Nowcasting U.S. state-level CO2 emissions and energy consumption

This paper proposes panel nowcasting methods to obtain timely predictions of CO₂ emissions and energy consumption growth across all U.S. states. This is crucial, not least because of the increasing role of sub-national carbon abatement policies but also due to the very delayed publication of the data. Since the state-level CO₂ data are constructed from energy consumption data, we propose a new panel bridge equation method. We use a mixed frequency set-up where economic data are first used to predict energy consumption growth. This is then used to predict CO₂ emissions growth while allowing for cross-sectional dependence across states using estimated factors. We evaluate the models’ performance using an out-of-sample forecasting study. We find that nowcasts improve when incorporating timely data like electricity consumption relative to a simple benchmark. These gains are sizeable in many states, even around two years before the data are eventually released. In predicting CO₂ emissions growth, nowcast accuracy gains are also notable well before the data release, especially after the current year’s energy consumption data are used in making the prediction.