Econometric Forecasting and the Science Court

Abstract

This paper examines the additional evidence produced by the seven scientists on each of the issues.The issues were: (1) Should econometricians use the method of multiple hypotheses rather than advocacy? (2) Do econometric provide the most accurate approach to short-range forecasting (Table 2 of versus Fact)? (3) Are complex econometric more accurate than simple econometric methods (Table 4 of versus Fact)? Comments Postprint version. Published in Journal of Business, Volume 51, Issue 4, October 1978, pages 595-600. Publisher URL: http://www.journals.uchicago.edu/ This journal article is available at ScholarlyCommons: http://repository.upenn.edu/marketing_papers/81 1 Published in the Journal of Business (1978), 51 (4), 595-600. Econometric Forecasting and the Science Court J. Scott Armstrong The Wharton School, University of Pennsylvania A man convinced against his will Is of the same opinion still. (Paraphrased from Samuel Butler, 1663) My paper, “Forecasting with Econometric Methods: Folklore versus Fact,” argued for the method of multiple hypotheses in econometrics. However, those who favor advocacy (e.g., Mitroff 1972) assert that ample opportunity exists for minority viewpoints to be aired. Furthermore, they feel that the marketplace for ideas will work efficiently. A recent proposal in Science (reviewed by Boffey 1976) called for a science court to improve the advocacy procedure (see also Callen 1976; Presidential Advisory Group 1976). Rather than wait for different viewpoints to appear in the literature, scientists would be called to act as advocates, much as in a court of law. Such a procedure should speed up the collection and examination of the evidence. The Journal of Business has, in effect, provided a science court to examine the issues discussed in “Folklore versus Fact.” Seven scientists were asked to participate in this endeavor. The Journal of Business assembled an impressive group of scientis ts, most of them well-known econometricians. These scientists also bring a variety of viewpoints with them. What should we expect of the science court? It should at least help to ensure that all relevant evidence on the issues has been presented. A second objective, though of lesser importance, is that it make recommendations on the basis of the evidence. The “Folklore versus Fact” paper obtained the relevant evidence by using the method of multiple hypotheses. I claimed that such an approach is efficient and relatively unbiased. What we have then, is a modest opportunity to contrast the method of multiple hypotheses with the most desirable form of the advocacy method – that is, with the science court. The next section of this paper examines the additional evidence produced by the seven scientists on each of the issues. The issues were: (1) “Should econometricians use the method of multiple hypotheses rather than advocacy?” (2) “Do econometric provide the most accurate approach to short-range forecasting” (Table 2 of “Folklore versus Fact”)? (3) “Are complex econometric more accurate than simple econometric methods” (Table 4 of “Folklore versus Fact”)? For simplicity, I refer to these issues as research strategies, short-range forecasting, and complex methods, respectively. Evidence Provided by the Seven Scientists If the advocacy method were superior to multiple hypotheses, it should generate substantially more evidence and the evidence should be more balanced. Did this happen? I examined each of the papers by the seven scientists (and 1 The editors of the Journal of Business made available to me a major work by Professor Victor Zarnowitz of the University of Chicago, An Analysis of Annual and Multiperiod Quarterly Forecasts of Aggregate Income, Output, and the Price Level,' which, though not directed at my versus Fact paper, nonetheless has implications for the issues discussed therein. 1 therefore refer to that paper as well. (The Zarnowitz paper appeared in the January 1979 issue of the Journal of Business.) 2 As in versus Fact, evidence was defined as published empirical evidence.