Validating a preliminary estimate when temporally disaggregating a time series

This work deals with the problem of obtaining an appropriate preliminary estimate of an unobserved high frequency (say monthly) time series. This problem arises when temporally disaggregating an observed series of low frequency (say quarterly) data. The preliminary estimated series should not only satisfy some basic criteria deduced from subject matter considerations, but it has to be valid in data. Thus, a test statistic is suggested for empirically validating the preliminary series. A real situation is described in which carrying out the monthly disaggregation of Mexico's Gross Domestic Product required a careful analysis of the preliminary estimate employed. A detailed exposition of the problem found as well as its possible causes and solutions is presented here.