Alignment of the Quarterly Financial Statistics to the Annual Financial Statistics data

Statistical data are often compiled at different frequencies. When analysing high and low frequency data on the same variable one often encounters consistency problems. In particular, the lack of consistency between quarterly and annual data makes it very difficult for time series analysis. This paper discusses the processes and challenges for the alignment of the quarterly and annual financial statistics surveys by industry. The process consists of three phases, the initial editing, to deal with large inconsistencies, a presentation of the methodology using the quarterly related series to interpolate the annual series, and an analysis of the results. In the initial editing phase the large differences are resolved by manually editing the input data and imputing for missing data. The temporal disaggregation/benchmarking technique used are based on the Fernandez optimisation method of allowing random drift in the error process. The main characteristic of this method is that quarter- to- quarter movements are preserved while quarterly-annual alignment is achieved. The diagnostics performed indicate that the Fernandez random walk model method produces plausible results.