Recent findings about the electoral cost of welfare state retrenchment challenge the view of the ‘New Politics’ literature that cutting welfare state entitlements is electorally risky. In fact, there seems to be no systematic punishment for governments retrenching the welfare state. At the same time, however, studies show that governments use numerous blame avoidance strategies when cutting welfare. Reflecting on this apparent contradiction we put forward two points. First, qualitative evidence from interviews with political leaders suggests that it is not the actual risk of being punished that entices politicians to use blame avoidance but the perception of this risk. This explains why blame avoidance strategies are widely used. Second, the existing studies showing that governments are not systematically punished for cutting the welfare state suffer from the lack of control for blame avoidance effects. We show that an experimental design could be a remedy for this problem.