In the first of this series of three papers we criticised the consumer price indexes based on unit values calculated from the unit records of the NSS Consumer Expenditure Surveys (NSS CES) which have been used to calculated new poverty lines for Indian states by Deaton and Tarrozi, 1999, Deaton, 2003a. This second paper examines the calculation of poverty lines using these Unit Value Consumer Price Indexes (UV CPIs). We suggest that using UV CPIs to account for temporal change and spatial variation in prices in the production of poverty lines does not appear to be a good strategy. Here we point out what we see as a flaw in the method used to calculate Poverty Lines for different states and sectors from a single base Poverty Lines. Further, we argue that neither UV nor official price indexes represent true cost of living indexes because they ignore “environmental” variables that differ between domains and affect the transformation of consumption into well-being. This results in problems of comparability suggesting that the PLs that can be calculated from household expenditure surveys such as the NSS CES do not correspond to the same level of well-being in different domains and thus do not generate poverty measures that compare differences in ill-being rather than differences in the yardstick by which well-being is assessed. A thorough overhaul of poverty line calculations is required, but welfare comparable poverty lines cannot be based on normative calorie requirements. In the third paper in this series we give our “best” CPIs and those that arise from “robust” methods of poverty comparison using stochastic dominance techniques. Unfortunately, on theoretical grounds neither our poverty calculations nor the use of robust methods in their usual form overcome the problems identified here, and there we give evidence in support for this contention in that other indicators of well-being are not well correlated with these poverty counts and comparisons.