Macroeconometric models can be used, for instance, to assess how the euro area responds to shocks or to make forecasts and projections for the euro area. At the same time, for policy purposes, such tools need to be associated with alternative—possibly judgmental—approaches, to the extent that a policy analysis has to be based on a full-information set. This paper addresses the particular issue of the ECB projection exercise, first in terms of its relation to monetary policy and the corresponding strategy, and second, with respect to the specific role on models in producing the projections, with a particular focus on which type of models are used and on the particular procedure followed to duly incorporate judgmental elements.
A number of leading experts participate in this conference. Most of them are modellers concentrating on the analysis of the euro area economy. A particular emphasis on this occasion has been put by the organisers on assessing how the euro area economy responds to a range of shocks, be they supply or demand, internal or external, etc. Some particular interest has also been expressed in forecasting tasks.
To address such wide range of issues, we know that a number of models are available, which are based on different approaches and cover a whole range of modelling possibilities. There are, for instance, single country as well as multi-country approaches to euro area modelling, varying degrees of theoretical content and differing weights put on the empirical basis underlying the models, etc. This variety of frameworks presumably translates into potentially different and even divergent interpretations of the results and, as a corollary, of the underlying mechanisms at work within the euro area. Last but not least, there are specific data issues for the euro area, which is a new entity, as yet without a fully complete set of harmonised historical data across the various countries comprising the area, so that it may be difficult to derive firm conclusions from a given model.
In such a context, the main issue in terms of monetary policy is therefore how model results can be validated and used in the overall assessment and finally in the decision-making process. In view of the quite model-dependent answers given to most questions posed, model results presumably need to be associated with what can be learnt from alternative approaches, some of them being judgemental; in other words, a non-model-based analysis is also required. Inevitably, a policy approach needs to be based on a ‘full information’ set, assigning appropriate weights to the different relevant pieces of information.
As a result, while models can certainly be used to give an idea of the euro area response to specific shocks and also to make forecasts and projections, they must not be used in a mechanistic manner. As regards projections, in practice, models are indeed combined with a range of other tools and expert judgement is added to what equations would deliver. Also, in terms of simulation results—especially for monetary policy simulations—there is such a diversity of possible frameworks, an apparent sensitivity to simulation environments, etc. that results cannot be taken at face value. These elements essentially demonstrate that the interaction we would like to capture, namely that between policy and the economic analysis, is very complex, which in some sense suggests that central bankers cannot be simply replaced by computers running model simulations.
These questions are key to any central bank, which is confronted with such issues, the ECB in this respect not being a particular case. In the remainder of this paper, I will take a specific example, in which models and modellers are involved, to illustrate how such issues are tackled in the context of the monetary policy decision-making process at the ECB. I will concentrate therefore on the particular case of the projection exercise that the staff conducts, and this will be done along two dimensions: first, the relationship with monetary policy and, in particular, the corresponding strategy, and, second, the specific role of macroeconometric models in the production of these projections.
Let me conclude by saying that, in view of our still relatively recent experience with the publication of projections, this step seems to have been positive in most respects—and in particular in the following two:
First, at the internal (i.e. Eurosystem) level, this exercise has provided important input to discussion at various levels within the system, on both technical and political issues. In the context of the Eurosystem staff projection procedures, I should mention that macroeconometric models appeared not only to provide important analytical elements, which trigger very stimulating discussions, but also to function as a powerful consensus-building tool and an important consistency device.
Second, at the external level, we find that the projections and their publication have improved communication, and have helped focus and shape policy discussion. As to the format employed, I find it interesting to note that the discussion on ranges seems to have come to an end, presumably in view of the uncertainty recently observed in economic prospects.
Finally, the decision to disclose the projection inputs to decision-making has not altered the fundamental principles which have governed our strategy since the inception of monetary union.
