This paper investigates whether inferences drawn about a population are sensitive to the manner by which those data are obtained. It compares information obtained using participatory appraisal techniques with a survey of households randomly drawn from a locally administered census that had been carefully revised. The community map tends to include household members who do not, in fact, reside in the enumerated locality. By contrast, the revised official census is slightly more likely to exclude household members who actually lived in the surveyed area. Controlling for the survey technique, we find that the revised official census produces higher estimates of average household size and wealth but lower estimates of total village size or wealth, than the community map. Pairwise comparison of the survey techniques, holding the households constant, shows that the community map leads, on average, to higher estimates of household size and lower estimates of wealth.