Diffuse support for democracy, as captured in mass surveys, tends to be treated as impervious to regime performance. Such a finding is often presented as confirmation of the basic distinction between ‘diffuse’ and ‘specific’ support as proposed by David Easton. This study argues that this line of argument stems from an incomplete reading of important aspects of Easton's theorisation about the relationship between system outputs and diffuse support. Using multilevel models, evidence from more than 100 surveys in close to 80 countries, and different measures of democratic support, it is shown that government effectiveness is the strongest macro-level predictor of such support. In democratic regimes, government effectiveness, understood as the quality of policy-making formulation and implementation, is linked to higher levels of support for democracy. Furthermore, in non-democracies, effectiveness and support for democracy are, under some model specifications, negatively related.