Michel Lubrano : michel.lubrano[at]univ-amu.fr
This paper proposes a new identification method for structural VAR models based on frequency interval restrictions. Applying this method over a frequency interval around zero allows to obtain consistent estimates of impulse responses and reliable confidence intervals in contrast to usual long-run restrictions. In so doing, we use the methodology of Carrasco and Florens (2000), the generalization of GMM for a continuum of moment conditions, in the case of the asymptotic least squares method, and we thus propose a new estimator, namely the continuum asymptotic least squares estimator (C-ALS). One first attractive feature of our method is that it allows to test the imposed restrictions. Moreover it others a data-driven procedure that can assess formally the relevance of the imposed identifying restrictions over a given set of frequencies. Finally, we provide some new results using Monte Carlo simulations and an application regarding the technology-hours debate.