effect.

[Anonymous]

Not just the passing of peak vol, leaving behind the spike in correlation induced by the liquidity crunch post Lehman means a whole lot of risk quota is going to be freed up around late November.