Energy Consumption and CO₂ Emissions

Numerous previous studies have focused on the long-term relationship between economic growth and energy consumption. Energy consumption relying on burning fossil fuels will lead to greenhouse gas emissions, which lead to climate change and environmental degradation [21]. An increase in energy use and population growth causes an increase in CO₂ emissions [22]. For economic development, energy will be consumed rapidly, and it also leads to a large number of greenhouse gas emissions, especially in China [23-24] showing that energy consumption has a positive impact on long-term CO₂ emissions. In Nigeria, carbon consumption has a significant negative impact on CO₂ emissions [8]. Energy usage intensity can increase CO₂ emissions [25]. Asumadu-Sarkodie & Owusu investigated the relationship between carbon dioxide, electricity production, and consumption in Ghana by using time series data from 1971 to 2012 [26]. Evidence from long-term resilience suggests that in the long run, a 1% increase in total energy production from combustible renewable energy and waste will lead to an increase of 307.9 kt in CO₂ emissions. In contrast, in the long run, a 1% increase in total energy production from hydropower will reduce 267.3 kt of CO₂ emissions.

The main approach to reducing CO₂ emissions is to develop renewable energy. Ghana’s nonrenewable energy generation and consumption have increased CO₂ emissions, and renewable energy can help mitigate climate change and its impacts [26]. Due to renewable energy and foreign trade in the world, traditional energy has decreased demand in recent decades, and CO₂ emissions have obviously increased [27]. Zhang & Zhao believe that investment in R&D and renewable energy plays an important role in reducing CO₂ emissions [28]. The development of hydropower, nuclear power, and other renewable energy has a negative impact on CO₂ emissions [29-30]. Per capita GDP and oil consumption are positively correlated with CO₂ emissions, while natural gas consumption has a negative impact on CO₂ [31]. Generally speaking, renewable energy consumption increases energy self-sufficiency, stimulates sustainable economic growth, and reduces CO₂ emissions [29-34]). Given the importance of renewable energy in reducing CO₂ emissions, policymakers need to deploy more renewable energy for final consumption to achieve long-term climate goals [35].

Energy Consumption and GDP

Most studies focus on the two-way causality between energy consumption and economic growth. Syzdykova et al. summarized the relationship between energy consumption and economic growth in the Commonwealth of Independent States (CIS) for the period of 1992-2018 [36]. According to the findings of the study, there is a two-way causality between energy consumption and economic growth in CIS countries. This shows that the feedback hypothesis is valid in these countries. In Italy, there is a long-term, two-way causal relationship between energy use and GDP, and energy is limited for GDP growth [37]. Using cross-sectional and panel data from ten emerging markets and G7 countries, Soytas & Sari found a two-way causal relationship in Argentina, a causal relationship between GDP and energy consumption in Italy and South Korea, and a causal relationship between energy consumption and GDP in Turkey, France, Germany, and Japan [38].

Morelli & Mele claimed that economic growth and energy consumption exist in a one-way causal relationship [39]. In 47 developing countries from 1970 to 1976, the elasticity of energy consumption to GDP remained stable, which was significantly higher than 1, especially in developing countries [40]. Ang donated that the energy elasticity of high-income developing countries ranged from 1.6 to 1.8, while that of industrial countries declined slightly [41]. Lescaroux showed similar conclusions in different sectors and regions of 101 countries [42]. In the long run, the inverted U-shaped relationship between energy intensity and income has not been supported by all economic sectors. Nillesen et al. indicated that there is no inverted U-shaped relationship in most countries [43]. LLANOS et al. indicated that the range of treatments that produce the best performance, whether it is energy consumption or carbon emission policies, is between -2% and 0.4% [44]. In addition, extreme policies such as significantly reducing or increasing energy consumption will have the worst consequences for economic development [45].

GDP and CO₂ Emissions

Following the Environmental Kuznets Curve (EKC) hypothesis, they first proposed this conjecture, i.e., that the positive relationship between environmental pollution and income reverses after the income reaches a certain threshold [46]. Most studies tested the inverted U relationship between economic growth and pollutant emissions. For example, Destek et al. found that an inverted U-shaped is confirmed only for the pre-1973 period in France, Italy, and the USA [47]. Holtz, Eakin, and Selden revealed that the marginal emission tendency of CO₂ emissions gradually decreased with the increase in per capita GDP [48]. Some authors [49-51] reported the existence of the EKC hypothesis. Of course, not all results support the EKC hypothesis. Cole et al. found evidence against the EKC hypothesis. Using the ARDL method, many scholars have proven the relationship between economic growth and pollutant emissions in many countries [52]. For example, Massagony & Budiono found that in the long run, CO₂ emissions will continue to rise simultaneously with income, implying that the EKC hypothesis is not valid for CO₂ emissions in Indonesia [53]. Saboori and Sulaiman found a one-way causal relationship between GDP growth and CO₂ emissions [54]. Raihan and Tuspekova revealed the coefficient of economic growth is positive and significant with CO₂ emissions, indicating a 1% increase in economic growth is related to a 0.34% rise in CO₂ emissions [55]. Saboori and Sulaiman supported the EKC hypothesis in Singapore and Thailand [56]. Jebli et al. found that there is a long-term bidirectional causal relationship between CO₂ emissions and economic growth in 25 OECD countries [57]. There is a significant long-term negative correlation between CO₂ emissions and GDP, while a short-term positive correlation indicates that the deterioration of environmental sustainability is due to long-term economic growth [58].

However, topics about environmental problems, poverty, and inequality are still rare. Some scholars believe that poverty is the decisive factor leading to environmental pollution, because economic pressure, population pressure, and environmental policies lead to the deterioration of environmental quality [59]. For example, both the low and the high-income are suffering from environmental pollution, but the poor are more affected, and the research on inequality and environmental problems is relatively lacking [60-61]. Baloch et al. attempted to investigate the relationship between income inequality, poverty, and CO₂ emissions in 40 sub-Saharan African countries between 2010 and 2016 [61]. The results obtained from the Driscoll Kray regression estimator indicate that the exacerbation of income inequality has led to an increase in CO₂ emissions [61].

Data

The study covered 35 years of electricity consumption, GDP, and CO₂ emissions in 15 countries from 1980 to 2014. Data include high-income countries (Canada, Germany, Australia, Italy, Japan, France, the United States, and the United Kingdom) and low-income countries (Burundi, the Central African Republic, Niger, Gambia, Mozambique, Malawi, and Madagascar). Electricity consumption is available from the International Energy Agency (IEA) (https://www. iea.org/), GDP is provided by the World Bank (website: https://data.worldbank.org/), and CO₂ emissions can be obtained from BP’s World Energy Statistical Review (website: http://www.bp.com/). Descriptive statistics are shown in Table 1.

As shown in Table 1, although the per capita GDP of poor and high-income countries varies greatly (73.5 times), the gap between CO₂ emissions (93 times) and electricity consumption (82 times) is even greater. In the era of energy-driven economic operation, high-income countries took advantage of resources and technology and achieved more GDP growth at the cost of CO₂ emissions, but low-income countries did not.

Model

In the panel vector autoregressive model (PVAR), the independent variable is related to the fixed effect. Therefore, it is necessary to use the forward mean difference method “Helmert process” when using the panel model to eliminate the state fixed effect. By using this method, the orthogonality between the transformed independent variables and the lagged independent variables can be maintained, so that the lag independent variables can use GMM system estimation coefficients more consistently, thus eliminating the forward average value of each individual (i.e., the average value of future observations in each period) and ensuring the orthogonality between the lagged variables and the transformation variables. Furthermore, in this way, the bias in coefficient estimation caused by the correlation between individual effects and regression elements can be avoided [62]. The estimation method and running program refer to Love & Zicchino [63] and their Stata program. The PVAR model has the following advantages: Compared with OLS (Ordinary Least Square), it can estimate the interaction between multiple variables at the same time. Compared with VAR (Vector Auto Regressive), it can deal with panel data and improve the sample capacity. Compared with VAR and OLS, it can control the problems of fixed effects and random effects in panel data. Compared with OLS, it can study the dynamic effects between variables through the impulse response function.

Panel Unit Root Test

Before proceeding in advance with the PVAR framework, the first step of the estimation process consists of examining the data properties of all the series in terms of stationarity.

Ronald Fisher first proposed that multiple statistical test evidence should be listed as one statistic; that is, a unit root test was performed on each individual to obtain the statistics of n individuals and their corresponding p values. After that, Choi proposed four methods to synthesize these p values into Fisher statistics [64]. The Fisher ADF test in Table 1 shows that all panels contain unit roots.

Selection Optimal Lags

Before carrying out PVAR, we should know the corresponding optimal lags for better estimation. In Table 1, we know that the corresponding panel of poor countries should choose five lag periods. For high-income countries, we prefer to choose three periods of lag.

PVAR-GMM Estimation

As shown in Table 3, the GMM (Generalized Moment Estimation Method) approach is used to estimate the relationship between electricity consumption, GDP, and CO₂ emissions. For low-income countries, the electricity consumption of the second lag period has a significant positive impact on CO₂ emissions, GDP has a shock attenuation impact on CO₂ emissions, and the impact direction is positive and negative alternately. Electricity consumption of two lag periods and CO₂ emissions of the third lag period have a significant positive impact on GDP; only CO₂ emissions of the first lag period will have a positive impact on electricity consumption.

According to the above simple analysis, it can be seen that the low-income countries in the world can obtain short-term economic returns at the cost of pollution (electricity consumption can promote economic growth in a short time), but they are unable to translate their economic achievements into technological advantages to reduce their reliance on polluting energy (GDP in any period does not have a positive impact on electricity consumption, but CO₂ emissions do, which shows that the cost of electricity consumption is high in terms of pollution).

Table 2 also shows that the electricity consumption of the second lag period and CO₂ emissions of the first lag period have a significant impact on CO₂ emissions in high-income countries; GDP of the first lag period, electricity of the second lag period, and the CO₂ emissions of the third lag period have a positive impact on GDP, while electricity consumption of the first lag period, CO₂ emissions and GDP of the second lag period have a negative impact on GDP; CO₂ emissions and electricity consumption of the first lag period have a positive effect on electricity consumption, while CO₂ emissions of the second lag period have a negative effect on electricity consumption.

Fig. 1.

Impulse responses and for low and high-income countries Part (a) is impulse responses for low-income countries, Part (b) is impulse responses for high-income countries, Part (c) id forecast-error variance decomposition (FEVD) for high-and low-income

The main results in Table 2 show that the electricity consumption, CO₂, and GDP of the first lag period in rich countries will further strengthen the current results, and the development mode relying on energy and pollution in the past is hindering the current economic development. It is gratifying that the economic development of rich countries has not continuously promoted power consumption and CO₂ emissions, but has inhibited power consumption and CO₂ emissions instead (though not significantly), which indicates that rich countries have begun to realize the importance of environmental problems and increase the development and application progress of alternative energy, but the effect may not be obvious.

Granger Causality Test

The Granger causality test further strengthens the relationship between CO₂, GDP, and electricity in Table 3. For poor and high-income countries, electricity consumption is a single Granger cause of GDP. However, for high-income countries, CO₂ emissions and electricity consumption are both Granger causes of domestic economic growth. However, it has not been tested in low-income countries, which indicates that CO₂ emissions and electricity consumption may not make a significant difference in low-income countries’ economic growth.

Impulse-Response Functions (IRFs)

This subsection presents and discusses the results of the impulse response functions (IRFs) for low and high-income countries. For low-income countries (Part (a) in Fig. 1), although the positive impact of GDP on CO₂ emissions is strengthened after the third period, the impact decreases rapidly. Electricity consumption has a short-term driving effect on CO₂ emissions, but it is not obvious in the long term. Both CO₂ emissions and GDP emissions have a positive impact on electricity consumption, but this impact is relatively stable. Although CO₂ emissions and electricity consumption promote economic development, they are stagnant in the long term.

However, for the high-income countries (Part (b) in Fig. 1), the impact of GDP on CO₂ emissions increases with time, but the impact of electricity consumption on CO₂ emissions weakens over time. This is shown by the continuous popularization of clean power technology in high-income countries. At the same time, the impact of GDP on electricity consumption increases with time, but the impact of CO₂ emissions and electricity consumption on GDP is not very strong for a long time (even for electricity, the impact is negative). The results show that the style of economic growth in high-income countries has undergone a fundamental change, i.e., from resource dependence to the technological revolution.

Variance Decomposition Analysis

Although impulse responses can give details regarding the influence of variations in one variable on other variables, they do not specify the magnitude and degree of these effects. As a result, the variance decomposition technique was performed to determine this. The variance decomposition provides information about the variation in percentages in the dependent series that is attributable not only to their own shocks but also to shocks generated by the other variables.

Part (c) of Fig. 1 shows the results of the variance decomposition obtained from the orthogonalized impulse response coefficient matrices. For low-income countries, the variance decomposition shows that CO₂ emissions explain approximately 66.7% of the variations in CO₂ emissions, 20.6% of the fluctuations in GDP, and 31.3% of the fluctuations in electricity. While GDP explains approximately 25.9% of the variations in CO₂ emissions, 65.4% of the fluctuations in GDP, and 17.2% of the fluctuations in electricity. Besides, electricity explains approximately 7.4% of the variations in CO₂ emissions, 13.9% of the fluctuations in GDP, and 51.5% of the fluctuations in electricity. For high-income countries, however, the variance decomposition shows that CO₂ emissions explain approximately 99.5% of the variations in CO₂ emissions, 12.4% of the fluctuations in GDP, and 9.1% of the fluctuations in electricity. Also, GDP explains approximately 2% of the variations in CO₂ emissions, 48.7% of the fluctuations in GDP, and 53.8% of the fluctuations in electricity. Besides, electricity explains approximately 0.3% of the variations in CO₂, 38.9% of the fluctuations in GDP, and 37.1% of the fluctuations in electricity consumption.