Beyond the Stars: a New Method for Assessing the Economic Importance of Variables in Regressions

Economists lack a systematic method to assess the economic importance of effects in regressions. In this article, I use experimental evidence to show that for a large majority of economists, the economic importance of an explanatory variable refers to its contribution to deviations in the level of the dependent variable. Existing statistics, such as standardized beta coefficients and the partial or semi-partial r2 and r, are only imperfect measures of the economic importance of explanatory variables: these statistics do not match with economists' common understanding of economic importance and are difficult to interpret. I therefore develop a new method, which consists in rescaling standardized beta coefficients such as to obtain the percentage contribution of each explanatory variable to deviations in the dependent variable. As an illustration, the method is applied to the study of the causes of long-run economic development.