The papers presented as chapters in this volume have been selected by the Program Committee from among those presented at the Eighth International Conference on Input-Output Techniques, held in August 1986 in Sapporo, Japan. The Conference was made possible by the United Nations Industrial Development Organization (UNIDO), which gave financial and organizational help, and the University of Hokkaido, which hosted the event. On behalf of all those who attended the Conference I would like to thank both for their efforts.
It is clear from the wide range of papers presented at the Conference and from the worldwide home bases of the participants that input-output analysis has become an essential tool of applied economics. Yet it would be misleading in this introduction to ignore the fact that, particularly in the developed market economies, the combination of the evolution of macroeconomic theory since the early 1970s and the methodological insistence that models must be based on analysis of the optimizing behavior of rational agents threatens to distance input-output analysis from "mainstream" economics. Since I believe that both of these fields would lose from such a development, it seems appropriate here both to discuss factors that have contributed to it and to suggest how input-output economics can learn from, and also contribute to, the more general development of empirical economics.
The field in which the danger of divorce between the input-output tradition and mainstream economics is most acute and most clearly visible is one that has occupied a number of researchers in input-output analysis for many years: the integration of a disaggregated input-output-based model of production and employment with a macroeconomic model explaining the evolution of final demand. These integrated models have been seen as the appropriate tool for detailed business forecasting, for the investigation of economic policy in the medium and long term, and for the analysis of changes in industrial structure brought about by exogenous supply shocks such as natural resource discoveries. Improvements in computing facilities and data availability have allowed the technical problems involved in constructing such models to be largely overcome, and the models are now in use on a regular basis in several developed countries (see, for example, Barker and Peterson [ 1987] for the United Kingdom, or Almon, Buckler, Horwitz, and Reimbold [ 1974] for the United States).
Yet, as suggested here, two important components of these models are at