Did Carbon Emission Trade Improve Resource Misallocation? Evidence from China

Abstract

Chinese Emissions Trading Scheme (ETS) has been a crucial policy for China, a major manufacturing and carbon-emitting country, in achieving its “carbon neutrality” goals and sustainable development. This paper examines the performance of the ETS in China through the perspective of resource misallocation, and to find out whether and how China’s ETS works cost-effectively. Using Propensity Score Matching and the Difference-in-Differences (PSM-DID) method, this paper empirically investigates the relationship between China’s ETS and resource misallocation. The results show that: (a) the capital misallocation in ETS pilot areas is higher than other areas, while the labor misallocation in ETS pilot areas is lower than other areas. (b) The estimation results of PSM-DiD show that the ETS in China aggravated the capital misallocation, while it significantly improved the labor misallocation. The post-treatment test shows that the aggravation of the capital misallocation may fade with time, and the improvement of labor misallocation remains unchanged. (c) The proxy variable “state-owned” did significantly improve both capital and labor misallocation, indicating that the ETS in China worked partly and effectively as a “command-and-control” instrument. All the robustness tests are constructed, proving that the main results remain stable and reliable. This paper may provide some marginal contributions to the ever-growing empirical literature about the policy effects and mechanism of the carbon emission policies. These results prove that although the ETS is theoretically considered to be a market-oriented and cost-effective instrument, the supplementary policies are still essential and effective for ETS. These results also show that it may require a combination of diversified policies to reduce carbon emissions while maintaining sustainable economic growth.

1. Introduction

Carbon emission trading (ETS) has been widely adopted as a market-based instrument to reduce greenhouse gas emissions while maintaining sustainable growth. As a global leader in both population and manufacturing, China has gradually implemented the Chinese Emissions Trading Scheme (ETS) since 2011 (https://www.ndrc.gov.cn/xxgk/zcfb/tz/201201/t20120113_964370.html?code=&state=123, accessed on 16 March 2025). This initiative, widely regarded as an effort to emulate European policies, represents a key national strategy in China’s attempt to reduce greenhouse gas emissions and achieve sustainable development. An efficient ETS should facilitate the optimal allocation of resources, ensuring that emission reductions are achieved at the lowest possible cost. Given that resource allocation efficiency is a key component of sustainable development, examining whether ETS improves resource misallocation is essential for understanding its broader economic and environmental implications. This study aims to provide empirical evidence from China, offering insights into whether carbon markets contribute to sustainable economic development.

The cost-effectiveness of ETS and its impact on economic sustainability have become a hot research topic. ETS is designed to decarbonize industrial production by pricing and trading the emissions among emitters in an emissions trading market [1,2] and considered to be a cost-effective instrument as emitters have different marginal emission abatement. An increasing number of studies have investigated the relationship between ETS and carbon emissions [3,4], economic growth [5,6], innovation [7,8,9], investment [10] and productivity [11,12]. Also, a considerable body of literature has examined the impact of China’s ETS on emissions reduction, economic growth, and productivity, concluding that China’s carbon market is also efficient [13,14,15,16].

However, the extent to which ETS improves resource allocation remains debated, particularly in the context of China. China’s policies are formulated by the central government, while regional development varies significantly. Additionally, a substantial number of Chinese enterprises are state-owned, and regional development is characterized by local government competition and resource contention [17,18]. These factors may contribute to an incomplete market mechanism of ETS, and lead to resources misallocation, ultimately reducing the cost-effectiveness of emissions reduction efforts [14,19,20,21].

In light of this, the study examines the performance and mechanisms of China’s Emissions Trading Scheme (ETS) by addressing two core questions: First, has the implementation of the ETS cost-effectively improved resource misallocation? Second, does the operational mechanism of the ETS lean more toward market-oriented principles or administrative “command-and-control” approaches?

Based on theoretical analysis, we propose the following hypotheses: ETS may guide firms toward better resource allocation through market price signals; however, local administrative interventions, characterized by significant governmental involvement, might exacerbate capital misallocation. To test these hypotheses, this study first calculates the labor and capital misallocations. Then the Propensity Score Matching - Difference-in-Differences (PSM-DiD) approach is employed to examines the performance and mechanisms of China’s ETS on the resource misallocation, when the “proportion of state-owned enterprises” as a proxy for administrative intervention is introduced.

The contributions of this study may lie in three aspects. First, this study evaluates the ETS’s effectiveness in China from the perspective of resource misallocations. Second, this study explores the mixed influence of administrative interventions, characteristic of China’s governance, on market-oriented mechanisms. Third, the results of this study identify the heterogeneous responses in capital and labor misallocation, elucidating the complex impacts.

The future and practical applications of this study include providing empirical evidence for the Chinese government and other developing countries to formulate and optimize carbon market policies, clarifying the balanced integration of market mechanisms and administrative interventions, and supporting sustainable economic development. Additionally, this research lays foundational groundwork for future investigations into the long-term effects of the COVID-19 pandemic on corporate carbon emissions behavior.

The rest of this paper is organized as follows: Section 2 is the related literature review. Section 3 introduces the methodology and Section 4 describes the data and variables. Section 5 presents the main empirical results. Section 6 provides an interpretation of the empirical results, elucidating their significance and positioning them within the context of existing theoretical frameworks and prior empirical research. Section 7 elaborates on the practical implications and applied value of the research findings in business decision-making and operational management. Section 8 gives the concludes.

2. Literature Review

2.1. Environmental Regulation and the Cost-Effectiveness of ETS

There are two well-known and opposing theories regarding the relationship between environmental regulation and economic growth. The “Pollution Haven” hypothesis suggests that stringent environmental regulations may force firms to relocate to regions with more lenient policies [22,23]. In contrast, the “Porter Hypothesis” argues that environmental regulations can stimulate firms to innovate in green technologies, thereby enhancing their competitiveness [24,25].

The impact of the European Union Emissions Trading System (EU ETS) on firms is also centered around these two opposing viewpoints. ETS is considered as a market-oriented cost effectively carbon emission reduction mechanism [26]. Studies on the European ETS suggest that a well-functioning carbon market can enhance resource allocation efficiency and improve technological innovation [6]. Yet theoretical studies suggest that the possible distortions can result from allocation of CO₂ allowances to existing facilities to new entrants [27]. The question of whether ETS promotes corporate performance and economic growth remains a subject of debate. Literature find that the implementation of European ETS had a significant impact on the emission reductions when EUA (EU emission allowance) shift from phase I to phase II, while it may cause a negative effect on the performance of participating firms [28]. Studies on the impact of the EU ETS on firm-level economic performance across the EU suggest that environmental policies may have short-term negative effects on factory output and employment. However, this research demonstrates that environmental policies can enhance environmental performance without necessarily compromising economic performance [29]. The innovation system of power sector are affected by EU ETS and the corporate culture and routines play an important role in inspiring the transition to a low-carbon sector-specific innovation system [30]. In German, the EU ETS constitutes a main driver for small-scale investments with short amortization times rather than large-scale investments or R&D [10].

2.2. China’s ETS and Resource Misallocation

In China, the Carbon-emission Trading Scheme (ETS), which was published in 2011 and put in force in 2013 (https://www.ndrc.gov.cn/xxgk/zcfb/tz/201201/t20120113_964370.html?code=&state=123, accessed on 16 March 2025). The ETS pilots in China requires 4 cities (Beijing, Tianjin, Shanghai, Chongqing) and 3 provinces (Hubei, Guangdong and Fujian) to launch trial carbon ETS. In September 2020, China put forward the schedules of “carbon neutrality” and “emissions peak” which plans to achieve CO₂ emissions peak before 2030 and hit carbon neutrality before 2060.

Literature examining the contribution of China’s ETS policy to sustainable development has gradually increased since 2012. Early studies based on the EKC (Environmental Kuznets Curve) find that ETS has a considerable potential for cost saving and carbon emission reduction [31]. After the establishment of ETS pilots, literature study the impact of the carbon market on carbon emissions [32], carbon intensity [33], technological progress [34], productivity [35], firms investment [16]. Most of these literature suggest that the ETS in China significantly reduced the carbon intensity in the pilot areas [33].

However, the extent to which ETS improves resource misallocation in China remains contested. It is found that China’s carbon-trading market is confronted with challenges such as the absence of a functional carbon-trading market, inaccuracy of the quota allocation, an imperfect trading mechanism, and lagging legislation [36]. The statistic in majority of China’s regional carbon markets is insignificant in any given period and China’s carbon markets are not weak-form efficient [37]. At the macro level, the impact of emission reductions on capital and labor misallocation appears inconsistent [38]. Additionally, in the Chinese context, the presence of state-owned enterprises and significant regional disparities may introduce market distortions. While some studies argue that ETS enhances sustainability by effectively reducing emissions [39], others highlight concerns regarding policy-driven distortions and potential resource misallocation. For example, most firms take the participation in the ETS as a gesture to improve relations with the government and win a good social reputation, rather than a cost-effective mechanism to reduce carbon emission [40]. Also the study shows the ETS regulation affecting the cost of capital differs between non-state-owned firms and state-owned firm: the inverted-U-shaped relationship appeared in non-state-owned firms only, and state-owned enterprises showed a traditional linear relationship [41]. Carbon emissions may come from resources misallocation, which cause productivity losses [20]. The above research suggests that China’s ETS market contains non-market factors, which may stem from or be reflected by the heterogeneity between state-owned and non-state-owned firms. These non-market factors could lead to a cost-ineffectiveness of the ETS, which may not necessarily be captured by economic growth or productivity, but could potentially be reflected in resources misallocation.

This paper aim to investigate the performance the ETS in China through resources misallocation. This paper first calculates the resources misallocation provided by [19] to quantify the cost-effectiveness of ETS. The PSM-DiD model than is set up to investigate the impact of the ETS in China on the resources misallocation. Especially, this paper also evaluate whether the ETS works “market-oriented” or “command-and-control” by introducing instrumental variable reflecting policy strength. The possible marginal contributions of this paper are: (a) Evaluating the impact of the ETS on resource misallocation using PSM-DiD model. (b) Considering the influence of administrative intervention when studying the impact of ETS on resource misallocation. The research on the relationship between carbon emissions and resources misallocation may offers a new perspective on explaining the success or failure of China’s emissions reduction efforts and sustainable development. It also provides a novel research angle for related studies.

3. Methodology

3.1. Misallocation

This paper uses the resources misallocation provided by [19]. Assume there is a single final output Y produced in a perfectly competitive market using capital (K) and labor (L). The production function is given by a Cobb-Douglas production function:

$$Y_i=TF{P}_i{K}^{{\beta }_{Ki}}{L}^{{\beta }_{Li}}$$

(1)

Here, ${\beta }_{Ki}$ and ${\beta }_{Li}$ are the output elasticity of capital and labor respectively. Cost minimization implies:

$$C_i=\left(1+{\tau }_{Ki}\right)R{K}_i+\left(1+{\tau }_{Li}\right)w{L}_i$$

(2)

Here, ${\tau }_{Ki}$ and ${\tau }_{Li}$ refer to the distortions that affect the price of capital and labor respectively. R refers to the interest on capital, and w refers to the wages of labor. Profit maximization is given by:

$${max}_{K_{i}Li}\left\{P_i{Y}_i-w\left(1+{\tau }_{li}\right)L_i-\left(1+{\tau }_{ki}\right)R{K}_i\right\}$$

(3)

$P_i$ refers to the product price. The optimal first-order condition for Equation (3) yields:

$$\begin{array}{c}\hfill {\beta }_{Ki}{p}_i·TF{P}_i·K^{{\beta }_{Ki}-1}{L}_i^{{\beta }_{Li}}=\left(1+{\tau }_{Ki}\right)R\end{array}$$

(4a)

$$\begin{array}{c}\hfill {\beta }_{Li}{p}_i·TF{P}_i·K^{{\beta }_{Ki}}{L}_i^{{\beta }_{Li}-1}=\left(1+{\tau }_{Li}\right)w\end{array}$$

(4b)

The right side of equation Equations (4a) and (4b) is the marginal revenue product of factors.

For further analysis, two types of distortion can defined:

(a) Based on Equations (4a) and (4b), the absolute price distortion of factors i is defined as:

$${\gamma }_{Ki}={\displaystyle \frac{1}{1+{\tau }_{Ki}}},{\gamma }_{Li}={\displaystyle \frac{1}{1+{\tau }_{Li}}}$$

(5)

where ${\gamma }_{Ki}$ and ${\gamma }_{Li}$ are the actual market price of factors required by the marginal revenue product of the factors. ${\gamma }_{Ki}$ and ${\gamma }_{Li}$ stand for the degree of absolute price distortions. For example, if ${\gamma }_{Ki}=1$, there is no absolute price distortion of capital; if ${\gamma }_{Ki}<1$, there is a negative distortion; if ${\gamma }_{Ki}>1$, there is a positive distortion.

(b) If the production functions of N industries are aggregated, the total output value of the whole economy is recorded as Y, the total capital is K, the total labor is L, and the output value share of industry i in the whole economy is $s_i={\displaystyle \frac{p_i{Y}_i}{Y}}$. The contributions of factors in production weighted by output are ${\overline{\beta }}_K={\sum }_{i=1}^N{s}_i{\beta }_{Ki}$ and ${\overline{\beta }}_L={\sum }_{i=1}^N{s}_i{\beta }_{Li}$ The relative price distortion is defined as follow:

$${\widehat{\gamma }}_{Ki}={\displaystyle \frac{{\gamma }_{Ki}}{{\sum }_{j=1}^N\left(\frac{s_j{\beta }_{Kj}}{{\overline{\beta }}_K}\right){\gamma }_{Ki}}},{\widehat{\gamma }}_{Li}={\displaystyle \frac{{\gamma }_{Li}}{{\sum }_{j=1}^N\left(\frac{s_j{\beta }_{Lj}}{{\overline{\beta }}_L}\right){\gamma }_{Li}}}$$

(6)

The relative price distortion reflects the distortion of resource prices in industry i compared with the whole economy. For example, if ${\gamma }_{Ki}<1$, the capital price of industry i is lower than that of the whole economy; if ${\gamma }_{Ki}>1$, it shows that the capital use cost of industry i is higher than that of the whole economy. The absolute price distortion can not be measured, but the relative price distortion can be restored:

$${\widehat{\gamma }}_{Ki}=\left({\displaystyle \frac{K_i}{K}}\right)/\left({\displaystyle \frac{s_i{\beta }_{Ki}}{{\overline{\beta }}_K}}\right),{\widehat{\gamma }}_{Li}=\left({\displaystyle \frac{L_i}{L}}\right)/\left({\displaystyle \frac{s_i{\beta }_{Li}}{{\overline{\beta }}_L}}\right)$$

(7)

When ${\gamma }_{Ki}<1$, the capital tends to be overused in industry i. Conversely, if ${\gamma }_{Ki}>1$, the capital is insufficient. The invisible factor price distortion coefficient can be expressed in a visible way, and establish the relationship between resource distortions and resource misallocation.

3.2. PSM-DiD Method

The Diffdrence-in-Difference method (DID) [42], propensity score method (PSM) [43], synthetic control method (SCM) [44] and regression discontinuity [45] are widely used when evaluating the effects of a policy or an event. The DiD method performs well when evaluating treatment effects where there are multiple individuals in treatment group and control group [46]. Also, the common trend between treatment group and control group is required in the DiD model. However, the DiD method may become unreliable where unobserved factors and heterogeneity exist. In these cases, PSM are often combined with DiD method to deal with the time-invariant and unobserved factors and to meet the common trend. The SCM is often used in policy evaluation where there is one or fewer individuals in the treatment group with a long time period. However, it demands high-quality data and the availability of appropriate control units, and it is primarily applicable in contexts involving a single unit [47]. The regression discontinuity is especially robust when clear policy thresholds or intervention points are present, enabling more precise causal inference. Yet it requires well-defined breakpoints, which is not suitable for this case as these pilot areas are not established at the same time. Due to significant heterogeneity in the economy across firms and the potential violation of the common trend assumption, this paper chooses the PSM-DiD method to evaluate the impact of the ETS in China.

The basic idea of Difference-in-Difference (DID) method is to evaluate the effect of a policy by comparing the changes before and after the implementation of the very policy between the treatment group and the control group [42]. The DiD model is constructed as follows:

$$\begin{array}{c}DID=∆{\widehat{Y}}_{treatment}-∆{\widehat{Y}}_{control}\hfill \\ =(∆{\widehat{Y}}_{treatment,t_1}-(∆{\widehat{Y}}_{treatment,t_0})-(∆{\widehat{Y}}_{control,t_1}-∆{\widehat{Y}}_{control,t_0})\hfill \end{array}$$

(8)

where $DID$ is the estimation variable of Difference-in-Difference. $treatment$ is the treated group and $control$ is the control group. $t_1$ represents the period after the implementation of the policy, $t_0$ represents the period before the policy.

Due to significant heterogeneity across provinces in China, this paper introduces the propensity score matching method (PSM) to eliminate the sample selection bias [43]. PSM method could create a control group which has the similar characteristics to the experimental group using propensity score, and can effectively solve some endogenous problems [48]. The propensity score matching method in this paper is presented as follows:

$$\begin{array}{c}\hfill {\tau }_{Kit}^{psm}={\beta }_0+{\beta }_{1}Tim{e}_i+{\beta }_{2}Trea{t}_{it}+{\beta }_{3}Tim{e}_i\times Trea{t}_{it}+{\beta }_4{X}_{it}+{\xi }_{it}\\ \hfill {\tau }_{Lit}^{psm}={\beta }_0+{\beta }_{1}Tim{e}_i+{\beta }_{2}Trea{t}_{it}+{\beta }_{3}Tim{e}_i\times Trea{t}_{it}+{\beta }_4{X}_{it}+{\xi }_{it}\end{array}$$

(9)

where i denotes provinces/cities, t denotes the years. ${\tau }_{Kit}^{psm}$ and ${\tau }_{Lit}^{psm}$ are the indicators of capital and labor misallocation respectively. $Tim{e}_i$ is the time dummy variable. $Tim{e}_i=0$ before the year 2013 and $Tim{e}_i=1$ after the year 2013. $Trea{t}_{it}$ is the dummy variable for treated group and control group. $Trea{t}_{it}=1$ for Beijing, Shanghai, Tianjin, Guangdong, Fujian, Hubei, and $Trea{t}_{it}=0$ for other regions. $X_{it}$ is a set of covariates that affects resource misallocation. ${\xi }_{it}$ is residual term. To be clearly, if ${\beta }_3$ (coefficient of $Tim{e}_i\times Trea{t}_{it}$) is positive, the ETS did not improve the resources misallocation. The ETS improved the resources misallocation only when the coefficient of $Tim{e}_i\times Trea{t}_{it}$ is significantly negative.

3.3. Placebo Test

The placebo tests are necessary for policy effects evaluation. The “false treated time or treated group” are generated in placebo test. If the results in placebo test is still significant, the policy effects obtained in the benchmark will be rejected. Otherwise the policy effects is reliable. In this paper, two different placebo tests are conducted: (a) Assuming that the implementation time of ETS is 1 or 2 years ahead of schedule, a false policy implementation time $Tim{e}_{false}$ is constructed and used test whether the implementation of ETS ($Treat\times Tim{e}_{false}$) will affect the resource misallocation. If there are other factors that significantly affect the resources misallocation between the control group and the treatment group, the coefficient will be significant. (b)Assigning the treatment to the provinces/cities which was not in the treatment group, a false treatment $Trea{t}_{false}$ is constructed and used to test whether $Trea{t}_{false}\times Time$ will significantly affect the resource misallocation. The coefficients will not be significant if there is no significant missing independent variables.

3.4. Post-Treatment Effect

Another issue discussed in this paper is whether the impact of the pilot ETS on the resources misallocations grows or fades by time. To answer that question, a post-treatment effect test is introduced with lag dummy variables:

$${\tau }_{it}={\upsilon }_0+{\upsilon }_{1}Trea{t}_{it}+{\theta }_j\sum _{j=1}^{2}Tim{e}_{-j}+{\upsilon }_{3}Trea{t}_{it}\times \sum _{j=1}^{2}Tim{e}_{-j}+{\beta }_4{X}_{it}+{\xi }_{it}$$

(10)

where $Tim{e}_{-j}$ is the time dummy variables. $Tim{e}_{-1}=1$ for the year 2014; $Tim{e}_{-1}=0$ for other years. $Tim{e}_{-2}=1$ for the year 2015; $Tim{e}_{-2}=0$ for other years. ${\xi }_{it}$ is residual term.

4. Data

The dependent variable of this paper is capital misallocation ${\tau }_{Ki}$ and labor misallocation ${\tau }_{Li}$ calculated with the provincial level panel data from 2008 to 2019. Data of capital, labor, and GDP in each province are collected from China National Bureau of Statistics. As there is no official data on capital stock at provincial level, follow [49,50], this paper uses the Perpetual Inventory Method (PIM) to obtain capital stock (K) in each province. Data were processed using Stata 16.

Follow [35,51], the covariates used in PSM are innovation ($innovation$, the number of patent granted), ratio of the secondary industry ($ind2$) and service industry ($ind3$), foreign direct investment ($FDI$), clean energy in electricity generation ($CE$) and real per capita GDP ($RPCG$). All covariates are collected from China National Bureau of Statistics.

Table 1. Descriptive statistics of covariates in the PSM-DiD model.

Variable	Mean	Std.	Min	Max
$ln\left(innovation\right)$	9.38	1.68	4.53	12.71
$ind2$	0.43	0.08	0.17	0.62
$ind3$	0.46	0.09	0.30	0.83
$ln\left(FDI\right)$	3.82	1.57	0.24	7.25
$ln\left(CE\right)$	4.34	2.02	−3.51	8.02
$ln\left(RPCG\right)$	0.28	0.40	−0.47	1.51

N = 310 for all the variables, “Mean”, “Std.”’, “Min”, and “Max” refer to mean value, standard deviation, minimum and maximum respectively.

lists the descriptive statistics of the covariates. The stochastic frontier method is employed to estimate the coefficients ${\beta }_{Ki}$ and ${\beta }_{Li}$ in Equation (1), which are used to calculate ${\widehat{\gamma }}_{Ki}$ and ${\widehat{\gamma }}_{Li}$ in Equation (7) and $ta{u}_{Ki}$ and $ta{u}_{Ki}$ in Equation (5) (see

Table 2. Estimation results of the relative price distortion.

Variable	Obs	Mean	Std.	Min	Max
${\tau }_k$	310	−0.10	0.79	−0.97	5.57
${\tau }_{k,T}$	50	0.91	0.35	0.46	1.51
${\tau }_{k,C}$	260	−0.29	0.02	−0.33	−0.28
${\tau }_l$	310	0.21	1.00	−0.96	4.18
${\tau }_{l,T}$	50	−0.29	0.03	−0.34	−0.27
${\tau }_{l,C}$	260	0.30	0.08	0.19	0.39

Note: This table presents the summary of statistics of variables. “Obs”, “Mean”, “Std.”, “Min”, and “Max” symbolize Observation, Mean, Std. Dev, Maximum and Minimum respectively.${\tau }_k$, ${\tau }_{k,T}$ and ${\tau }_{k,C}$ are the capital misallocation for the whole sample, treatment group and control group respectively. ${\tau }_l$, ${\tau }_{l,T}$, ${\tau }_{l,T}$ and ${\tau }_{l,C}$ are the labor misallocation for the whole sample, treatment group and control group respectively.

and Figure 1).

The agency of “command-and-control”. This paper equates “command-and-control” with administrative intervention. Therefore, indicators reflecting the intensity of administrative intervention can also be used as indicators of command-and-control. As most of the emitter regulated in the ETS pilots and most of the participants in the carbon market are “state-owned” enterprices [15]. Compared with “non-state-owned” enterprises, the local governments have stronger control ability over state-owned enterprises [13]. Although there is no official document, based on interviews with local pilot ETS regulators and experts, the performance evaluation of state-owned enterprise executives in the pilot areas does take into account [52]. So this paper uses the number of state-owned enterprises ($state-owned$) as the agency of “command-and-control” administrative intervention. The descriptive statistics of $state-owned$ are listed in Table 1. FDI, clean energy in electricity generation, real per capita GDP and the number of state-owned enterprises are taken the natural logarithm to eliminate the possible heteroscedasticity in the estimation.

5. Empirical Results

5.1. Resources Misallocation

Based on the framework proposed by [19], this paper calculates capital and labor misallocation using provincial level data from 2008 to 2017. Table 2 lists the estimation results of the resources misallocation. Figure 1 shows the trends of ${\tau }_{Ki}$ and ${\tau }_{Li}$ between the treatment group and the control group. In Table 2, the mean value of ${\tau }_k$ for treatment group(${\tau }_{k,T}=0.91$) is larger than the control group(${\tau }_{k,T}=-0.29$), and the difference between the two has been continuously expanding since 2012 (Figure 1). This result indicates that the relative price of capital in ETS areas is overvalued and the shortage of capital exists, while the opposite occurs in the control group. On the contrary, the mean value of ${\tau }_k$ for treatment group (${\tau }_{k,T}=-0.29$) is smaller than the control group (${\tau }_{k,T}=0.30$) in Table 2, which indicating that the relative price of labor in ETS areas is undervalued.

The possible reasons for the difference of ${\tau }_k$ between treatment group and control group may comes from the “Race-to-the-Bottom” for local government in environmental regulation [53] and local government competition under “promotion tournament model” [17,54]. Under the “promotion tournament model”, regions with low productivity and competitiveness are more likely to use land and tax policies to attract and retain investment and industry. After the implementation of ETS, regions without ETS also tend to adopt lax environmental standards. Capital flows may flee to regions with lax environmental regulations, which aggravates the capital misallocation. These results may also reflect the fact that the ETS pilot areas are more industrialized and their industrial outputs are higher than other areas on average, which is important for understanding the impact of the ETS on resources misallocation.

5.2. Propensity Score Matching (PSM)

The propensity score Matching (PSM) is constructed to eliminate the selection bias and the endogeneity. The effectiveness of PSM depends on whether the common trend assumption is fulfilled, that is to say, there is no significant difference in covariates between the treatment group and the control group.

Table 3. Characteristics of covariates before and after matching.

	Unmatched	Mean		t-Test		V(T)/V(C)
Variable	Matched	Treated	Control	t	$\mathit{p}>\left|\mathit{t}\right|$
$ln\left(innovation\right)$	U	10.53	9.17	4.66	0.00	0.40
	M	10.24	10.17	0.19	0.85	0.32
$ind2$	U	0.45	0.43	1.59	0.11	0.14
	M	0.45	0.45	−0.04	0.97	0.09
$ind3$	U	0.48	0.45	1.56	0.12	0.37
	M	0.48	0.48	0.08	0.94	0.25
$ln\left(FDI\right)$	U	4.93	3.62	4.83	0.00	0.32
	M	4.57	4.57	0.01	0.99	0.17
$ln\left(CE\right)$	U	4.72	4.28	1.23	0.22	3.27
	M	4.40	4.40	−0.01	1.00	3.76
$ln\left(RCPG\right)$	U	0.45	0.21	3.85	0.00	1.40
	M	0.31	0.32	−0.12	0.91	0.48

Note: the statistics in line with “U” and “M” in column 2 (Unmatched/Matched) represent the results of covariates before and after matching respectively. “Treated” and “Control” in column 3 and 4 represent the mean value of treatment group and control group respectively.

shows the t-test of the covariates before and aftermatching. Figure 2 describes the standardized bias across covariates (left in Figure 2), kernel density before and matching (middle and right in Figure 2).

From the results of t-test in Table 3, it is clear that all covariates between the treatment group and control group are significantly different before matching, while there is no significant difference after matching. The standardized bias across covariates of Figure 2 show there is no significant bias across covariates between the two groups. The results of kernel density in Figure 2 show there are more common trend after matching. All these results show that the possible endogeneity and selection bias can be effectively solved.

5.3. Policy Effect

Based on the results of PSM test, this paper uses PSM-DiD method to investigate the impact of the ETS on resources misallocation.

Table 4. Estimation results of the PSM-DiD.

	Model (1)		Model (2)		Model (3)
Variables	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$
$Treat\times Time$	1.114 ***	−0.126 **	0.087 **	−0.057 *	0.082 **	−0.065 *
	(9.40)	(−2.09)	(2.31)	(−1.79)	(2.39)	(−1.73)
$Treat$	0.791 ***	−0.716	0.459 ***	−0.368 **	0.379 ***	−0.561 *
	(4.60)	(−1.23)	(3.17)	(−2.13)	(2.62)	(−1.93)
$Time$	−0.327	0.156	−0.044 *	0.115 *	−0.044 ***	0.113 **
	(−1.14)	(1.08)	(−1.79)	(1.78)	(−2.86)	(2.19)
$ln\left(innovation\right)$			−0.017 ***	0.005 ***	−0.010 ***	0.016 **
			(−3.32)	(3.13)	(−4.73)	(2.40)
$ind2$			0.871 ***	−2.064 ***	0.155	−3.232 ***
			(3.37)	(−2.96)	(0.80)	(−4.28)
$ind3$			0.508	−3.144 ***	−0.047	−3.918 ***
			(1.54)	(−3.53)	(−1.15)	(−4.32)
$ln\left(FDI\right)$			0.018 ***	−0.025 ***	0.020 ***	−0.023 **
			(2.96)	(−3.54)	(3.47)	(−2.35)
$ln\left(CE\right)$			0.050 ***	−0.004 **	0.045 ***	−0.013 ***
			(5.11)	(−2.13)	(5.00)	(−3.47)
$ln\left(RCPG\right)$			0.031	−0.150 ***	0.043	−0.095 **
			(1.47)	(−2.66)	(1.46)	(−2.32)
$ln(state$-$owned)$					−0.114 ***	−0.116 ***
					(−4.04)	(−3.39)
i.year	NO	NO	YES	YES	YES	YES
Constant	−0.299 ***	1.192	−1.047 ***	2.557 ***	−0.972 ***	2.052 **
	(−4.09)	(0.90)	(−3.83)	(3.35)	(−3.31)	(2.23)
Observations	281	281	281	281	281	281
Number of id	30	30	30	30	30	30

Note: t statistics in parentheses; *** p < 0.01, ** p < 0.05, * p < 0.1.

reports the results of policy effect. In Table 4, ${\tau }_k$ and ${\tau }_l$ are capital and labor misallocation respectively. Model (1) does not include the covariates; Model (2) includes all the covariates and Model (3) includes all the covariates and control variable “state-owned”, which is the proxy variable for the “command-and-control” instrument.

Most of the coefficients of $Treat\times Time$ for Model (1)–(3) are significant at 5% level. Specifically, the coefficients of $Treat\times Time$ for ${\tau }_k$ are significantly positive, which indicates that the ETS tends to worsen the capital misallocation.

On the other hand, the coefficients of $Treat\times Time$ for ${\tau }_l$ are significantly negative, which indicates that the establishment of China’s ETS tends to alleviate the labor misallocation. The ETS regulation may promote the upgrading of the industrial structure in ETS pilots areas [14], which also affects the flow of migrant workers. Environmental regulations also have significantly improve the job quality, which driven by the labor productivity and the positive adjustment of employment structure [55].

Further more, the coefficients of $ln(state-owned)$ are significantly negative at 1% level (Model (3)) for both capital misallocation and labor misallocation, indicating that $ln(state-owned)$ improved capital and labor misallocation significantly. As $ln(state-owned)$ is the proxy for “command-and-control” instrument, this result also indicates that the ETS in China worked effectively as a “command-and-control” instrument. It remains uncertain whether the ETS in China worked mainly a market-oriented instrument, a command-control instrument or both without confliction.

In general, China’s ETS, while improving labor resource misallocation, simultaneously exacerbates capital resource misallocation, highlighting the imperfect cost-effectiveness of China’s ETS. The deterioration of capital misallocation under China’s ETS can be seen as a manifestation of the “pollution haven” hypothesis, albeit with distinct Chinese characteristics. The“race-to-the-bottom” environmental regulation [53] and local government competition under “promotion tournament model” [54] may be responsible for the high capital misallocation in the ETS pilot area. Similar perspectives supporting the conclusions of this study can also be found in other related literature. It is found that the carbon policy may go against intensive firms and raise the capital cost [56]. Also, researchers found there is an inverted-U-shaped relationship between the ETS regulation and capital cost for non-state-owned enterprises in China, while state-owned enterprises showed linear relationship that ETS regulation lowers the capital cost [41]. The results in Table 4 indicate that the ETS in China may aggravate the capital misallocation, which is consistent with these conclusions mentioned above. The empirical results for “state-owned” and “non-state-owned” firms suggest that China’s ETS functions, to some extent, as a “command-and-control” tool. This might explain why the ETS exacerbates capital misallocation. On the other hand, China’s ETS significantly improves labor resource misallocation, demonstrating its cost-effectiveness and contribution to sustainable development. The improvement in labor resource misallocation and the exacerbation of capital misallocation may stem from the higher mobility of labor compared to capital in China. However, this remains a theoretical assumption that requires solid empirical validation.

5.4. Robust Test

5.4.1. Placebo Test

Table 5. Placebo Test.

	Model (4)		Model (5)		Model (6)
Variables	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$
$Treat\times Tim{e}_{false-1}$			0.075	−0.085
	(2.16)	(−0.93)
$Treat\times Tim{e}_{false-2}$			0.073	−0.111
			(2.01)	(−1.16)
$Trea{t}_{false}\times Time$					−0.043	−0.02
					(−0.99)	(−1.39)
$Time$					−0.684	0.209 *
					(−1.52)	(1.85)
$Time1$	−0.014	0.268 ***
	(−0.38)	(2.65)
$Time2$			−0.010	0.264
			(−0.27)	(2.64)
$Treat$	0.464 ***	−0.344	0.464 ***	−0.321
	(3.21)	(−0.59)	(3.20)	(−0.55)
$Trea{t}_{false}$					−0.043	−0.023
					(−0.99)	(−1.39)
$Covariates$	YES	YES	YES	YES	YES	YES
$i.year$	YES	YES	YES	YES	YES	YES
Observations	267	267	267	267	267	267
Number of id	29	29	29	29	267	267

Note: t statistics in parentheses; *** p < 0.01, * p < 0.1.

lists the results of placebo test. The difference-in-difference interaction variables in Model (4) and Model (5) are generated with forged year dummy variables. $Tim{e}_{false-1}$ equals 1 for the year 2012, and 0 for other year in Model (4). $Tim{e}_{false-2}$ equals 1 for the year 2011, and 0 for others year Model (5). The difference-in-difference interaction variable in Model (6) is generated with forged treatment group. $Trea{t}_{false}$ equals 1 for the province randomly selected from control group, and 0 for others.

In Table 5, all the six coefficients of difference-in-difference interaction variables are not significant and t-statistics are far less than critical value. Specially, the results in Model (4) and Model (5) mean that the resources misallocation for the treatment group are not significantly different from control group between 2011 and 2012. The results in Model (6) indicate that the resources misallocation in the forged treatment group are not significantly different from control group. The results of placebo test demonstrates the previous arguments is reliable and the ETS did affect the resources misallocation in these pilot areas.

5.4.2. Winsorized Test

Fat-tailed distributions are found in both ${\tau }_{Ki}$ and ${\tau }_{Li}$. To find out whether the policy effects are driven by particularly influential outliers, this paper constructs a winsorized test and re-evaluate the policy effects using PSM-DiD method. The estimation results are reported in

Table 6. Robust Test of the PSM-DiD model (Winsorized).

	Model (7)		Model (8)
Variables	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$	${\mathit{\tau }}_{\mathit{k}}$	${\mathit{\tau }}_{\mathbf{l}}$
$Treat\times Time$	0.092 ***	−0.097 **	0.047 **	−0.097 **
	(10.26)	(−2.14)	(2.03)	(−2.14)
$Treat$	0.810 ***	−0.683	0.332 ***	−0.683
	(4.95)	(−1.36)	(2.64)	(−1.36)
$Time$	−0.296	0.098	−0.023	0.098
	(−1.37)	(0.90)	(−0.41)	(0.90)
$Covariates$	YES	YES	YES	YES
$i.year$	YES	YES	YES	YES
Observations	281	281	276	276
Number of id	30	30	30	30

Note: t statistics in parentheses; *** p < 0.01, ** p < 0.05,

.

The samples are winsorized by 1% in Model (7) and 5% in Model (8). The coefficient signs are consistent with the benchmark regression results. The values of ${\tau }_k$ change from 0.092 (winsorized by 1%) to 0.047 (winsorized by 5%) indicating that there is a visible decrease of capital misallocation associated with the outliers. It is found that the influential outliers in these cases come from Shanghai. The coefficient values of ${\tau }_l$ stay the almost the same. The results of winsorized test indicate that there are influential outliers when detecting the policy effect on capital misallocation.

5.5. Post-Treatment Effect

A post-treatment test is conducted to further investigate the trend of ETS policy effects over time. The results are shown in

Table 7. Post-treatment Test.

Variables	${\mathit{\tau }}_{\mathit{k}}$		${\mathit{\tau }}_{\mathit{l}}$
$Treat\times Tim{e}_{-j}$	0.037 **	0.127 **	−0.047 **	−0.046 **
	(2.63)	(2.52)	(−2.63)	(−2.62)
$Tim{e}_{-1}$	−0.002		0.058 **
	(−0.38)		(2.10)
$Tim{e}_{-2}$		−0.005		−0.001
		(−0.27)		(−0.02)
$Treat$	0.691 ***	0.680 ***	−0.279 ***	−0.256 ***
	(4.66)	(4.59)	(−3.55)	(−3.50)
$Covariates$	YES	YES	YES	YES
$ln(state-owned)$	−0.064 **	−0.062 **	0.015	0.008
	(−2.44)	(−2.40)	(0.31)	(0.17)
Constant	−0.470	−0.486	0.878	0.922
	(−1.16)	(−1.20)	(1.22)	(1.27)
Observations	296	296	296	296
Number of id	30	30	30	30

Note: t statistics in parentheses; *** p < 0.01, ** p < 0.05,

. The coefficients of $Treat\times Tim{e}_{-j}$ for both ${\tau }_k$ and ${\tau }_l$ are significant are 5% level. The sign of these coefficients is consistent with benchmark results in Table 4. These results indicate that the policy effects obtained in are stable. The possible reasons for these results have been explained, that is “race-to-the-bottom” environmental competition, local government competition under “promotion tournament model” and possibly increased cost in heavy industry for non-state-owned enterprises.

However, the impact of the ETS on capital misallocation (${\tau }_k$) is gradually enhancing, from 0.037 to 0.127. On the other hand, the impact of labor misallocation (${\tau }_l$) remains almost unchanged. These results indicate that the establishment of ETS has constantly improved labor misallocation in the pilot area, while the policy effect for capital misallocation is aggravated.

6. Discussion

There are literature theoretically proved merit of the ETS as a cost-effective market-oriented instrument in reducing carbon emission. However, the ETS can not be as perfect as theory in practice. To study this issue, this paper to empirically test whether China’ s ETS improved the resources misallocation using PSM-DiD method.

The results show that: firstly, the capital misallocation in ETS is higher than that in other regions, while the labor misallocation is lower than that in other regions. Secondly, the ETS did not improve the capital misallocation, but significantly improved the labor misallocation, indicating that the ETS are not perfectly cost-effectively. The results of placebo test and winsorized test show that the policy effects are stable and reliable. The post-treatment test shows the aggravation of capital misallocation caused by the ETS fade with time, while the improvement of labor misallocation stay unchanged. Finally, “command-and-control” instruments have significantly improved both capital and labor misallocation, indicating that the local government’s supplementary policies have a significant effect on China’s carbon emission reduction policies.

The findings of this study highlight the nuanced impact of China’s Emissions Trading Scheme (ETS) on resource allocation, revealing both improvements and inefficiencies. While the ETS has successfully mitigated labor misallocation, it has simultaneously exacerbated capital misallocation. These results may have complex and dynamic implications for sustainable development.

7. Managerial Implications

This study’s findings provide valuable managerial insights for enterprises navigating the complexities of China’s Emissions Trading Scheme (ETS). Given the dual characteristics of the ETS as both a market-oriented mechanism and a “command-and-control” instrument, enterprises—particularly state-owned enterprises—should carefully balance compliance strategies with market-driven innovative initiatives. Effective managerial practices should integrate proactive emissions reduction measures, such as investments in green technologies and optimization of resource allocation. Firms must also closely monitor policy changes and “command-and-control” interventions, adapting their operational and strategic planning accordingly to prevent exacerbating capital misallocation and to capitalize on labor allocation efficiencies. Ultimately, aligning managerial practices with these dual characteristics of ETS can significantly enhance corporate sustainability and competitive advantage amid evolving policy landscapes. Efficient resource allocation is a fundamental component of sustainable development, ensuring that capital and labor are optimally utilized to achieve both economic and environmental objectives. The study’s findings indicate that China’s ETS has improved labor allocation but worsened capital misallocation, which suggests an uneven impact on overall resource efficiency. The observed improvement in labor allocation suggests that the ETS has incentivized the reallocation of labor towards more productive and environmentally efficient sectors. This aligns with sustainable development goals, as optimized labor allocation enhances productivity while reducing emissions. However, the worsening capital misallocation presents a significant concern for long-term sustainability. This inefficiency may stem from structural factors such as state-owned enterprises’ dominance, financial market distortions, and regional economic disparities. Without addressing these capital misallocation issues, the full potential of ETS as a market-based instrument for sustainability may remain constrained.

8. Conclusions

This study empirically examines whether China’s ETS improved resource misallocation using the PSM-DiD method. Results indicate significant improvements in labor allocation but worsening capital misallocation, suggesting an uneven impact on resource efficiency. Further analysis demonstrates that administrative interventions through “command-and-control” instruments significantly influence the effectiveness of ETS policies. Policymakers should carefully integrate market efficiency and administrative intervention to enhance sustainable economic development and emissions reduction.

9. Limitation and Future Research Recommendations

The data used in this study is from before 2020, which is admittedly somewhat regrettable. However, there are reasons for this choice. First, since 2020, China’s carbon reduction policies have been implemented nationwide, altering the fundamental assumptions of the PSM method. The second and more critical reason is the impact of the COVID-19 pandemic. From 2020 onward, China faced severe disruptions due to the COVID-19 pandemic, leading to strict social control measures that lasted until the end of 2022. These measures significantly changed the behavior of most enterprises. On one hand, high-carbon industries essential for maintaining societal operations, such as coal-fired power generation, were relatively less affected. On the other hand, the operations of most enterprises were disrupted, leading to substantial changes in their emissions. This shift also affected the data foundation of this study.

In fact, the COVID-19 crisis and subsequent economic recovery measures have reshaped industrial behaviors and emission patterns, necessitating further investigation into the resilience of ETS mechanisms under external shocks. Future research should incorporate post-2020 datasets to evaluate the China ETS’s performance amid evolving policy frameworks. Additionally, longitudinal analyses comparing pilot and national phases could disentangle the long-term effects of market maturation versus administrative adaptations. In these cases, the findings of this study provide important empirical and theoretical support for future research.