Nonstationary panels and an application

Panel data estimation methods have become increasingly popular because of the improved availability of this type of data coupled with the ability of panel data studies to allow analysis that is not possible from either cross-section or time series data alone. Along with this popularity, the number of studies dealing with both empirical and theoretical panel data has increased dramatically. One of the most popular areas of research in panel data econometrics currently is nonstationary panel data models. As distinct from conventional panel data analysis, time components are the source of nonstationarity in panel data with large cross-section and time dimensions. Furthermore, adding the cross-sectional dimension to the time series dimension provides a significant contribution in testing the non-stationarity and co-integration.In this thesis, the methods of estimation often encountered in the analysis of nonstationary panel data are studied. To illustrate the use of panel unit root and co-integration tests employed, the hypothetical positive relationship between the real wage rate and average productivity of labor suggested by the Efficiency Wage Theory is investigated for the Turkish Manufacturing Industry. The results obtained do not reject the Efficiency Wage Theory for the Turkish Manufacturing Industry.In addition to the data analysis, the study comparatively presents various test results. Firstly, following Choi (2001) and Maddala and Wu (1999), Fisher type panel unit root tests that do not take into account the cross-sectional correlation were applied. After that, to compare, the series were cleaned of the cross section correlation as much as possible, and Fisher type panel unit root tests were applied again. Furthermore, Pesaran (2003, 2007) panel unit root tests, which consider cross-section correlation within the the regression equation, were used for additional comparison purposes. In order to test the existence of panel co-integration, a new test proposed by Hanck (2007) that extends the panel unit root tests of Choi (2001) and Maddala and Wu (1999) to the panel co-integration case was used. This test was also applied to the data both before and after the series were cleaned of the cross-sectional correlation. Besides the panel unit root and co-integration tests presented in this study; ADF, CADF and Engel-Grenger tests were performed for individual series.