Reconciling VAR-based and Narrative Measures of the Tax-Multiplier

The currently available empirical evidence shows remarkable differences between various estimates of the effects on U.S. output of an exogenous shift in Federal tax liabilities. Shocks identified via the narrative method imply a multiplier of about three over an horizon of three years. Tax shocks identified in fiscal VAR models deliver a much smaller multiplier of about one. Is this heterogeneity real, or is it simply the result of different approaches to the identification of exogenous shifts in taxes? Or of different specifications of the empirical model used to estimate the tax multiplier? In this paper we reconcile this apparently contradictory evidence by showing that the large multiplier obtained via the narrative identification methods is generated by the choice of a limited information approach in their estimation and not by the different nature of the shocks. Using the shocks identified by a Narrative methods in a multivariate dynamic model delivers estimates of the tax multiplier very much in line with those obtained in the traditional fiscal VAR approach.