with the computation of the generalized impulse response function. Empirical results are reported in Section 4. Finally, Section 5 concludes. 2 Noncausalautoregression The starting point of our analysis is the noncausal AR model of Lanne and Saikkonen (2011) that can be described as follows.
2 An alternative formulation is proposed by Breidt et al. (1991). However, as Lanne and Saikkonen (2011) point out, their model has the advantages that it is straightforward to test for the specified number of leads and lags and inference on the autoregressive parameters is
future and past errors, implying that future errors are predictable given the realized observations of the variable in question. An early discussion of noncausalautoregressions is provided by Breidt et al. (1991). Recently, Lanne and Saikkonen (2011b) introduced a useful reparametrization of the noncausal AR process allowing for explicit dependence on both leads and lags of the variable in question. A stationary noncausal AR( r,s ) process y t , depending on r lags and s leads (with r and s both positive integers), is defined by: with ϕ ( L )=1– ϕ 1 L – … ϕ