Probability seminar: Dynamic factor models, cointegration, and error correction mechanisms
SM3, School of Mathematics
Matteo Barigozzi, London School of Economics
In this paper we study non-stationary Dynamic Factor Models when the vector of common factors is I(1) and singular, while the idiosyncratic components may or may not be I(1). We first extend the Granger Representation Theorem to the case of singular vector. In particular, we prove that for generic values of the parameters there exists a finite Error Correction representation for the common factors where the number of error corrections terms is higher than in the standard case due to singularity. We then discuss estimation of the common factors, an information criterion for determining the number of common trends, and estimation of impulse responses. Finally, we apply our model to a large panel of US quarterly data to study the effects of monetary policy shocks, as well as the effects of technology shocks. Results support the use of our model.
Joint work with Marco Lippi and Matteo Luciani.