The Regrettable Loss of Mathematical Molding in Econometrics

Abstract

Although most accounts on causality discuss the specific role statistics and theory should have, it is taken for granted that they at least have a role in finding causal structures. The role for mathematics is not so obvious. However, before what is called the Probabilistic Revolution in econometrics, identification of causal relations was not a matter of economic-theoretical and statistical significance alone. Mathematical molding was considered as an essential tool in finding significant causal factors. In the 1940s, mathematical molding disappeared in the changeover from methods to specify causal mechanisms of business cycles to methods to identify economic structures, that is, invariant relationships underlying the workings of an economy. Mathematical molding could fulfill its role in modeling business cycle mechanisms because of the assumed close connection between mathematical representations of the business-cycle phenomenon and those of the explanatory mechanism. When the econometric program shifted its focus from mechanisms explaining phenomena to uncovering structural relationships, direct feedback from the phenomenon to the mechanism was lost and the role of mathematical molding ceased to exist.