Structural break, unit root, and the causality between government expenditures and revenues

Abstract

Government expenditures and revenues are shown to be trend stationary with a break in the deterministic trend. This is in contrast to the claim that these two series are non-stationary and integrated of order 1. It is therefore inappropriate to first difference data to achieve stationarity. Instead, data is appropriately de-trended using endogenously determined break dates. Tests using the de-trended data show that causality is unidirectional, with expenditures causing revenues.