Let's get serious about Caprice

Abstract

It is often asserted that economic science is predictive. Our textbooks are full of such claims. Predictive success, in fact, is given the central role in what serves as the conventional methodology of orthodox economics-Friedman’s “methodology of positive economics.” According to this popular view, theories rise or fall based on their records of predictive success (Boland, 1979; Friedman, 1953; dz Stanley, 1985). Wouldn’t the majority of economists agree with the view expressed by Christ (1951) that “the ultimate test of an econometric model . . . comes with checking its predictions” (Armstrong, 1984, p. 19)? With such importance placed on prediction, why do economic models predict so poorly? Specifically, why do naive time series methods typically outperform theoretically well-supported, macroeconomic forecasting models? After decades of distinguished development, what explains the meager record of successful econometric accomoplishments? In spite of the folklore that dismisses this