Assessing the functional relationship between CO2 emissions and economic development using an additive mixed model approach

Researchers have suggested that the relationship between the emission of carbon dioxide per capita and the real gross domestic product per capita follows an inverted-U-shaped (so-called environmental Kuznets) curve. Studies have generally used polynomial regression (quadratic or cubic form) to investigate this relationship. It has been recognised that polynomials are not that flexible and that, by choosing the degree of the polynomial, researchers make a priori assumptions. In this paper, we investigate the environmental Kuznets curve hypothesis using a flexible approach from additive mixed models. Such models are well-suited to handle nonlinear covariate effects flexibly and to simultaneously deal with temporal error structure. We consider the following countries: Australia, Austria, Canada, Denmark, Finland, France, Italy, Spain and Switzerland. Our results show the existence of the classic environmental Kuznets curve for France and Switzerland, and of a nonlinear (increasing) relationship for Australia, Italy and Spain. For Austria, the evidence reveals a weak N-shaped relationship. New nonlinear shapes are found for Finland (inverted-L-shape relationship), Canada (a special case of the inverted-L-shape relationship), and Denmark (M-shape relationship). Our findings are complemented by the calculation of the elasticity of the carbon dioxide emission per capita as a percentage of each level of real gross domestic product per capita.

► We investigate the EKC hypothesis using additive mixed models. ► RLRT test suggests that classic polynomial regression is not an adequate approach. ► Classic shapes found: inverted-U-shape, N-shape, nonlinear (increasing) relationship. ► New shapes found: M-shape, inverted-L-shape.