Professional Documents
Culture Documents
Page 050
Page 050
(j=2), and further assume that x1 and x2 are I(1), so dmax=1, and satisfying a
VAR(k=1). 11 The next step is to estimate the augmented model (p=k+dmax) via
Ordinary Least Squares (OLS) as in Dolado and Lütkepohl (1996), that is VAR(2):
𝑒1𝑡
Where the expected value of the error is 𝐸(𝑒𝑡 ) = [𝑒 ] = 0 and 𝐸(𝑒𝑡 𝑒𝑡 ′ ) = Σ.
2𝑡
Testing the parameter restriction 𝑎12 (1) = 0, is testing that x2 does not Granger cause
Let us expand the bivariate example above to the case of testing Granger non-
causality of volatility on exports. 12 Let xt be the vector that contains the five
t=p+1,...,T
11
See Rambaldi and Doran (1996)
12
We adapt Rambaldi and Doran (1996) example to our case.
37