Pollution is a common byproduct of economic activity. Although policymakers should account for both the benefits and the negative externalities of polluting activities, it is difficult to identify those who are harmed and those who benefit from them. To overcome this challenge, our paper uses a novel dataset on the mid-20th century expansion of the U.S. power grid to study the costs and the benefits of coal-fired electricity generation. The empirical analysis exploits the timing of coal-fired power plant openings and annual variation in plant-level coal consumption from 1938 to 1962, when emissions were virtually unregulated. Pollution from the burning of coal for electricity generation is shown to have quantitatively important and nonlinear effects on county-level infant mortality rates. By 1962, it was responsible for 3,500 infant deaths per year, over one death per thousand live births. These effects are even larger at lower levels of coal consumption. We also find evidence of clear tradeoffs associated with coal-fired electricity generation. For counties with low access to electricity in the baseline, increases in local power plant coal consumption reduced infant mortality and increased housing values and rental prices. For counties with near universal access to electricity in the baseline, increases in coal consumption by power plants led to higher infant mortality rates, and lower housing values and rental prices. These results highlight the importance of considering both the costs and benefits of polluting activities, and suggest that demand for policy intervention may emerge only when the negative externalities are significantly larger than the perceived benefits.