John Helliwell, University of British Columbia.
More than in any other country model building has been an open collaborative effort in Canada. A general understanding of the various models has consequently developed.
The process has been open from the beginning. While the origin of model-building in Canada was official, that is within government departments and agencies, the work was documented and made
publically available. The large required investments such as for data bases were undertaken by government departments and made available to all as a public good..
The collaboration has been a very positive factor. It has prevented modellers from assuming exaggerated positions, but not at the cost of stifling a rich diversity among models. The
diversity among Canadian models, in terms of disaggregation and theoretical structures, is about the same as in other countries where there has been less of a collaborative effort. Another aspect of the collaborative effort in Canada is that many of the people now having principal responsibility for a model were at one time assistants in one of the earlier efforts. Many people currently involved have been associated with more than one modelling project.
All modellers should continue this process of openness in allowing others access to the details of their model and in providing complete documentation. This is not only necessary for
the modelling community to advance, but should also be beneficial for the individual organizations.
Advances in computer software and hardware have made models accessible to a larger number of people. Some onus is on modellers to provide good documentation so that proper use of their models can be made. When mapping out model development strategies an important
consideration should be the accessibility of the model to outsiders.
With proper documentation modellers can empirically test each others' equations and theoretical structures. This will permit a better understanding.
Individual modelling gains will be made by ensuring that the "modelling community" operates openly, consistently, and scientifically.
Consistency in models is important. When imposed, theoretical properties must be consistent throughout a model and should be imposed at the maximum degree of transparency to users of the model. The "black box" phenomenon should be avoided, or at the very least, not encouraged. Logical structures should be kept trim. Models should be transformable into reduced forms.
Mike McCracken, Informetrica.
We should not make our models simple just to accomodate "intellectually lazy" people. To improve the state of models they will probably have to get more sophisticated. It is also not apparent that a small size is inherently desirable.
Modellers should not be overly concerned to downsize and simplify their models so that they can be accomodated by micro-computers. The capabilities of computers restricted the development and uses of models in the 1960s. A shift to modelling on micro-computers would just return all of those restrictions. However, micro-computer versions may be useful in giving a larger number of people an understanding of modelling and macroeconomics.
Helliwell's call for openness in the modelling process deserves support. All modellers have benefitted from the openness and have a public responsibility to contribute from their own work.
David Rose, Bank of Canada
The state of documentation can be as important as the size of a model and its inherent complexity. A model may be complex, but it should be explained simply and clearly.
disaggregating models and making them more complex.
Claude Simard, Bank of Canada.
Helliwell noted the extensive efforts that went into the initial preparation of a database for models. Unfortunately, the "common database" is slipping in quality. In particular, the degrees of freedom are being reduced because when series are altered consistent historical data are often not created. Consequently, at present there is nothing like a 30-year base of historic data.
Mike McCracken, Informetrica.
When the initial databases were constructed they were built by modellers for modellers. This focus has been lost. Since 1969-70 when CANSIM was introduced, the database has become a warehouse of data. There has not been an attempt to provide consistency over time.
Krishna Sahay, Department of Finance.
Disaggregation and complexity often arise from the inability to empirically verify hypotheses at an aggregate level. This raises the issue of at what level we should impose constraints on
models. Another point worth bearing in mind is that most models that have desirable theoretical responses do not forecast well.
Jean-Pierre Aubry, Bank of Canada.
Policy-makers had high expectations of models at the outset. It is clear that models have not fulfilled these expectations. We now have a sense of reality of our models. We know the limitations and the risks involved in their use. We should not try to hide these limitations.
Ron Bodkin, University of Ottawa.
David Rose is right in saying that clarifying models does not so much mean simplifying them as explaining them. Hence, models and the research going into their development should be documented in a public forum. This has not been the case in the past. It was very difficult
obtaining information on the early Canadian modelling efforts for the purpose of writing a survey of Canadian
Ernie Stokes, Conference Board.
Account must be taken of the objective function of private firms offering model services. They usually can not afford to devote a lot of resources to model documentation.