The study provides new evidence on the sustainability debate concerning the US budget balance. Existing studies are often only able to reject non-cointegration between government expenditure and revenue once structural breaks or regime changes are allowed for. Using a testing procedure advocated by Bierens, this study provides a different perspective on this issue. Government expenditure and revenue are both found to be stationary around a nonlinear deterministic trend, and they are co-trended insofar as they share a common nonlinear deterministic trend.