At least since the work of Kydland and Prescott (1977), it is acknowledged that the ability for policy makers to commit to future policies often is of key importance for outomces and welfare. An example is capital income taxation, where the commitment solution typically involve zero capital income taxes after some initial periods, while no commitment may imply empirically much too high taxes. A more realistic intermediate outcome can be supported by trigger strategies where a deviation from a "good" equilibrium is punished by a possibly infinite revertion to a "bad" Nash-equilibrium. In a political economy setting, the degree of coordination between voters of different generations required to sustain such an equilibrium is large, arguably unreasonably large. We propose loss-aversion as an alternative explanation for how a commitment-like equilibrium can arise. We set up a politico-economic OLG-model where individuals make investments and dynamically form reference points for future consumption, around which they are loss-averse. Without loss-aversion, the only Markov equilibrium involves 100 % taxation of any investments and trigger strategies are required to sustain lower taxes and positive investments. With loss-aversion, we find a Markov equilibrium with positive investments. In contrast to the case of trigger strategies, this equilibrium is renegotiation proof and independent of discounting, surviving also for arbitrarily high rates of discounting.