Reflections on Econometric Methodology

General issues about the methodology of empirical econometric research are discussed. It is argued that the most successful paradigms for applied work are the ones that have a capacity to survive and to evolve into more useful forms as these are needed Paradigms that embrace progressive modelling principles, such as those espoused by David Hendry, seem most amenable to this criterion. It is also argued that econometric theory has a large role to ploy in helping us to understand the strengths and the weaknesses of a methodology and to codify what its prescriptions entail. The time‐series methodology of David Hendry is considered in some detail. It is shown that the Hendry methodology comes remarkably close to achieving an optimal inference procedure for long‐run structural relationships even though it is conducted on a single‐equation basis. The findings indicate that the methodology may be unproved further to achieve results that are equivalent to optimal estimation. Detailed responses are provided to the panel discussions on econometric methodology by Dennis Aigner, Clive Granger, Edward Learner and Hashem Pesaran that were presented at the 1988 Australasian Meetings of the Econometric Society in Canberra Some personal reflections are offered on the many issues that arise from this panel discussion, including the merits of Bayesian and classical approaches, asymptotic theory, experimental and non‐experimental data, model evaluation, diagnostic testing sharp prior hypotheses, the use of graphics, and the role of economic theory in empirical modelling.