Why do individuals support the public policies they do? We argue that individuals can have quite sophisticated policy preferences and that not correctly modeling those preferences can lead to critically misspecified empirical models. To substantiate this position we derive and test a decision-theoretic model that relies upon three critical assumptions: (1) policies affect the provision of multiple goods about which individuals care; (2) individuals have diminishing returns to scale in those goods; and (3) preferences over at least some subset of those goods are correlated. Using this model, we demonstrate that arbitrarily small secondary policy effects can confound predictions over primary policy effects. Thus, not considering even arbitrarily small policy effects can cause one to conclude that evidence is consistent with one's theory when in fact it is inconsistent or vice versa. Testing this theory on support for forming a European common defense, we find evidence consistent with our model.