Are Econometricians Useful? Folklore Versus Fact
Stephen K. McNees · The Journal of Business · 1978
Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.
The Journal of Business · University of Chicago Press · article · 1978
Abstract
Armstrong's paper raises conventional question of usefulness of econometrics in forecasting along with provocative question of whether opinions of expert econometricians on this subject be trusted. Econometricians were virtually unanimous in opinion that econometric methods provide forecasts that are either significantly or more than competitive methods for short-term forecasting in social sciences. A vast majority believe that more complex econometric methods provide at least somewhat more accurate forecasts than can be obtained from less complex econometric methods for forecasting in social sciences.
Although one quibble about whether Armstrong posed most interesting questions in most informative manner, his survey is, to best of my knowledge, only information available and must therefore be accepted as current state of our knowledge of econometrician folklore. Armstrong's description and interpretation of the facts, however, are at best incomplete and at worst misleading. On most basic level, Armstrong's classification of evidence-from significantly more to significantly less accurate-puzzles me.
I know of no test of statistical significance of differences in summary error measures of different sets of ex ante forecasts. (If I am mistaken and such a test exists and has been successfully applied to these data, that central point should be developed more fully.) In addition, richness of broad array of evidence cited is not conveyed (and probably cannot be conveyed in only two dimensions). Extrapolation, for example, is used to encompass a wide range of techniques with varying amounts of economic and statistical sophistication. Time series models, for example, are very different from extrapolations drawn from
Open the resolved locator · 10.1086/296019
Metadata licence: OpenAlex metadata (CC0)