When agents, due to incomplete preferences, fail to have well-defined marginal valuations for goods, a great many government policies will maximize social welfare or achieve efficiency. Welfare economics then becomes useless as a practical guide to decision-making. For example, the values agents assign to increases in a public good will be discretely smaller than the values they assign to decreases. For society as a whole, a large valuation gap will form and a wide range of quantities of the public good will therefore qualify as optimal. Applied welfare economics and cost–benefit analysis bypass this obstacle by paying attention only to agents’ smallest valuations, thus slanting policymaking against public goods. The multiplicity of preferences that agents view as reasonable also neuters Pareto efficiency as a policy guide: virtually any policy change is likely to harm some of the preferences agents deem reasonable.