This dissertation develops a new approach that enables policy-makers to analyze welfare gains from improvements in the quality of infrastructure services in developing countries where data are limited and supply is subject to interruptions. With the tight budgetary constraints that usually exist, it is important to be able to prioritize public sector investments on the basis of expected benefits. However, policy analysts are rarely able to measure the benefits of improving the quality of infrastructure services, even though they may yield large welfare benefits. The most frequently cited reason for failures to carry out such welfare analysis is the scarcity of data on service quality. The main contribution of this dissertation is a new model of welfare evaluation of changes in the quality of infrastructure services. This model is estimated using the existing data from household energy surveys or data from the energy sections of multi-purpose household surveys. Potential applications of this model range from ex-ante reform evaluation to ex-post monitoring of policy outcomes, which makes this approach a useful contribution to policy analysis and to the literature on welfare evaluation of quality changes in infrastructure. An application of the proposed model in the former Soviet Republic of Azerbaijan demonstrates how this approach can be used in welfare assessment of energy sector reforms. The planned reforms in Azerbaijan include a set of measures that will result in a significant improvement in supply reliability, accompanied by a significant increase in the prices of energy services so that they reach the cost recovery level. Currently, households in rural areas receive electricity and gas for only a few hours a day because of a severe deterioration of the energy infrastructure following the collapse of the Soviet Union. The reforms that have recently been initiated will have far-reaching poverty and distributional consequences for the country as they result in an improvement in supply reliability and an increase in energy prices. The new model of intermittent supply developed in this dissertation is based on the household production function approach and draws on previous research in the energy reliability literature. Since modern energy sources (network gas and electricity) in Azerbaijan are cleaner and cheaper than the traditional fuels (fuel wood, etc.), households choose modern fuels whenever they are available. During outages, they rely on traditional fuels. Theoretical welfare measures are derived from a system of fuel demands that takes into account the intermittent availability of energy sources. The model is estimated with the data from the Azerbaijan Household Energy Survey, implemented by the World Bank in December 2003/January 2004. This survey includes an innovative contingent behavior module in which the respondents were asked about their energy consumption patterns in specified reform scenarios. Estimation results strongly indicate that households in the areas with poor supply quality have a high willingness to pay for reliability improvements. However, a relatively small group of households may incur substantial welfare losses from an electricity price increase even when it is combined with a partial reliability improvement. Unlike an earlier assessment of the same reforms in Azerbaijan, analysis in this dissertation clearly shows that targeted investments in improving service reliability may be the best way to mitigate adverse welfare consequences of electricity price increases. Hence, policymakers should focus their attention on ensuring that quality improvements are a central component of power sector reforms. Survey evidence also shows that, although households may incur sizable welfare losses from indoor air pollution when they rely on traditional fuels, they do not recognize indoor air pollution as a factor contributing to the high incidence of respiratory illness among fuel wood users. Therefore, benefits may be greater if policy interventions that improve the reliability of modern energy sources are combined with an information campaign about the adverse health effects of fuel wood use.