Since 1994 the Federal Planning Bureau has been using the annual version of the econometric model modtrim as a central tool to produce its short-term macroeconomic forecasts. At the origin of the project, and as its name indicates, this annual version was meant to be short-lived and quickly replaced by a quarterly version. Unfortunately, the lack of quarterly national accounts prevented from doing so for several years. In 1998, the Institute for National Accounts published official quarterly accounts for the first time and the construction of the quarterly version of the model started in Spring 2000. On that occasion, the opportunity was taken to reassess all behavioural equations of the model. The more limited availability of quarterly data, in comparison with annual data, implied that a more aggregated version of the accounting framework of the yearly model had to be constructed.
The choice to develop a quarterly model was of course not only motivated by its original name, but seemed very well suited to business cycle analysis and short-term forecasting. In this context, three elements are worth mentioning:
Working with a quarterly instead of a yearly model makes it possible to integrate explicitly all the quarterly information available when preparing the forecasts. Of course this ‘advantage’ of relying entirely on recent information may be a drawback if data are subject to major revisions over time.
Working with quarterly data means that carry-over effects are taken into account much more precisely as compared with yearly data. In fact, the average yearly growth rate of a variable is influenced by the quarterly pattern of that variable during the course of the previous year. Working exclusively with annual data means ignoring this phenomenon completely.
- Business cycles are simply better described with quarterly data: specific dynamics within a single year and the lag structure in economic relationships can be captured more precisely.
The aim of this working paper is not to provide a complete model user’s guide, but to present the specification and estimation results of the behavioural equations, and to give an insight into the overall accounting structure of the model. The simultaneous responses of the complete model to exogenous shocks are examined through different technical simulations and scenario analyses.