Temporal disaggregation: an alternative multivariate methodology

In this paper we propose a new extension of Di–Fonzo (1990)'s methodology for multivariate temporal disaggregation. We assume that the errors of the high–frequency series follow a VAR(1) model instead of a white noise process. Additionally, an extensive review of different univariate and multivariate disaggregation methods is presented. Finally, we carry out a multivariate application to obtain Colombia's monthly national accounts from quarterly data. The results obtained using the proposed methodology are similar to those with Di–Fonzo's method. However, our resulting series are less volatile.