696 Panel Methods to Test for Unit Roots and Cointegration
policy, especially for countries that previously had less credible policy regimes.
Especially in countries with previously rather less successful monetary policy, the
perceived stabilization of monetary policy may well have induced the producers
to change their pricing strategies and would thus be expected to have an influence
on the long-run ERPT. Moreover, the adoption of a common currency has changed
competitive conditions by increasing the share of goods denominated in the (new)
domestic currency. Finally, virtually all the currencies were depreciating against the
US dollar in the period 1995–2000, and especially since 1996. Thereafter, following
a short period of a stable euro–dollar exchange rate, the euro started appreciating
till the end of our sample. This asymmetry of exchange rate developments may
have different implications for ERPT.
Panel cointegration tests
There would essentially be three ways of proceeding in order to construct pan-
els from the datasets – (1) creating country panels of industry cross-sections, (2)
industry panels with country cross-sections, and (3) a pooled panel in which every
country and industry combination constitutes a separate unit. In search of the
existence of a cointegrating relationship in the series we try to maximize the dimen-
sions of our panel, and thus will focus on (3). Results for (1) and (2) are available
from us upon request.
In Table 13.9 we present the results of the modified Pedroni tests due to Banerjee
and Carrion-i-Silvestre (2007), allowing for structural breaks (as described in section
13.3.1.3) and allowing for both structural change and cross-sectional dependence
(as described in section 13.3.1.4). As noted earlier, results are presented for the
longest available panel which includes all countries. Throughout the analysis it
is assumed that in each cross-section member at most one break occurs, with,
depending upon the model, this break occurring at the same time in all breaking
components (intercept, trend, cointegrating vector).
The panel headed “Cross-sectionally independent” reports the results from the
modified Pedroni (pooled panelt-) tests on the idiosyncratic components under
the assumption that these are cross-sectionally independent. The results for all
six model specifications concerning the deterministic components as outlined in
section 13.3.1.3 – for all of which heterogeneous break dates are permitted – are
reported in this panel.
The panel headed “Cross-sectionally dependent” provides the results for the tests
of Banerjee and Carrion-i-Silvestre (2007), where cross-sectional dependence is also
allowed. The test results are again reported for the idiosyncratic components and
we display the results for both cross-sectionally homogeneous and heterogeneous
break dates. For Models 3 and 6, we are restricted to imposing a cross-sectionally
homogeneous (but unknown) break point. The remaining models allow for both
heterogeneous and homogeneous breaks. The maximum number of factors allowed
is six, the column labeledˆrprovides estimates of the number of common factors,
and under the headingˆr 1 the number of integrated common trends detected from
the MQ statistic is also reported. The break dates detected by the cross-sectionally
dependent test, when homogeneous breaks are imposed, are also given.