Equations

1.
1 Empirical Estimation Techniques
1.2.1 Diagnostics Tests (IPS & CIPS unit root, CD test)
This study employed the most advanced econometric approach to test the cross-sectional
dependence (CD) between the variables, as conventional econometric approaches cannot address
that issue. Therefore, we adopted (Pesaran, 2004) proposed cross-sectional dependence test
which is important before unit root test. We have used following equations for cross-sectional
dependence and Lagrange multiplier (LM) and cross-sectional dependence:
√ ( )
N−1 N
2T
CD=
N ( N −1 )
∑∑ ρij Eq. (A. 2)
i=0 j=i+1
y ¿ =α ¿ + β i x ¿ + µ¿ Eq. (A. 3)
Further, to confirm the stationarity of the variables we employed second generation IPS(i.e.
CADF) and CIPS(i.e. CIPS) unit root test proposed by (Pesaran, 2007). The unit root test helps
to confirm the integration order, either variable is stationary at level i(0) or first difference i(1).
Following equations are used to estimate CADF and CIPS unit root test:
n
y ¿ =α ¿ + β i x ¿−1 ρi T + ∑ ϴ¿ △ x i ,t− j+ µ¿ Eq. (A. 4)
j=0
N
1
CIPS= ∑ CADFi Eq. (A. 5)
N i=1
3.2.1 Cross-sectional dependency (CD) test

Panel base study has chances of cross-sectional dependent issues. Overlooking or without taking corrective
action issues of cross-sectional dependency data can forecast biased results and misleading information. Therefore,
like (Saud, Chen, et al., 2018), before probing stationarity properties of the framework, conduct the cross-sectional
dependence test. This empirical study using the Lagrange multiplier (LM) and Cross-sectional approach suggested
by (Pesaran, 2004) which is represent as:
√ (∑ ∑ )
N−1 N
CD= 2T (10)
ρij
N ( N −1 ) i=0 j=i+1
Here CD represents the cross-sectional dependence, T indicates the time and N is the cross-sectional
correlation. Moreover, between i and j, cross-sectional correlation of errors is defined by ρij. To investigate the
cross-sectional dependence, we use this equation (LM test):
y ¿ =α ¿ + β i x ¿ + µ¿ (11)
Here i represent the cross-sections and t indicates time. For both methods null hypothesis indicates the cross
sections between the variables are independent, and the alternative hypothesis explains cross-sections are dependent
on each other.
3.2.2 Panel Unit Root Test
After confirming the cross-sectional between the variables through the cross-sectional dependence test, the next step
is to examine the variables' integration order. Therefore we are using the second generation CIPS and CADF unit
root test by (Pesaran, 2014). These test address cross sectional issues while examine the unit root order of variables.
CIPS test equation is as follow:
n
y ¿ =α ¿ + β i x ¿−1 ρi T + ∑ ϴ¿ △ x i ,t− j+ µ¿ (12)
j=0
Here x ¿ and µ¿ represent variables and residual, moreover, i and t is the cross-section in and time in the panel data.
Null hypothesis for CIPS and CADF explains the data series have unit root and the alternative hypothesis indicates
the stationarity of the data. Cross-sectional augmented Dickey-fuller statistic (CADF) is used to estimate the CIPS
test as follow:
N
1
CIPS= ∑ CADFi
N i=1
(13)
3.2.3 Panel Causality Test
To conduct the causality analysis, we adopted Dumitrescu and Hurlin, (2012), which is based on the 1969
granger test (individual wald statistic averaged non-causality across the cross-section). Empirically we estimate this
test as follow:
J J
y ¿ =α i +∑ λij yi (t− j )+ ∑ β ij X i(t − j) + µ¿ (14)
j−1 j −1
X and y represent the number of observations β ji ∧λ capture autoregressive parameters and regression
coefficients. The null hypothesis shows no causal relationship between the variables and alternative
hypothesis indicates the relationship between the selected variables.
3.2.4 Panel FE-OLS and D-K Estimation techniques

we employed (Pesaran, 2004) fixed effect ordinary least square (FE-OLS), which have individual
intercepts and also allows for heterogeneous serial correlation across panel data variables . Traditionally
to estimate the results for panel data, we use a fixed or random-effects model. Although, Random effect
model is suitable for unobserved heterogeneity between cross-sections are due to the constant variables
over the time period but vary among the cross-sections. Thus, we consider the random effect model is
appropriate for large cross sections N and cross sections are randomly use for the given sample (Hadri,
2000) . However, the fixed effect model addresses omitted variables and keeps constant over the time
period for cross sectional heterogeneity, so the fixed effect model is suitable for small cross sections N
(Arellano, 2003). Our study consists of (N=19) cross-sections, having 6 explanatory variables (k=6) with
33 observations (T=33). Hence, like (Anwar et al., 2021) our sample data indicates that the, size of cross
sections is less than observations N<T, so we go for Fixed effects OLS.
Moreover for confirmation of our findings, like (Ullah et al., 2021) we also used (Driscoll & Kraay,
1998) standard error approach for G20 countries to gage the impact of explanatory variables on carbon
emission (with the role of moderation and mediation). D-k technique addresses the issue of cross-
sectional dependence for robust estimator. Moreover, we can take an average of products between
explanatory variables and residuals and then use the weighted HAC estimator values with a stranded error
that was recently incorporated. It considers one of the best techniques to address the heteroscedasticity
and spatial and serial dependency in data (Jalil, 2014; Özokcu & Özdemir, 2017). Moreover, DK also
handles balance and unbalanced panel and missing values. It captures general form of cross-sectional and
temporal dependence. The estimation of equation for pooled ordinary least square as follow:
¿
CO 2¿ =α o + X ¿ β+ µ¿ (15)
Where co2 is dependent variable, X indicates set of control variables, i = 1,2,…19 (19 countries of G20
block), t represent time period (t= 33) .
1.3.2 Pairwise Dumitrescu and Hurlin Causality test
After confirmation of CD, this study employed the causality test proposed by (Dumitrescu &
Hurlin, 2012) based on the individual Wald statistic of Granger (1969) test. Empirical
representation of test is as follows:
J J
y ¿ =α i +∑ λij yi (t− j )+ ∑ β ij X i(t − j) + µ¿ Eq. (A. 6)
j−1 j −1
Where, y and x indicate the numbers of observations, λ∧¿ β ji report the coefficients of
regression and autoregressive parameters. The null hypothesis demonstrates no causal
association among the variables and shows the causal relationship between the variables.

Equations

Uploaded by

Copyright:

Available Formats

Equations

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Equations

Uploaded by

Copyright:

Available Formats

1.

1 Empirical Estimation Techniques

1.2.1 Diagnostics Tests (IPS & CIPS unit root, CD test)

dependence and Lagrange multiplier (LM) and cross-sectional dependence:

3.2.1 Cross-sectional dependency (CD) test

3.2.2 Panel Unit Root Test

3.2.3 Panel Causality Test

3.2.4 Panel FE-OLS and D-K Estimation techniques

representation of test is as follows:

regression and autoregressive parameters. The null hypothesis demonstrates no causal

You might also like