Poverty, Income Distribution and CGE Modeling: Does the Functional Form of Distribution Matter?

In this paper, we provide an overview of approaches used to model income distribution and poverty in CGE models. CGE models have started to use income distribution functional forms such as the lognormal, Pareto, beta distribution and Kernel non-parametric methods to apply GFT poverty indices. None of the authors of these papers have gone into much detail to justify the use of one method or functional form over the other, within the context of this type of work. Extensive literature exists on the choice of functional forms to estimate income distribution; however it has not been utilized in the CGE context. Given the fact that the desegregation of groups of households can be important in CGE analysis and the fact that the impact on income of policy simulations are often small in CGE models, we investigate the importance of othe choice of the functional form used to estimate the income distribution of groups of households. We compare six functional forms with parametric estimation and on a non-parametric method. Results show that no single form is more appropriate in all cases or groups of households. The characteristics of samples and subgroups play an important role and the choice shoudl be guided by the best fitting distribution.