AbstractWe provide a theorem that lists the necessary and sufficient dominance conditions for the poverty comparisons of the bivariate distributions function when considering an asymmetric treatment of attributes. The normative justification for an asymmetric treatment is based on the compensation principle proposed by Muller and Trannoy (2012), under which it makes sense to use one attribute to compensate another. The formulation results in a generalization of the needs approach in poverty analysis proposed by Atkinson (1992). The dominance conditions we found lie between those obtained by Bourguignon and Chakravarty (2002) and Duclos, Sahn, and Younger (2006a) when attributes are symmetric and those obtained within the needs framework by Atkinson (1992) and Jenkins and Lambert (1993) when attributes are asymmetric, but one is of discrete nature.