Palgrave Handbook of Econometrics: Applied Econometrics

Anindya Banerjee and Martin Wagner 673

Table 13.5 Results of the tests of Bai and Carrion-i-Silvestre

(2007) for unit roots in non-stationary panels with common

factors and structural breaks

Euro-area CEEC Industrial Worldwide

Z and p-value tests

Z 6.36 −1.61 2.36 −1.55

BCN −2.33 1.44 −0.24 −0.04

BCχ 2 6.54 31.57 55.46 113.37

Simplified Z and p-value tests

Z 17.51 3.03 15.96 0.51

BCN −2.61 0.49 −1.08 0.05

BCχ 2 4.69 25.25 46.40 114.81

Number of breaks 9 5 11 8

Notes: The number of factors is chosen according to the information cri-

terion BIC 3. The test results reported allow for at most one break in both

the intercept and linear trend, except for the euro-area dataset where up

to two breaks are allowed for. The critical value (at the 5% nominal level)

is given by –1.645 for theZtest, by 1.645 for theBCNtest and by 32.92

(N=11), 76.78 (N=29) and 139.92 (N=57) for theBCχ 2 test. Number

of breaks indicates the number of countries in which breaks have been

detected. The tests are as described in section 13.2.3.

without non-stationary common components. All in all, the findings show that a
careful modeling of cross-sectional dependence that leads to an appropriate choice
of panel unit root test is of key importance.
Let us finally turn to the issue of structural change, neglected up to now. Table
13.5 contains the results obtained when applying the Bai and Carrion-i-Silvestre
(2007) tests. Bai and Carrion-i-Silvestre also propose simplified approximate tests
whose limiting distributions are independent of the break fractions. The results
obtained with these tests are also included in Table 13.5.
The major observation that emerges is that the unit root null hypothesis is
not rejected for any of the datasets by any of the tests. Thus, even when allow-
ing for structural breaks and common factors, no evidence for stationarity of the
RER panels emerges. In particular, contrasting the findings obtained by allowing
for multiple factors (and breaks), with no evidence for stationarity of the RER,
with the findings obtained with first-generation tests or more restrictive second-
generation tests, where seemingly more evidence for PPP prevails, indicates that
the resurrection of PPP due to the usage of panel methods may not yet have been
accomplished.

13.2.4.2 The environmental Kuznets curve

The empirical example, based on Wagner (2008b), considered in this sub-section
is based partly upon using the Groningen dataset mentioned in the introduction.
Since the seminal work of Grossman and Krueger (1995) many econometric stud-
ies of the relationship between measures of economic development (typically per