The objective of this paper is to propose an early warning system that can predict the likelihood of the occurrence of financial stress events within a given period of time. To achieve this goal, the signal extraction approach proposed by Kaminsky, Lizondo and Reinhart (1998) is used to monitor the evolution of a number of economic indicators that tend to exhibit an unusual behaviour in the periods preceding a financial stress event. Based on the individual indicators, we propose three different composite indicators, the summed composite indicator, the extreme composite indicator and the weighted composite indicator. In-sample forecasting results indicate that the three composite indicators are useful tools for predicting financial stress events. The out-of-sample forecasting results suggest that for most countries, including Canada, the weighted composite indicator performs better than the two others across all criteria considered.