Robustness procedures in economic growth regression models

A central question for empirical economics, particularly economic growth, is which explanatory variables to include and exclude in the regressions. This paper aims to identify variables strongly correlated with provincial income growth in the Philippines by applying robustness procedures in determining which variables are strongly correlated with income growth. The extreme bound analysis and the Bayesian averaging of classical estimates were applied to 15 determinants of income growth from a data set consisting of 74 Philippine provinces for the period 1985-2003 to test which among the explanatory variables are strongly correlated to growth. The tests show that among the 15 variables, five variables stand out as being robust. The log of initial income, the Autonomous Region in Muslim Mindanao (ARMM) indicator, the expenditure Gini and its square, and the proportion of young dependents are all considered as strongly correlated to growth.