Forerunning programmes of the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA), which were designed as poverty elimination programmes, took notice of geographical pockets of poverty and incorporated formula-based fund allocation mechanisms to poorer states and regions. The MGNREGA programme, in contrast, used a right-based ‘self-selection’ approach— relying on the initiative of households’ demand-driven strengths—to allocate need-based resources to states and regions within states. This article examines how well the demand-driven, right-based programme with self-selection allocated resources to states and regions according to their respective needs, and to what extent the benefits reached the poverty pockets and catered to the poorest, weakest and neediest households. We find that adequate resources did not reach the poorest states and regions, substantial numbers of poor households remained outside the programme or were deemed underserved, and there was a pronounced programme capture by elite states. The article explores causes and consequences of capacity limitations and low absorption pulls among states, and points to policy implications and ways forward.