An empirical analysis of the determinants of stock market development in the Southern African Development Community (SADC)

Abstract

This study examines the determinants of stock market development in SADC for the period 1990 to 2013. The study applies the modified version of the Calderon-Rossell model to a GMM estimator to assess the major determinants of stock market development. We estimate models based on the Arellano and Bond (1991) difference GMM estimator and the Blundell and Bond (1995) system GMM estimator. The validity of these models is confirmed by the results of the Sargan test and the second order test for serial correlation. The results of the Arellano and Bond estimation show that stock market development is explained by the level of income, stock market liquidity, savings, macroeconomic stability measured through the real interest rate and current inflation and private capital flows. Lagged stock market development is found to be insignificant. Meanwhile, the Blundell and Bond estimation shows that stock market development is explained by the level of income, stock market liquidity savings, macroeconomic stability measured through the real interest rate and private capital flows. Current inflation is found to be insignificant. The study therefore concludes that there is a consensus between the two estimations that financial intermediary development, savings, national income and private capital flows and the real interest rate are significant predictors of stock market development in SADC. While financial intermediary development, savings, capital flows and the real interest rate have a positive influence, the national income has a negative influence. In both cases the results show a negative and significant relationship between stock market development and financial intermediary development for high levels of financial intermediary development. Literature suggests that the system GMM produces more efficient estimates. Specifically, Blundell and Bond (1995) have revealed a weakness with the difference GMM estimator. Their argument is that the lagged levels implied by the difference GMM estimator are rather poor instruments for first differenced variables. This modification therefore gives the system GMM estimates an edge over the difference GMM estimates. Hence, in line with this reasoning the current study adopts the estimates from the system GMM estimator.