1. Introduction

The literature identifies that energy and long-run economic growth are closely associated in countries around the globe (Baz et al., 2019; Rahman & Velayutham, 2020; Salisu & Ogbonna, 2019). However, sustainable growth is not independent of environmental problems. Therefore, it is imperative that the empirical research explicitly includes both energy consumption and environmental indicators (Gorus & Aydin, 2019).

Admittedly, addressing economic growth only in the context of energy and the environment can have misleading results, while the complexity of the real world requires certain indicators that may be directly related to growth. The impact of foreign direct investment (FDI) on growth in relation to liberalization policies is also important (Anwar & Nguyen, 2011; Omri & Kahouli, 2014), given the profound role of FDI in new investments, up-to-date technologies, and managerial skills for the host countries (Pegkas, 2015). International trade, along with associated liberalization developments, seem to be also important for economic growth (Frankel & Romeri, 1999; Musila & Yiheyis, 2015). The literature emphasizes that economic growth also heavily depends on trade openness (Grossman & Helpman, 1991; Spilimbergo, 2000). In contrast, as the share of the state in the economy is gradually shrinking, the role of government expenditures in economic growth is another important determinant, although its impact has generated mixed results (Kelly, 1997; Knoop, 1999).

The goal of this paper is to explore the nexus between energy consumption (EC), carbon dioxide emissions (CO₂), gross capital formation (GCF), labor force (L), FDI, government expenditures (GE), trade openness (TO) and economic growth (EG) across 57 countries, spanning the period 1990-2018. The study contributes in three ways. First, the energy-environment-growth nexus is considered along with GCF, L, FDI, and GE. Second, the analysis focuses on 57 countries from the European and Asian regions (22 countries), Latin American and Caribbean regions (15 countries), and Middle Eastern and African regions (20 countries). Finally, the analysis adopts a panel quantile regression model. The advantage of this method is that it allows not only heterogeneity across countries, but also provides different estimates, with respect to the mean, at the tails of the distribution.

2. Literature review

In the literature that uses panel-country studies, Saidi and Hammami (2015) reach the conclusion that for 58 countries (i.e. European and North Asian, Latin American and Caribbean, and Middle Eastern, North African and Sub-Saharan), there is a positive impact of EG on EC, while CO₂ emissions cause higher energy consumption at a global level. In the case of European and North Asian, and Latin American and Caribbean countries, higher levels of EG and CO₂ emissions contribute to stronger EC. Similar country groups have also been employed by Acheampong (2018), who finds that EC negatively affects EG at the global level. Chen et al. (2016) show that energy use negatively impacts GDP. Gorus and Aydın (2019) consider the case of MENA countries. Their results signify the absence of any causal relationship between EG and environmental pollution, while Muhammad (2019), in the case of emerging and MENA countries, illustrates that EG improves with EC in both developed and emerging countries, while it declines in the case of MENA countries. He finds that $\text{CO}_{2}$ emissions increase across all countries due to stronger EC.

3. Model and Methodology

The empirical analysis makes use of a basic production function accounting framework:

$$GDP = f(EC,\ \text{CO}_{2},\ FDI,\ L,K,GE,TO)$$

where $\textit{GDP}$ denotes real $\textit{GDP}$.

To the best of our knowledge, this paper is the first attempt to apply panel quantile regressions to explore the association between CO₂, EC and EG. A special advantage is these estimators are robust to outliers and skewed distributions. Moreover, allowing for unobserved heterogeneity, this method enables the exploration of differences in the growth rates among slow, moderate and fast-growing cases. In our sample, certain countries can grow faster than others; by contrast, OLS regressions provide estimates that represent the effect from an independent variable on the ‘average country’. The analysis employs the method proposed by Koenker (2004), which uses the idea of penalized least squares interpretation of the classical random-effects estimator.

4. Data

The analysis uses annual data, spanning the period 1990 to 2018 for 57 countries belonging to three groups, i.e. the European and Asian region (22 countries), the Latin American region (15 countries), and the Middle Eastern, North African, and the Sub-Saharan region (20 countries). Data on real GDP (constant 2010 US dollars), CO₂ emissions (metric tons per capita), EC (in kg of oil equivalent per capita), FDI net inflows as percentage of GDP, GCF (in constant 2010 US dollars), TO (the percentage of the sum of imports and exports to GDP), GE as percentage of GDP, and L measures (total workers) are obtained from the World Development Indicators database. All data are in logarithm form except those already in percentage terms.

5. Empirical analysis

Table 1 reports the estimates for the full sample per quantile (i.e., low-growth-10-25%; medium-growth-50%; and high-growth-75-90%). Focusing on EC, the results show that EG is heavily dependent on EC across all growth quantiles. The coefficient appears positive and statistically significant only up to medium-growth levels, where then turns out to be negative. Similarly, in terms of CO₂ emissions, the findings illustrate a positive effect on EG up to medium-growth levels, while at high levels of EG the coefficient turns out to be negative. In other words, at low growth levels, the growth process seems capable of sustaining environmental pollution. By contrast, at high levels, the evidence shows that countries do not invest in technologies that are environmentally friendly, implying that countries cannot achieve sustainable growth. Accordingly, although it is expected that strong EG records can be high-energy efficient, compensating for the presence of increased emissions, does not hold in our full-country case.

The findings come in conflict with the literature that performs analysis with respect to the mean of the distribution, indicating the negative effect of EC on EG at high growth levels, thus, putting in jeopardy the validity of the growth hypothesis at these levels.

Table 2 reports the estimates per geographical region. CO₂ and EC reveal a negative effect on EG at high levels, implying that European countries cannot exert an efficient control over sustainable growth targets at high EG rates. In the Latin American countries, the estimates document that CO₂ (at all levels) and EC (at the medium and high levels) exert a negative impact on EG, i.e. the inability of these countries to achieve sustainable EG. Finally, the results for the Middle East and African countries highlight that EC cannot support EG targets at high levels due to the presence of detrimental environmental effects.

6. Conclusion

This paper examines the impact of EC and CO₂ emissions on EG for 57 countries from three different geographical regions over the period 1990-2018. The results with respect to the full sample and the European and Asian regions suggest that EG is positively influenced by CO₂ and EC up to the medium growth levels, whereas it turned negative. The results imply that these economies witness deviations from sustainability targets at high growth rates. The Middle East and African countries can achieve sustainability up to high levels of growth, but they are affected by EC and CO₂ after the medium growth levels. The evidence from the Latin American countries indicates a negative effect from CO₂ emissions across all growth levels, i.e. they fail to achieve more sustainable growth.

The findings imply that European countries have come up with pioneering policies in terms of climate change and sustainability, but not at high growth levels. They need some long-run policies towards environmental investments. This is similar for the Asian, Middle East and African country groups. It is clear that more fundamental reforms are needed by the Latin American countries. They should encourage more environmentally friendly activities to mitigate carbon emissions.

Acknowledgement

The authors would like to thank the Editor and anonymous referees of Energy RESEARCH LETTERS.

Appendix

Country samples.

European and Asian (22 countries): Albania, Belgium, Bulgaria, Denmark, France, Germany, Greece, Hong Kong, Hungary, Iceland, Ireland, Italy, Japan, the Rep. of Korea, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, the UK.

Latin American (15 countries): Argentina, Bolivia, Brazil, Chile, Costa Rica, Ecuador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, Uruguay, Venezuela.

Middle East and African (20 countries): Algeria, Botswana, Cameroon, the Rep. of Congo, Cote d’Ivoire, Egypt, Ethiopia, Gabon, Ghana, Iran, Jordan, Kenya, Morocco, Mozambique, South Africa, Senegal, Sudan, Togo, Tunisia, Zambia.