Estimating a Production Function for Iran with Emphasis on Energy and Expenditure on Research and Development: An Application of Genetic Algorithm Method

This paper aims to estimate various production functions, with emphasis on energy and investment in R& D, in Iran over the period 1979-2010. Following estimation of different production functions including CES (Constant Elasticity of Substitution), GPF (Generalized Production Function), Cobb-Douglas, Transcendental, Translog, and GLPF (Generalized Linear Production Function), a proper production function is selected. The functions are mainly non-linear and their estimation requires large sample sizes. The conventional econometric techniques estimate regression parameters through minimizing residual sum of squares (RSS). However, this approach is less efficient than minimization the Least Absolute Deviation (LAD). Moreover, the conventional nonlinear techniques cannot minimize absolute deviation of errors from their expected values. In order to overcome this problem, we use Genetic Algorithm (GA) method with LAD to estimate six non-linear production functions. The results suggest that the Translog function is the most appropriate production function for the Iranian economy. According to our findings, a 10 percent increase in energy consumption, raises the output by 7.3 percent. However, a 10 percent increase in R&D expenditure only increases the output by 2.6 percent. Finally, the results show that the production function in Iran exhibits increasing return to scale after the end of Iran-Iraq war. Thus, it seems constructing growth models for Iran by assuming constant returns to scale production technology needs to be reexamined.