Between-Group Dependence in PPP Equations and its Causes: A Principal Components Approach

Principal components of the residuals are proposed for analysing between-group dependence when estimating PPP equations in large-T, large-N panels. If this dependence arises because omitted variables are correlated with included variables, the appropriate response to between-group dependences may not be GLS. Instead estimating the factors that account for the covariances between groups may help to suggest relevant omitted variables that can be included in the model. The approach is applied to four PPP panels of 17 OECD countries for the 1973:1-1998:12 period.