We construct and derive the properties of an estimator of welfarewhich takes advantage of the detailed nature of information about living standards available in small household surveys and the comprehensive coverage of a census. By combining the strengths of each, our estimator can be used at a disaggregated level. It has a clear interpretation; it can be expanded in a consistent way to any welfare measure;and can be assessed for reliability usingstandard statistical theory. Because unit record census data present some computational hurdles, we explore simulation and numerical integration approaches, as well as the use of distributional approxima-tions. For non-separable inequality measures we derive specific formulas to allow the use of`shortcut' computational methods. Using data from Ecuador we obtain estimates of welfare measures which are very reliable for populations of 5,000 house-holds, 'town's', and in many cases for those as small as 500. We provide simple illustrations of their use. In the longer run, such estimates open up the possibility of estimating and testing, at a more convincing intra-country level, the many recent models relating welfare distributions to growth and a variety of socioeconomic and political outcomes.