Econometric Theory and Methods

Answers to Starred Exercises

137

Solution to Exercise 13.18

13.18 Consider the GARCH(1, 1) model with conditional variance given by equation (13.78). Calculate the unconditional fourth moment of the stationary distribution of the series ut generated as ut = t t with t NID(0, 1). It may be advisable to begin by calculating the unconditional fourth moment of the stationary distribution of t . What is the necessary condition for the existence of these fourth moments? Show that, when the parameter 1 is zero, this condition becomes 32 1 < 1, as for an ARCH(1) process.

Since ut = t t , and the fourth (central) moment of t is 3, the unconditional 4 4 fourth moment of ut must be 3E(t ). Thus, we need to calculate E(t ), the unconditional fourth moment of the stationary distribution of t , which we will call m4 . From equation (13.78), we see that
2 2 m4 = E(0 + 1 u2 t1 + 1 t1 ) .

(S13.23)

The quantity of which we wish to take the expectation here is

2 2 4 2 4 2 2 2 0 + 1 ut1 + 1 t1 + 2 0 (1 u2 t1 + 1 t1 ) + 2 1 1 ut1 t1 . 4 4 Evidently, E(t 1 ) = E(t ) = m4 . Since ut = t t , we have that 4 4 u4 t1 = t1 t1 2 2 and u2 t1 = t1 t1 .

Therefore,

E(u4 t1 ) = 3 m4

2 and E(u2 t1 t1 ) = m4 .

2 2 Moreover, the unconditional expectations of both u2 t1 and t1 are just . Thus equation (S13.23) becomes 2 2 2 m4 = 3 1 m 4 + 1 m4 + 2 1 1 m4 + 0 + 2 0 (1 + 1 ) 2.

Rearranging this equation and using the result (13.79), we obtain

2 2 2 2 m4 (1 3 1 1 2 1 1 ) = 0 + 2 0

1 + 1 . 1 1 1

A bit more algebra yields

2 2 m4 (1 (1 + 1 )2 2 1 ) = 0

1 + 1 + 1 . 1 1 1

Therefore, m4 =

2 0 (1 + 1 + 1 )/(1 1 1 ) . 2 1 (1 + 1 )2 2 1

(S13.24)

Copyright c 2003, Russell Davidson and James G. MacKinnon

Econometric Theory and Methods

Answers to Starred Exercises

138

As we saw earlier, the unconditional fourth moment of ut is just 3 times this quantity. A necessary condition for the existence of both fourth moments is that the denominator of the expression for m4 in equation (S13.24) be positive. When 1 = 0, this condition becomes
2 2 2 1 1 2 1 = 1 3 1 > 0, 2 which is another way of stating the condition 31 < 1 that is required for an ARCH(1) process to have an unconditional fourth moment.

Copyright c 2003, Russell Davidson and James G. MacKinnon