Public policies typically involve trade-offs across multiple goals, such as efficiency, equity, freedom, security, and sustainability. While existing research documents how specific beliefs or values correlate with policy support, much less is known about the decision processes by which individuals aggregate competing considerations into an overall judgment. This paper studies how citizens evaluate policies when gains along some dimensions come with losses along others. I develop a model of policy evaluation that allows for both compensatory reasoning, where benefits can offset costs across goals, and non-compensatory reasoning, where support is constrained by minimum standards on particular goals. To test the model, I propose a vignette experiment in which respondents evaluate hypothetical policies described solely by their effects on core policy goals, with information revealed sequentially to generate explicit trade-offs. This design enables the identification of discontinuities and goal-specific constraints in policy support. I then link experimentally identified evaluation rules to beliefs and support for real-world policies across major policy domains. The analysis assesses whether accounting for evaluation processes improves the prediction of policy support beyond beliefs alone. The project contributes to the study of policy preferences by shifting attention from isolated attitudes to the mechanisms through which individuals resolve trade-offs between competing policy goals