CHAPTER FOUR

PRESENTATION OF RESULTS AND FINDINGS

In this chapter, the results of the research are presented and systematically analyzed using the
econometric software Stata 17. All tests and estimations were conducted in line with the
objectives and hypotheses outlined in Chapter One. These procedures ensure that the findings are
objective, robust, and reliable, while also allowing for verification of whether the results conform
to the a priori expectations specified in Chapter Three.

4.1 Descriptive Statistics

Table 4.1 Descriptive Statistics Results

Variables Obs Mean Std. Dev. Min Max

GHG 132 47.904 89.848 0.844 360.086
GTAX 132 0.931 0.739 0.012 3.207
RQ 132 -0.579 0.341 -1.156 0.269
FDI 132 3.539 3.794 -11.192 18.828
INDUS 132 6310.597 13814.201 86.477 64895.895
INF 132 5.968 7.777 -3.233 47.643
GDPPC 132 1471.65 886.137 469.69 4192.401
URB 132 44.329 12.581 16.208 67.982
NB: GHG is Green House Gas Emissions in Million tons of carbon dioxide equivalent emitted
per year., GTAX is Green Tax revenue as a percentage of GDP, RQ is Regulatory Quality
Estimate, FDI is Foreign Direct Investment as a percentage of GDP, INDUS is Industrial Output
in millions of US Dollars, INF is Inflation Rate, GDPPC is Per Capita Income Output in
millions of US Dollars and URB is Urban Population as a percentage of total population.

Source: Researcher’s construct using results obtained from Stata 17

From Table 4.1, substantial variation across the variables can be noted. Greenhouse Gas
Emissions (GHG) range from a minimum of 0.844 (Cote d'Ivoire) to a maximum of 360.086
(Nigeria), with an overall mean of 47.904 and a relatively high standard deviation of 89.848,
indicating wide disparities in emission levels across countries and years. Green Taxation (GTAX)
varies between 0.012 and 3.207, with an average of 0.931 and a standard deviation of 0.739,
reflecting uneven adoption of environmental fiscal measures. Regulatory Quality (RQ) ranges
from –1.156 to 0.269, with a mean of –0.579 and a standard deviation of 0.341, suggesting that
institutional performance is generally weak but differs significantly across the sample. Foreign
Direct Investment (FDI) has a minimum of –11.192, a maximum of 18.828, and an average of
3.539. Industrial Output (INDUS) ranges from 86.477 to 64,895.895, with a mean of 6,310.597
and a large standard deviation of 13,814.201. Inflation (INF) ranges from –3.233 to 47.643, with
an average of 5.968 and a standard deviation of 7.777. GDP per capita (GDPPC) demonstrates
sharp contrasts, with a minimum of 469.69 and a maximum of 4,192.401, yielding an average of
1,471.65 and a standard deviation of 886.137. Finally, Urban Population (URB) ranges from
16.208 to 67.982, with a mean of 44.329 and a standard deviation of 12.581.

4.2 Correlation Matrix

Before estimation, it is important to assess whether the regressors in the model exhibit
collinearity. In other words, this involves determining whether two or more explanatory variables
are exerting similar influences on the dependent variable. Collinearity is indicated by a
correlation coefficient of ±0.8 or higher.

Table 4.2 Correlation Matrix

Variables GHG GTAX RQ FDI INDUS INF GDPPC URB

GHG 1.000

GTAX -0.334 1.000

RQ -0.344 -0.123 1.000

FDI -0.249 0.090 0.077 1.000

INDUS 0.936 -0.348 -0.174 -0.200 1.000

INF 0.309 0.185 -0.438 0.004 0.309 1.000

GDPPC 0.252 -0.382 0.481 0.097 0.432 0.011 1.000

URB 0.081 -0.128 0.294 0.016 0.192 0.058 0.639 1.000

 Source: Researcher’s construct using results obtained from Stata 17

The correlation matrix in Table 4.2 shows that none of the regressors have the issue of
collinearity ±0.8. The strongest relationship is between GHG and INDUS (0.936), however, this
collinearity is required because GHG is the dependent variable.

4.3 Pre-Estimation Tests

Pre-estimation tests are carried out prior to model estimation to ensure that the variables satisfy
the fundamental requirements for economic analysis. By applying these tests, the study
safeguards against biased inference and strengthens the credibility of the empirical findings. In
this panel study, these tests include the cross-sectional dependency test, unit root test and the
co-integration test.

4.3.1 Cross-Sectional Dependency Test

The Pesaran Cross-Sectional Dependence (CD) test was applied to examine whether residuals
across panel units are correlated. This test is particularly relevant for the scope of the study
where countries in the region of West Africa may be influenced by common shocks or spillover
effects. Detecting cross-sectional dependence is critical because ignoring it can lead to biased
standard errors and unreliable inference.

Test Hypothesis:

H 0: There is cross-sectional independence (residuals across panel units are not correlated).

H 1: There is cross-sectional dependence (residuals across panel units are correlated).

Decision Rule: Reject H 0if the CD statistic is greater than the CD critical value in absolute
terms ( ∣C Dcal ∣>∣C Dtab ∣ at α=5 % ); otherwise, do not reject. Alternatively, reject H 0if the
p-value of the CD statistic is less than 0.05; fail to reject if otherwise.

The result of the Pesaran Cross-Sectional Dependence (CD) test is shown below:

Table 4.3.1: Pesaran Cross-Sectional Dependence (CD) Test Results

 CD -Statistic -0.845
Probability 0.3983
Source: Researcher’s construct using results obtained from Stata 17

Since the p-value (0.3983) is greater than the 5% significance level, we fail to reject the null
hypothesis of cross-sectional independence. This implies that there is no evidence of
cross-sectional dependence in the panel dataset, and the residuals across countries are not
significantly correlated

4.3.2 Unit Root Test

Stationarity is crucial because non-stationary data can produce spurious regressions, where
results look statistically significant but are actually meaningless. To investigate the stationarity
properties of the variables, the study employed the Im–Pesaran–Shin (IPS) panel unit root test.
This test extends the Augmented Dickey–Fuller (ADF) framework to panel data by allowing for
heterogeneity across cross sections, thereby providing a more flexible and powerful approach to
detecting unit roots.

Test Hypothesis:

H 0: The variable has unit root (non-stationary).

H 1: The variable has no unit root (stationary).

Decision Rule:

Reject H 0if the IPS statistic is greater than the IPS critical value in absolute terms
( ∣ IP S cal ∣>∣ IP Stab ∣ at α =5 % ) ; otherwise, do not reject. Alternatively, reject H 0if the p-value of
the IPS statistic is less than 0.05; fail to reject if otherwise.

The results of the IPS panel unit root test are computed in Table 4.3.1 below
Table 4.3.2: IPS Panel Unit Root Test Results

Variables IPS Statistic 5% Critical IPS Statistic 5% Critical Order Of

at Levels Value at 1st Value Integration
Difference

GHG 2.2664 0.9883 -2.6961 0.0035 I(1)

GTAX -3.1053 0.0010 - - I(0)

RQ 0.3713 0.6448 -3.1783 0.0007 I(1)

FDI 0.5811 0.7194 -2.6016 0.0046 I(1)

LINDUS 3.7495 0.9999 -4.8948 0.0000 I(1)

INF 3.1632 0.9992 -1.9479 0.0257 I(1)

LGDPPC 0.9920 0.8098 -2.6922 0.0035 I(1)

LURB 1.6629 0.9518 -1.6977 0.0448 I(1)

Source: Researcher’s construct using results obtained from Stata 17

From table 4.3.1 above, only one variable (Green Taxation) is stationary at levels became
stationary after their first differencing. The above results indicate a possible cointegrating
relationship among the variables; thus, there is need to conduct a cointegration test. Also, the
mixture of the orders of the integration entails that there is not direct long run relationship and
Cointegration will have to be tested. Therefore, the observation of the order of stationarity in the
above result satisfies the condition upon which this study was conducted.

Note: GTAX*RQ is an I(1) variable as it is made up of the interaction of an I(1) and I(0)
variables.

4.3.3 Co-Integration Test

When the variables in the model are not stationary at levels but are instead integrated of order
one, a cointegration test is applied to determine whether a long-run relationship exists among
them. This is done by examining whether their linear combination is stationary at levels. The
Kao panel cointegration test was conducted to determine whether the variables share a long-run
equilibrium relationship. This test builds on the Engle–Granger two-step methodology but adapts
it to panel data by imposing homogeneity on the cointegrating vector across cross sections.

Test Hypothesis:

H 0: No cointegration (panel residuals contain a unit root)

H 1: Cointegration (panel residuals are stationary)

Decision Rule:

Reject H 0if the Kao statistic is greater than the Kao critical value in absolute terms
( ∣ Ka o cal ∣>∣ Ka otab ∣ at α =5 % ) ; otherwise, do not reject. Alternatively, reject H 0if the p-value of
the Kao statistic is less than 0.05; fail to reject if otherwise..

The results of the Kao Cointegration test are computed in Table 4.3.2 below:

Table 4.3.3: Kao Cointegration Test Results

Test Statistic Value p-value Decision (5%)

Modified Dickey–Fuller t 3.4175 0.0003 Reject H₀

Dickey–Fuller t 4.2799 0.0000 Reject H₀

Augmented Dickey–Fuller t 5.0007 0.0000 Reject H₀

Unadjusted Modified DF t -4.6226 0.0000 Reject H₀

Unadjusted DF t -4.8807 0.0000 Reject H₀

Source: Researcher’s construct using results obtained from Stata 17

The Kao panel cointegration test results indicate that all test statistics are significant at the 5%
level (p-values < 0.05). Therefore, the null hypothesis of no cointegration is rejected, confirming
the presence of a long-run equilibrium relationship among the variables. With this, a long run
analysis can be conducted safely.

4.4 Presentation and Analysis of the Regression Results

As earlier stated, the objectives of this study are to estimate the moderating impact of regulatory
quality on the green taxation – greenhouse gas emission nexus in West Africa; to estimate the
unconditional impact of green taxation on greenhouse gas emission in West Africa; and to
investigate the unconditional impact of regulatory quality on greenhouse gas emission in West
Africa. In line with these objectives, the model was designed and estimated using the Pooled
OLS, the Driscoll–Kraay standard errors, and the High-Dimensional Fixed Effects. It’s results
are presented in Table 4.4 showing the output of the estimations.
 Discroll and Kraay High-Dimensional Fixed
Variables Pooled OLS
Standard Errors Effect (HDFE)

GTAX -60.88*** -5.017** -4.4861***

(16.51) (1.868) (1.385)
RQ -126.1*** -2.584 -2.469
(24.42) (3.304) (0.301)
GTAX*RQ -21.82 -3.886** -3.550*
(18.89) (1.736) (1.959)
FDI -3.305** 0.1014** 0.0861
(1.455) (0.0407) (0.0559)
LINDUS 42.35*** 1.453 1.903
(4.821) (3.751) (2.734)
INF -0.0856 -0.146** -0.1506***
(0.842) (0.0638) (0.0535)
LGDPPC -6.405 21.83** 26.81***
(22.89) (9.732) (5.9452)
LURB 51.69** 4.566 27.69**
(25.49) (7.914) (12.28)
C -1,015*** -153.06** -286.05***
(95.04) (58.21) (38.56)
Observations 132 132 132
R-squared 0.788 0.5107 0.999
Adj R-squared - - 0.9988
F-test 17.81 5323.55 20.41
Prob > F 0 0 0
Time Effect YES YES YES
Country Effect NO NO YES
Table 4.4: Output from Pooled OLS, the Driscoll–Kraay standard errors, and the
High-Dimensional Fixed Effects estimation – (Dependent variable: GHG)
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Source: Researcher’s construct using results obtained from Stata 17

4.5. Evaluation of Result

4.5.1 Evaluation Based on Economic Criteria

In line with the ordered estimation strategy outlined in Chapter 3, the evaluation of results
proceeds from the baseline Pooled OLS estimator, through Driscoll–Kraay robust inference, and
culminates in the High-Dimensional Fixed Effects (HDFE) specification. This progression
ensures that magnitudes and signs are first observed in a transparent manner, then subjected to
conservative inference checks, and finally identified within a framework that absorbs country
and time heterogeneity. The below are detailed interpretations of the coefficients in terms of their
signs, magnitudes, and conformity with economic expectations.

 Green Taxation (GTAX): Green taxation consistently shows a negative and significant
effect across all estimators. Economically, this conforms to expectation: higher
environmental taxes discourage polluting activities, incentivize cleaner production, and
reduce emissions. The magnitude is much larger in the baseline pooled OLS (–60.88)
than the HDFE and Driscoll–Kraay estimate absorbs country and year effects, conducting
robust inference. This similarity of the two estimates suggests that the effect of green
taxation is robust and not driven by spurious variation. The HDFE coefficient of –4.4861
means that, on average, a one-unit increase in green taxation reduces emissions by about
4.5 units within West African countries in the long run, holding other factors constant.
The statistical significance at the 1% level confirms that this effect is both strong and
credible.

 Regulatory Quality (RQ): Regulatory quality exhibits a negative coefficient across

specifications, indicating that stronger institutions are associated with lower emissions.
This conforms to theory, as effective regulation enhances enforcement of environmental
standards and curbs harmful practices. However, significance weakens under robust
inference, suggesting institutional effects vary across West African states. The large
negative OLS estimate (–126.1) reflects uncontrolled heterogeneity, while the Driscoll–
Kraay and HDFE result with the range of -2.6 to -2.5 shows the effect weakens once
 robust inference is applied. The variance across estimators highlights that institutional
quality matters, but its impact is uneven across West African states. The economic
implication is that while governance quality matters in the reduction of GHG emission,
alone, its impact is uneven due to differences in enforcement capacity, corruption levels,
and institutional maturity. Under HDFE, regulatory quality has a coefficient of –2.469,
negative but statistically insignificant. This implies that, on average, a one-unit
improvement in regulatory quality reduces emissions by about 2.5 units within West
African countries, though the effect is not significant.

 Green Tax–Regulatory Quality Nexus (GTAX × RQ): The interaction between green
taxation and regulatory quality is negative and significant in robust specifications. This
implies that taxation policies are more effective in reducing emissions when implemented
in countries with stronger regulatory institutions. While pooled OLS shows
insignificance, the Driscoll–Kraay and HDFE results confirm significance once
heterogeneity is absorbed. The variance across estimators reflects that the interaction
effect is masked in pooled models but becomes evident when country-specific
governance differences are controlled. Economically, this result highlights the
complementarity between fiscal instruments and governance: taxation alone may be
insufficient, but when combined with high regulatory quality, it produces stronger
environmental outcomes. The HDFE coefficient is –3.550, meaning that the joint effect
of taxation and regulatory quality reduces emissions by about 3.6 units within West
African countries in the long run.

 Foreign Direct Investment (FDI): FDI shows mixed results: negative in the baseline but
positive under robust inference. The variance reflects sectoral composition: pooled
estimates may capture cleaner inflows, but robust inference shows that FDI in West
Africa is often directed toward extractive industries, raising emissions. Economically, this
reflects the dual nature of foreign investment. On one hand, FDI can introduce cleaner
technologies and management practices, reducing emissions. On the other, investment in
resource-intensive sectors may increase pollution. The discripancies in significance of the
Discoll and Kraay and HDFE suggests that while FDI may appear to worsen emissions
 when comparing countries (as Driscoll–Kraay shows), within individual countries (as
HDFE shows) over time the effect is weak or inconsistent. indicating that FDI does not
exert a consistent within-country influence on emissions in West Africa. This underscores
the heterogeneous nature of foreign investment across the region, with its environmental
impact depending heavily on sectoral composition and institutional context. The HDFE
coefficient is 0.0861, positive but insignificant, implies that a one-unit increase in FDI
inflows raises emissions by about 0.09 units within West African countries, holding other
factors constant.

 Industrialization (LINDUS): Industrialization has a strong positive and significant effect

in the baseline, consistent with theory that industrial expansion increases emissions
through higher energy use and resource extraction. Under both robust inferences, the
effect remains positive but loses significance, reflecting heterogeneity across countries.
The pooled OLS estimate (42.35) was much larger, reflecting uncontrolled heterogeneity.
Driscoll–Kraay and HDFE both show smaller magnitudes between 1.4 – 1.9, indicating
that once country effects are absorbed, the industrialization effect is more moderate.
Economically, this conforms to theory: industrial expansion increases emissions. The
HDFE coefficient is 1.903, positive but insignificant. This means that a 1% increase in
industrialization raises emissions by about 1.9 units within West African countries in the
long run.

 Inflation (INF): Inflation shows a negative and significant effect under robust
specifications. Economically, this implies that rising prices dampen consumption and
production, thereby reducing emissions. Although inflation is not traditionally considered
an environmental variable, its impact reflects macroeconomic constraints: higher inflation
reduces demand for energy-intensive goods and services, indirectly lowering emissions.
The pooled OLS result (–0.0856) was insignificant, but robust inference strengthens the
effect. The HDFE specification further confirmed the effect, showing that once country
and year heterogeneity are absorbed, inflation exerts a robust negative impact on
emissions. Using the HDFE coefficient, on average, a one-unit increase in inflation
reduces emissions by about 0.15 units within West African countries in the long run.
 Per Capita Income (GDPPC): In the robust models, per capita income is positive and
significant under robust inference, indicating that higher income levels increase
emissions. This conforms to the scale effect in environmental economics: as incomes rise,
consumption and production expand, leading to greater environmental stress. The result
aligns with the early stages of the Environmental Kuznets Curve hypothesis, where
growth initially worsens environmental outcomes before improvements occur at higher
income levels. The HDFE coefficient is 26.81, positive and highly significant. Since
LGDPPC is logged, this means that a 1% increase in GDP per capita raises emissions by
about 26.8 units within West African countries in the long run. The pooled OLS estimate
is, however, negative and insignificant. This is because countries with lower GDP per
capita often also have lower emissions (e.g., less industrialized economies), which is a
between-country bias. When pooled together, this cross-sectional pattern dominates,
producing a negative association.

 Urbanization (LURB): Urbanization has a positive and significant effect, particularly in

the preferred HDFE specification. Economically, this reflects the environmental costs of
rapid urban expansion: increased energy demand, transportation emissions, and
infrastructure development. The HDFE coefficient is 27.69, positive and significant.
Since LURB is logged, this means that a 1% increase in urbanization raises emissions by
about 27.7 units within West African countries in the long run. The pooled OLS result
(51.69) was larger, while Driscoll–Kraay (4.566) was smaller and insignificant. The
variance reflects that pooled models exaggerate the effect, while robust inference shows a
credible positive impact. However, the HDFE was significance, suggesting that urban
expansion is a consistent driver of emissions within West African states. While
urbanization can support efficiency gains in the long run, in West Africa it currently
contributes to higher emissions due to inadequate planning, reliance on fossil fuels, and
weak regulatory oversight.

 Constant Term (C): The HDFE constant is –286.05, large and negative. This captures
baseline emission levels when explanatory variables are held constant, reflecting
 structural factors such as historical energy dependence, baseline industrial activity, and
unobserved country characteristics. The variance across estimators (–1,015 in OLS; –
153.06 in Driscoll–Kraay) shows that once heterogeneity is absorbed, the baseline effect
stabilizes.

The conformity to a priori expectation is summarized in the table below:

Table 4.5.1: A Priori Expectation Table– (Dependent variable: GHG)

Variables A Priori Output Signs Remarks

Expectation

GTAX Negative (–) Negative (–) Conforms

RQ Negative (–) Negative (–) Conforms

GTAX × RQ Negative (–) Negative (–) Conforms

FDI Positive (+) Positive (+) Conforms

LINDUS Positive (+) Positive (+) Conforms

INF Negative (–) Negative (–) Conforms

LGDPPC Positive (+) Positive (+) Conforms

LURB Positive (+) Positive (+) Conforms

Source: Researcher’s Construct

4.5.2 Evaluation Based on Economic Criteria (First-Order Tests Results)

[Link] T-test

The t-test is a two-tailed test conducted to check the individual statistical significance of each
predictor variable in the HDFE specification. The test follows the t-distribution, and the
probability value (p-value) is used as the basis of decision.

 Null hypothesis (H₀): Variable is not statistically significant.

 Alternative hypothesis (H₁): Variable is statistically significant.

 Decision rule: Reject H₀ if p-value < 0.05 at the 5% level of significance; otherwise, fail
to reject H₀.
Because the HDFE estimator absorbs country and year fixed effects, the t-test results reported
here reflect within-country variation over time, making them more reliable than pooled OLS or
Driscoll–Kraay inference.

Table 4.5.2: T-test Results for Statistical Significance (HDFE Specification)

Variables p-value Decision Rule Conclusion

GTAX 0.0000 Reject H₀ @ 5% Statistically significant

RQ 0.3010 Do not reject H₀ Statistically insignificant

GTAX × RQ 0.0959 Reject H₀ @ 10% Marginally significant

FDI 0.3270 Do not reject H₀ Statistically insignificant

LINDUS 0.1840 Do not reject H₀ Statistically insignificant

INF 0.0000 Reject H₀ @ 5% Statistically significant

LGDPPC 0.0000 Reject H₀ @ 5% Statistically significant

LURB 0.0120 Reject H₀ @ 5% Statistically significant

Constant 0.0000 Reject H₀ @ 5% Statistically significant

Source: Researcher’s construct from HDFE output.

Note: FDI is significant using the Discroll and Kraay estimate implying that its apparent effect
on emissions is driven by cross-country differences and regional shocks rather than by robust
within-country dynamics.

[Link] F-test

The F-test examines the joint significance of all explanatory variables in the model.

 Null hypothesis (H₀): Variables are not jointly statistically significant.

 Alternative hypothesis (H₁): Variables are jointly statistically significant.

 Decision rule: Reject H₀ if p-value < 0.05.

Across all three estimation methods the probability value of the F-statistic is 0.0000, which is
below the 5% threshold. This consistent outcome leads to the rejection of the null hypothesis in
each case, confirming that the explanatory variables are jointly statistically significant. The F-test
results across all three methods reinforce the robustness of the model and validate the ordered
estimation strategy adopted in the study.

[Link] Multiple Coefficient of Determination (R²)

The results show clear differences in explanatory power across the three estimation methods. In
the Pooled OLS specification, the R² is 0.788, meaning that the explanatory variables account for
about 78.8% of the variation in emissions. While this is relatively strong, it still leaves a sizeable
portion unexplained, reflecting the limitations of pooled estimation that does not control for
heterogeneity. Under the Driscoll–Kraay estimator, the R² falls to 0.5107, indicating that the
explanatory variables explain only about 51% of the variation once inference is corrected for
heteroskedasticity, serial correlation, and cross-sectional dependence. This lower value suggests
that pooled associations are weakened when robust error structures are applied. By contrast, the
High-Dimensional Fixed Effects (HDFE) specification reports an R² of 0.9999, showing that the
explanatory variables account for virtually all of the variation in emissions. The much higher R²
under HDFE arises because the model absorbs country and year fixed effects, stripping away
unobserved heterogeneity and isolating within-country variation. This demonstrates that once
heterogeneity is controlled, the predictors exert a very strong explanatory power on emissions,
making HDFE the most credible specification.

[Link] Adjusted R²

Only the HDFE specification reports an adjusted R², which is 0.9988. The closeness of this value
to the multiple R² of 0.9999 confirms that the model is parsimoniously specified and robust. It
shows that the inclusion of country and year fixed effects does not reduce explanatory power but
instead strengthens the model’s credibility by ensuring that the explanatory variables explain
almost all of the within-country variation in emissions. The adjusted R² therefore validates the
robustness of the HDFE specification and supports its selection as the preferred estimator in this
study. This reinforces the credibility of HDFE as the preferred estimator for capturing
within-country dynamics in emissions.
4.5.3 Evaluation Based on Econometric Criteria (Second-Order Tests Results) / Post
Estimation Tests

The following tests are necessary to ensure that the estimated models meet econometric
standards. They tests are designed to validate the reliability of the results. The tests results
include:

[Link] Autocorrelation Test

Baltagi (2008) emphasizes that while heteroskedasticity can be handled with robust estimators,
autocorrelation must be explicitly tested because it signals deeper problems in the error structure.
In this study, the Wooldridge test for autocorrelation was applied, which regresses residuals on
their lagged values to detect first-order serial correlation. Where autocorrelation is present,
reliance is placed on the Driscoll–Kraay estimator, which is specifically designed to correct for
both heteroskedasticity and autocorrelation in panel data.

Hypothesis:

 Null Hypothesis (H₀): Residuals are not serially correlated (no first-order
autocorrelation).

 Alternative Hypothesis (H₁): Residuals are serially correlated (presence of first-order

autocorrelation).

Decision Rule: Reject H₀ if the probability value of the Wooldridge test statistic is less than 0.05
at the 5% level of significance, and conclude that the residuals are serially correlated. Otherwise,
fail to reject H₀ and conclude that the residuals are not serially correlated.

The results of the Wooldridge Test for Autocorrelation are computed in Table 4.3.1 below

Table 4.5.3: Wooldridge Test for Autocorrelation

F-Statistic 4.471
Probability 0.0581
Source: Researcher’s construct using results obtained from Stata 17

Since the probability value (0.0581) is greater than 0.05 at the 5% level of significance, the null
hypothesis of no first-order autocorrelation is not rejected. This implies that the residuals are not
serially correlated across time, and the model satisfies the independence of errors assumption.

[Link] Multicollinearity Test

Multicollinearity inflates the variances of estimated coefficients, making it difficult to isolate the
individual effect of each explanatory variable. As Naluba et al. (2023) explain, the Variance
Inflation Factor (VIF) is a widely used diagnostic tool for detecting multicollinearity in
econometric models, since it quantifies how much the variance of a regression coefficient is
increased due to collinearity among regressors. While robust estimators like HDFE can absorb
unobserved heterogeneity, they cannot fully eliminate collinearity among regressors. This
ensures the standard errors are not inflated.

Hypothesis:

 Null Hypothesis (H₀): No serious multicollinearity (VIF ≤ 10).

 Alternative Hypothesis (H₁): Presence of serious multicollinearity (VIF > 10).

Decision Rule: Reject H₀ if any VIF value exceeds 10, and conclude that the model suffers from
multicollinearity. Otherwise, fail to reject H₀ and conclude that multicollinearity is not a major
concern.
Table 4.5.4: Variance Inflation Factor (VIF)

Variable VIF 1/VIF

GTAX*RQ 12.75 0.078417
GTAX 12.36 0.080919
LGDPPC 9.4 0.106374
LURB 5.76 0.173646
RQ 4.57 0.218581
LINDUS 2.7 0.370613
INF 2.24 0.446657
FDI 1.23 0.810777

Mean VIF 3.92

Source: Researcher’s construct from Stata 17 output.

The VIF results reveal that while the overall model does not suffer from serious multicollinearity
(mean VIF = 3.92), two variables (GTAX*RQ and GTAX) exceed the threshold of 10,
suggesting strong collinearity. This is common among interaction terms as they are mechanically
correlated with their components (GTAX and RQ). This inflates their standard errors; however, it
does not bias coefficients. Therefore, while the model remains valid for joint inference.

4.6 Evaluation of Research Hypotheses

For this study, one hypothesis were specified and formulated to capture the main objectives.
While the other two were specified to capture the two specific objectives. The hypotheses
include:

H01: Regulatory quality has no significant impact on the green taxation – greenhouse gas
emission nexus in West Africa.

H02: Green taxation has no significant impact on greenhouse gas emission in West Africa.

H03: Regulatory quality has no significant impact on greenhouse gas emission in West Africa.
Hypothesis One:

Using the p-value to check for statistical significance, we reject the null hypothesis if the p-value
is less than or equal to 0.05 at the 5% level of significance. The HDFE estimation shows that the
interaction term between Green Taxation and Regulatory Quality (GTAX*RQ) is statistically
significant, with a p-value below 0.05. Therefore, we reject the null hypothesis and conclude that
the Regulatory Quality has a significant impact on the green taxation – greenhouse gas emission
nexus in West Africa.

This finding implies that taxation policies alone are not sufficient; their effectiveness are
supported by the quality of regulatory institutions. Strong regulatory frameworks ensure that
revenues from green taxes are properly enforced, monitored, and directed toward environmental
protection. In weak regulatory environments, taxation may fail to achieve sustainability goals
due to corruption, poor enforcement, or policy inconsistency. This result is consistent with
Obomeghie (2025), who found that fiscal instruments combined with institutional quality
significantly enhance sustainability outcomes in West African states. It also aligns with Addai et
al. (2022), who emphasized that regulatory quality moderates the effectiveness of environmental
policies in emerging economies.

Additionally, the HDFE results reveal that Green Taxation exerts a stronger direct effect on
greenhouse gas emissions than the moderating impact of Green Taxation and Regulatory Quality.
This suggests that fiscal instruments are the primary driver of environmental sustainability in
West Africa, while regulatory quality plays a complementary role. Although the interaction term
is statistically significant, its smaller magnitude indicates that regulatory quality enhances but
does not overshadow the effect of green taxation. This finding is consistent with studies such as
Ezeilo et al. (2024) and Sadiq & Durowaiye (2025), which emphasize the central role of green
taxation in promoting sustainability, while also acknowledging the importance of institutional
quality in ensuring effective implementation.

Hypothesis Two (H₀₂):

Applying the same decision rule, the HDFE results show that the coefficient of Green Taxation
(GTAX) is statistically significant with a p-value less than 0.05. Hence, we reject the null
hypothesis and conclude that Green Taxation has a significant impact on greenhouse gas
emission in West Africa.

This finding highlights the role of fiscal instruments in discouraging environmentally harmful
practices and incentivizing sustainable production. Green taxes increase the cost of pollution,
thereby encouraging firms and households to adopt cleaner technologies and reduce emissions.
In the West African context, where rising affluence and urbanization are accelerating, taxation
provides a direct mechanism to internalize environmental costs. This result is supported by
Ezeilo et al. (2024), who demonstrated that environmental taxation promotes corporate
sustainability practices in Nigeria. Similarly, Sadiq and Durowaiye (2025) argue that green
taxation provides incentives for sustainable practices across energy, transportation, and industrial
sectors, making it a vital tool for achieving environmental goals.

Hypothesis Three (H₀₃):

The HDFE estimation shows that Regulatory Quality (RQ) has a negative coefficient but is not
statistically significant at the 5% level. Therefore, we fail to reject the null hypothesis and
conclude that RQ does not exert a significant impact on greenhouse gas emissions in West
Africa.

This outcome suggests that weak enforcement capacity and institutional inefficiencies limit the
effectiveness of regulatory frameworks in driving sustainability outcomes. Although theoretically
important, regulatory quality, alone, in practice appears insufficient to influence emissions
reduction in the region. This finding contrasts with Asongu (2018), who emphasized that
institutional quality is a critical determinant of sustainable development in Africa.

This reflects institutional weaknesses rather than irrelevance. Strengthening governance

frameworks remains essential to complement fiscal instruments like green taxation in achieving
sustainability in West Africa.