Making policy choices is unavoidable. The standard economic approach to guiding policy consists of two steps: estimate policy effects on individuals’ utility, and then use a social choice rule that aggregates across individuals to generate a policy recommendation. Traditionally, the policy-effect estimates come from market data. However, market data are insufficient in some contexts involving externalities, public goods, and non-market goods, and may not reliably reveal preferences in contexts where people are uninformed or make systematic mistakes. For these reasons, economists have increasingly been exploring survey-based approaches to estimating policy effects on utility. For our purposes, such surveys have two key features: they are unincentivized, and they generally work best for eliciting local preferences at the status quo (because respondents find it harder to introspect about how they would feel in counterfactual, unfamiliar situations). In this paper, we propose a social choice rule for aggregating local ordinal preferences elicited from surveys into a local policy improvement, while reducing the incentive for survey respondents to misreport their preferences. We have in mind three examples of survey-based preference measurement to which our method could be applied. First, contingent valuation surveys often elicit respondents’ marginal rates of substitution between different policies or between a policy and money (i.e., willingness to pay). Second, much research on subjective well-being (SWB) treats the response to a SWB survey question, typically regarding happiness or life satisfaction, as a proxy for utility and estimates how policies affect it. Unlike contingent valuation or policy referendums, this approach sidesteps respondents’ policy misconceptions or lack of information; and relative to traditional market-based indicators such as GDP, SWB data may capture a broader range of aspects of well-being (e.g., Stiglitz, Sen, and Fitoussi, 2009). However, evidence indicates that people make choices that systematically deviate from what they believe would maximize their responses to commonly-used SWB measures (Benjamin, Heffetz, Kimball, and Rees-Jones, 2012, 2013), suggesting that a single-question SWB measure is inadequate as a utility proxy. The primary application we have in mind is estimating the effects of policy on responses to a range of SWB questions that capture distinct aspects of well-being, including, for example, own and family happiness, health, security, and freedoms. The effect of policy on utility can be estimated as a weighted average of the effect of policy on these responses, with the weights derived from a separate survey that elicits individuals’ marginal rates of substitution across aspects of well-being.1 The theory and methods for combining responses to multiple SWB questions into a utility proxy are developed in a companion paper (Benjamin, Heffetz, Kimball, and Szembrot, 2012).