We study loss aversion in majority voting. First, we show a status quo
bias. Second, loss aversion implies a moderating effect. Third, in a dynamic setting, the effect of loss aversion diminishes with the length of the planning horizon of voters; however, in the presence of a projection bias, majorities are partially unable to understand how fast they will adapt. Fourth, in a stochastic environment, loss aversion yields a significant distaste for risk, but also a smaller attachment to the status quo. The application of these results to a model of redistribution leads to empirically plausible implications.