Revisiting the Narrative Approach of Estimating Fiscal Multipliers

We re-examin the notion of identifying macroeconomic effects using the narrative approach taking as an application the estimation of tax multipliers. We point out to a test for the checking the adequacy of regressing the narrative measure directly on the outcome variable. This test straightforwardly reconciles the reduced form narrative approach with the instrumental variable approach. We consider estimating tax multipliers using six different recently published narrative tax series. The results of our test suggest that none of the considered measures can be regressed directly on output growth. This finding entails questioning the exogeneity of these variables or, in case that exogeneity is maintained, rethinking the appropriate specification. We argue in favor of using an IV procedure and correcting the confidence bands for the weak instrument problem.