Assessing the CO₂ Emissions and Energy Source Consumption Nexus in Japan

Abstract

This research investigates the variation in the impact of different energy sources on carbon dioxide (CO₂) emissions in Japan during the period from January 2019 to March 2023. The results of the Autoregressive Distributed Lag (ARDL) model suggest that a 1% increase in energy consumption produced through the photovoltaic (PV) decreases carbon emission by 0.053% in the short-run. Conversely, a 1% increase in coal, oil, and liquefied natural gas (LNG) leads to an increase in CO₂ emissions by 0.317%, 0.038%, and 0.214%, respectively. The study also reveals an inverted-U-shaped relationship between CO₂ emissions and economic growth, represented by the Nikkei stock index. The research emphasizes the critical need for Japan to prioritize investments and incentives in renewable energy technologies such as the PV systems, which have been demonstrated to effectively reduce CO₂ emissions in Japan. This is essential to uphold long-term ecological balance and to proactively support the ongoing reduction in CO₂ intensity, a key objective outlined in the Paris Agreement.

1. Introduction

Renewable energies play a crucial role in reducing long-term global carbon dioxide (CO₂) emissions. The Energy Information Administration (EIA) highlights that renewable energy sources are the fastest growing among all energy sources, with their consumption increasing by 3% annually [1]. According to the EIA, both developed and developing economies are experiencing significant growth in adopting renewable energy sources to mitigate greenhouse gas (GHG) emissions [2].

Japan is among the nations prioritizing the production of renewable energy, aiming to augment its share in the energy mix relative to non-renewable sources as a part of its long-term strategic vision. After the Fukushima accident, GHG emissions peaked in 2013, but by 2018, they had returned to 2009 levels, as reported by the International Energy Agency (IEA) in 2021. However, World Population Review reported that Japan is ranked as the world’s third-largest economy and the fifth-largest emitter of GHGs [3]. However, their efforts towards decarbonization faced significant setbacks after the Fukushima nuclear disaster in 2011. As a result, Japan shifted away from nuclear power and increased its reliance on fossil fuels. In this regard, the government of Japan plans to promote both renewable and nuclear power, but it is also considering the construction of new coal power plants. As of 2018, Japan set a target to lower its GHG emissions by 26% below the levels recorded in 2013 by 2030 [4]. Over the past decade, Japan has made considerable progress in developing an efficient, resilient, and sustainable energy system. The nation’s strong foundation in innovation and technology will be crucial in achieving its goal of becoming carbon-neutral by 2050 [5]. Japan’s energy policy is guided by four main principles: ensuring energy security, promoting economic efficiency, fostering environmental sustainability, and prioritizing safety [5]. In recent years, Japan has diversified its energy mix, improved the efficiency of fossil fuel usage, and reduced energy demand.

To achieve carbon neutrality by 2050, Japan must significantly accelerate its adoption of low-carbon technologies, address regulatory and institutional barriers, and foster competition in its energy markets. However, looking into the Japanese energy mix, as of 2022, Japan still heavily relies on fossil fuel energy. Although the share of fossil fuel energy sources in total energy consumption has been stagnant, it still covers over 70% of the total energy consumption in Japan (see Figure 1).

Generally, in Japan, the energy demand is met through a diverse range of sources, comprising renewable options such as photovoltaic (PV), hydroelectric power (HYDRO), and wind power (WP), as well as non-renewable sources like coal, oil, and liquefied natural gas (LNG). Some scholars have examined the interconnections between the consumption of specific energy resources, carbon emissions, and economic growth to inform policy decisions [7,8]. While studies have investigated the contribution of renewable energy consumption to CO₂ emissions using the Environmental Kuznets Curve (EKC) model [9,10,11], these examinations typically rely on annual or aggregated data, often due to a lack of data sources at a monthly level. Additionally, most prior research utilizes panel data, which fail to capture the nuanced impact of changes in the adoption of specific types of renewable energy on the CO₂ emissions of individual countries. Bulut [7] stands out as one of the few studies that focuses on the United States (US), concluding that increased renewable energy consumption correlates with reducing CO₂ emissions. However, even this study relies on aggregated renewable energy consumption data, leaving uncertainty regarding how distinct types of renewable energy sources may have differing impacts on CO₂ emissions. The present study is significant because it focuses on a local area, providing detailed insights into the specific situation in Japan. This study contributes to a broader understanding of how CO₂ emissions are linked to energy source consumption within the country, offering valuable data and analysis that can inform policy decisions, environmental strategies, and efforts to mitigate climate change. The study also ascertains how the current energy mix in Japan influences CO₂ emissions.

The overall goal of this research is to examine the intricate interplay between various sources of energy consumption, both renewable and non-renewable, alongside the dynamics of economic growth, and their collective impact on CO₂ emissions in Japan. Through an extensive analysis, we seek to elucidate the complex relationships governing the nation’s energy mix and its environmental footprint. This study seeks to investigate this objective under the EKC model. The study will have great significance in the context of sustainable development and environmental concerns regarding the nexus between CO₂ emissions and economic growth in Japan. This study stands out as the pioneering work delving into the existence of the EKC concept and the impact of both renewable and non-renewable energy consumption on CO₂ emissions in Japan. Moreover, it scrutinizes the connections between CO₂ emissions and renewable and non-renewable energy consumption (highlighting the top three contributors from each category) using monthly datasets. Finally, this approach offers a more comprehensive understanding of the intricate relationships between specific types of energy consumption and their corresponding effects on CO₂ emissions.

2. Literature Review

Since our study focuses on the relationship between CO₂ emissions, sources of energy consumption, and economic growth in Japan, we only covered studies that investigate such relationships along with the EKC hypothesis here. Nevertheless, numerous studies have investigated the global connections between CO₂ emissions and energy consumption. In one such study [12], researchers utilized the Autoregressive Distributed Lag (ARDL) and Granger causality approach in the vector error correction model (VECM) with panel data from Southeast Asian countries. They found a positive relationship between energy consumption and CO₂ emissions in the short and long-run. Additionally, the study observed a gradual increase in CO₂ emissions corresponding to a rise in energy usage. In a separate time-series study focusing on the United Arab Emirates (UAE), conducted by Sbia et al. [13], it was revealed that energy consumption and economic growth exhibited a positive association. Conversely, CO₂ emissions showed a negative relationship with the demand for energy use. On the other hand, a positive relationship between non-renewable energy consumption and CO₂ emissions was examined by Boontome et al. [14]. Table 1 summarizes all related papers discussed in the Literature Review Section.

In discussions centered around achieving sustainable growth while reducing CO₂ emissions, many studies have analyzed the connection between economic growth and CO₂ emissions. In their study, Azam et al. [15] observed a direct association between CO₂ emissions and economic growth in China, Japan, and the US. Likewise, Li et al. [16] and Pao and Tsai [17] provided compelling evidence, indicating that in terms of CO₂ emissions, energy consumption has a statistically significant positive impact in the long run for BRIC (Brazil, Russia, India, and China) countries. Similar positive correlations between CO₂ emissions and economic growth have been found in the case of Iran [18] and Algeria [19].

Table 1. Recent research on CO₂ emissions, energy consumption, and economic growth.

Authors	Period	Countries	Methods	EKC
Ahmed et al. [20]	1971–2014	India	ARDL	Presence
Ali et al. [8]	1971–2012	Malaysia	ARDL and Dynamic Ordinary Least Square (DOLS)	Presence
Azam et al. [21]	1980–2012	Indonesia, Malaysia, and Thailand	Linear multiple regression model	-
Boontome et al. [14]	1971–2013	Thailand	Cointegration and Granger causality test	-
Bouznit and Pablo-Romero [19]	1970-2010	Algeria	ARDL	Presence
Dissanayake et al. [22]	1990–2019	152 countries	Granger causality	-
Halicioglu [23]	1960–2005	Turkey	ARDL	Presence
Kasman and Duman [24]	1992–2010	European Union (EU) countries	Panel unit root and panel cointegration tests	Presence
Li et al. [16]	2000-2019	BRICS countries	ARDL	-
Munir et al. [25]	1980–2016	ASEAN-5 a countries	Panel unit root and panel cointegration tests	Presence
Ozturk and Acaravci [26]	1968–2005	Turkey	ARDL	Presence
Pao and Tsai [17]	1990–2005	BRIC b countries	Panel unit roots and panel cointegration tests	Presence
Saboori and Sulaiman [12]	1980-2009	Malaysia	ARDL and VECM Granger causality technique	Presence
Salari et al. [27]	1997–2016	USA	Generalized Method of Moments (GMM)	Presence
Sbia et al. [13]	975Q1–2011Q4	UAE	ARDL	-
Shahbaz et al. [28]	1975Q1–2011Q4	Indonesia	ARDL and VECM Granger causality technique	-
Yousefi-Sahzabi et al. [18]	1994–2007	Iran	Pearson product–moment correlation coefficients (PMCC)	-

Notes: ^a Association of Southeast Asian Nations; ^b Brazil, Russia, India, and China.

Table 1. Recent research on CO₂ emissions, energy consumption, and economic growth.

Authors	Period	Countries	Methods	EKC
Ahmed et al. [20]	1971–2014	India	ARDL	Presence
Ali et al. [8]	1971–2012	Malaysia	ARDL and Dynamic Ordinary Least Square (DOLS)	Presence
Azam et al. [21]	1980–2012	Indonesia, Malaysia, and Thailand	Linear multiple regression model	-
Boontome et al. [14]	1971–2013	Thailand	Cointegration and Granger causality test	-
Bouznit and Pablo-Romero [19]	1970-2010	Algeria	ARDL	Presence
Dissanayake et al. [22]	1990–2019	152 countries	Granger causality	-
Halicioglu [23]	1960–2005	Turkey	ARDL	Presence
Kasman and Duman [24]	1992–2010	European Union (EU) countries	Panel unit root and panel cointegration tests	Presence
Li et al. [16]	2000-2019	BRICS countries	ARDL	-
Munir et al. [25]	1980–2016	ASEAN-5 a countries	Panel unit root and panel cointegration tests	Presence
Ozturk and Acaravci [26]	1968–2005	Turkey	ARDL	Presence
Pao and Tsai [17]	1990–2005	BRIC b countries	Panel unit roots and panel cointegration tests	Presence
Saboori and Sulaiman [12]	1980-2009	Malaysia	ARDL and VECM Granger causality technique	Presence
Salari et al. [27]	1997–2016	USA	Generalized Method of Moments (GMM)	Presence
Sbia et al. [13]	975Q1–2011Q4	UAE	ARDL	-
Shahbaz et al. [28]	1975Q1–2011Q4	Indonesia	ARDL and VECM Granger causality technique	-
Yousefi-Sahzabi et al. [18]	1994–2007	Iran	Pearson product–moment correlation coefficients (PMCC)	-

Notes: ^a Association of Southeast Asian Nations; ^b Brazil, Russia, India, and China.

Moreover, due to distinct methodological approaches, types of data employed, and time frames taken into account, the scientific literature presents conflicting results concerning the connection between renewable and non-renewable energy consumption and its impact on economic growth. It is also important to note that most of the studies applied the ARDL model [8,13,16,19,20,23,26,28], cointegrations, the Granger causality test [12,14,22], and so on. The majority of the research has been carried out using annual datasets, and a portion of these studies took into account the existence of the EKC theory, as demonstrated by Salari et al. [27], Halicioglu [23], Kasman and Duman [24], and Munir et al. [25]. The current study also employed the ARDL model to examine the relationships between economic growth and CO₂ emission, along with the conditional error coefficient estimation among the modeled variables. In addition, a recent study by Dissanayake et al. [22] investigated the connections between renewable and non-renewable energy consumption, CO₂ emissions, and economic growth across developed, developing, least developed countries (LDCs), and economies in transition using the Granger causality technique. The results of previous research on the relationship between the use of renewable and non-renewable energy sources and economic growth are inconsistent, suggesting that a definitive link between these factors may or may not exist. Moreover, Hashmi et al. [29] found that energy consumption leads to increased emissions both in the short and long- term worldwide. Additionally, the scientific literature shows mixed results regarding the relationship between renewable energy consumption (REC) and economic growth due to varying methodologies, types of data, and time periods considered. For instance, gross domestic product (GDP) per capita, renewable energy, and non-renewable energy each have distinct effects on CO₂ emissions in BRICST countries (Brazil, Russia, India, China, South Africa, and Turkey), according to Syed et al. [30].

This literature review shows that much research has been conducted on the relationships between CO₂ emission, economic growth, and energy consumption. However, to the best of our knowledge, no study considers individual-level renewable and non-renewable energy source types and their relationships with CO₂ emission based on the monthly level dataset of Japan. Hence, our study makes a significant contribution to the existing literature by examining how various types of renewable energy distinctly impact the CO₂ emissions of a single country. We contend that estimating the individual effects of different types of renewable energy is more valuable for formulating energy policies and developing plans to alter the energy mix of a given country. Furthermore, gaining a sophisticated comprehension of the interaction between energy consumption and economic growth can aid in effectively allocating resources, optimizing energy efficiency, and fostering innovation within Japan’s energy sector. As a result, the empirical findings from this study hold considerable potential as a valuable tool for policymakers in crafting well-informed and sustainable energy policies that combat climate change and bolster the nation’s long-term economic prosperity. Our study is also among the few that capture the impact of renewable energy sources on CO₂ emissions using relatively high-frequency time-series data—specifically, monthly data. This contrasts with the prevailing reliance on annual data in most previous studies. We believe the significance of assessing the effects of renewable energy consumption at the monthly level is increasing, given the accelerating influence of climate change on energy consumption. As the pace of the impact of climate change on the energy sector accelerates and becomes more severe and unprecedented, there is an increasing necessity for further research to comprehend the effects of renewable energy consumption on CO₂ emissions using higher frequencies than annual intervals.

3. Materials and Methods

3.1. Sources of Data and Descriptive Statistics of the Modeled Variables

This study analyzes the relationship between CO₂ emissions, energy consumption (renewable and non-renewable), and economic output using monthly data from January 2019 to March 2023. The daily CO₂ emissions per capita are measured in metric tons (MT) and are obtained from the Carbon Monitor website. Monthly data for energy consumption (renewable and non-renewable consumption sources) are measured in thousands of kWh and obtained from the Agency for Natural Resources and Energy’s website. As a proxy of an economic indicator, this study considered the Nikkei stock index. Monthly Nikkei 225 datasets were obtained from the Nikkei indices website. The variables used in this study are described in

Table 2. Data and source of the parameters.

Variables	Symbol	Measurement Unit	Sources
Hydroelectric power	HYDRO	1000 kWh	[31]
Photovoltaic	PV
Wind power	WP
Coal	COAL
Oil	OIL
LNG	LNG
Nikkei index 225	NIKK		[32]
CO2 emission	CO2	MtCO2/day	[33]

.

It is essential to mention that thirteen sources of energy consumption, renewable (seven) and non-renewable (six), were identified. From these sources, the top three highly consumed sources were taken from renewable (HYDRO, PV, and WP) and non-renewable (coal, oil, and LNG) energy sources. These six energy source consumption data are converted into natural logs. The descriptive statistics of these variables are summarized in

Table 3. Descriptive statistics.

Criteria	LNCO2	LNNIKK	Renewable			Non-Renewable
			LNHYDRO	LNPV	LNWP	LNCOAL	LNOIL	LNLNG
Mean	4.489	10.124	15.575	14.125	13.257	16.960	14.115	17.135
Median	4.481	10.188	15.612	14.083	13.292	16.976	14.050	17.127
Maximum	4.711	10.294	16.084	14.661	13.829	17.210	15.011	17.460
Minimum	4.159	9.849	15.010	13.426	12.741	16.600	13.318	16.793
Std. Dev.	0.129	0.132	0.288	0.310	0.317	0.152	0.462	0.170
Skewness	−0.087	−0.438	−0.219	−0.341	−0.079	−0.625	0.214	0.003
Kurtosis	2.540	1.786	1.830	2.460	1.908	2.595	2.077	2.286
Jarque–Bera	0.514	4.765	3.315	1.606	2.589	3.674	2.198	1.083
Observations	51	51	51	51	51	51	51	51

.

Comparing the mean, median, and standard deviation among the model variables, it is discernible that the share of energy consumption is higher from LNG than WP. Figure 2 plots our modeled fixed regressors with CO₂ emissions and the Nikkei 225 stock index for Japan from January 2019 to March 2023.

Based on the kurtosis values, none of the variables are mesokurtic, and all are considered platykurtic. The Jarque–Bera test [34] results indicate that the residuals of all variables are not normally distributed, which enables us to apply the ARDL model accurately. It is apparent from Figure 2 that except for the LNNIKK variable, which is considered a proxy of Japan’s economic outcome, all renewable and non-renewable energy sources, along with CO₂ emission, show very strong up-and-down trends. However, after March 2020, there was a sharp increase in the Nikkei index up to February 2021, which remained constant [35]. This decrease might be due to the COVID-19 pandemic.

3.2. Model Specification

This research investigates how energy usage, economic growth, and both renewable and non-renewable energy resources impact CO₂ emissions, along with considering the presence of the EKC. The EKC is a widely used approach to evaluating environmental performance. Originating from a concept introduced by Kuznets in 1955 [36], the EKC is depicted as an inverted U-shaped curve. Initially, it was developed to examine the correlation between income per capita and income inequality. However, the EKC gained prominence when this inverted U-shape began to be applied to environmental research. In the early 1990s, it was widely and intensely used as a theoretical framework to study the relationship between yield and environmental degradation [37]. Since the early 1990s, numerous empirical studies have shown that environmental pressures decrease once economic growth reaches a certain level, especially when fossil fuel-intensive industries are replaced by those less dependent on fossil fuels [38,39,40]. Additionally, our model is supported by the theory of EKC developed in previous studies [7,8,41]. To assess the presence of the EKC and the influence of GDP growth on CO₂ emissions in Japan, the study employs the following empirical model:

$${CO}_2={\beta }_0+{\beta }_{1}Nikk+{\beta }_2{Nikk}^2+{\epsilon }_t$$

(1)

where ${CO}_2$ signifies the carbon emissions per capita, $Nikk$ represents the real GDP per capita, and ${Nikk}^2$ stands for the square of the real GDP per capita. Furthermore, ${\beta }_0$ is a constant and ${\epsilon }_t$ corresponds to an error term. The existence of the EKC is tested by including the squared term of $Nikk$ (${Nikk}^2$).

The study inspects the effect of economic growth through the Nikkei stock index on CO₂ emissions and the existence of EKC. Moreover, it is observed that non-renewable energy consumption can induce an increase in CO₂ emissions [14,20], while renewable energy sources have a negative relationship with CO₂ emissions [27]. Thus, the model can be written as follows:

$${CO}_2={\beta }_0+{\beta }_{1}Nikk+{\beta }_2{Nikk}^2+{\beta }_{3}REC+{\beta }_{4}NREC+{\epsilon }_t$$

(2)

where $REC$ and $NREC$ stand for renewable and non-renewable energy consumption, respectively.

Nevertheless, three REC sources—hydroelectric power (HYDRO), photovoltaic (PV), and wind power (WP)—alongside three NREC sources, such as coal, oil, and LNG, were included in our final model based on their major shares in the Japanese energy sector. Therefore, the structure of the model is written as follows:

$${CO}_2={\beta }_0+{\beta }_{1}Nikk+{\beta }_2{Nikk}^2+{\beta }_{3}HYDRO+{\beta }_{4}PV+{\beta }_{5}WP+{\beta }_{6}COAL+{\beta }_{7}OIL+{\beta }_{8}LNG+{\epsilon }_t$$

(3)

3.3. Estimation Techniques

Before employing the econometric approaches, we conducted several unit root tests. The study employed the Phillips–Perron (PP) [42], the Augmented Dickey–Fuller (ADF) [43], the Kwiatkowski–Phillips–Schmidt–Shin (KPSS) [44], and the Zivot–Andrews (ZA) [45] tests to check the stationarity. After checking the data, this research utilizes the ARDL model to explore the relationship between CO₂ emission and economic growth. The ARDL method, introduced by Pesaran et al. [46], is chosen due to its advantages over other cointegration tests. The ARDL model offers a notable advantage in handling variables with different stationarity levels, accommodating integrated order zero I (0), integrated order one I (1), or a combination of both types. This makes it a robust choice regardless of the stationary properties of the variables. The ARDL approach also offers a suitable framework for modeling by capturing the data-generating process effectively by incorporating a sufficient number of lags. This “general to specific” modeling approach aids in drawing meaningful conclusions from the data. Incorporating both short-run adjustments and long-run equilibrium, the ARDL approach allows for deriving the error correction mechanism through a simple linear transformation without losing information about the long-run relationship. This feature enhances the model’s interpretability and practicality. Moreover, the ARDL approach outperforms the Johansen and Juselius approach regarding small sample properties [46]. Another advantage of the ARDL approach is that it avoids issues of residual correlation, eliminating endogeneity problems that could arise from inappropriate lag selection [47]. Furthermore, the ARDL approach distinguishes between dependent and independent variables, offering a more comprehensive analysis of the relationships between variables. However, it is essential to note that the computed F-statistics provided in the table by Pesaran et al. [46] are not valid for I (2) variables, as indicated by Ouattara [48].

Furthermore, in essence, the variables “summer” and “winter” are used to represent seasonal patterns. The “summer” variable encompasses the months of June to September, while the “winter” variable includes the months from December to March. Thus, the model takes the form as follows:

$$\begin{gathered} {CO}_2={\beta }_0+{\beta }_{1}Nikk+{\beta }_2{Nikk}^2+{\beta }_{3}HYDRO+{\beta }_{4}PV+{\beta }_{5}WP+{\beta }_{6}COAL+{\beta }_{7}OIL+{\beta }_{8}LNG+{\beta }_{9}summer+ \\ {\beta }_{10}winter+{\epsilon }_t \end{gathered}$$

(4)

Grossman and Krueger [49] have shown a non-linear relationship between GDP and CO₂ emissions that can be reflected by the following Equation (all variables are translated into their log forms):

$$\begin{gathered} {LnCO}_{2t}={\beta }_0+{\beta }_1{LnNikk}_t+{\beta }_2{LnNikk}_t^2+{\beta }_3{LnHYDRO}_t+{\beta }_4{LnPV}_t+{\beta }_{5}Ln{WP}_t+{\beta }_6{LnCOAL}_t+ \\ {\beta }_7{LnOIL}_t+{\beta }_8{LnLNG}_t+{\beta }_9{summer}_t{+\beta }_{10}{winter}_t+{ϵ}_t \end{gathered}$$

(5)

where ${LnCO}_{2t}$ indicates a logarithmic form of ${CO}_2$ within the time frame t (January 2019 to March 2023). ${LnNikk}_t$ and ${LnNikk}_t^2$ denote the Nikkei indices as a proxy for economic indicators. The logarithmic forms of hydroelectric power plants ($HYDRO$), photovoltaic ($PV$), wind power ($WP$), coal ($COAL$), oil ($OIL$), liquefied natural gas ($LNG$), $summer,\mathrm{a}\mathrm{n}\mathrm{d}winter$ are considered fixed regressors.

The study used the Akaike information criterion (AIC) to choose the lag length. The AIC criteria graph is presented in Appendix A. After finding the long-run association between variables, the study used a conditional error correction model (ECM) to investigate the short-run dynamics of the respective variables along with the short-run adjustment rate towards the long-run rate. The ECM general form of Equation (5) is formulated below in Equation (6).

$$\begin{array}{ll}∆Ln{CO}_{2t}={\beta }_0& +{\beta }_1{LnCO}_{2\left(t-1\right)}+{\beta }_{2}Ln{Nikk}_{\left(t-1\right)}+{\beta }_3{LnNikk}_{\left(t-1\right)}^2\\ & +\sum _{i=1}^{n1}{\beta }_{4i}∆{LnCO}_{2(t-i)}+\sum _{i=0}^{n2}{\beta }_{5i}∆Ln{Nikk}_{(t-i)}+\sum _{i=0}^{n3}{\beta }_{6i}∆Ln{Nikk}_{(t-i)}^2\\ & +{\beta }_7{LnHYDRO}_t+{\beta }_8{LnPV}_t+{\beta }_9{LnWP}_t+{\beta }_{10}{LnCOAL}_t+{\beta }_{11}{LnOIL}_t\\ & +{\beta }_{12}Ln{LNG}_t+{\beta }_{13}{summer}_t+{\beta }_{14}{winter}_t+\mathsf{\lambda }{e}_{t-1}+{ϵ}_t\end{array}$$

(6)

where $∆$ is the first difference operator. If the empirical findings reported a cointegration relationship among the variables (${CO}_2,Nikk,$ and ${Nikk}^2$), the short-run dynamic would be adjusted through the error correction $“\mathsf{\lambda }{e}_{t-1}”$ movement. In addition, the coefficient sign of $\mathsf{\lambda }{e}_{t-1}$ is assumed to be negative and significant to achieve long-run equilibrium if devaluations exist in the model.

After verifying the integration order, the focus shifts to examining the cointegrating properties of the variables. Bounds cointegration tests, utilizing F- and t-statistics, are employed to assess potential cointegration among the variables. These test methods rely on upper and lower critical bounds, as Pesaran et al. [46] defined. If the actual F- and t statistic values surpass the upper bound, the alternative hypothesis is accepted, suggesting that the variables are cointegrated. On the other hand, if the values are below the lower bound, it indicates that the variables are not cointegrated. Meanwhile, the test outcome remains inconclusive if the values fall within the critical bounds. The empirical results show that both the F- and t-statistic values are greater than the upper bound values, indicating a long-run relationship between variables (

Table 4. ARDL bound test.

Test Statistics	Model Output
F-statistic	69.408 ***
K-value	2
Critical value bounds (N = 47)
Significance	I (0) bound	I (1) bound
10%	2.63	3.35
5%	3.1	3.87
1%	4.13	5
Decision	Long-run relationship presence

Note: *** denotes significance at 1% level.

).

To assess the presence of serial correlation, heteroskedasticity, and normality distribution in the models, the study performed the Breusch–Godfrey autocorrelation test [50,51], the Breusch–Pagan test for heteroskedasticity [52], and the Jarque–Bera test for normality [34].

We also assessed the stability of the parameters estimated using the ARDL model using cumulative sum (CUSUM) and cumulative sum of squares (CUSUMSQ) tests. As observable from the Breusch–Godfrey (BG) test results presented in

Table 5. Serial correlation, heteroskedasticity, normality, and model stability tests.

Tests	Problem	F-Statistic	p-Value	Decision
Breusch–Godfrey serial correlation LM test	Serial correlation	0.871	0.430	No serial correlation
Breusch–Pagan–Godfrey F-stat	Heteroskedasticity	1.863	0.066	Heteroskedasticity
Jarque–Bera test	Normality	-	0.924	Normally distributed
CUSUM test	Model stability	-	-	Model is stable
CUSUMSQ test	Model stability	-	-	Model is stable

, our model contains no serial correlation based on the 5% significance level.

However, our model had a low heteroskedasticity concern at a 10% level. To solve this issue, the ARDL model was estimated using the Newey–West heteroskedasticity and autocorrelation consistent (HAC) estimator.

The Lagrange Multiplier (LM) and Breusch–Pagan–Godfrey tests’ p-values for the model were 0.430 and 0.066, respectively, indicating that the model lacks serious serial correlation problems but has heteroscedasticity issues. On the other hand, the p-value of the Jarque–Bera test is greater than the significance level (0.05), and the decision is made to retain the null hypothesis. This means that our data are normally distributed (Appendix B.1). The stability of the predicted parameter is validated when the CUSUM and CUSUMSQ lines stay within the higher and lower limits of the corresponding graphs. Our model’s stability is evidenced by the blue line of the upper and lower bounds staying within the red line of the CUSUM and CUSUMSQ graphs (Appendix B.2). The ARDL model was executed using EViews 12.

4. Results

ARDL bounds testing emerges as a versatile option as it can be applied regardless of whether the data are integrated of order zero I (0), one I (1), or even fractionally integrated. On the other hand, the Johansen and Juselius [45] approach, another standard cointegration test, is suitable only when the data are stationary at the first difference, i.e., I (1). The study conducted unit root tests to ascertain the order of integration of the variables. Based on the outcomes of these tests, it was confirmed that all endogenous variables in the study exhibit either order zero or one integration I (0) or I (1) (

Table 6. Unit root tests.

Energy Types	Variables	Levels				First Differences
		ADF	PP	KPSS	ZA	ADF	PP	KPSS	ZA
Renewable	LNHYDRO	−5.803 ***	−3.282 *	0.039	−6.38 ***	−4.269 ***	−5.549 ***	0.037	−7.688 ***
	LNPV	−1.494	−2.979	0.044	−7.775 ***	−8.081 ***	−5.310 ***	0.033	−10.270 ***
	LNWP	−1.761	−3.208 *	0.033	−6.179 ***	−5.708 ***	−4.935 ***	0.035	−6.472 ***
Non-renewable	LNCOAL	−1.845	−3.452 *	0.071	−3.539	−7.233 ***	−3.988 **	0.347 ***	−8.498 ***
	LNOIL	−5.434 ***	−3.072	0.146 **	−6.536 ***	−6.234 ***	−4.432 ***	0.500 ***	−7.652 ***
	LNLNG	−1.666	−3.370 *	0.044	−3.610	−6.420 ***	−6.462 ***	0.350 ***	−7.146 ***
LNNIKK		−2.301	−1.993	0.135 *	−4.911 **	−5.515 ***	−5.361 ***	0.061	−8.127 ***
LNCO2		−2.110	−2.884	0.067	−4.597 *	−9.628 ***	−5.844 ***	0.250 ***	−9.829 ***

Note: All the unit root tests include both a constant and a linear trend. ***, **, and * denote significance at the 1%, 5%, and 10% levels, respectively. ADF, PP, KPSS, and ZA represent the Augmented Dickey–Fuller, Phillips–Perron, Kwiatkowski–Phillips–Schmidt–Shin, and Zivot–Andrews test statistics.

).

Consequently, the order of integration is a mixture of I (0) and I (1), rendering it appropriate to solely utilize the ARDL bounds test while the Johansen and Juselius [53] approach is not applicable in this scenario.

4.1. ARDL Long-Run Results

Table 7. ARDL long-run estimations.

Variable	Coefficient	t-Statistic
LNNIKK	32.660 ***	2.935
SQLNNIKK	−1.612 ***	−2.928
Constant	−171.766	−3.008

Note: *** denotes significance at the 1% level.

presents the long-run estimates of the ARDL approach. The findings indicate that the Nikkei index, considered a proxy variable of GDP growth, positively and significantly impacts CO₂ emissions in the long-run. This suggests that an increase in GDP growth will further contribute to higher concentrations of CO₂ emissions. Conversely, the squared term of Nikkei, utilized as a proxy for economic growth in Japan, negatively impacts CO₂ emissions in the long-run. This result supports the EKC hypothesis. As the economy grows, environmental degradation initially increases but eventually declines as societies invest more in sustainable practices and adopt environmentally friendly technologies. However, the squared term of Nikkei represents that a 1% increase in economic growth leads to a 1.61% decrease in CO₂ emissions and vice versa. This finding aligns with the fact that in conjunction with GDP growth, government policies concerning pollution regulation have been progressively increasing and improving [8,27]. We also calculated the turning point, and it was around 25,092.

4.2. ARDL ECM Model Results

Table 8. ARDL conditional error correction model and seasonal dummy estimation.

Variable	Coefficient	t-Statistic
Δ (LNCO2(−1))	−0.135 ***	−2.735
Δ (LNCO2(−2))	−0.251 ***	−6.255
Δ (LNNIKK)	3.069	0.518
Δ (LNNIKK (−1))	−12.552 **	−2.260
Δ (SQLNNIKK)	−0.156	−0.532
Δ (SQLNNIKK (−1))	0.617 **	2.238
LNHYDRO	−0.026	−1.391
LNPV	−0.053 ***	−3.155
LNWP	0.003	0.139
LNCOAL	0.317 ***	8.913
LNOIL	0.038 ***	3.108
LNLNG	0.214 ***	6.947
SUMMER	−0.012	−0.766
WINTER	0.048 ***	3.846
ECT	−0.782 ***	−17.476

Note: *** and ** denote significance at the 1% and 5% levels, respectively.

presents the ARDL conditional error correction model and seasonal dummy estimation. According to these results, coal, oil, and LNG significantly and positively influence CO₂ emissions, with a significance level of 1%. The findings show that a 1% increase in coal, oil, and LNG leads to an increase in CO₂ emissions by 0.317%, 0.038%, and 0.214%, respectively. The findings are similar to the studies of [24,54,55]. Among non-renewable energy sources, coal emitted more CO₂ than oil and LNG sources, as examined by the World Nuclear Association [56]. Conversely, renewable energy sources such as PV energy have negative and significant effects on CO₂ emissions in Japan, while the coefficients of HYDRO show a negative influence on CO₂ emission. To be precise, a 1% increase in the PV rate decreases carbon emission by 0.053% in the short-run. These results support studies such as Güney [57], Yuaningsih et al. [58], and Zhang et al. [59], which stated that solar energy reduces CO₂ emissions. The coefficient for PV is negative and significant, while the coefficients for other renewable energy sources, such as HYDRO and WP, are not significant at the 5% level. This is due to the Feed-in Tariff (FIT) program, which has not significantly increased the supply of hydroelectric and wind power compared to PV. The higher installation rate of PV is attributed to its higher profitability. Additionally, the development of wind energy has been stagnant in Japan due to delays in environmental assessment, lack of investment, regulatory challenges, lack of suitable land, and public opposition to wind farms due to concerns about noise and visual impact. The significant and positive impact of non-renewable energy sources may be attributed to the fact that non-renewable energy consumption sources are a primary contributor to environmental pollution, and still, the country’s economic growth heavily relies on non-renewable energy.

In analyzing the coefficient estimates, we can infer that renewable energy significantly reduces CO₂ emissions, while non-renewable energy sources positively generate more CO₂ emissions in Japan. This finding aligns with the work of Inglesi-Lotz [2], who also observed that adopting renewable energy sources can significantly reduce GHG emissions. Moreover, in Japan, the winter season (December to March) typically sees increased energy consumption due to heating needs. These findings are consistent with Matsumoto’s [60] study, which observed that household CO₂ emissions in Japan increase by less than 10% during winter, even with the addition of one family member. Moreover, higher energy consumption often correlates with higher CO₂ emissions, especially when fossil fuels are a major energy source [60]. Lastly, the statistically significant error correction term (ECT) with a negative sign and a value of 0.782 documents the long-run relationship between the study variables.

Additionally, we have used changes in COVID-19 case numbers as a proxy to assess the model’s robustness regarding the impact of COVID-19. Our results show that COVID-19 does not affect the relationship between CO₂ emissions and energy source consumption in Japan. These robust findings are detailed in the Supplementary File (see Tables S1 and S2 in the Supplementary Materials).

5. Discussion

Our study’s results, demonstrating that an increase in photovoltaic consumption correlates with a decrease in CO₂ emissions, are in line with previous research by Jebli and Youssef [61], which indicated that renewable energy has decelerated Tunisia’s CO₂ emissions rate. Similarly, our findings are consistent with another study by Salari et al. [27], which reveals similar results using both static and dynamic models for the US. It suggests that CO₂ emissions are positively influenced by total, non-renewable, industrial, and residential energy consumption, while there exists a negative correlation between CO₂ emissions and renewable energy consumption.

Since 2012, the Japanese government has been fostering the development of renewable energy plants through the implementation of a FIT system. Additionally, a Feed-in Premium (FIP) system has been in place since 2022, aiming to progress towards a carbon-neutral society [62]. The FIT program mandates power companies to procure electricity from certified renewable sources such as solar, wind, hydro, geothermal, and biomass at fixed government-set prices for a specific duration. This mechanism ensures stable revenue, thereby encouraging investments in renewable power generation [63]. As a result of the FIT policy, solar PV installations in Japan experienced rapid growth, establishing the country as one of the global leaders in solar energy capacity. While solar power gained more momentum than wind power initially, it is crucial to acknowledge that Japan has also been promoting wind energy as part of its renewable energy portfolio [63]. This study’s findings support the notion that WP has yet to be entirely overlooked despite its slower development compared to solar PV. Our findings endorse the benefits of the government’s FIT measures that promote the uptake of renewable energy by creating renewable energy markets and implementing renewable energy portfolio standards. These policies aim to enhance environmental conditions and yield positive outcomes from a macroeconomic perspective.

As a result, our findings offer significant insights into the directional predictability among energy consumption, economic growth, and CO₂ emissions. First, the findings indicate that individual-level non-renewable energy consumption sources lead to increased CO₂ emissions compared to renewable energy sources. It is established that clean and sustainable energy sources are recognized for their environmentally friendly nature and lack of direct CO₂ emissions. When the utilization of renewable energy grows, it substitutes the necessity for fossil fuel-derived energy sources (like coal, oil, and LNG) in generating electricity [64]. This substitution of fossil fuels leads to a reduction in their usage and, consequently, a decrease in CO₂ emissions, aligning with the observations of declining CO₂ emissions linked to increased adoption of renewable energy.

In contrast, coal, oil, and LNG represent fossil fuels that emit CO₂ and other pollutants upon combustion for energy production. These fossil fuels have historically been the primary contributors to human-induced CO₂ emissions, significantly driving climate change [65]. The amplified consumption of coal, oil, and LNG corresponds to heightened emissions of GHG, including CO₂ [66]. An escalation in the utilization of these fossil fuels would likely correlate with increased CO₂ emissions, as underscored by our findings. The study primarily focuses on the relationship between energy consumption, GDP growth, and CO₂ emissions, overlooking a short period of time due to the limitations of the monthly time-series datasets for both renewable and non-renewable specific energy types. While the study identifies relationships between variables, establishing causality can be challenging. We believe the significance of assessing the effects of renewable energy consumption at the monthly level is increasing, given the accelerating influence of climate change on energy consumption. As the pace of the impact of climate change on the energy sector accelerates and becomes more severe and unprecedented, there is an increasing necessity for further research to comprehend the effects of renewable energy consumption on CO₂ emissions using higher frequencies than annual intervals. Moreover, understanding these changes in the energy mix is crucial, especially as wind energy is expected to grow in Japan and contribute to reducing CO₂ emissions in the future.

6. Conclusions

The current study examined how different renewable and non-renewable energy consumption sources and GDP growth affect CO₂ emissions in Japan. It used monthly datasets from January 2019 to March 2023 to investigate the EKC hypothesis in the presence of the variables described above. The cointegration analysis using the ARDL model by Pesaran et al. (2001) [46] was utilized to analyze the data. The ARDL long-run findings demonstrated the existence of the EKC relationship in Japan. The coefficient values from the results also indicated that coal had a greater impact on CO₂ emissions compared to the other two primary sources of fossil fuel energy consumption (oil and LNG) in Japan. More specifically, a 1% increase in the use of coal, oil, and liquefied natural gas (LNG) leads to increases in CO₂ emissions by 0.317%, 0.038%, and 0.214%, respectively. Therefore, the findings of this study provide a valuable perspective, suggesting that replacing coal with alternative energy sources would be desirable unless significant advancements in clean coal technology are introduced.

The findings of our study, which demonstrate a link between fossil fuel energy consumption and the escalation of CO₂ emissions, are likely parallel to the present situation of Japan’s energy mix. The country still relies heavily on fossil fuels as its primary energy source, while the utilization of renewable energy sources remains notably limited.

However, the study also revealed that an increase in PV consumption has the potential to reduce CO₂ emissions by 0.053%, highlighting the importance of ongoing efforts to promote the adoption of renewable energy sources. Given the study’s findings, we recommend that Japan prioritize replacing coal with alternative energy sources to reduce CO₂ emissions. Therefore, increasing investment and incentives are recommended for renewable energy sources, particularly PV, which has been shown to effectively reduce CO₂ emissions in Japan.