Simultaneous Nowcasting of Netherlands’ Macroeconomic Trends and Seasonal Patterns Based on Fourier Analysis

An objective method is developed for providing fast initial estimates and corresponding uncertainty of large macroeconomic indicators, including the aggregate gross domestic product (GDP). Furthermore, a new approach to modelling trend and seasonal effects within time series based on Fourier analysis is explored. Underlying latent factors are extracted from a large panel of auxiliary time series, in a two-step procedure containing principal component analysis and dynamic factor state space analysis. This procedure is applied in a real-time framework to predict the present state of the target macroeconomic indicators, using the extracted factors.
The information flow of the auxiliary time series is simulated in a real-time setting over the period 2013-2020, obtaining a target series estimate each quarter. These estimates, often called nowcasts, are compared with the realized figures of the target series to measure the performance of the present procedure. The results indicate an improvement on earlier studies and the current method used by Statistics Netherlands to provide so-called ’flash estimates’.