Table 1: Definition of variables and data sources.
Variable
|
Description
|
Expected sign
|
Source
|
CO2
|
CO2 emissions (kg per 2010 US$ of GDP)
|
N/A
|
WDI
|
EC
|
Energy consumption, million tonnes oil equivalent
|
Positive
|
BP Statistical Review
of World Energy
|
OPEN
|
Trade openness computed as composite trade intensity introduced by Squalli & Wilson (2011) capturing trade effect
|
Positive or negative
|
WDI, Authors
|
SE
|
Real GDP per capita capturing scale effect
|
Positive
|
WDI
|
TE
|
Real GDP per capita squared capturing technique effect
|
Negative
|
WDI, Authors
|
FDI
|
Foreign direct investment, net inflows (% of GDP)
|
Positive
|
WDI
|
TECH
|
Technological innovation measured by total patent applications
|
negative
|
WDI
|
IGDP
|
Industry, value added (% of GDP)
|
Positive or negative
|
WDI
|
N/A: Not available; WDI: World Development Indicator
Table 2: Descriptive statistics.
Variables
|
Mean
|
Median
|
Maximum
|
Minimum
|
Std. Dev
|
Skewness
|
Kurtosis
|
J-B Stat
|
Probability
|
CO2
|
0.264
|
0.238
|
0.477
|
0.084
|
0.120
|
0.217
|
1.652
|
4.682
|
0.196
|
SE
|
7.706
|
7.959
|
8.984
|
6.073
|
0.843
|
-0.511
|
2.156
|
4.102
|
0.129
|
TE
|
60.316
|
63.754
|
80.717
|
36.880
|
12.663
|
-0.387
|
2.082
|
3.422
|
0.181
|
OPEN
|
6.060
|
6.512
|
7.665
|
2.745
|
1.329
|
0.636
|
2.077
|
5.757
|
0.156
|
EC
|
4.220
|
4.422
|
4.840
|
3.177
|
0.527
|
-0.558
|
1.921
|
5.621
|
0.160
|
FDI
|
13.203
|
13.286
|
14.659
|
11.913
|
0.738
|
0.056
|
2.463
|
0.702
|
0.704
|
IGDP
|
3.513
|
3.580
|
3.813
|
3.258
|
0.161
|
-0.215
|
1.697
|
4.474
|
0.107
|
TECH
|
9.360
|
9.255
|
10.545
|
8.210
|
0.766
|
0.082
|
1.634
|
4.499
|
0.105
|
Source: Authors’ calculations
Table 3: Unit root analysis.
Variable
|
Dickey-Fuller GLS
|
Phillips-Perron
|
Augmented Dickey-Fuller
|
Kwiatkowski-Phillips-Schmidt-Shin
|
Narayan and Pop (2010) Unit Root Test
|
|
(DF-GLS)
|
(PP)
|
(ADF)
|
(KPSS)
|
Model 1
|
Model 2
|
|
Level
|
Test – Statistics value
|
|
|
|
Break-Year
|
ADF-stat
|
Break-Year
|
ADF-stat
|
InCO2
|
-0.570
|
-0.464
|
-1.152
|
0.966
|
1982:1985
|
-3.132
|
1987:1994
|
-8.160***
|
InSE
|
-0.116**
|
-0.079
|
-1.308
|
0.833***
|
1979:1988
|
-2.914
|
1982:1990
|
-7.601***
|
InTE
|
-0.112*
|
-0.076
|
-1.268
|
0.848***
|
1979:1990
|
-1.939
|
1982:1994
|
-6.791***
|
InOPEN
|
-0.072
|
-0.082
|
-1.335
|
1.080*
|
1996:2001
|
-3.053
|
2003:2009
|
-7.318***
|
InEC
|
-0.011
|
-0.014
|
-0.366
|
1.300***
|
1982:1989
|
-4.372**
|
1985:1991
|
-8.521***
|
InFDI
|
-0.032*
|
-0.001
|
-0.012
|
0.640
|
2001:2006
|
-2.021
|
2004:2010
|
-8.362***
|
InTECH
|
-0.254***
|
-0.284***
|
-2.999
|
0.255***
|
1995:2000
|
-4.318
|
2008:2011
|
-7.821***
|
InIGDP
|
-0.046
|
-0.071*
|
-1.718
|
1.060**
|
1972:1985
|
-3.815
|
1982:1991
|
-7.521***
|
First difference
|
|
|
|
Critical value (1%, 5%, and 10%)
|
ΔInCO2
|
-0.995***
|
-0.996***
|
-7.176***
|
0.705***
|
1999:2005
|
-4.801**
|
1980:1991
|
-5.832***
|
ΔInSE
|
-0.695***
|
-0.707***
|
-5.319***
|
0.585***
|
1983:1997
|
-5.831***
|
1985:1995
|
-6.831***
|
ΔInTE
|
-0.694***
|
-0.707***
|
-5.316***
|
0.589***
|
1991:2000
|
-8.531***
|
1987:1996
|
-5.893***
|
ΔInOPEN
|
-0.935***
|
-0.938***
|
-6.699***
|
0.626***
|
1996:2004
|
-6.842**
|
2001:2007
|
-8.942***
|
ΔInEC
|
-1.105***
|
-1.121***
|
-8.142***
|
0.586***
|
1985:1993
|
-5.921***
|
1989:1997
|
-7.942***
|
ΔInFDI
|
-0.207**
|
-0.209**
|
-6.443***
|
0.609***
|
2005:2008
|
-6.831***
|
2001:2008
|
-6.973***
|
ΔInTECH
|
-1.023***
|
-1.034***
|
-7.473***
|
0.424***
|
1999:2003
|
-4.841**
|
2006:2010
|
-5.983***
|
ΔInIGDP
|
-0.799***
|
-0.801***
|
-5.878***
|
0.431***
|
1975:1990
|
-7.742***
|
1988:1992
|
-7.892***
|
Source: Authors’ calculations
Note: *, ** and *** denote statistical significance at 10%, 5% and 1% levels, respectively. MacKinnon’s (1996) one-sided p-values. Lag Length based on SIC and AIC. Probability-based on Kwiatkowski-Phillips-Schmidt-Shin (1992). The critical values for Narayan-Popp unit root test with two breaks are followed by Narayan and Pop (2010). All the variables are trended.
Table 4: Lag length criteria.
Lag
|
LogL
|
LR
|
FPE
|
AIC
|
SC
|
HQ
|
0
|
178.453
|
NA
|
3.2e-12
|
-6.594
|
-6.331
|
-6.493
|
1
|
607.095
|
857.28
|
1.5e-18
|
-21.195
|
-19.094*
|
-20.390*
|
2
|
661.093
|
108
|
1.4e-18
|
-21.388
|
-17.448
|
-19.877
|
3
|
719.755
|
117.32
|
1.2e-18*
|
-21.759
|
-15.981
|
-19.544
|
4
|
784.113
|
128.72*
|
1.3e-18
|
-22.350*
|
-14.733
|
-19.430
|
Source: Authors’ calculations
Note: * indicates lag order selected by the criterion.
Table 5: ARDL bounds test analysis.
Test statistics
|
Value
|
K
|
H0 |
H1 |
F-statistics
|
14.341
|
7
|
No level relationship
|
Relationship exists
|
t-statistics
|
-8.752
|
|
|
|
Kripfganz &Schneider (2018) critical values and approximate p-values y
|
Significance
|
F-statistics
|
t-statistics p-value F
|
|
1(0)
|
1(1)
|
1(0)
|
1(1)
|
1(0)
|
1(1)
|
10%
|
2.12
|
3.23
|
-2.57
|
-4.04
|
0.000***
|
0.000***
|
5%
|
2.45
|
3.61
|
-2.86
|
-4.38
|
p-value t
|
1%
|
3.15
|
4.43
|
-3.43
|
-4.99
|
0.000***
|
0.002**
|
Note: *, ** and *** respectively represent statistical significance at 10%, 5% and 1% levels. The respective significance levels suggest the rejection of the null hypothesis of no cointegration. The optimal lag length on each variable is chosen by the Schwarz's Bayesian information criterion (SBIC).
Table 6: Diagnostic statistics tests.
Diagnostic statistics tests
|
X2(P values)
|
Results
|
Breusch Godfrey LM test
|
0.3812
|
No problem of serial correlations
|
Breusch-Pagan-Godfrey test
|
0.2610
|
No problem of heteroscedasticity
|
ARCH test
|
0.6837
|
No problem of heteroscedasticity
|
Ramsey RESET test
|
0.5183
|
Model is specified correctly
|
Jarque-Bera Test
|
0.2715
|
Estimated residual are normal
|
Source: Authors’ calculations
Table 7: Dynamic ARDL simulations analysis.
Variables
|
Coefficient
|
St. Error
|
t-value
|
Cons
|
-1.1814
|
1.2878
|
-0.92
|
InSE
|
0.2139***
|
0.1829
|
4.56
|
ΔInSE
|
0.3962***
|
0.2721
|
2.77
|
InTE
|
-0.6657**
|
0.8605
|
-2.34
|
ΔInTE
|
-0.7441
|
0.1387
|
-1.79
|
InOPEN
|
0.1883***
|
0.0487
|
5.39
|
ΔInOPEN
|
-0.3043**
|
0.0570
|
-2.53
|
InEC
|
0.2713***
|
0.1762
|
3.98
|
ΔInEC
|
0.5906*
|
0.1719
|
1.98
|
InFDI
|
0.9064
|
0.0810
|
1.12
|
ΔInFDI
|
0.2846**
|
0.2657
|
2.59
|
InTECH
|
-0.7358***
|
0.5941
|
-3.24
|
ΔInTECH
|
-0.2274
|
0.0738
|
-0.31
|
InIGDP
|
0.3429**
|
0.1577
|
2.17
|
ΔInIGDP
|
0.5308
|
0.2309
|
0.23
|
ECT(-1)
|
-0.8243***
|
0.1396
|
-3.04
|
R-squared
|
0.7845
|
|
|
Adj R-squared
|
0.7693
|
|
|
N
|
55
|
|
|
P val of F-sta
|
0.0000***
|
|
|
Simulations
|
1000
|
|
|
Source: Authors’ calculations
Note: *, ** and *** denote statistical significance at 10%, 5% and 1% levels, respectively.
Table 8: Frequency-domain causality test
Source: Authors’ calculations
Note: *, ** and *** denote statistical significance at 10%, 5% and 1% levels, respectively.