Welfare Implications of Political Patronage in Electricity

Political involvement in electricity provision is a common occurrence, with politicians favoring their voters. This benefits needy consumers, while hurting provider revenues. I combine confidential data on the universe of billing records from a large state utility in India with data on elections and nighttime lights. I demonstrate new evidence on the extent of, and the mechanisms behind, political patronage in electricity provision. Using a close-election Regression Discontinuity Design (RDD), I find higher satellite-measured electricity consumption, but systematically lower bill collections in constituencies aligned with the ruling party. I show evidence of under-reporting of consumption and also of data manipulation in billing records. Exploiting the rich variation and detail in the data, I estimate net welfare to be unambiguously lower as a result of these actions. The producer losses of about $57 million are more than double the gains to consumers. To measure welfare changes, I develop a novel method that estimates the price elasticity of electricity demand in the presence of data manipulation, combining policy changes as instrumental variables with machine learning for model selection. My estimates are more robust than prior work which relies on aggregated data and on data that is likely to have been manipulated. Consistent with a model of patronage politics, I find that politicians target voters with greater influence and those who benefit more from subsidies.