The challenges faced in calibrating poverty and welfare measures to objective data have long been recognized. Until recently, most economists have resisted a seemingly obvious solution, namely to ask people themselves: "Do you feel poor?" The paper studies the case for and against this approach. It is argued that, while one would not want to use self-assessments as welfare metrics in their own right, there is scope for using such data to help calibrate multidimensional measures. Indeed, the idea of a "social subjective poverty line" (below which people tend to think they are poor, but above which they do not) is arguably the most conceptually appealing way of defining poverty. However, the paper points to a number of concerns that have received insufficient attention, including the choice of covariates, survey design issues, measurement errors, frame-of-reference effects, and latent heterogeneity in personality traits and personal tradeoffs. Directions for future research are identified.