13.08.2022 Views

advanced-algorithmic-trading

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

146

11.5.2 Why Does This Model Volatility?

I personally find the above "formal" definition lacking in motivation as to how it introduces

volatility. However, you can see how it is introduced by squaring both sides of the previous

equation:

Var(ɛ t ) = E[ɛ 2 t ] − (E[ɛ t ]) 2 (11.8)

= E[ɛ 2 t ] (11.9)

= E[wt 2 ] E[α 0 + α 1 ɛ 2 t−1] (11.10)

= E[α 0 + α 1 ɛ 2 t−1] (11.11)

= α 0 + α 1 Var(ɛ t−1 ) (11.12)

Where I have used the definitions of the variance Var(x) = E[x 2 ] − (E[x]) 2 and the linearity

of the expectation operator E, along with the fact that {w t } has zero mean and unit variance.

Thus we can see that the variance of the series is simply a linear combination of the variance

of the prior element of the series. Simply put, the variance of an ARCH(1) process follows an

AR(1) process.

It is interesting to compare the ARCH(1) model with an AR(1) model. Recall that the latter

is given by:

x t = α 0 + α 1 x t−1 + w t (11.13)

You can see that the models are similar in form with the exception of the white noise term.

11.5.3 When Is It Appropriate To Apply ARCH(1)?

What approach can be taken in order to determine whether an ARCH(1) model is appropriate

to apply to a series?

Consider that when we were attempting to fit an AR(1) model we were concerned with the

decay of the first lag on a correlogram of the series. However if we apply the same logic to the

square of the residuals and see whether we can apply an AR(1) to these squared residuals then

we have an indication that an ARCH(1) process may be appropriate.

Note that ARCH(1) should only ever be applied to a series that has already had an appropriate

model fitted sufficient to leave the residuals looking like discrete white noise. Since we can only tell

whether ARCH is appropriate or not by squaring the residuals and examining the correlogram,

we also need to ensure that the mean of the residuals is zero.

Crucially ARCH should only ever be applied to series that do not have any trends or seasonal

effects, which are those that have no evident serial correlation. ARIMA is often applied to such

a series (or even Seasonal ARIMA) at which point ARCH may be a good fit.

11.5.4 ARCH(p) Models

It is straightforward to extend ARCH to higher order lags. An ARCH(p) process is given by:

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!