Most EU member states will adopt fiscal rules that refer to cyclically-adjusted borrowing limits. Under the standard cyclical adjustment procedure, trend increases in public debt based on cyclical components are prevented if the real-time output gaps used to calculate cyclical components balance over time. We analyse real-time output gaps for EU-15 countries over the 1996-2011 period as estimated by the EU, the IMF and the OECD. Compared to each institution's final estimate, we find that real-time output gaps in our sample period are negatively biased. This bias is observed (i) irrespective of the source of the data, (ii) in all real-time vintages, (iii) basically across the entire cross-section of countries. The magnitude of the bias is considerable: on average, real-time cyclical components as a percentage of GDP are biased downwards by about 0.5 percentage point per year. Our results suggest that fiscal rules should incorporate ex-post checks of the unbiasedness of the cyclical components used within the rule. Potential biases would then decrease or increase future borrowing limits.