Real‐time estimates of potential output are used for the calculation of the cyclically adjusted budget balance, one of the main indicators in the assessment of the fiscal performance of EU member states. The estimation of potential output involves a decomposition of actual output into a cyclical and a structural component based on arbitrary assumptions about the statistical properties of the two unobserved components. With a very high degree of smoothing, variations in GDP are mostly taken to be temporary, as are the ensuing changes in the budget deficit. Conversely, with a low degree of smoothing, variations in GDP are mostly taken to be permanent, leading to different policy conclusions. Our paper examines whether and how different potential output estimates would have supported different decisions in the EU budgetary surveillance in terms of both timing and substance. The results show that only a very high degree of smoothing of potential output would significantly reduce the reliability of the surveillance indicators. We conclude that a higher degree of smoothing compared with current practice would not be harmful for EU fiscal surveillance, while it could contribute to more cautious policies by signalling larger and longer periods of economic ‘good times’.