Energy consumption structural adjustment and carbon neutrality in the post-COVID-19 era

Abstract

Since the spread of COVID-19 pandemic all over the world, a significant recession has broken out with no precedent. China has brought up a new voluntary contribution target that achieving carbon neutrality by 2060. How to achieve climate change mitigation targets without heavily hindering economic development is of great importance in the future. In this study, a Markov chain model is employed to forecast primary energy consumption (PEC) structure and verify whether the carbon intensity target would be achieved under three scenarios with different economic growth rates, such as 6.1%, 4.2%, and 2.3%, respectively. A multi-sector dynamic stochastic general equilibrium (DSGE) model is employed to simulate and evaluate economic development, fossil and non-fossil energy consumption, and CO₂ emissions under three scenarios using data calibration according to the Markov chain prediction result. The prediction results from the Markov chain show that energy structural adjustment can help us achieve the carbon intensity target of 2030 under both steady and mid-speed development scenarios. As long as the economic growth rate is higher than 4.2%, the carbon intensity target can be achieved mainly through energy consumption structural change, which provides a way to achieve the carbon neutrality target of 2060. The simulation results from the DSGE model show that energy structural adjustment can also smooth the volatility of the economic fluctuation when exogenous stochastic shocks happened.

1. Introduction

Aiming at containing the global warming no more than 2 °C, the Paris Agreement was signed by 175 countries around the world in 2016. Such an ambitious target requires reaching zero emissions (also referred to as carbon neutrality) in the second half of this century. Carbon neutrality has a range of definitions with a shared understanding (Salvia et al., 2021). Carbon neutrality refers to achieving a state of net zero anthropogenic (man-made) carbon emissions into the atmosphere.1 Pursuing net zero carbon emissions means having a balance between emitting carbon and absorbing carbon from the atmosphere in carbon sinks. Carbon sinks denote any systems that emits less than it absorbs. The main natural carbon sinks usually refer to forests, soil, and oceans. Carbon neutrality differs from climate neutrality because it does not consider other greenhouse gasses (Berndes et al., 2016). The concept is useful in determining and reducing carbon emissions associated with transportation, energy, production, and agriculture.

China was by far the biggest driver of energy, accounting for more than three quarters of net global growth result in the high-speed economic growth. The share of both non-fossil fuels and natural gas has increased in energy consumption in China recently, however, coal consumption dominates the PEC has resulted in severe environmental issues and led to total carbon emissions ranking first in the world(Wu et al., 2020a). The issue of CO₂ emissions has attracted much attention from the Chinese government. During the APEC summit in November 2014, the Chinese government promised to achieve a peak of carbon emissions and raise the ratio of non-fossil energy in its energy mix to at least 20% no later than 2030. In the 75th UN General Assembly on 22nd September 2020, Chinese President Xi Jinping announced that China would deliver a stronger emissions reduction target, peak emissions before 2030, and strive to reach the carbon neutrality before 2060. Some further commitments were announced on 12nd December 2020 that China will lower its CO₂ emissions per unit of GDP by over 65 percent comparing to the 2005 level, increase the share of non-fossil fuels in PEC to around 25 percent.2

Since 2020, the COVID-19 pandemic has been creating panic and chaos all over the world. The outbursts of pandemics usually ensued in a huge number of mortalities and a great loss in global economy (Razzaq et al., 2020). Including China, many countries have employed "lockdown" measures to forbidden its rapid propagation earlier last year (Wang et al., 2020). Government policies during the COVID-19 pandemic, such as closed international borders, populations were confined to their home, restrictions in travel or gathering, have drastically altered the patterns of energy demand around the world (Le Quéré et al., 2020). In its new Global Energy Review 2020 report, the International Energy Agency (IEA) describes how the global energy demand affected by economic turmoil, travel restrictions, and lockdowns(International Energy Agency, 2020). It is noted that the global energy demand was 3.8% lower in the first quarter of 2020, or 150 million tonnes of oil equivalent (Mtoe), relative to the first quarter of 2019.3

Lockdown interrupted the global supply chains and caused worldwide economic chaos, a significant decrease in carbon emissions at the same time (Nikolopoulos et al., 2020; Guan et al., 2020). Existing studies have shown that the forced confinement methods during COVID-19 pandemic has an important impact on global CO₂ emissions change (Le Quéré et al., 2020). Wang et al. (2020) estimated that energy-related CO₂ emissions decreased by 18.7% in the first quarter of 2020 compared to the same period in 2019 in China. Han et al. (2020a) estimated China's CO₂ emissions decreased 257.7 Mt caused by COVID-19 mitigation measures, equivalent to approximately 11% compared to 2019 during the same period.

However, a retaliatory growth of CO₂ emissions was found in a post-COVID-19 era through the short- and long-term analysis (Wang and Wang, 2020). A rebound in the consumption of motor gasoline was found in May 2020(Tian et al., 2021) . Since the emission reductions associated with the pandemic are only temporary, the CO₂ emissions may skyrocket again in reviving economies. Since China is the first major economy to show recovery after a slowdown induced by COVID-19 pandemic (Wang and Zhang, 2021), hereafter, balancing economic growth with energy use and CO₂ emissions reduction is still an important issue confronting China.

Two approaches that can be used to obtain the carbon neutrality target were found to be carbon emissions balancing with carbon removal, or using renewable energy that does not produce carbon emissions (Berndes et al., 2016). Also, the Ford company had mentioned the same method to achieve their carbon neutrality target 2050.4 Comparatively, directly reducing CO₂ emissions is safer and more reliable in reality. Therefore, both carbon emissions reduction policy and energy consumption structure adjustment are of utmost importance in achieving the goal.

For the record of the energy industry, to achieve the fourteenth five years plan and head for green sustainability development, critical restrictive measures on coal ratio in the total energy consumption should pay more attention. The primary method for the government is to lower the ratio of coal input and promote the proportion of windy, nuclear, solar, biomass, and hydroelectricity. Fig. 1 plots the share of three fossil energy (coal, oil, and natural gas) and non-fossil energy to the PEC from 2000 to 2019.

Fig. 1.

Trend in primary energy consumption structure and CO₂ emissions in China

Notes: The data on the ratio of three representative fossil energy and renewable energy ratio to the total primary energy are calculated by the authors. The original data are from the China Statistical Yearbook from 2000 to 2019. The total CO₂ emissions are from the BP statistical review 2020 (available at https://www.bp.com/en/global/corporate/energy-economics/statistical-review-of-world-energy.html).

A remarkable observation of Fig. 1 is that a simultaneous growth trend in renewable energy and natural gas consumption, and a decrease in the share of coal in the PEC. The share of renewables has lifted for almost 2 times period from 2000 to 2015. Fig. 2 depicts the rate of change in carbon intensity compared to the level of 2005.

Fig. 2.

Trends in energy consumption structure and carbon intensity from 2005 to 2019, in China.

Notes: In Fig. 2, CI represents carbon intensity. Letters after underline, i.e., "cal", "BP", and "IEA" denote different original data sources, where "CI_cal" in column four means the carbon intensity result is calculated according to the original data from the China Statistical Yearbook. Accordingly, "CI_BP" and "CI_IEA" is calculated according to the BP statistical energy review 2019 and the IEA database, respectively. Numerical values in the lower right subgraph represent the changing rate of each year's carbon intensity compared to the level in 2005.

As for three different data sources are employed to calculate carbon intensity (the lower right subgraph in Fig. 2), a significant change in the PECS (primary energy consumption structure) with a lower share of coal has led to about 50% decrease in carbon intensity, which confirmed an ahead of schedule in 2020 carbon intensity target. Another remarkable observation of Fig. 2 is that the share of coal consumption in the PECS is in a downward trend from 72.4% in 2005 to 57.7% in 2019, the share of oil consumption is mainly smooth among all these years, and the share of natural gas, however, has been increased almost 3 times compared to the level in 2005, i.e., 2.4% in 2005 to 8.1% in 2019. Similarly, the share of non-fossil fuel consumption has been escalated almost two times since 2005. Therefore, a significant amount of change has been taking place in the Chinese energy consumption structure all these years. The heading direction is more low-carbon, cleaner, and pluralistic.

The rest of the paper is organized as follows: Section 2 reviews the relevant and informative literatures. Section 3 presents the methodology for both the Markov chain model and the construction of the DSGE model. Section 4 provides parameter calibration. Section 5 analyzes the impulse response of the macroeconomy, energy consumption, and carbon emissions under different exogenous uncertainties. Section 6 concludes the study and propose essential and useful implications.

2. Literature review

The debate between cap-and-trade and carbon tax (the two major carbon emission reduction mechanisms to deal with global warming) has been going on for years unsettled (Hu et al., 2020). Carbon emissions trading scheme (CETS) is considered to be an effective means of carbon emissions reduction that is based on carbon market transactions and minimizes abatement costs. Zhang et al. (2020) assessed the impact of carbon trading on economic output and carbon emissions by applying China's industrial data during 2006–2015. The results showed that carbon trading can not only help the industrial sectors to reduce emissions, but also achieve the target of reducing carbon intensity. Yu et al. (2020) using provincial panel data demonstrated that every 10% promotion in renewable energy generation would lower the carbon intensity by approximately 0.84%—1.49%.

From the forecasting perspective, the Markov chain is a discrete-time random process with Markov properties in mathematics (Han et al., 2020b). In this process, given the current knowledge or information, the past historical state is irrelevant to predicting the future state. Notably, a real-time energy management system was also in line with the Markov Chain features (Li and Lu, 2019). Hence, it is an excellent choice to use this method in forecasting energy structural adjustment in the future.

Many studies have investigated the relationship between energy consumption structure and carbon emissions reduction. Some studies have focused on the relationship between renewable energy use and low-carbon transition, such as passing out subsidies (Monasterolo and Raberto, 2019), renewable energy cooperation in electricity (Gullberg et al., 2014), uncertain investment decisions in renewable energy use towards low-carbon transition (Zhang et al., 2019b). Using a dynamic threshold panel regression model, Wu et al. (2020b) evaluated the nonlinear relationship between energy consumption and carbon emissions. And the vital role of energy consumption played in surging carbon emissions were also confirmed. By applying the panel quantile regression model, Khan et al. (2020) examined the heterogeneity of renewable energy consumption, carbon emissions and financial development. The negative impact of renewables on CO₂ emissions was confirmed.

The existing studies have learned about the relationship between energy consumption structure and CO₂ emissions, most of them were standing from a static perspective with an ignorance of the dynamic stochastic exogenous shocks during the economic development process. Since Kydland and Prescott (1982) proposed the Real Business Cycle (RBC) theory, a body of techniques has been developed that now allows one to build quite complicated microeconomic-based macroeconomic models. The main characteristics of DSGE models are that it is “micro-founded,” i.e., agents optimize their behavior based on well-specified and time-invariant preference and technology parameters. Since then, the DSGE models have become mainstream in the macro-economy analysis. A conventional RBC model not only captures the characteristics of economic growth, but also efficiently analyzes the economy based on assumptions about preference, endowments, and technology (Zhang et al., 2019a).

Few studies have employed the DSGE models to environmental issues recently (Fischer and Springborn, 2011; Heutel, 2012; Fischer and Heutel, 2013). These efforts mainly represented the US economic fluctuation after the shock occurred, and ignore the energy consumption structural adjustment impact on carbon emissions with consideration of economic development. In addition, when assessing carbon emission reduction policies using the DSGE model, many studies focusing on its emission reduction effect or economic effect without consideration of exogenous shocks can also affect their policy effect. Besides, employing DSGE model research on China's emission reduction regarding energy structure is rarely reported. Also, embedded a Markov chain result into a DSGE model have not yet been reported, either.

On top of the evolving studies in energy structural change and CO₂ emissions reduction, this paper contributes to the current literature on both methodologies and research perspectives. To begin with, the research target of this paper is to evaluate whether the 2030 carbon intensity target would be achieved, even under a lower speed development scenario simply through energy structural change. After that, we simulate and evaluate how the economy would react to energy structural change confronting different exogenous shocks, which cannot be simulated perfectly in other types of models except for a DSGE model. Therefore, we model the substitution between fossil energy and non-fossil energy in a DSGE framework by embedding the prediction results according to the Markov model. This is for the first time a combination of a Markov chain model of a DSGE model, which successfully solved the difficulty of introducing energy consumption historical change into a DSGE model. After that, a DSGE model consists of three sectors (households, firms, and the government) were constructed to make the whole economy an integration, where we can see it as a macroeconomic laboratory and simulate different exogenous shocks to see the macroeconomic variables reactions, so as to forecast the impact of energy structural change on the economy in the next four decades through dynamic convergence paths.

3. Methodology

3.1. Markov chain model

According to Wang et al. (2010), the variation of the energy consumption structure in China identified as a Markov process. As aforementioned, the Chinese energy consumption structure is affected by multiple different factors, such as economic growth, energy imports reliability, energy price, environmental and policy factors. Under the co-effect of these factors and budget constraints, the energy consumption structure can be regarded as a collection of best choices for all households and firms. The higher the probability a specific type of energy is, the higher the share of it to the primary energy consumption. And each year's primary energy consumption structure can be regarded as energy distribution among households and firms. Therefore, it is reasonable to use the Markov chain model as a way to forecast future energy consumption structure with full consideration of the historical impact of multiple factors. And energy primary consumption structure in each year can be regarded as the best choices' distribution for all the households and firms in the economy.

3.1.1. Basic principles

The Markov chain model presents a stochastic process. A stochastic process is a family of random variables $\left\{\mathrm{X}\left(\mathrm{t}\right),\mathrm{t}\in \mathrm{T}\right\}$ defined on a given probability space S, indexed by the parameter t, where t is in an index set T. Specifically, the Markov chain is defined as follows:

Assuming that the random sequence $\left\{{\mathrm{X}}_{\mathrm{n}},\mathrm{n}\in \mathrm{T}\right\}$ has a discrete parameter set $T=\left\{0,1,2,\dots \right\}$ and a discrete state-space $\mathrm{S}$, where S is a collection of all possible values that the random variables of the stochastic process may assume. If $\mathrm{P}\left\{{\mathrm{X}}_{n+1}={\mathrm{s}}_{n+1}|X_0=s_0,X_1=s_1,X_2=s_2,\dots ,X_n=s_n\right\}>0$ is applicable for any positive integer $\mathrm{n}\in \mathrm{N}$ and any state ${\mathrm{s}}_1,s_2,\dots ,s_n\in S$, the transfer process is called a Markov chain when the conditional probability satisfies Eq. (1) (Han et al., 2020b).

$$\begin{array}{ccc}& & \mathrm{P}\left\{{\mathrm{X}}_{n+1}={\mathrm{s}}_{n+1}|X_0=s_0,X_1=s_1,X_2=s_2,\dots ,X_n=s_n\right\}\hfill \\ & & =P\left\{{\mathrm{X}}_{n+1}={\mathrm{s}}_{n+1}|{\mathrm{X}}_{\mathrm{n}}={\mathrm{s}}_{\mathrm{n}}\right\}\hfill \end{array}$$ (1)

The values assumed by a random variable ${\mathrm{X}}_{\mathrm{n}}$ are called “states”. Eq. (1) denotes that the state at moment n + 1 is only concerned with the state at moment n. It is a significant feature of a Markov process that the future state is merely associated with the current state and does not rely on the previous one (Li and Lu, 2019). In other words, this characteristic is also known as the "non-aftereffect" property (Li et al., 2018). Following Li and Lu (2019), for all $\mathrm{i},\mathrm{j}\in {\mathrm{S}}_{\mathrm{n}}$, the Markov chain one-step transition probability matrix is defined in Eq. (2), which denotes a state transition from moment n to moment n + 1.

$$\mathrm{P}\left\{{\mathrm{X}}_{n+1}=j|X_n=i\right\}\genfrac{}{}{0pt}{}{\Delta }{=}{p}_{\text{ij}}\left(n\right)\left({\mathrm{p}}_{\text{ij}}\left(n\right)\ge 0;\sum _{j\in S}{p}_{ij}\left(n\right)=1\right)$$ (2)

When the transition probability is independent of n, the Markov chain becomes a homogeneous Markov chain, which is the most commonly used random process. Hereafter, the one-step transition probability matrix is defined in Eq. (3), and $p_{ij}\left(n\right)$ becomes $p_{ij}$.

$$P=\left(p_{ij}\right)=\left[\begin{array}{cccc}{p}_{11}& p_{12}& \cdots & p_{1n}\\ p_{21}& p_{22}& \cdots & p_{2n}\\ ⋮& ⋮& & ⋮\\ p_{n1}& p_{n2}& \cdots & p_{nn}\end{array}\right]$$ (3)

Under this assumption, the k step transition probability is calculated by Eq. (4) as follows

$${\mathrm{P}}_{\text{ij}}^{\left(k\right)}=P\left\{M_{n+k}=j|M_n=i\right\}$$ (4)

According to the assumption of Chapman-Kolmogorov equation,5 if the initial probability distribution vector is ${\mathrm{P}}_0$, the probability distribution of the state after k periods is calculated by Eq. (5)

$$\mathrm{P}\left(\mathrm{k}\right)=P_0{P}^k$$ (5)

3.1.2. Markov-based energy consumption structure prediction model

The variation of the energy consumption structure is identified as a stochastic process (Wang, 2011; Li and Lu, 2019). The Markov chain model is employed to forecast the possible energy consumption structure in China and containing the impact of the COVID-19 pandemics. Define its state space as S and the initial energy consumption structure state vector can be described as ${\mathrm{S}}_{\mathrm{i}}=\left\{S_{coal}\left(i\right),S_{oil}\left(i\right),S_{gas}\left(i\right),S_{re}\left(i\right)\right\}\left(S_i\in S\right)$. The four elements $S_{coal}\left(i\right),S_{oil}\left(i\right),S_{gas}\left(i\right)$, and $S_{re}\left(i\right)$ in ${\mathrm{S}}_{\mathrm{i}}$ represents the initial ratio of coal, oil, natural gas, and renewable energy consumption to the primary energy consumption. The transition probability matrix calculations identified as the most critical procedure during the prediction process (Li and Lu, 2019; Li et al., 2018). The one-step transition probability matrix is shown in Eq. (6)

$$P_i=\left[\begin{array}{cccc}{p}_{coal\to coal}\left(i\right)& p_{coal\to oil}\left(i\right)& p_{coal\to gas}\left(i\right)& p_{coal\to re}\left(i\right)\\ p_{oil\to coal}\left(i\right)& p_{oil\to oil}\left(i\right)& p_{oil\to gas}\left(i\right)& p_{oil\to re}\left(i\right)\\ p_{gas\to coal}\left(i\right)& p_{gas\to oil}\left(i\right)& p_{gas\to gas}\left(i\right)& p_{gas\to re}\left(i\right)\\ p_{re\to coal}\left(i\right)& p_{re\to oil}\left(i\right)& p_{re\to gas}\left(i\right)& p_{re\to re}\left(i\right)\end{array}\right]$$ (6)

Where the main diagonal elements denote the original share probability of the four types of energy consumption subcategories, the row elements indicate the conversion probability for each subcategory, and the column elements imply the switching probability from other subcategories.

This study intends to forecast energy consumption structural change from 2020 to 2060 based on the historical data of 2011–2019.6 The prediction is processed over three steps: First, suppose the initial state of energy consumption structure the one at 2011; Second, the one-step transition probability matrix is calculated annually with one year as a time interval; Third, the average probability transition matrix can be calculated accordingly. The average transition matrix can be obtained by Eq. (7)

$$P={\left[\underset{i=1}{\overset{n}{\Pi }}P\left(i\right)\right]}^{\frac{1}{n}}$$ (7)

The main diagonal element equals 1 when a type of energy has a higher ratio of primary energy consumption at time t + 1 comparing to time t. In other words, the conversion probability of this type of energy is 1. The calculation result of transition probability matrix for each year are shown in Table 1 .

Table 1.

The one-step transition probability matrix for 2011–2019.

Notes: The main diagonal elements are called reversion probability. When the share of a specific type of energy to the primary energy consumption at moment t + 1 is not less than it is at time t, the reversion probability equals one. Suppose the sum of the row equals one, when the diagonal element equals one, the rest element in the same row equal to zero.

Hence, the average probability transition matrix is shown in Eq. (8).

$$P=\left[\begin{array}{cccc}0.9760& 0.0063& 0.0062& 0.0115\\ 0& 1& 0& 0\\ 0& 0& 1& 0\\ 0& 0& 0& 1\end{array}\right]$$ (8)

3.2. DSGE model construction

To simulate the impact of energy consumption structural change on the macroeconomy development and carbon emissions under different exogenous shocks, we employ a DSGE model with energy structural substitution. The parameters refer to energy structural change is obtained from the forecasting result in Section 3.1.

We develop a DSGE model with three sectors: households, the government, and the final goods firms. To investigate the energy substitution effect on carbon emission reduction, the substitution rate achieved by the ratio of renewable energy taken into the total energy consumption is under consideration. A function is employed to relate fossil energy consumption to carbon emissions. This function allows us to examine the effect of energy structural adjustment on carbon emissions and its impact on economic development. In order to simplify the model, we only consider carbon emissions emanated from fossil energy combustion.

3.2.1. Households

We adopt a separable instantaneous logarithmic utility function. It is assumed that there is a typical household sector with unlimited life span in the economy. The real business cycle (RBC) model with indivisible labor suggests that a representative consumer can earn income by supplying labor to the final goods firms and get utility through the enjoyment of consumption and leisure time. Representative household allocates their time endowment (which simplify to be one) between work and leisure time, each of them maximizes the expected value of an intertemporal utility function in Eq. (9).

$$E_0\sum _{t=0}^n{\beta }^t\left[\ln C_t+\eta \ln \left(1-L_t\right)\right]$$ (9)

where β denotes the discount factor, η is the relative preference of the leisure time compared to the consumption. ${\mathrm{C}}_{\mathrm{t}}$ and ${\mathrm{L}}_{\mathrm{t}}$ represent consumption and working hours, respectively. Besides, to simplify the model calculation, following Fischer and Heutel (2013), we presumed a constant population and normalized it to be 1.

Representative household maximizes their expected utility under certain constraints. The intertemporal budget constraint is as follows in Eq. (10).

$$w_t{L}_t+R_t^k{K}_t=C_t+I_t$$ (10)

where ${\mathrm{w}}_{\mathrm{t}}$ indicate the nominal wages, and $R_t^k{K}_t$ is the nominal return of renting capital during time t, respectively. ${\mathrm{I}}_t$ is the investment. The capital accumulation equation of the household can be described as follows:

$$K_{t+1}=\left(1-\delta \right)K_t+I_t$$ (11)

where $\delta$ denotes the depreciation rate. The maximization of the utility function under certain constraint can yields the first-order conditions for consumption and working. The household's first-order conditions of the optimal decision problems are as follows:

$$w_t=\frac{\eta C_t}{1-L_t}$$ (12)

$$1=\beta E_t\left[\frac{C_t}{C_{t+1}}\left(R_{t+1}^k+1-\delta \right)\right]$$ (13)

3.2.2. Final goods firms

Assuming that the final goods firms use energy as input during the production process and CO₂ emissions as a byproduct. Hence, it is to choose from $\left\{{\mathrm{L}}_{\mathrm{t}},{\mathrm{K}}_{\mathrm{t}},{\mathrm{E}}_{\mathrm{t}}\right\}$ to achieve the goal of profit maximization. Meanwhile, they should bear the carbon permits when the total amount of CO2 emissions beyond the free carbon permits a firm is allocated. Following our previous study (Zhao et al., 2020), the output is supposed to be a constant-to-scale Cobb-Douglas function in Eq. (14).

$$Y_t=Z_t{K}_t^{\alpha }{L}_t^{\gamma }{E}_t^{1-\alpha -\gamma }$$ (14)

where $\alpha$ and $\gamma$ denote the share of income of capital and labor, respectively. ${\mathrm{Z}}_{\mathrm{t}}$ represents the technology level, which following an AR (1) process as follows:

$$\ln Z_t=\ln \overline{Z}+{\rho }_z\ln Z_{t-1}+{\epsilon }_t^z{\epsilon }_t^z\sim i.i.d.N\left(0,{\sigma }_z^2\right)$$ (15)

In the production function, we add the energy use as input and the total energy use can be divided into two parts, fossil energy use and renewable energy use.

$$E_t=E_t^{trad}+E_t^{rene}$$ (16)

where ${\mathrm{E}}_{\mathrm{t}}^{\text{trad}}$ and ${\mathrm{E}}_{\mathrm{t}}^{\text{rene}}$ represent fossil energy use and renewable energy use, respectively. And we assumed that carbon dioxide emissions are only generated from fossil energy use. From the perspective of energy efficiency in reality, the price difference between new energy and conventional fossil energy is slight. By assuming the ratio of renewable energy use in the total energy consumption equals to $\xi$, the total $\mathrm{C}{\mathrm{O}}_2$ emissions equal to

$$\mathrm{E}{\mathrm{M}}_{\mathrm{t}}=\left(1-{\xi }_t\right){\varphi }_t^{trad}{E}_t$$ (17)

where ${\varphi }_{\mathrm{t}}^{\text{trad}}$ is a comprehensive carbon emissions factor calculated according to each type of fossil energy and their corresponding ratio in the total energy use, and ${\xi }_{\mathrm{t}}$ is the ratio of renewable energy to the total energy consumption during period t.

3.2.3. Government

Taking the lead in the new trend of carbon intensity target and energy consumption structure target, China has explored a cap-and-trade system since 2013 with seven pilot cities and provinces (Hu et al., 2020). This system features a free allocation of initial carbon permits to enterprises by their historical amount of emissions. When the current emissions beyond the free allocation, the enterprise needs to pay for excessive emissions, and the price of the carbon permits is determined by the carbon market. The pilot cities that have implemented the cap-and-trade system often use a combination of both free allocation and auction for the initial permit distribution. Hence, we assumed that the government collect and use all of the revenue from the auction, and the constraint budget can be expressed as

$$G_t=\left(1-\theta \right)P_t^{em}E{M}_t$$ (18)

where $\theta$ is the ratio of free permits and ${\mathrm{P}}_{\mathrm{t}}^{\text{em}}$ represents the auction price of the rest permits, which assumed following a stochastic autoregression process as follows:

$$\ln P_t^{em}=\ln {\overline{P}}^{em}+{\rho }_{em}\ln P_{t-1}^{em}+{\epsilon }_t^{em}{\epsilon }_t^{em}\sim i.i.d.N\left(0,{\sigma }_{em}^2\right)$$ (19)

3.2.4. The market cleaning condition

Market cleaning condition is given by:

$$Y_t=C_t+I_t+G_t$$ (20)

We linearize the equilibrium conditions around the steady-state, and conduct the impulse response simulation on the obtained linear equation by the Dynare software version 4.5.1.7

4. Parameters

The parameters in our DSGE model can be divided into two parts. On the one hand, the parameter determines the steady-state of the model and is determined by the calibration method. The other parameters are determined by data from realistic energy consumption, carbon emissions and prediction results in Section 3.1.2.

4.1. Parameter calibration

To calibrate the static parameters in our model, we refer to the abundant literatures and choosing the value which are suitable for the model. The values of steady-state parameters in our DSGE model are shown in Table 2 .

Table 2.

The values of steady-state parameters in the DSGE model.

Parameters	Interception	Value
$\beta$	Subjective discount factor	0.985
$\alpha$	Capital share on output	0.493
$\mathrm{\gamma }$	Labor share on output	0.349
$\mathrm{\delta }$	Capital Depreciation rate	0.025
$\eta$	Relative preference for leisure time	0.77

Following previous studies, the calibration of the household subjective discount factor using quarterly consumer price index from 2000 to 2018 and the value is set to be 0.985 (Song et al., 2019; Balke and Brown, 2018). Also, parameter $\alpha$ and $\gamma$ are set to be 0.493 and 0.349 (Fan et al., 2016; Niu et al., 2018).

4.2. Parameter estimation and calculation

Parameter ${\varphi }_{\mathrm{t}}^{\text{trad}}$ in our model is a time variation parameter that need calculates according to corresponding energy consumption, especially the ratio of three representative fossil energy to the total energy consumption. It is determined by Eq. (21):

$${\varphi }_t^{trad}=\sum _{i=1}^3{e}_i{\phi }_i\left(i=1,2,3\right)$$ (21)

In Eq. (20), the value of i equals to 1,2, and 3, which indicated coal, oil, and natural gas, respectively. Variable ${\mathrm{e}}_{\mathrm{i}}$ and ${\phi }_{\mathrm{i}}$ denote i type of energy consumption ratio in the total energy consumption and its corresponding carbon dioxide emissions factor, respectively. The value of the parameter ${\varphi }_t^{\text{trad}}$ is different when the energy substitution rate change. To simulate the impact of energy consumption structural change on economic development and carbon dioxide emissions, three scenarios are set by different energy substitution rates, which is $\xi$ in our model.

According to the prediction result based on the Markov chain, the renewable energy consumption is approximately 21.8% in 2030. With no other change, this percentage will be 32.7% in 2060. Hence, we can set up three different scenarios by different free carbon permit ratio and the ratio of renewable energy consumption.

4.3. Scenario settings

Most importantly, in order to figure out the impact of both the energy consumption structural change and the CETS on energy consumption (both fossil energy and non-fossil energy), CO₂ emissions, and economic development, we consider three different scenarios for comparison; each of them corresponds to the present, 2030, and 2060. The specific scenarios and key variable settings are presented in Table 3 .

Table 3.

The specific scenarios and key variable settings.

Scenario	Description	Key variable
Benchmark (BAU)	Corresponding to present	$\xi =0.15,\theta =0.9$
Mid-term (MT)	Corresponding to 2030	$\xi =0.218,\theta =0.9$
Long-term (LT)	Corresponding to 2060	$\xi =0.327,\theta =0.1$

Notes: Three different scenarios are considered during the DSGE model simulation process. Two key variables are substitution rate of non-fossil energy (i.e., $\xi$) and free share of carbon permit in the total permits (i.e., $\theta$). Values of $\xi$ are prediction results according to the Markov chain model.

5. Results and discussion

5.1. PECS prediction results and carbon intensity in different scenarios

According to Eqs. (4)–(7), the prediction result of energy consumption structural change from 2020 to 2060 can be achieved and plotted in Fig. 3 .

Fig. 3.

Markov chain prediction results for Primary Energy Consumption Structure (PECS).

Since there is no big difference among carbon intensity results with different original data sources of total carbon dioxide emissions, a prediction of carbon intensity from 2020 to 2030 can be calculated with energy structural prediction result according to the Markov chain model. The specific calculation results are shown in Table 4 .

Table 4.

Forecast the actual contribution of the carbon emission reduction target in 2020–2030 by scenario.

Year	Primary energy consumption structure (PECS)				Carbon intensity			Contribution to the 2030 carbon intensity target
					(tonnes/million yuan)
					Steadyscenario	Post-COVID-19 scenario		Steady scenario	Post-COVID-19 scenario
	Coal (%)	Oil (%)	Natural Gas (%)	Non-fossil fuels (%)		4.20%	2.30%		4.20%	2.30%
2020	56.3152	19.2635	8.4577	15.9636	6.80	6.93	7.06	80.03%	78.69%	83.73%
2021	54.9636	19.6183	8.8069	16.6112	6.46	6.70	6.95	83.79%	81.21%	85.03%
2022	53.6445	19.9646	9.1477	17.2433	6.13	6.47	6.84	87.35%	83.64%	86.29%
2023	52.3570	20.3025	9.4803	17.8602	5.82	6.26	6.73	90.71%	85.98%	87.53%
2024	51.1005	20.6324	9.8049	18.4623	5.53	6.05	6.63	93.90%	88.23%	88.73%
2025	49.8741	20.9543	10.1217	19.0499	5.25	5.85	6.53	96.92%	90.40%	89.91%
2026	48.6771	21.2685	10.4309	19.6235	4.98	5.66	6.43	99.77%	92.48%	91.05%
2027	47.5088	21.5752	10.7327	20.1833	4.74	5.47	6.34	102.48%	94.49%	92.17%
2028	46.3686	21.8745	11.0273	20.7296	4.50	5.29	6.25	–	96.43%	93.26%
2029	45.2558	22.1666	11.3148	21.2629	4.28	5.12	6.16	–	98.29%	94.32%
2030	44.1696	22.4517	11.5953	21.7833	4.06	4.96	6.07	–	100.09%	95.36%

Notes: Numbers in the second column are displaying in each type of energy's percentage from 2020 to 2030 to the primary energy consumption based on the Markov-chain prediction model. "Steady scenario" in the third column means the economic growth rate stays at 6.1% (the same as 2019), and the "post-COVID-19 scenario" in the fourth column means a hypothesis downward trend happened to economic growth since the unexpected shock of the COVID-19. There are two hypothesis scenarios in the post-COVID-19 scenario, where the economic growth rate is downward to 4.2% and 2.3%, correspondingly. The contribution to the 2030 carbon intensity target means the ratio of carbon intensity reduction in the corresponding year to the reduction when meeting the 2030 target.

In order to evaluate the PEC structural change's contribution to the 2030 carbon intensity target, three different scenarios are considered, each of which corresponds to a different GDP growth rate. During the pandemic of COVID-19 all over the world, GDP growth rates are set to be 6.1%, 4.2%,8 and 2.3%, respectively, and each of them corresponding to the steady-, middle-, and low-speed growth scenario. The PEC is calculated according to the annual average growth rate of 1.86% from 2000 to 2019. Real GDP is calculated at constant prices in 1978. This section analyzes the corresponding carbon emission intensity under the two scenarios, and the actual contribution of each scenario to achieve the goal of reducing carbon intensity by 60% - 65% in 2030 compared with 2005.

In Table 4, the contribution of energy consumption structure adjustment to the carbon intensity target under two economic growth scenarios is calculated based on a 65% decrease in carbon intensity compared with the 2005 level. Under the scenario of steady economic development, when coal, oil, natural gas, and non-fossil energy account for 47.5%, 21.6%, 10.7%, and 20.2% of the PEC, it is predicted that the carbon intensity in 2027 will decrease by 65% compared with the 2005 level. When the growth rate of the economy slows down to 4.2% results from the COVID-19 pandemic, it is predicted that the carbon intensity will be reduced by 65% compared with 2005 around 2030. According to the results, we can infer that the structural change in PEC will lead to the carbon intensity target of 2030 achieved as long as the growth rate of GDP is higher than 4.2%. Hence, it can also be the right choice to achieve the carbon neutrality target in 2060 by energy consumption structural change and promote the ratio of non-fossil energy. The non-fossil energy mainly consists of windy-, solar-, biomass-, and tidal energy. Among them, the combustion of biomass energy may generate carbon emissions. However, this type of emissions belong to cyclical carbon flow, compared with linear carbon flow generated by fossil-based energy use (Berndes et al., 2016).

5.2. Impulse responses analysis

In this section, we analyze the dynamic economic impacts of the energy structural change from the perspective of multiple different exogenous shocks we considered during the DSGE model construction in Section 3.2. The impulse responses of the economy to exogenous stochastic shocks under three different scenarios and CETS constraint are shown in Fig. 4, Fig. 5, Fig. 6 . In the following, differences in the impulse responses of each variable among three scenarios.

Fig. 4.

Impulse responses of carbon permit price shock.

Fig. 5.

Impulse responses of energy price shock.

Fig. 6.

Impulse responses of technology shock.

5.3.1. Environmental policy uncertainty

We evaluate the effect of environmental policy uncertainty by simulating the effect of a positive carbon permit price shock on the economic system. Fig. 4 represents the results of environmental policy uncertainty in three scenarios, viz.: the benchmark (BAU), the midterm (MT), and the long-term (LT). All the impulse responses are reported as percentage deviations from the steady-state over a 10-quarterly period.

When the economy is positively affected by a carbon permit price shock, output, energy consumption, and CO₂ emissions will negatively deviate from their steady-state, while an upward trend is shown in consumption, labor supply, and capital stock. According to the current settings relating to the CETS, a positive shock of carbon permit price can be regarded as a rise in enterprises cost if no adjustment happened to the previous inputs' portfolio. Hence, enterprises will reduce fossil energy inputs and invest more labor force and capital stock instead at the beginning when the energy substitution rate is low (i.e., 15% and 23.7% corresponding to scenarios BAU and MT, respectively). And at the general equilibrium condition, more demand in the labor force generates more labor supply, and result in higher income for households after a long working hour. Households can afford more products than they used to, and that can explain the rise of consumption and labor force supply since the shock happens.

All of the variables show differences at the beginning of shocks that happened to the three scenarios, and their impulse response curves overlap with each other when they convergence their steady-state. One remarkable observation of Fig. 4 is that the impulse responses under the LT (yellow curve) have a big deviation compared to the BAU and the MT. This fact indicates that when the share of free permits takes approximately 90% of the initial allocation at the first stage of CETS implementation, the policy constraint is relatively small on the CO₂ emissions and has little impact on the economy. And a significant decrease in the share of free permits give the economy more constraints when there is a rise in carbon permit price.

With a sudden decrease in the share of free permits, and a more than half percent of renewable energy substitution rate under the LT scenario, most variables show a significant deviation at the beginning and gradually reach a convergence with steady-state after ten periods. This outcome indicates that CETS has a short duration of impact and when a rise happens to carbon permit price, enterprises with emissions exceeding the allocated free permits need to bear an additional cost to get more carbon permits, which will lead to a fall in energy use for pursuing a maximum profit, especially in fossil energy use. This will lead to a significant recession in output (Fig. 4a) since the level of technology is not advanced enough for enterprises to increase the energy substitution rate in the benchmark scenario. The enterprises have to compress the output for cost-saving purposes under the limitation of carbon emissions.

5.3.2. Energy price uncertainty

Fig. 5 depicts the results of output, consumption, capital stock, labor supply, energy consumption, and carbon emissions’ impulse responses to an energy price shock. All the impulse responses are reported as percentage deviations from the steady-state of variables over a 100-quarter period. Owing to a one-unit positive energy price shock, fluctuations in output, energy consumption, and carbon emissions decrease, whereas consumption, capital stock, and labor supply increase.

When the economy is positively impacted by an energy price shock, a simultaneous decrease showed in total output and energy consumption, stimulating the carbon emissions to fall in response to the shock. When a rise happens to energy price, especially concerning the international oil price, all the variables show bigger deviations from their steady-state. For the output, it shows a significant downward trend in scenario 1 compared to the scenario LT in subplot Fig. 5a, indicating that when a higher ratio of renewable substitution in the total energy consumption, all the economic losses caused by the decrease in the ratio of free carbon permits can be smooth to some degree. Besides, an approximately reverse trend happens to the consumption and labor supply, which are shown in Fig. 5b and c. This reverse trend indicates that under the LT (compare to the BAU) a higher energy substitution with lower free initial carbon permits allocation would lead to a relatively higher growth rate in residential consumption and a slower growth rate in labor supply. However, compared with scenario 1 alone, an increase in the share of renewable energy in the input portfolio of the production process under the LT makes the economy maintaining the same level of output without excessive labor supply. Therefore, the growth rate of labor supply is slower compare to the BAU.

5.3.3. Technological progress uncertainty

Fig. 6 depicts the impulse responses of the economy to a TFP shock. All the impulse responses are reported as percentage deviations from the steady-state of variables over a 100-quarter period. Owing to a one-unit positive TFP shock, fluctuations in output, energy consumption (both renewables and fossil energy) increase, whereas consumption, capital stock, CO₂ emissions and labor supply decrease.

When the economy is positively impacted by a TFP shock, which can be regarded as unexpected technology progress, most variables react positively. Intuitively speaking, a technology shock means an enhancement in productivity, and enterprises can maintain the same level of output with fewer inputs. Under the constraint of the CETS, however, enterprises may prefer to adjust their inputs portfolio with more non-fossil fuels instead of fossil energy regarding to energy usage. An expansion in output (see Fig. 6a) will lead to an increase in energy usage (both in renewable energy and fossil energy). One remarkable observation of Fig. 6 is that both capital and labor supply show a decrease since the technology shock happens, which implies that enterprises more prefer to expand their output scale with energy inputs, especially with technical progress, enterprises get access to more choices. In reality, the cost differences between fossil energy and renewable energy are getting smaller and the technology is no longer the biggest hinderance to the large-scale use of renewables. Enterprises change their inputs portfolio by applying renewables more than a half in their energy mix under the LT, although there is a sharp decrease in the share of free permits, the output maintains the status quo (compare to the BAU and the MT). This observation indicates that energy structural change can smooth volatility to some degree.

6. Conclusions and policy implications

Mitigation of carbon emissions without heavily hindering economic development is of great importance, especially in the post-COVID-19 era. Since the impact of the COVID-19 on energy consumption and CO₂ emissions is only temporary, balancing economic recovery, curb CO₂ emissions, and reach the carbon neutrality target in a post-COVID-19 era is urgent and necessary. This paper contributes to the literature from both methodologies and research perspectives. The Markov-based energy structure prediction model built in this paper gives us a clear picture of whether the carbon intensity target of 2030 would be achieved under three economic growth scenarios with the PEC structural adjustment. The dynamic stochastic general equilibrium (DSGE) model constructed in this paper provides a comprehensive analytical framework for how would the structural change in PEC affect the economic development and carbon emissions under carbon trading scheme constraint and different exogenous shocks. Regarding to the methodology, this is for the first time a Markov-based energy structure prediction model combined with a DSGE model. The results of this paper indicate the following.

First, the energy consumption structure has changed a lot in China. Compared to 2000, the share of non-fossil energy to the PEC is improved almost twice in 2019. Calculation results based on three original data resources showed that with a significant increase in the share of non-fossil energy and a decrease in the share of coal consumption, the carbon intensity in 2019 has decreased by 50.24% approximately compared to the 2005 level.

Second, the Markov-based energy structure prediction model built in this paper gives us a clear blueprint that energy structural change can help us achieve the carbon intensity reduction target in 2030 both under steady and lower-speed development scenarios. As long as the growth rate of GDP higher than 4.3%, the carbon intensity target would be achieved through structural change in the PEC, which provides a way to achieve the carbon neutrality target by 2060.

Third, the simulation result of the DSGE model we constructed indicates that the energy price shock has a great impact on the economy compared to the carbon permit price shock. A rise in carbon permit price can lead to a more significant effect on both carbon emissions and the economy when the free initial carbon permits decrease. When the economy is confronting technological progress, enterprises prefer to use more energy than labor force or capital stock to expand their output scale. And the results of the simulation also confirmed that structural change in the PEC could smooth the volatility of economic development to some degree.

Finally, both climate change and the COVID-19 crisis are sharing in common: generating human tragedies and economic catastrophes. Rather than invest in fossil-fueled intensive industries, a greener and more sustainable way for economic recovery is an optimal choice. Encouraging the development of a low-carbon and clean energy consumption mix is also essential. Also building climate-resilient infrastructure and supporting renewable energy expansion is of great necessity. For example, we should take full advantage of the latest technological progress of renewable energy such as photovoltaic power generation and biomass energy to promote the energy structure transition. The cyclical carbon emissions generated by biomass energy will not increase the burden of atmospheric carbon emission.