What STIRPAT Tells About Effects of Population and Affluence on Environmental Impact?

Abstract

In the literature of STIRPAT application to environmental impacts of population and affluence, controversial results are obtained from different studies. One example is the effects of population size, which is concluded to be unity in some studies (e.g., York et al. 2003) while far from unity in some others (e.g., Shi 2003). Another problem that may arise in the STIRPAT application is multicollinearity, which is encountered in several studies (e.g., Fan et al. 2006; Lin et al. 2009). To solve the multicollinearity problem, various strategies are adopted, e.g., excluding correlated variables, ridge regression, and PLS regression. How can we interpret these results? I offer a consistent framework facilitating to understand these results in the present paper. This is done by deriving a latent model equivalent to STIRPAT, which explicitly specifies the different role of technology (T) in the STIRPAT formulation from that in IPAT accounting model. By the latent model, I conclude that the different specification of STIRPAT formulation can be used to explain the controversial results on environmental impacts of population and affluence. Moreover, methods like ridge regression and PLS regression are plausible only if the common components of high correlated variables have the same explanatory power on the dependent variable in the model.

Keywords: IPAT, STIRPAT, Carbon emissions, Energy consumption, GDP, Population, Ridge regression, PLS regression, ecological elasticity

JEL Classification: Q43, Q56