Palgrave Handbook of Econometrics: Applied Econometrics

Anindya Banerjee and Martin Wagner 669

Looking at (13.16), it becomes clear that there are several potential sources of
cross-sectional dependence in RER panels. First, all nominal exchange rates are
coupled to each other by no arbitrage restrictions which appear to hold quite well
in liquid foreign exchange markets. Given that nominal exchange rates, if not fixed
to the base country currency, typically fluctuate more than prices, this dependence
will not be wiped out by price level movements across countries. Second, all RERs
contain the same (non-stationary) base country price indexp∗t. Since the goods
contained in the consumer baskets generally differ across countries, it is realistic
to assume that the permanent components in (ei,t+pi,t)andp∗tdo not exactly
coincide. If they do not perfectly coincide, then in general theqi,tseries contain a
common permanent component related top∗t. Third, the no arbitrage (in the goods
markets) arguments underlying the law of one price and PPP rely upon economic
interaction of one form or the other. The world economy becomes ever more inte-
grated and thus shocks can also be expected to be transmitted more strongly across
countries. The type of short-run dependence considered by O’Connell (1998) and
discussed above may not be sufficient since, as we have just argued, RER panels
may be prone to the presence of common non-stationary components.
Lyhagen (2000) studies a special case of this situation with one common stochas-
tic trend and shows that several first-generation tests, including those of Levin, Lin
and Chu (2002) and Im, Pesaran and Shin (2003), are severely affected in this case.
In particular, Lyhagen shows that, in the presence of one common stochastic trend,
the size of the tests tends to one with increasing cross-sectional dimension.
To assess the impact of cross-sectional dependence when testing for PPP, Wagner
(2008a) applies a battery of first- and second-generation panel unit root tests to
four monthly RER panels. The sample periods, as well as the cross-section and
time dimensions of the panels, are contained in Table 13.1. Details are given in
Appendix A.
The euro-area dataset (the data are taken from the IMF IFS, OECD MEI and ECB
databases) consists of 11 of the 12 countries that were starting members of the euro-
area in January 1999, with Ireland missing due to constraints on data availability.
The Central and Eastern European countries (CEEC) dataset consists of 11 transition

Table 13.1 Time periods and panel

dimensions of the four monthly datasets

considered

Start End T N

Euro-area 1980/1 1998/12 228 11

CEEC 1993/1 2004/6 138 11

Industrial 1980/1 1998/12 228 29

Worldwide 1981/1 2004/4 280 57

Note: The number of observations over time is

denoted byTwhileNdenotes the cross-sectional

dimension.