Comparing the ANN and Linear Regression in Estimation of the Growth Model (The Case of MENA)

Economic convergence is one of the important topics of new macroeconomics. It refers to tendency of income per capita of countries (regions) to converge to their steady-state value. There are two kinds of convergence: conditional and absolute convergence. This paper examines income convergence between 22 MENA countries during the period of 1970-2003 by using the neoclassical growth model of Barro- Salla-i-Martin for both kinds of convergence. Non-linearity of the underlying relationships, the restrictiveness of assumptions of functional forms and econometric problems in the estimation and application of theoretical models advocate for the use of Artificial Neural Networks (ANN) algorithms. We show that by changing the quantitative tools of analysis and using ANN, the results become more precise. Results show that absolute convergence and conditional convergence are significant but the rate of convergence is low.