Empirical Macroeconometric Modelling∗

The paper describes an approach to macroeconometric modelling which has roots back to Trygve Haavelmo’s classic contribution to econometrics (Haavelmo (1944)). Building onto a long tradition in the field, we suggest an approach to macroeconometric modelling which is based on fundamen-tal statistical concepts like the joint distribution function of all observable variables for the whole sample period. Users of macroeconomic models often demand a detailed description of the economy, and in order to accommodate that demand, realistic macroeconomic models invariably become too large to be specified simultaneously. The suggested methodology therefore relies on valid conditioning and marginalisation of this function in order to arrive at tractable subsystems, which can be analysed with statistical methods. The objective is to obtain data congruent sub-models that represents partial struc-ture, which can be combined into a model for the entire economy. There is however a Catch 22: a general theory for the procedure will contain criteria and conditions which are formulated for the full system. Two case studies-the modelling of the household sector and of wages and prices in the Norges Bank RIMINI model- highlight these issues