**4. Results and Discussion**

As outlined above, the first step in our method is to observe whether series are generating common shocks in the long run. To this aim, we test for cross-section dependence in our panel time-series data. The outcomes of the Pesaran cross-sectional dependency test [39] are shown in Table 3. The test results rejected the null hypothesis and confirmed the presence of cross-country dependency, which is not unexpected because the European Union countries share a common market and economic policy. A number of the conducted studies point to systematic economic convergence between these countries in recent years, for example, see Józwik [ ´ 48] or Bernardelli et al. [49]. Because of this convergence, one country's economic and environmental transformations can easily be transferred to its neighboring countries. Therefore, we need to use a proper stationarity approach to circumvent the common effect and provide reliable results [45].

In the second step, we identify the order of integration of the variables by employing the Im–Pesaran–Shin panel unit root test. We subtracted the cross-sectional averages from the series and requested that the number of lags of the series be chosen in such a way that the AIC for the regression is minimized (max AIC is four). The stationarity test results in Table 4 confirm that the data series is unstable at this level. However, after considering the first difference, the test confirmed that the series became stationary at the 1% significance.


**Table 3.** Results of cross-sectional dependency Pesaran test.

Note: Under the null hypothesis of cross-section independence CD ~ *n*(0,1). \*\*\* denotes statistical significance at the 1% level.


**Table 4.** Im–Pesaran–Shin Panel unit root test (W-t-bar statistics).

Notes: H0: All panels contain unit roots. Ha: Some panels are stationary. Cross-sectional means removed. Max AIC is 4. The number of lags of the series is chosen in such a way that the AIC for the regression is minimized. \*\*, \*\*\* denote statistical significance at the 5% and 1% levels, respectively.

In addition, we employed the panel second generation unit root test in the presence of cross-section dependence proposed by Pesaran [41]. We assume that the serial correlation order to be tasted with the Breusch–Godfrey Lagrange multiplier test in each regression is one, and the number of lags is four. Table 5 displays results for two deterministic models' specifications: with individual-specific intercepts and incidental linear trends. The test results confirm that the variables are stationary at the first difference, almost all at the 1% significance.

**Table 5.** Pesaran panel unit root test in the presence of cross-section dependence.


Notes: critical values: at \*\*\*—1% significant level is −2.32 and at \*\*—5% is −2.15; the serial correlation order to be tasted with the Breusch–Godfrey Lagrange multiplier test in each individual regression is 1; the number of lags is 4.

As noted previously, environmental degradation can be proxied in various ways. We selected CO2 emissions as a proxy for environmental degradation in the model I, but for a robustness check, we also used the greenhouse gas emissions per capita variable. In this respect, we checked the cointegration (long-run relationship) between variables using the Pedroni and Westerlund tests. The tests have a common null hypothesis of no cointegration. The alternative hypothesis of the Pedroni tests is that the variables are cointegrated in all panels. In the version of the Westerlund test in which the AR parameter is panel specific, the alternative hypothesis is that the variables are cointegrated in some of the panels. In the version of the Westerlund test in which the AR parameter is the same over the panels, the alternative hypothesis is that the variables are cointegrated in all the panels. In the Pedroni tests, we subtracted the cross-sectional averages from the series and requested that the number of lags of the series be chosen in such a way that the AIC for the regression is minimized (max AIC is four), as in the panel unit root tests calculations. Table 6 reports results for cointegration where six out of seven Pedroni tests confirm cointegration in Model I and Model II. The results of the Westerlund tests indicate that the variables are cointegrated in some of the panels.

**Tests Model I Model II** Pedroni test AR parameter: Same Modified variance ratio −3.2956 \*\*\* −3.5891 \*\*\* Modified Phillips–Perron t 0.9245 1.1932 Phillips–Perron t −8.0501 \*\*\* −6.8419 \*\*\* Augmented Dickey–Fuller t −10.4652 \*\*\* −10.1033 \*\*\* Pedroni test AR parameter: Panel specific Modified Phillips–Perron t 3.0399 \*\*\* 3.1541 \*\*\* Phillips–Perron t −8.6913 \*\*\* −7.3555 \*\*\* Augmented Dickey–Fuller t −12.4472 \*\*\* −11.0408 \*\*\* Westerlund test AR parameter: Same Variance ratio −1.2798 −1.3705 \* Westerlund test AR parameter: Panel specific Variance ratio −2.5575 \*\*\* −2.6808 \*\*\*

**Table 6.** Pedroni and Westerlund panel cointegration tests results.

Notes: Westerlund test AR parameter: Same. Ha: All panels are cointegrated; Panel specific. Ha: Some panels are cointegrated. \*, \*\*\* denote statistical significance at the 10% and 1% levels, respectively.

In the final step, we estimated the coefficients of Equations (1) and (2). Table 7 provides the FMOLS test results. To test the robustness of the estimated results, we used the Pesaran and Smith Mean Group estimator, the Pesaran Common Correlated Effects Mean Group, and the Augmented Mean Group estimators. These tests, which concern with correlation across panel members (cross-section dependence), were introduced by Eberhardt and Teal [50] and Bond and Eberhardt [51]. The advantage of using these estimators is that they are designed for 'moderate-T and moderate-N' macro panels. These results are presented in Table 8.

**Table 7.** Panel FMOLS test results.


Notes: The number of observations is 532. *p*-value for two-tailed hypothesis. \*\*\* denotes statistical significance at the 1% level.


**Table 8.** Mean Group (MG), Augmented Mean Group (AMG), and Common Correlated Effects Mean Group (CCEMG) estimation results.

Notes: numbers in parentheses are *p*-value. \*\*\*, \*\*, \* denote statistical significance at the 1%, 5%, and 10% level, respectively.

As can be seen from Tables 7 and 8, in both models, the significant coefficients of the real GDP per capita are positive, whereas those of the squared GDP per capita are negative. It means that the long-run linkage between CO2 emissions per capita and GDP per capita is an inverted U-shape implying that the environmental Kuznets curve concept is verified for the whole group of the European Union countries. The economic growth and development are supportive of carbon emissions in that region. Similar results of studies for the group of European countries were recently obtained by Destek et al. [20], Maneejuk et al. [9], and Verbiˇc et al. [19], as well as by Balsalobre-Lorente et al. [18]. Despite significant technological advances in the European Union countries, energy consumption still positively influences carbon emissions per capita. Our results are similar to the papers mentioned earlier, as well as to those that have been published recently, namely, Khezri et al. [17], Kar [52], and Mohsin et al. [53].

Interestingly, all results from Tables 7 and 8 indicate that urbanization negatively impacts carbon emissions per capita. However, there are differences in the significance level of the coefficients depending on the method used. Nevertheless, this shows that urbanization has an essential effect on environmental protection in the European Union area, nowadays. For example, the results of FMOLS for model I show that a 1% increase in a share of the urban population decreases emissions per capita by 5.02% if all other variables remain the same. We want to highlight that our study on the urbanization process' effect on carbon dioxide emissions points to different results than many studies mentioned in the literature review section. As we remember, the significant results indicate that urbanization positively impacted the carbon dioxide emissions in different groups of European countries. To give an example, in an article by Ali et al. [15], coefficients are positive and equal to 0.188 and 0.011; in the study by Balsalobre-Lorente et al. [18], between 0.44 and 6.36; while in the studies by Destek et al. [20] and Amin et al. [21] there was no relationship. This difference probably results from a few reasons, some of which include research methods, samples, and periods. Another reason can be related to trespassing the threshold after which both increased income per capita and the coefficient of urbanization give rise to improved quality of the environment, which was indicated by Gierałtowska et al. [11]. This effect can be enhanced by the deindustrialization process we wrote about in the literature review. Dong et al. [54] highlighted that from the perspective of income level, industrialization contributes to the growth in carbon emissions. The effect of industrialization on CO2 emissions gradually increases in the low- and intermediate-income

levels. Azam et al. [55] also state that the industrialization process in OPEC economies increases environmental pollution, while the impacts on income are the opposite. However, the effect of industrialization begins to weaken at the high-income level according to research conducted by Dong et al. [54]. Probably this effect is observed in the European Union countries with a high-income level. Furthermore, economic development supports human capital, which significantly improves environmental quality [56]. Thus, our results indicate that studies in this area should be extended to different research models and methods.

#### **5. Conclusions and Recommendations**

In our research, we took into consideration two trends. First is the urbanization process, which increased the urban population in most European Union countries in years 2000–2018, and the second trend is a decrease in carbon dioxide emissions, which is indirectly the consequence of technological advances and the applied European climate policy. Considering these two trends, our research goal was to answer the following question: is there a long-run relationship between urbanization, energy consumption, economic growth, and carbon dioxide emissions, and what roles do urbanization and energy consumption play in the concept of the environmental Kuznets curve in European Union countries? We used the data from 28 European Union countries to assess the relationships. Our findings confirmed the long-run relationship between variables. We validated the environmental Kuznets curve hypothesis, indicating that economic growth has an inverted U-shaped effect on CO2 emissions.

However, energy consumption still positively influences carbon emissions per capita, even though European Union countries have made significant economic and technological progress. At the same time, urbanization has a highly negative impact on carbon dioxide emissions per capita. If all other variables remain the same, a 1% increase in a share of the urban population decreases CO2 emissions per capita by 5.02%. The result of our study is different from the results in the majority of earlier published articles. This difference probably arises from a few reasons. One of them may be the fact that the threshold after which both an increase in income per capita and urban population causes a decrease in carbon dioxide emissions in European Union countries has been trespassed in recent years.

Our results provide new insights for policymakers in European Union institutions. The findings suggest that the European policy should support the process of urbanization in a complex manner to fulfill the European Green Deal and achieve the Sustainable Development Goals related to improving environmental quality, especially by promoting urbanization with a low-carbon infrastructure and transport (smart technology and energyefficient hybrid vehicles). A positive coefficient associated with energy consumption indicates that local authorities should support the development of home renewable energy infrastructure, for example, energy-efficient electric appliances and solar energy. Another important practical implication is related to human capital. The urban population can be motivated to adopt a sustainable lifestyle, including energy-saving, renewable energy sources, and public transportation [56]. It is very important in this context that urbanization be carried out according to environmental norms, possibly without social compromises in this respect. In addition, modern technological solutions enable the development of intelligent cities that are environmentally neutral.

However, we only conducted a preliminary empirical analysis of the relationship between environmental degradation and urbanization, and our study has a few limitations. The first limitation refers to sample size. The sample covers the period 2000–2018, this means we should be cautious in generalizing the findings. Second, although we have robust results using an alternative measure of environmental degradation, the two proxies (CO2 and greenhouse emissions) might limit the ecological degradation effects. Additionally, it would be interesting to examine the consumption environmental impacts offshored to other countries and the deindustrialization processes. Third, we did not divide the European Union countries, for example, into less developed countries (Central European countries) and developed countries (Western European countries) to make a comparative analysis. These limitations could be addressed in future research.

Undoubtedly, in further research, we must also remember that climate neutrality is a global challenge. This means that it requires international dialogue and cooperation between the states. Although the pressure applied usually refers to particular countries and their economic structures, international activity is also an issue that plays a predominant role. It is especially important due to the necessity of creating a synergy between the European and international climate initiatives. For this reason, understanding that adaptation to climate changes is important; however, this is not in itself the aim, but rather a principle. It should, however, be a component of properly functioning and developing countries and societies.

**Author Contributions:** Conceptualization, B.J. and K.G.; methodology, B.J. and A.-V.G.; software, B.J.; formal analysis and investigation, B.J., K.G. and A.-V.G.; writing—original draft preparation, B.J., K.G. and A.-V.G.; writing—review and editing, B.J.; visualization, B.J.; project administration, B.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** The APC was funded by the John Paul II Catholic University of Lublin.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Publicly available datasets were analyzed in this study. This data can be found here: https://data.worldbank.org and https://ec.europa.eu/eurostat.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

**Table A1.** CO2 emissions per capita and urban population in the European Union countries.


Source: World Development Indicators.
