1
$\begingroup$

Assume we have a GARCH-M(1,1) process as follows:

$y_t = \beta y_{t-1} + \delta h_t + \epsilon_t, \quad \epsilon_t \sim N(0, h_t) $

$h_t = a_0 + a_1 \epsilon^2_{t-1} + b_1 h_{t-1}.$

If we scale $y_t$ to create $x_t = 100\times y_t $, I recently found out in this question Do parameters stay unchanged when GARCH is scaled? that $a_0$ would scale quadratically with the scaling value. How would the $\delta$ parameter change though in the GARCH-M process?

$\endgroup$
1
$\begingroup$

$\delta$ would scale inversely proportionally to the scaling of $y_t$.
Take the original model

\begin{aligned} y_t &= \beta y_{t-1} + \delta h_t + \epsilon_t, \\ h_t &= a_0 + a_1 \epsilon^2_{t-1} + b_1 h_{t-1}, \\ \epsilon_t &\sim i.i.N(0,h_t) \\ \end{aligned}

and scale $y_t$ with a positive constant $c$. This turns the model into

\begin{aligned} (cy_t) &= \beta (cy_{t-1}) + \frac{\delta}{c} (c^2 h_t) + (c\epsilon_t), \\ (c^2 h_t) &= (c^2 a_0) + a_1 (c\epsilon_{t-1})^2 + b_1 (c^2 h_{t-1}), \\ (c\epsilon_t) &\sim i.i.N(0,c^2 h_t) \\ \end{aligned}

which can be expressed as

\begin{aligned} \tilde y_t &= \beta \tilde y_{t-1} + \tilde\delta \tilde h_t + \tilde\epsilon_t, \\ \tilde h_t &= \tilde a_0 + a_1 \tilde\epsilon_{t-1}^2 + b_1 \tilde h_{t-1}, \\ \tilde\epsilon_t &\sim i.i.N(0,\tilde h_t) \\ \end{aligned}

where

\begin{aligned} \tilde y_t &= c y_t, \\ \tilde\delta &= \frac{\delta}{c}, \\ \tilde\epsilon_t &= c\epsilon_t, \\ \tilde a_0 &= c^2 a_0 \quad \text{and} \\ \tilde h_t &= c^2 h_t. \end{aligned}

The second line of the equation block above is what you are interested in.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.