On the Design of Experiments when Data are Collected by Passive Observation

Some recent developments in the macroeconometric model analysis of time series are discussed in the light of the monograph by Haavelmo (1944). A brief overview of some important works in this area during the last three decades is given, and the importance of the widespread use of personal computers for the realization of Haavelmo's ideas is pointed out. The notion of design of experiments in econometrics is discussed and related both to the case of controlled experiments and the case of passive observation. The latter case is discussed in particular with reference to recent developments in the analysis of nonstationary data. Some of the open questions addressed by Haavelmo are finally discussed in order to find out whether they can be answered more fully today.