Abstract-Previous research on dynamic pricing finds evidence of social benefits through peak-shaving channels; high peak periods reduce the maximum grid load and allow utilities to defer construction of costly peaker plants. However, the literature ignores a second channel through which dynamic pricing influences behavior: reallocation of consumption within a single day from high price periods to low price periods called "load shifting". This paper estimates the hourly marginal emission- and cost-changes that result from changes in the daily profile of household demand using a randomized control trial of household consumption under dynamic pricing as well as system wide observations of power generating units and grid level demand. We find that load shifting reduces hourly emissions of pollutants SO2, CO2, and NOx, as well as reducing production costs for utilities. Averted production costs per household are very small, but tightly estimated.

Bio-Matthew Harding is an Econometrician and Data Scientist who develops machine learning and artificial intelligence techniques to answer Big Data questions related to individual consumption and investment decisions in areas such as health, energy, and finance. He is an Associate Professor of Economics and Statistics and UC Irvine. He holds a PhD in Economics from MIT and an MPhil in Economics from Oxford University. He directs the Deep Data Lab which conducts research into cutting edge econometric methods for the analysis of “deep data”, large and information-rich data sets derived from many seemingly unrelated sources to provide novel economic insights. He works at the intersection of machine learning and statistics with a focus on deep learning and causal inference. Professor Harding advised a number of companies and agencies on economics and data science problems, including Apple, US Commodity Futures Commission, World Bank, Electric Power Research Institute (EPRI), and the Department of Justice.