Association for Public Policy Analysis & Management

Environmental policy makers need to assess the negative externalities generated by activities such as electricity generation, industrial process, and driving. A key challenge to quantify the negative externalities is to understand the feedback and interactions among the externalities, human behaviors and technologies (Sun et al., 2017). This paper demonstrates how such feedback can change the assessment of negative externalities in the context of particulate matter (PM) air pollution.

Using customer-level daily electricity data in Arizona, we show that the interactions of consumer avoidance behaviors, clean technology performance, and particulate air pollution generate extra-damage. Specifically, we find that when average daily PM10 concentration increases by one standard deviation of 30 ug/m3, the daily electricity consumption of residential households increases by 15% (because consumers stay indoors more often) and electricity generated by solar panels reduces by 21% (because solar irradiance is impacted by PM); when daily PM2.5 concentration increases by one standard deviation of 5ug/m3, the daily electricity consumption increases by 14%; on a dust storm day, electricity consumption goes up by 4% and electricity generated by solar panels reduces by 2%. Based on marginal damage factors of electricity supply, we estimate that the co-damages from a one standard deviation increase in both PM10 and PM2.5 concentrations are $0.29/customer/day. Our data is obtained from an electric utility company in Arizona. The number of total residential consumers of the utility company is about 690,200, so the total co-damages of this single utility company’s service territory is more than $0.2 million per day.

The consumer-level dataset includes daily electricity consumption data for 4,313 residential consumers as well as daily electricity generated by each of the 300 residential distributed solar consumers. We use a fixed effects panel instrumental variable (IV) approach, in which thermal inversion and wind direction are used as the IVs for PM pollution. In addition, we conduct a regression discontinuity in time to analyze the impact of dust storm. Various confounding factors such as temperature, solar irradiance, electricity price, and a rich set of fixed effects are controlled for in the regression.

In terms of contributions to existing literature, this paper provides the first empirical evidence about the impact of consumers’ air pollution avoidance behavior on electricity consumption. In addition, there is a lack of studies using large sample of actual distributed-solar panel generation data to evaluate empirically the impact of air pollution on solar electricity generation (Bergin et al., 2017) and our study fills this gap.

Our estimated co-damages are both statistically and economically significant, providing several key policy implications. First, most existing studies on air pollution damage do not explicitly consider these co-damages (Ebenstein et al., 2017). Policy makers should be aware of the magnitude of these co-damages. Second, our findings show that to reduce the associated co-damages, cleaner fuels for electricity generation is essential. Third, our results imply that policy makers should further encourage the efforts to reduce air pollution. Lastly, our results are relevant to other regions especially for highly polluted regions such as China and India.