Confronting collinearity in environmental regression models: evidence from world data

Abstract

Despite the evidence, the correlation between environmental impact factors has mostly been neglected in econometric environmental models or treated with traditional methodologies such as ridge regression, which are recommended when the goal is prediction and the estimated parameters are not interpreted as causal effects. This paper addresses the existing collinearity with alternative methodologies, not only to mitigate the problem mechanically, but also to isolate the effects of the environmental impact factors with the main objective of designing better policies for countries. The methodologies are applied to analyze the CO2 emissions of 114 countries covering the thirteen most recent years with available data, and the results from the empirical and methodological perspectives are compared. The treatment of collinearity with the residualization or raise regression procedures allows the researcher to obtain a global vision of the relationship between the different factors affecting CO2 emissions, thus reaching alternative conclusions to those from traditional methodologies.