A statistical approach for disaggregating mixed-frequency economic time series data

Abstract

The problem of mixed-frequency time-series data arises from changing the observation frequency. For example, we may have a time series with quarterly observations in the first portion and annual figures in the remainder. We shall call that quarter-year mixed-frequency data. In this paper we suggest a method to disaggregate the annual observations to quarterly values. The proposed method can easily be generalized to the year-quarter, quarter-month, year-month, and other mixed-frequency situations; it may avoid difficulties of time-series modeling and is easy to implement. A step-by-step algorithm of the method is given so that econometricians not expert in this area can still perform the procedure. The proposed method is illustrated through two real examples. We also conduct a small-scale Monte Carlo experiment to compare the proposed procedure with two existing alternative methods. Finally, some concluding remarks are given.