This paper reports the result of an experiment designed to understand the interaction of institutions and trust. The experiment had two treatments. Both shared a first phase, which determined levels of trust and trustworthiness of the participants. In a second phase we made it possible for participants to create institutions which, though costly, enforced a minimum respect for trust. In the first treatment, insurance is purchased, while in the second, participants vote for it. We find that the participants in the experiments purchase, or vote, for insurance quite often, and the decision depends on the level of trustworthiness of the group to which one belongs, given that demand is higher in the low trustworthiness group. Demand for institutions is the same when purchasing them or when voting them, but only if senders are the only subjects whose vote counts. If receivers can vote, the demand for institutions is lower. Risk aversion or cognition measures do not explain demand for insurance, trust or trust-worthiness; but social preferences measures are very correlated to these decisions. Interestingly, the possibility of choosing institutions seems to crowd-out civic behaviour, given that individuals in the high trustworthiness group return less in the second part of the experiment when we introduce institutions.