EWMA

[Razor425]

Hi Guys,

Quick question in relation to using EWMA in practice - I know Risk Metrics uses alpha factors of 0.94 and 0.97 but does anyone know the reasoning for this? Also does anyone using EWMA use a different alpha factor for some reason - if so would you mind explaining the reasoning behind it?

Thanks

 

[ShaktiRathore]

Please visit: https://forum.bionicturtle.com/threads/garch-vs-ewma.7374/#post-26666
The alpha chosen is more subjective u can choose either .94 or .97. You can solve for alpha on an empirical data ,allow some judgement on your part and adjust the calculated alpha accordingle. Normally alpha is near .94 as might be justified by empirical results un most cases. There is nothing special formula for alpha i think its just an empirical outcome of some data.
Thanks

 

[Colin_Edwards]

The choice of decay rate will vary by what you cover. For example, if you risk manage a natural gas trading desk, you might feel that [ recent changes in fracking / politics in the Ukraine / transport of liquified natural gas by tanking (instead of just by pipeline), or whatever ] might mean that historic data isn't as valuable as more recent data in looking at risk... and you could use a higher decay rate. If the Ukraine deteriorates, you want that in your VaR now; not 20 days from now. Data from 1-2 years ago might not be very meaningful. If you cover residential mortgages in northern europe: data from older periods is still relatively meaningful (so, less decay.) There are no 'right' or 'wrong' answers.

 

[David Harper CFA FRM]

@Razor425 Here is what Carol Alexander says about the lambda parameter (aka, alpha in RM), emphasis mine:

"So which is the best value to use for the smoothing constant? How should we choose λ? This is not an easy question (By contrast, in GARCH models there is no question of ‘how’ we should estimate parameters, because maximum likelihood
estimation is an optimal method that always gives consistent estimators). Statistical methods may be considered: for example, λ could be chosen to minimize the root mean square error between the EWMA estimate of variance and the squared return. But more often λ is often chosen subjectively. This is because the same value of has to be used for all elements in a EWMA covariance matrix, otherwise the matrix is not guaranteed to be positive semi-definite. If the value of lambda is chosen subjectively the values usually range between about 0.75 (volatility is highly reactive but has little persistence) and 0.98 (volatility is very persistent but not highly reactive)." -- Carol Alexander, Market Risk Analysis, Volume II, http://www.amazon.com/Market-Analysis-Practical-Financial-Econometrics/dp/0470998016