Panel unit root tests under the null hypothesis of stationarity and confirmatory analysis with applications to PPP and the convergence hypothesis

My dissertation, which is comprised of four main parts, is about the econometric analysis of unit-root nonstationary (I)(1)) observations in panel data. Most existing (and popular) panel unit-root tests are built under the null hypothesis that all individuals in the panel are I(1). A potential pitfall in the exclusive use of these tests lies in the interpretation of rejections of the null hypothesis since the rejection is consistent both with the idea that the panel is stationary (I(0)) and also with the idea that the panel is a mixture of I(0) and I(1) series. Developing a strategy to overcome this problem is the main focus of my dissertation. The particular strategy that I adopt is to conduct confirmatory analysis by combining panel unit root tests under the I(0) null and tests under the I(1) null. The first problem, however, is that not many panel unit root tests under the I(0) null are available and little is known about their small-sample properties.;Thus in Part 1 of my dissertation, I develop two panel tests under the null hypothesis that all individuals in the panel are I(0). They are the panel G-test and the panel KPSS test as multivariate extensions of the univariate tests proposed by Park and Choi (1988) and Kwiatkowski et al. (1992), respectively. Both their asymptotic and finite sample properties are investigated. Part 2 studies how these tests can be productively employed in conjunction with existing panel unit root tests under the I(1) null in order to conduct confirmatory analysis. Specifically, I ask which combination of tests under the I(0) null and under the I(1) null provides the most power in detecting mixed panels where I(0) series coexist with I(1) series. In part 3, I investigate long-run PPP by confirmatory analysis. In contrast to earlier studies, I find little evidence for PPP with the panel-G test in a panel of 20 OECD countries over the post-Bretton Woods period. In part 4, I perform confirmatory analysis to test the convergence hypothesis in growth theories. I find evidence of convergence neither in international data nor in U.S. states data.