Log-growth rates of CO2: An empirical analysis

We study the parametric distribution of log-growth rates of CO2 and CO2 per capita emissions for 207 countries and territories taking data from 1994 to 2010. We define the log-growth rates for different duration periods, from one year apart to fifteen years apart. The considered probability distributions have been the following: the normal (N), the asymmetric double Laplace normal (adLN), the exponential tails normal (ETN) and a mixture of two normal (2N) or three normal (3N) distributions. The main result is that the best one is different depending on the period considered, in such a way that there is not a systematically dominant distribution. Thus, the behavior may change from one year to the next one, and possibly this is influenced by policy measures such as the Kyoto protocol or the Clean Development Mechanism. Moreover, a policy measure that can be derived from this paper is that some countries can still reduce their emissions of CO2 compared with others, as seen by the non-uniformity of the preferred probability distribution for each period. We also model a stochastic differential equation whose associated Fokker–Planck equation has as a solution the observed time-dependent probability density function.