Rural Haiti is usually taken as the least likely place on earth for voluntary cooperation to work, which is precisely why White and Runge choose it. They recast the upkeep of a small watershed as a problem of public goods, where a landholder's contribution of labor amounts to a conditional promise to cooperate rather than defect, then test statistically which farmers actually turned out.

The results unsettle a familiar assumption. Insecure land tenure made no difference to participation, and neither wealth nor poverty predicted much; check dams went up on defectors' land as readily as on cooperators', and roughly a third of the labor came from people holding no land in the watershed at all. What did predict cooperation was prior membership in a groupman and the habit of exchanging labor, reciprocity-based institutions already in place. Jointly held ravines, effectively open access, were the one arrangement local assurance could not overcome.