Electronic International Standard Serial Number (EISSN)
1435-8921
abstract
In this paper, we analyse the relationship between the rate of growth of CO2 emissions and economic activity in Spain from 1964 to 2020. We explain CO2 emissions by fitting a structural regression model with selected indicators of economic activity augmented with dynamic common factors extracted from a large macroeconomic database, as explanatory variables. The variables to include in the regression are selected using Machine Learning procedures while we use alternative supervised and non-supervised procedures to extract the factors. We find that, regardless of the procedure used for variable selection, private consumption and maritime transportation have the highest explanatory power for the rate of growth of emissions. We also show that the way the common factors are extracted is crucial to exploit their information content. The common factors extracted by partial least squares add valuable information on top of that of the selected individual indicators, while they are not significant when extracted by two-step-Kalman filter (2SKF).
Classification
subjects
Economics
Environment
Statistics
keywords
dynamic factor models; lasso; partial least squares; variable selection