Palgrave Handbook of Econometrics: Applied Econometrics

676 Panel Methods to Test for Unit Roots and Cointegration

Table 13.6 Results of Bai and Ng (2004) analysis

Factors BNN BNχ 2 MQc(m) MQf(m) CFADF

SO 2 2 −0.70 180.22 1 2 –

GDP 1 2.89 250.96 ––−2.30

Notes: Bold entries indicate rejection of the unit root null hypothesis at

the 5% critical level.

CFADF: ADF test on single common factor as in section 13.2.2.2.

Also see notes to Table 13.3.

Table 13.7 Results of second-generation tests

MPa MPb CIPS Cp CZ CL∗ NL

Fixed effects

SO 2 –14.70 –7.56 –1.78 9.45 –5.83 –6.22 4.05

GDP –17.72 –9.05 –1.64 1.75 –0.56 –0.79 8.29

Fixed effects and linear trends

SO 2 –0.81 –0.49 –2.44 1.81 3.89 4.91 6.69

GDP 1.38 1.41 –2.46 –3.01 6.30 7.26 5.82

Notes: Bold entries indicate rejection of the unit root null hypothesis at

the 5% critical level. Also see notes to Table 13.4.

The other tests are included to show again that by a strategic choice of panel unit
root test any desired conclusion can be supported.
Neglected up to now and potentially important for the case of emissions series is
the issue of structural change (for example, due to technological or legal changes).
In this respect the short time dimension of the panels represents a major limitation,
since the panel unit root tests that allow for structural change in the determinis-
tic component, as described in section 13.2.3, require, when the break dates are
assumed to be unknown, the individual specific estimation of these break dates.
Nevertheless, applying the Bai and Carrion-i-Silvestre (2007) tests, see Table 13.8,
provides evidence for structural change in about a third to a half of the countries
included in the panel. All the unit root test results lead to non-rejection of the unit
root null hypothesis even when allowing for structural breaks. The results in the
table refer to the case with at most one structural break per unit, since even when
allowing for more structural breaks only one break is detected.^23
This example is continued in section 13.3.5 where, in particular, the estimation
of (13.17) is discussed.

13.2.5 Some concluding remarks

We have continued to bring together the key strands of the testing strategies within
the panel integration framework. Important considerations have included: