4.1. Univariate Analysis
Based on
Table 1 and
Table 2, below, renewable energy consumption is on average higher for oil-exporting countries (23.25) than for oil-importing countries (12.00). This could be true because oil-exporting countries have higher budgets and may be more able to invest in and consume new costly technologies, compared with oil-importing countries that usually have energy import cost burdens affecting their budgets. The minimum renewable energy consumption for exporting countries is zero, which indicates that not all countries have equal renewable energy consumption. This is also evident with the high standard deviation (29.803) of REC for oil exporter countries.
Moreover, the oil rents for oil-exporting countries (14.22) are higher than the oil rents for oil-importing countries (0.287), which reflects the fact that most exporters of oil benefit from oil rents. Furthermore, the net energy import value for oil exporter countries is negative 216.679, which indicates an export value of energy of 216.679. For non-exporters of oil, the net energy import average value is a positive 57.202, indicating an import of energy of 57.202. The standard deviation of the net energy import variable for exporter countries is 249.799, indicating a large discrepancy in the export power of the sampled countries. The average economic growth rates for exporters and non-exporters countries are close, which indicates that the two samples are comparable. In addition, the average energy use per capita for exporters and non-exporters countries is also very close, which also indicates that the countries on average have similar behaviors in terms of energy use and consumption, making the sample comparable.
For the years between 1990 and 2018, the average price of crude oil was USD 47.533, reaching a high of USD 99.670 in the year 2008 and a minimum of USD 14.42 in the year 1998.
The correlation matrices for both samples are provided in
Table 3 and
Table 4, below.
Table 3 shows the correlation matrix for the oil-exporting countries. It was evident that per capita growth rate was highly correlated with GDP growth rate; thus, it was dropped from the model.
For the oil-importing countries, the correlation matrix shows that per capita growth rate highly correlates with GDP growth rate; thus, it was dropped from the model. On the other side, urban population also highly correlated with several variables including OR, NEIs, and EUPC; thus, it was also dropped from the model.
Panel cointegration testing was performed for both samples and based on the augmented Dickey–Fuller results, it was concluded that there was no long-term cointegration between the variables for the sample of oil exporters, with a p-value of 0.1365; the null hypothesis that there is no cointegration is retained. Similarly for the oil importers sample, the augmented Dickey–Fuller test resulted in a p-value of 0.7961, also indicating no cointegration between the variables.
4.2. Unit Root Tests
Unit root testing was performed for all the variables in the model, for each of the samples.
The results for oil-exporting countries are shown in
Table 5, below.
The results for oil-importing countries are shown in
Table 6, below.
Table 7, below, shows the result of the regression tests applied using Equation (1) for the two samples, oil-exporting and oil-importing countries.
Oil-Exporting Countries:
The least squares estimation was performed; the R-squared value was very small (0.049) and the coefficients of all the variables were insignificant at a 5% level of significance.
A cross-section dependence test was performed; the Breusch–Pagan LM (0.01) and Pesaran CD (0.00) p-values were lower than the 5% level of significance; thus, the null hypothesis that there is no cross-sectional dependence is rejected. Cross-sectional SURs (seemingly unrelated regressions) were selected to correct for that.
Using cross-sectional SUR, most of the variables are now significant with an improved R-squared value as seen in
Table 8.
The cross-sectional dependence test was performed again, and the Breusch–Pagan LM (0.99) and Pesaran CD (0.31) p-values were higher than the 5% level of significance; thus, the null hypothesis that there is no cross-sectional dependence is accepted.
Due to insufficient instruments, GMM with random effects could not be performed. Since the cross-sectional dependence affects the estimation with random effects, as the unobserved individual-specific effects are assumed to be uncorrelated with the independent variables, which in this case is not valid, then estimating the model with fixed effects allows us to correct for cross-sectional dependence.
Table 9, below, provides the panel regression results with fixed effects.
All the coefficients were found to be insignificant with very small R-squared values. The cross-sectional dependence test was performed, and the Breusch–Pagan LM (0.00) and Pesaran CD (0.00) p-values were lower than the 5% level of significance; thus, the null hypothesis that there is no cross-sectional dependence is rejected. Cross-sectional SURs (seemingly unrelated regressions) were selected to correct for that.
To ensure that there was no cross-sectional dependence, the test was performed and the Breusch–Pagan LM (1.00) and Pesaran CD (0.36)
p-values were higher than the 5% level of significance as seen in
Table 10; thus, the null hypothesis that there is no cross-section dependence is accepted.
For oil-exporting countries, the fixed-effect model results are provided in the below equation:
The equation shows that, for oil-exporting countries, there was a positive significant relationship between energy use per capita (EUPC) and renewable energy consumption (REC). Renewable energy consumption (REC) increases by 0.00022 for each energy use per capita (EUPC) unit increase. In addition, the results of the regression showed that there was a negative significant relationship between GDP growth (G) and renewable energy consumption (REC). Renewable energy consumption (REC) decreased by 0.0163 for each 1% economic growth. Net energy imports (NEIs) had a negative significant relationship with renewable energy consumption (REC). Renewable energy consumption (REC) decreased by 0.00378 for each unit increase in net energy imports (NEI). Moreover, the regression results showed a negative significant relationship between oil rents (OR01) and renewable energy consumption (REC). Renewable energy consumption (REC) decreased by 0.0095 for each dollar increase in oil rents (OR01).
Crude oil average closing price (COP) and inflation (I) had an insignificant impact on the level of renewable energy consumption (REC) for oil-importing countries.
Oil-Importing Countries:
The least squares estimation was performed; the R-squared value showed a low percentage (0.29), and the model was significant at a 5% level of significance.
A cross-sectional dependence test was performed, in which the Breusch–Pagan LM (0.00) and Pesaran CD (0.09)
p-values were mostly lower than the 5% level of significance as seen in
Table 11; thus, the null hypothesis that there is no cross-sectional dependence is rejected. Cross-sectional SURs (seemingly unrelated regressions) were selected to correct for that.
For oil-importing countries,
Table 12 provides the least squares regression results after correcting for cross-sectional dependence followed by the cross-sectional dependence test.
Growth, inflation, and NEIs were significant while the other variables were not. The R-squared value was calculated to be 0.41, and the cross-sectional dependence test was performed, in which the Breusch–Pagan M (1.00) and Pesaran CD (0.66) p-values were mostly lower than 5% level of significance; thus the null hypothesis that there is no cross-sectional dependence is rejected.
Regression with fixed effects was also performed, and the results are shown in
Table 13, below. The R-squared value was 0.31, and only growth rate, inflation, and NEIs were significant.
The cross-sectional dependence test was performed, and the Breusch–Pagan LM (0.0002), Pesaran scaled LM (0.0000), and Pesaran CD (0.0804)
p-values were mostly lower than the 5% level of significance; thus, the null hypothesis that there is no cross-sectional dependence is rejected.
Table 14, below, provides the regression results after correcting for cross-sectional dependence.
The cross-sectional dependence test was performed, and the Breusch–Pagan LM (1.000), Pesaran scaled LM (0.0000), and Pesaran CD (0.7673) p-values were mostly higher than the 5% level of significance; thus, the null hypothesis that there is no cross-section dependence is accepted.
The results of the corrected fixed effects model indicated that EUPC, G, NEIs, and ORs were significant with R-squared of 0.52. The results are provided in the below equation:
The equation shows that, for oil-importing countries, there is a negative significant relationship between energy use per capita (EUPC) and renewable energy consumption (REC). Renewable energy consumption (REC) decreases by 7.75 × 10−5 for each energy use per capita (EUPC) unit increase. In addition, the results of the regression show that there is a negative significant relationship between GDP growth (G) and renewable energy consumption (REC). Renewable energy consumption (REC) decreases by 0.05 for each 1% economic growth. Net energy imports (NEIs) have a negative significant relationship with renewable energy consumption (REC); renewable energy consumption (REC) decreases by 0.03 for each net energy imports (NEIs) unit increase. Moreover, the regression results show a positive significant relationship between oil rents (OR01) and renewable energy consumption (REC). Renewable energy consumption (REC) increases by 0.04 for each dollar increase in oil rents (OR01).
Crude oil average closing price (COP) and inflation (I) have insignificant impact on the level of renewable energy consumption (REC) for oil-importing countries.
Comparative discussion of Oil-Exporting Countries vs. Oil-Importing Countries:
For both oil-exporting and oil-importing countries, crude oil average closing price (COP) and inflation (I) have insignificant impact on the level of renewable energy consumption (REC). Our findings are contradictory to the findings of Slav [
3] and Deniz [
4], who found a positive significant relationship between the crude oil average closing price and renewable energy consumption. The result is also inconsistent with Boyer and Filion [
5], Le [
70], and Deane [
71]’s findings on decrease in fuel dependency in Europe because of crude oil price increases.
Our findings show that for oil-importing countries, there is a negative significant relationship between energy use per capita (EUPC) and renewable energy consumption (REC), but there is a positive significant relationship between energy se per capita (EUPC) and renewable energy consumption (REC) for oil-exporting countries. This result is divergent from studies conducted in Africa [
7]; there is a negative and significant correlation between oil revenues and renewable energy consumption in the most developed and populous African countries, such as Nigeria, Democratic Republic of Congo, and South Africa, which, while not defined as such by the authors, are considered oil-exporting countries. Similarly, in the works of Eder, Provornaya, Filimonova, Kozhevin, and Komarova [
6] and Boyer and Filion [
5] in Canada, and Makki, Mawad, and Khalili [
9], where no distinction was made between the two types of countries, the results were also contradictory to studies in the Balkan countries [
8], importers of energy where no significant relationship was found between oil rents and renewable energy consumption.
In addition, our findings show that there is a negative significant relationship between GDP growth (G) and renewable energy consumption (REC) for oil-exporting countries and oil-importing countries. Economic growth is significantly related to renewable energy consumption at the 5% significance level; a 1% increase in economic growth leads to a decrease in renewable energy consumption by 0.01 for oil-exporting countries, and 0.05 for oil-importing countries. The direction of the relationship is similar for both samples, which indicates similar conduct and behavior for countries in terms of growth whatever their net import of energy status. This result contradicts the findings of Makki, Mawad, and Khalili [
9] and Doytch and Narayan [
10].
Moreover, the regression results show a significant relationship between oil rents (OR01) and renewable energy consumption (REC) for both oil-importing countries and oil-exporting countries, but in different directions. Renewable energy consumption (REC) increases by 0.04 for each dollar increase in oil rents (OR01) for oil-importing countries but decreases by 0.0095 for each dollar increase in oil rents (OR01) for oil-exporting countries. Our findings match those of the work of Olanrewaju, Olubusoye, Adenikinju, and Akintande [
7] and Eder, Provornaya, Filimonova, Kozhevin, and Komarova [
6] in African oil-exporting countries.
Finally, and most importantly, net energy imports are significantly related to renewable energy consumption at a 1% significance level, similar to the findings of Wei, Ni, and Shen [
11], Solangi, Islam, Saidur, Rahim, and Fayaz [
12], and Schmidt, Gruber, Klingler, Klöckl, Camargo, Regner, Turkovska, Wehrle and Wetterlund [
15] and contrary to the findings of York [
13] and York [
14]. An increase in net energy imports of 1% decreases renewable energy consumption by 0.00378 for oil-exporting countries and by 0.03 for oil-importing countries, which is ten times more.