Despite frequent blackouts and brownouts, extension of the central grid remains the Indian government's preferred strategy for the country's rural electrification policy. This study reports an assessment that compares grid extension with distributed generation (DG) alternatives, based on the subsidies they will necessitate, and costs of service interruptions that are appropriate in the rural Indian context. Using cross-sectional household expenditure data and region fixed-effects models, average household demand is estimated. The price elasticity of demand is found to be in the range of -0.3 to -0.4. Interruption costs are estimated based on the loss of consumer surplus due to reduced consumption of electric lighting energy that results from intermittent power supply. Different grid reliability scenarios are simulated. Despite the inclusion of interruption costs, standalone DG does not appear to be competitive with grid extension at distances of less than 17 km. However, backing up unreliable grid service with local DG plants is attractive when reliability is very poor, even in previously electrified villages. Introduction of energy efficient lighting changes these economics, and the threshold for acceptable grid unreliability significantly reduces. A variety of polices to promote accelerated deployment and the wider adoption of improved end-use efficiency, warrant serious consideration. (C) 2014 Elsevier Ltd. All rights reserved.