A panel data set covering 145 countries between 1960 and 2010 has been investigated closely by using models of parameter heterogeneity. The Functional Coefficient Model (FCM) introduced by Cai, Fan and Yao (2000) allows estimated parameters of growth determinants to vary as functions of one or two status variables. As a status variable, coefficients depend on the level of development, measured by initial per capita GDP. In a two-dimensional setting, time is used as an additional status variable. At first, the analysis is restricted to bivariate relationships between growth and only one of its determinants, dependent on one or both status variables in a local estimation. Afterwards, the well-known Solow (1956) model serves as a core setting of control variables, while functional dependence of additional explanatory variables is investigated. While some constraints of this modeling approach have to be kept in mind, functional specifications are a promising tool to investigate growth relationships, as well as their robustness and sensitivity. Finally, a simple derivation of FCM called local mean values provides a suitable way to visualize macroeconomic or demographic development patterns in a descriptive diagram.
economic growth cross-country growth regression functional coefficient model varying parameter parameter heterogeneity kernel regression panel data local mean value