In several recent studies unit root methods have been used in detection of financial bubbles in asset prices. The basic idea is that fundamental changes in the autocorrelation structure of rel-evant time series imply the presence of a rational price bubble. We provide cross-country evi-dence for performance of unit-root-based early warning systems in ex-ante prediction of finan-cial crises in 15 EU countries over the past three decades. We find especially high performance for time series that are explicitly related to debt, which issue signals a few years in advance of a crisis. Combining signals from multiple time series further improves the predictions. Our results suggest an early warning tool based on unit root methods provides a valuable accessory in financial stability supervision.