Nonparametric and semiparametric generalized panel data analysis of convergence and growth

The purpose of this dissertation is to devise the new econometric technique for generalized panel data (where there is more than one cross sectional component of the panel data). The new estimators for generalized panel data for fixed and random effects and for nested (when one cross sectional component is nested in the other) and nonnested cases (when the cross sectional components are independent of each other) are developed in parametric, semiparametric and nonparametric frameworks. The asymptotic properties of these new estimators are also analyzed. Another objective is to implement the new estimators and to examine the issue of convergence of economic growth in different sectors for OECD countries and a combined set of developed and underdeveloped countries. Generalized panel data take care of the sectoral linkages as well as country sector and time specific heterogeneity. Nonparametric and semiparametric frameworks take care of the misspecification bias problem. Also since nonparametric is a local point-wise estimation method, it enables us to obtain individual sector-wise convergence results after taking care of the sectoral linkages. For the OECD countries the lack of convergence is observed at the sectoral level but convergence is dominant at the aggregate level. This is in consonance with the trade induced convergence theory. Results for the combined set of developed and underdeveloped countries show lack of convergence in the sectoral as well as aggregate growth rate. It also shows that there is a threshold level of income beyond which countries start converging.