Looking ahead: Forecasting total energy carbon dioxide emissions

In recent years, the international community has been increasing its efforts to reduce the human footprint on air pollution and global warming. Total carbon dioxide (CO2) releases are a crucial component of global greenhouse gas emissions, and as such, they are closely monitored at the national and supranational levels. This study presents different models to forecast energy CO2 emissions for the US in the period 1972-2021, using quarterly observations. In an in-sample and out-of-sample analysis, the study assesses the accuracy of thirteen forecasting models (and their combinations), considering an extensive set of potential predictors (more than 260) that include macroeconomic, nature-related factors and different survey data and compares them to traditional benchmarks. To reduce the high-dimensionality of the potential predictors, the study uses a new class of factor models in addition to the classical principal component analysis. The results show that economic variables, market sentiment and nature-related indicators, especially drought and Antarctic wind indicators, help forecast short/medium-term CO2 emissions. In addition, some combinations of models tend to improve out-of-sample predictions.

Automated summary

In recent years, efforts to reduce air pollution and global warming have been increasing globally. This study focuses on forecasting energy CO2 emissions for the US from 1972 to 2021 using different models and extensive predictors. The results show that economic variables, market sentiment, and nature-related indicators help improve short to medium-term CO2 emission predictions, with certain model combinations performing better in out-of-sample predictions.