The Levin, Lin and Chu (2002) and Breitung (2000) panel unit root tests specify a
null hypothesis of a unit root against the alternative of no unit root (across
all cross-sections). The Im, Pesaran and Shin (2003) test employs the null
hypothesis of a unit root against the alternative that some cross-sections do not contain a unit root. Table D1 shows the
results of these tests with fixed effects, time trends and two lags.
Table D1: Panel Unit Root Tests
LLC
Breitung
IPS
8.23
0.93
−4.73***
GDPjt
−3.04***
13.53
−11.03***
GDPit
−1.97**
15.31
−12.95***
Xit
13.82
10.17
10.12
(GNE–M)it
−5.80***
22.66
−0.08
popjt
34.63
31.63
15.10
popit
29.32
36.24
22.66
Notes: *, **, *** represent statistical significance at the 10, 5 and 1
per cent levels respectively. Data cover bilateral trade between 777 country-pairs
for the period 1987–2008, with all variables in logs.
Pedroni (1999, 2004) proposes several panel cointegration tests that extend the Engle-Granger
(Engle and Granger 1987) cointegration framework. Table D2 reports the panel
Augmented Dickey-Fuller (Dickey and Fuller 1979) test statistics for two of
these tests: first, the null hypothesis of no cointegration against the alternative
of cointegration with common autoregressive coefficients for all country-pairs;
and second, the null hypothesis of no cointegration against the alternative
of cointegration with individual autoregressive coefficients for all country-pairs.
The Pedroni tests are implemented with fixed effects and two lags. The Kao
(1999) test also extends the Engle-Granger two-stage cointegration framework,
testing the null hypothesis of no cointegration against the alternative of
cointegration with common autoregressive coefficients for all country-pairs.
Table D2: Panel Cointegration Tests
Pedroni
Pedroni
Kao
Common autoregressive coefficients
Individual autoregressive coefficients
Common autoregressive coefficients
Equation (1)
−5.59***
−10.08***
−0.56
Equation (2)
−3.91***
−8.24***
−4.33***
Notes: *** represents statistical significance at the 1 per cent level.
Data cover bilateral trade between 777 country-pairs for the period 1987–2008,
with all variables in logs.
Table D3: Gravity Model Results for Resource Imports
Equation (1) in log differenced form
Baseline
China effect
No fixed or time effects
Time effects only
No fixed or time effects
Time effects only
GDPjt
0.16
0.21
0.16
0.21
GDPit
1.66***
1.80***
1.60***
1.75***
GDPit×Chinai
−0.03
−0.05
popjt
1.80***
1.79***
1.79***
1.78***
popit
0.23
0.20
0.27
0.23
Adjusted R2
0.002
0.004
0.002
0.004
Observations
66782
66782
66782
66782
Notes: Model estimated with robust standard errors. *, **, *** represent
statistical significance at the 10, 5 and 1 per cent levels respectively.
Data cover bilateral trade between 180 economies for the period 1980–2008.
Resource imports are excluded from aggregate imports (M). For
regressions with China dummy interactions, the China
dummy was also included on its own (because country-pair fixed effects
are not employed), but was found to be insignificant (not reported).
Table D4: Gravity Model Results for Resource Imports
Equation (2) in log differenced form
Baseline
China effect
No fixed or time effects
Time effects only
No fixed or time effects
Time effects only
GDPjt
0.25*
0.25*
0.26*
0.25*
Xit
0.32***
0.38***
0.27***
0.33***
Xit×Chinai
–0.13
–0.13
(GNE–M)it
0.08**
0.07*
0.06
0.06
(GNE–M)it×Chinai
0.26
0.22
popjt
1.60***
1.71***
1.57***
1.67***
popit
0.91*
1.04*
0.98*
1.10**
Adjusted R2
0.001
0.003
0.001
0.003
Observations
62545
62545
62545
62545
Notes: Model estimated with robust standard errors. *, **, *** represent
statistical significance at the 10, 5 and 1 per cent levels respectively.
Data cover bilateral trade between 180 economies for the period 1980–2008.
Resource imports are excluded from aggregate imports (M). For
regressions with China dummy interactions, the China
dummy was also included on its own (because country-pair fixed effects
are not employed), but was found to be insignificant (not reported).
