The in-principle inconclusiveness of causal evidence in macroeconomics

Abstract

The paper analyzes the methods that macroeconomists can use to provide evidence in support of causal hypotheses: the instrumental variable (IV) method and econometric causality tests. It argues that the evidence that macroeconomists provide when using these methods is in principle too inconclusive to support the hypothesis that X directly type-level causes Y, where X and Y stand for macroeconomic aggregates like the real interest rate and aggregate demand. The evidence provided by the IV method is too inconclusive because it derives from conditions requiring that there be no confounders of I and X and X and Y (where I is an instrumental or intervention variable that typelevel causes X), and because in macroeconomics, confounders that cannot be controlled for or measured are likely to be present. The evidence provided by econometric causality tests is too inconclusive because they can be shown to rely on the conditions of the IV method at least tacitly.