Journal of Statistical and Econometric Methods

A method for clustering panel data based on parameter homogeneity

• [ Download ]
• Times downloaded: 9386

• Abstract

  Panel data models assume that parameters are common to each subject, that assumption is not satisfied in many cases. The slope heterogeneity problem may be solved by obtaining groups where the slope parameters are heterogeneous across groups but homogeneous within groups, followed by panel data theory within each group. In this paper, an algorithm to determine clusters of subjects is discussed; the clustering is achieved by checking whether confidence intervals from different subjects overlap or not. The number of groups is determined based on the data variability. The clusters are useful by themselves to analyze the similar behavior of subjects. Monte Carlo simulations were performed to examine the properties of the methodology considered. Finally, clusters of countries with similar GDP per capita trend were obtained.

  Mathematics Subject Classification: 62F03, 62F25, 62H30, 91G70
  Keywords: Panel data models, Clustering of subjects, Parameter homogeneity test, Confidence intervals