A new version of the MODTRIM model
This Working Paper describes the new version of MODTRIM II, the quarterly macroeconomic model developed at the FPB for short-term forecasting. The FPB’s short-term macroeconomic forecasts are published three times a year and are usually referred to as the “economic budget”, as they are used by the Belgian federal government to set up its budget and to perform budgetary control exercises. While the model has retained the same overall architecture and underlying structure since the early 1990s, the main differences relative to the 2003 version, described in Working Paper 6-03, are highlighted.
The model aims to produce forecasts for the main categories of expenditures in the quarterly national accounts (with a separate modelling of prices and volumes) as well as for the accounts of the institutional sectors. The objective of this Working Paper is not to provide a complete user’s guide for the model, but to focus on the specification and estimation results of the behavioural equations for the private sector. As the forecasting horizon of the model is six to eight quarters, fluctuations in economic activity are assumed to be driven predominantly by changes in aggregate demand. The main behavioural equations for the different expenditure categories (private consumption, housing investment,
business investment, exports and imports), for wage-earning employment as well as for the main deflators, are presented in the Working Paper.
MODTRIM is a structural model, which has the advantage (compared to a purely statistical approach) that the forecasts result from the interaction of economic mechanisms. This makes the forecasts of the key variables easier to explain. If possible, long-run structural relationships between the variables as well as short-term dynamics are established econometrically using an error-correction mechanism, so that variables react to past disequilibria. To illustrate this adjustment
process, tables are provided which present, for each (semi-)elasticity, the transition path to its long-term value.
Compared to the 2003 version of the behavioural equations, priority has been given to empirical validation. For instance, the strong assumptions on labour, capital demand and the deflator of value added implied by the use of a Cobb-Douglas production function have been abandoned. Another change concerns export and import prices, for which a distinction is now made between energy and non-energy products. Moreover, due to the increasing importance of financial variables in the real economy, the impact of interest rates on investment decisions has been carefully tested and taken into account when empirically validated. Finally, in order to limit the sample to officially published national accounts and to be able to obtain coefficients over only one monetary policy regime, behavioural equations have been estimated on a sample starting in 1995Q1 at the earliest.
The structural approach also makes the model appropriate for risk scenario and sensitivity analyses, as required by the new European Directive on budgetary frameworks. A few simulation results are presented in the paper to examine the response of the complete model to exogenous shocks or policy adjustments.