We study the implications of state-specific mismatch costs in a model of political agency with (possibly) biased politicians. We stress the importance of the interaction between the direction of the bias and the direction of the asymmetry in mismatch costs, both in determining the likelihood of a pandering equilibrium and its effect on voters' welfare.

Our model builds on the political agency literature. We introduce one important modification, letting the cost of mismatch, i.e. the difference in voters’ payoff between choosing the right and the wrong policy, be state-specific. This allows us to show two sets of results, none of which would apply with symmetric policy mismatch costs.

First, we show that pandering equilibria are more likely to be incentive compatible when the biased politician advocates ``high damage'' policies. Up to the point that, in our model, pandering can be ``benevolent'', i.e. an equilibrium even for purely policy motivated politicians, without office rents. Secondly, we show that voters may prefer to face biased politicians supporting the policy producing, if implemented when not appropriate, more extensive damages.

We discuss the implications of those results for populism, environmental policies and special interest groups and the equilibrium incentives to over- or under-provide lockdowns or other mitigation measures.