The spurious effect of unit roots on vector autoregressions. An analytical study (Q1314477)

This paper considers specifically the spurious inference problem for ``exogeneity tests''. The inclusion of a statistically independent random walk (RW) in a vector autoregression (VAR) was raised originally by \textit{L. E. Ohanian} [ibid. 39, 251-266 (1988)]. This paper provides a full analytical study of the above problem using the methodology developed by \textit{J. Y. Park} and the second author [Econ. Theory 4, 468-497 (1988); ibid. 5, 95-131 (1989)], and it gives the derivation of the asymptotic distribution of the Wald statistic. Two cases are considered in this study: the case in which the genuine variables are \(I(1)\), i.e., integrated of order one, and the case in which they are \(I(0)\), i.e., stationary. The results show some differences in each of the two cases. The Wald statistic always converges to a nonstandard distribution if the variables are \(I(1)\), and do not have a deterministic trend. The inclusion of an independent RW variable in a stationary VAR estimation does not cause any bias in the test, at least asymptotically. The paper is accompanied by some simulated critical values for the nonstandard distributions.