Why do high-income earners support higher levels of income redistribution in some countries than in others? I argue that differences in the social insurance design have consequences for fairness considerations and that this matters for preference formation. Flat-rate systems provide social benefits in equal amounts to everyone in need, while earnings-related systems provide benefits in relation to previous earnings. In the case of income loss, earnings-related systems maintain unfair income differences, while flat-rate systems equalize unfair income differences between the rich and the poor. Cross-national patterns reveal that support for redistribution among the rich is higher in income-maintaining welfare states. For a strict test of my fairness argument, I conduct a laboratory experiment and show that participants reduce inequality more if given endowment differences are maintained in the case of loss.