Abstract
Climate change has substantially impacted the environment and human society. Carbon dioxide (CO2) emissions are one of the primary causes of the climate crisis, and fossil fuel consumption for numerous economic activities has been a major factor leading to those emissions. On the one hand, the “treadmill of production” theory posits that economic growth inevitably leads to environmental damage. On the other hand, the “ecological modernization” theory argues that the association between the two will decouple over time. To evaluate these competing perspectives, this study investigates panel data from 30 provinces in China over a 25-year period, from 1997 to 2021. The country's CO2 emissions almost quadrupled during this period, currently accounting for one-third of global emissions, and its gross domestic product (GDP) increased by around 15 times. The Prais–Winsten regression results show that GDP is positively and significantly associated with increased CO2 emissions. However, the magnitude of the association varies, intensifying from 1997 to 2002 and then decoupling for the subsequent two decades, which supports the ecological modernization argument. The findings advance the environmental sociology literature and carry policy implications for China's sustainable development. The expanded deployment of renewable energy, more energy-efficient manufacturing, and the promotion of electric vehicles could fuel the transition toward a low-carbon economy.
Introduction
Climate change has led to unpredictable and sometimes irreversible changes to weather patterns and poses challenges to human society (IPCC, 2023). In 2024, the average temperature reached a level unmatched for over 100,000 years, making it the hottest year in human history (NOAA, 2025). Global surface temperature surpassed 1.5°C above preindustrial levels for the first time, marking a failure to avoid crossing the critical threshold set by the Paris Agreement. Climate change has already resulted in substantial impacts on the economy and social well-being (Carleton and Hsiang, 2016; Nordhaus, 2013). The climate crisis is mainly driven by carbon dioxide (CO2) emissions from fossil fuel combustion, primarily for economic activities such as manufacturing, transportation, and power generation. As a result of its rapid economic growth, China's emissions have surged dramatically and currently account for around one-third of global emissions (Guan et al., 2021). Against this backdrop, rapid and deep reductions in emissions in order to arrive at net-zero carbon emissions by 2050 are essential to realizing the Paris Agreement goal (IPCC, 2023).
Two foundational arguments of the environmental sociology scholarship, the “treadmill of production” (TOP) theory and the “ecological modernization” (EM) theory, are at loggerheads over the economy's impact on climate change (Bohr and Dunlap, 2018; Fisher and Jorgenson, 2019; Rudel et al., 2011). On the one hand, the TOP theory posits that economic growth leads to environmental damage because expansion requires continuous extraction of resources and unavoidably generates different types of pollution. On the other hand, the EM theory argues that while the expanded economy can cause environmental issues, the initially strong association between the two in the early stage of development will eventually become decoupled due to technological advancement, especially when modernization reaches a certain level.
Previous cross-national research has tested these arguments using country-level indicators (Dietz and Rosa, 1997; Jorgenson and Clark, 2012). Despite these contributions, the corpus of macro-level analyses should be complemented by micro-level, subnational studies. This is because significant variations occur within countries, and universal national measures could mask those differences. Therefore, this line of research requires assessment to reveal those variations, especially for countries like China, which has a vast territory and massive domestic differences. Spatially, the coastal areas exhibit greater economic size and higher carbon emissions than inland regions. These asymmetrical patterns can only be captured through provincial-level metrics that account for regional nuances and variations. In contrast, the national metrics will merely reflect the general trends in China but overlook the unique characteristics of individual provinces and obscure the underlying drivers of carbon emissions. Longitudinally, trends in economic growth and rising emissions levels differ for the 2000s, 2010s, and 2020s. Also, China's substantial contribution to global CO2 emissions and gross domestic product (GDP) underscores the importance of studying this country in greater depth, with implications for other countries, especially those in the Global South with similar backgrounds.
Using data from 30 provinces in China, this study aims to bridge the gaps and examine the effect of economic growth on the environment and the magnitude of this effect. Panel data from 1997 to 2021 are used to examine the changing size of the economy and the unfolding climate consequences. Fixed-effects regression will estimate how the time-varying predictor (GDP) impacts the climate crisis (measured by CO2 emissions). The interactions between GDP and the year included in the model will inform whether this GDP–CO2 relationship has intensified or decoupled. An intensified pattern means that the magnitude of GDP's effect on carbon emissions has amplified over time, consistent with the predictions of the TOP theory. Conversely, a decoupled relationship refers to the effect's magnitude having weakened, aligning with the EM theory. The findings will help evaluate the competing theoretical arguments and thus carry policy consequences. The former scenario suggests China's economy is becoming more carbon-intensive, necessitating extensive government intervention to shift the growth model. The latter scenario signifies the effectiveness of climate policies, demonstrating that the country is progressing toward sustainable development.
The subsequent section clarifies the theoretical mechanisms and presents the hypotheses. This is followed by an introduction to the data and measurements. Next, the modeling approach is detailed, and the results are outlined. The discussion and conclusion sections will summarize and interpret the findings, emphasize contributions to the literature and policy implications, and suggest directions for future research.
Literature review
Existing literature in environmental sociology and sister disciplines has contributed to the understanding of anthropogenic activities that lead to carbon emissions (Davidson, 2022; Jorgenson et al., 2019; Rosa et al., 2015; Rosa and Dietz, 2012). Among various factors, economic growth has been identified as the primary driving force of climate change (Jorgenson et al., 2025). This section will begin by examining the theoretical frameworks of the TOP and EM theories regarding the economy–environment interaction, followed by empirical analyses using national data and the research hypotheses that this study seeks to evaluate.
Theoretical framework
The TOP theory is derived from the Marxist tradition and contends that unfettered economic expansion is antithetical to biophysical viability. The treadmill is a self-reinforcing machine fueled by endless capital accumulation. Instead of deploying profits to innovation or social welfare, the increased margins are reinvested to keep the treadmill moving even faster. The scale and intensity of manufacturing are expanding while causing unending environmental pollution (Schnaiberg, 1980; Schnaiberg and Gould, 1994). Ecological problems are inevitable because rapid economic growth requires continually withdrawing resources and burning fossil fuels at an accelerating pace, and such energy-intensive activities generate waste that often exceeds natural limits (Clark and York, 2005; Foster, 1999). As a result, the infinite demand for finite natural resources makes the ever-growing economy unsustainable and will lead to escalating environmental degradation (Gould et al., 2004, 2008). From an economic standpoint, the environmental impact of economic production is classified as an externality. When these externalities are not properly considered, market inefficiencies emerge. The failure of the market to accurately measure and regulate the actual cost of production will lead to greater environmental harm (Dées, 2020).
The macroeconomic perspective argues that economic growth is essential for national progress. Consequently, unchecked and continuous growth is often associated with pollution and various forms of environmental degradation. However, traditional analyses of the theory tend to narrowly focus on national contexts, treating each country as an isolated unit of analysis when examining the relationship between economic activity and environmental impact (e.g. Knight and Schor, 2014; Thombs, 2018). Thus, it is critical to acknowledge that the theory overlooks the impact of an increasingly interconnected global economy, where significant shifts occur in production and consumption patterns through international trade. The details of the limitations are further presented in the discussion section.
The TOP theory aligns with the “Jevons paradox” (Jevons, 1866), which suggests that while improvements in production efficiency can theoretically reduce the environmental impact by lowering material input without affecting output, they often lead to an overall expansion in production. As a result, greater efficiency paradoxically accelerates resource depletion and waste generation (Clement, 2011). The economic concept known as the “rebound effect” originates from the Jevons paradox, which highlights substantial rebounds in energy consumption connected with rising energy efficiency. In this case, enhanced efficiency does not reduce fossil fuel use but instead enables expansion in further production and consumption (York et al., 2022). Lange et al. (2021) proposed a coherent typology for the rebound effect, distinguishing between the mechanism—how changes in energy consumption occur—and the effect, which refers to the magnitude of those changes. Their framework allows analysis of both aspects across four economic levels (micro, meso, macro, and global) and over two timeframes (short term and long term). Empirical research has observed the rebound effect in various settings. For instance, one cross-national study covering 131 countries from 1965 to 2010 found that more efficient nations often experience faster growth in energy use and CO₂ emissions (York and McGee, 2016). Similarly, when analyzing state-level data for the United States, another study revealed that energy efficiency policies did not correlate with energy conservation (Adua et al., 2021). These findings suggest that technological advancements and efficiency gains may not yield the expected decreases in consumption, and their actual effects should instead be examined within broader economic, political, and social frameworks.
Different from the claims of the TOP theory, the EM theory does not perceive environmental pollution and the emission of greenhouse gases as inherent characteristics of economic growth. In contrast, the theory follows the neoclassical economic tradition and suggests that continued growth promotes technological advancement and environmental governance. Thus, over time, modernization will improve the environment and alleviate the stress from the economy (Brock and Taylor, 2005; Mol, 2003; Mol and Spaargaren, 1993; Spaargaren and Mol, 1992).
The argument can be represented as the environmental Kuznets curve, an inverted U-shape relationship between economic growth and environmental harm (Dinda, 2004; Grossman and Krueger, 1995). The modern economy is self-referential, and the impact (e.g. CO2 emissions) might worsen in the early stages of development when economic rationality is adopted as the sole measure of progress. Further modernization will help address these problems and promote sustainability when ecological rationality becomes the core principle and environmental damage is not considered a minor consequence. Manufacturers and giant corporations engage in modernization and integrate environmental concerns into their operational decisions through program requirements or independent initiatives (Mol, 1995).
Despite the classic view on the inverted U-shape association, alternative perspectives suggest that the relationship between economic growth and environmental pollution is nonlinear and fluctuating, resembling a W-shaped pattern. Specifically, environmental damage tends to increase with initial economic growth and decrease later because of technological improvements, environmental regulations, and structural shifts, all related to modernization. Next, pollution could rise again due to the rebound effects of increased consumption and the outsourcing of pollution through international trade (Acheampong and Opoku, 2023; Karsch, 2019; Wang et al., 2024). In other words, a variety of economic, political, and global factors interact in complex ways, causing the environmental impact to rise and fall repeatedly, rather than trending in a smooth curve.
Empirical analyses
The TOP theory adopts a critical perspective, focusing on the damage caused by unhindered expansion, whereas the EM theory is derived from an optimistic perspective, highlighting the compatibility between the economy and the environment (Lidskog et al., 2015). Some studies support the TOP argument by presenting that countries experiencing speedy economic growth produce more carbon emissions over the course of development. For example, a survey of 106 countries from 1990 to 2016 found that the effect of economic growth on carbon pollution was significant (Jorgenson et al., 2023). Knight and Schor (2014) examined a balanced dataset of 29 affluent countries and revealed economic growth's positive effect on both territorial and consumption-based emissions.
In comparison, other studies reveal findings that align with the EM theory and identify a positive but decoupled association between economic growth and carbon emissions. For instance, a survey of 86 countries from 1960 to 2005 suggests a relative decoupling between total carbon emissions and economic development (Jorgenson and Clark, 2012). Another study of 70 countries also shows that, while the effect of GDP per capita on total CO2 emissions remained constant from 1965 to 2010, a long-term decoupling trend was observed in developed countries (Thombs, 2018). Similarly, a mini cycle of decoupling was identified in a study of a group of Global North nations from 1870 to 2014 (Thombs and Huang, 2019). In a study of 34 high-income countries, Huang (2024) examined four emissions components to understand the multidimensionality of the economy–emissions relationship. The findings show that affluence is strongly coupled with CO2 emissions related to trade activities, such as those embodied in imports and exports, which offers partial support for the TOP theory. Meanwhile, national affluence is decoupled from emissions generated by domestic-oriented supply chain activities and direct emissions of end-user activities, aligning with some perspectives from the EM theory.
Previous empirical analyses predominantly adopted a cross-national design and estimate models using country-level measures to evaluate the contrasting arguments between TOP and EM theories. This is largely because organizations like the World Bank compile historical records of CO₂ emissions, GDP, population, trade, and other relevant indicators for nearly all countries dating back decades. Such types of analyses are not common within countries due to the lack of data to cover subnational districts over an adequate period, making it difficult to construct panels with sufficient observations. An exception to this trend is research on the United States, where state-level carbon emissions and related measures have been accessible for decades. As a result, numerous studies have investigated various determinants of CO2 emissions, including fossil fuel production, income inequality, and working hours, by analyzing panel data for states over time (Adua et al., 2022; Fitzgerald et al., 2018; Hao and Van Brown, 2019; Jorgenson et al., 2017; Thombs, 2025). Building on this paradigm and leveraging newly available provincial-level carbon emissions data for China, this study will evaluate the TOP/EM arguments within the context of Chinese provinces.
Research hypotheses
Overall, the two aforementioned theories explain the pendulum of the economy–environment relationship. The TOP theory argues that a society driven by economic expansion conflicts with the environment. Economic development is likely to increase the use of resources, leading to concomitant environmental degradation. On this basis, this study will test the following hypothesis using panel data from Chinese provinces.
The EM theory was formulated in response to the TOP theory. The literature and empirical evidence predict that economic development can happen in tandem with environmental improvement. To evaluate this argument, the subsequent hypothesis will be assessed.
Data and measurements
This study uses balanced data covering 25 years (1997–2021) from 30 data-available provinces in China. “Province” here refers to the province-level administrative division, which in the case of this study covers 22 provinces, four municipalities, and four autonomous regions (see Table 1). The analysis was performed during this period because the CO2 emissions data are only available for these years. The measurements for all variables (CO2 emissions, GDP, and population) are described below, and Table 2 shows the summary statistics (mean, standard deviation, minimum, and maximum).
Province-level administrative divisions of China included in the study (N = 30).
Descriptive statistics.
Dependent variable
The amount of CO2 emissions is the dependent variable that gauges the level of environmental impact and the severity of the climate crisis. The emissions are generated from the burning of key energy sources, including raw coal, crude oil, and natural gas. The data are extracted from Carbon Emission Accounts and Datasets, a project funded by the National Natural Science Foundation of China, the Ministry of Science and Technology of China, and Research Councils UK. It provides accurate and up-to-date carbon emissions data for China and other emerging economies. The computations are based on advanced emissions-accounting methods following the IPCC approach with a territorial administrative scope (Guan et al., 2021; Xu et al., 2024). The inventories include energy-related emissions (from 17 fossil fuels in 47 sectors) and process-related emissions (such as those from cement production). Shan et al. (2018) documented the details and the calculation methodology. The annual change in CO2 emissions is displayed in Figure 1, which shows that China's emissions soared from nearly 3 billion tons in 1997 to over 10 billion tons in 2021, representing a 3.5-fold jump. Its share of global emissions, drawn from the Global Carbon Atlas (2025) database, also rose from 12% to 29% during this period.
China's CO2 emissions, 1997–2021.
A bar chart displaying the differences in emissions across provinces is presented in Figure 2. Among all provinces, Shanxi, Shandong, and Inner Mongolia lead the list with more than 1 billion tons of emissions in 2021. On the other side of the table, Qinghai and Hainan rank at the bottom with less than 100 million tons of emissions. Provinces with high emissions usually have large populations and carbon-intensive industries such as manufacturing, coal mining, and petroleum refining. In comparison, provinces with low emissions tend to have a smaller population size, and industries such as tourism produce fewer emissions. Additionally, Figure 3 presents the percentage change in emissions over time. CO2 grew by less than 100% for provinces like Yunnan, Jilin, and Heilongjiang in 2021 compared to 1997. In contrast, emissions surged by over 500% for provinces such as Inner Mongolia, Shanxi, and Fujian. Although the total emissions for Ningxia and Hainan are relatively low, they top the list in terms of growth, with increases of approximately 1000% and over 2000%, respectively.
Provincial CO2 emissions (million tons), 2021.
Provincial CO2 emissions growth, 1997–2021.
Independent variables
To measure economic growth, GDP data are obtained from the China Statistical Yearbook. Figure 4 shows that China's economy experienced rapid expansion, with its GDP increasing by around 15 times, from 8 trillion yuan to 115 trillion yuan over the 25 years covered by this study. Figure 5 illustrates the growth in GDP across provinces. The three northeastern provinces rank at the bottom, with their GDP growing by less than 1000% between 1997 and 2021. GDP jumped between 1000% and 2000% for most provinces, such as Shandong (∼1200%), Guangdong (∼1500%), and Jiangxi (∼1800%). Guizhou and Shaanxi recorded the highest growth, exceeding 2000%, despite the relatively low overall GDP levels.
China's GDP (trillion yuan), 1997–2021.
Provincial GDP growth, 1997–2021.
Interaction terms are created between GDP and dummy variables for time points, and 1997 is the reference year. The interactions help gauge whether the association between GDP and CO2 emissions becomes lessened (the interactions are negative and significant), intensified (the interactions are positive and significant), or remains constant when the interactions are insignificant (Jorgenson and Clark, 2012).
Population is included as a control variable, and the data are also from the China Statistical Yearbook. Unlike CO2 emissions and GDP, which have data for all years, population measures are only available since 2000. Prior research on China's environmental issues used similar indicators for analyses (Hao et al., 2018). The analyses include two predictors due to the data availability of provincial-level measures during the period of investigation. Other variables that impact emissions and are included in cross-national research are unavailable; this limitation is further detailed at the end of the article.
Regression method
This study employs the Prais–Winsten regression for statistical estimation. The regression model is used to estimate a dependent variable that evolves over time in response to changes in predictors. Specifically, this study examines the influence of GDP on CO2 emissions across 30 Chinese provinces from 1997 to 2021 after controlling for the population effect. The disturbances exhibit heteroskedasticity and are contemporaneously correlated across all panels. The panel-corrected standard error correction is applied, preventing overconfidence in the feasible generalized least squares estimator when the number of panels exceeds the number of time periods (Beck and Katz, 1995).
The models incorporate both province-specific and year-specific intercepts. The province-specific intercept accounts for unobserved heterogeneity unique to each province but consistent over time, while the year-specific intercept captures unobserved heterogeneity specific to each year but constant across provinces. Previous research on related topics has employed the same regression approach (Hao, 2023; Jorgenson and Clark, 2012). All variables have been converted into logarithmic form, making the model estimate elasticity coefficients. Using elasticity coefficients compensates for skewness and allows for a straightforward interpretation and more intuitive understanding of the relationship between variables. This means that any coefficient of independent variables represents the expected net percentage change in the dependent variable resulting from a 1% increase in the independent variable (Allison, 2009). The approach has been widely used in existing studies of analogous subjects (Jorgenson et al., 2023; Thombs, 2022).
The coefficients are presented in Table 3, and standard errors are in parentheses. Model 1 estimates the impact of GDP on CO2 emissions, where the data are available for all 25 years (n = 30 provinces × 25 years = 750 observations). The population measure is added in Model 2, and the number of observations is reduced because of unavailable population data before 2000 (n = 30 provinces × 22 years = 660 observations). Also, due to the strong correlation between total GDP and population (r = 0.674), the per capita indicators are used for the analyses.
1
Model 3 extends Model 1 by including interactions between GDP and year. The equations for these models are displayed below by following the literature on fixed-effects regression (Allison, 2009; Beck and Katz, 1995):
Regression results.
*p-value < 0.05; **p-value < 0.01; ***p-value < 0.001.
The subscript i represents province, and the subscript t represents year. For example, CO2 emissionsi,t is the outcome variable for province i at year t. Similarly, GDPi,t is the economic measure for province i at year t. The coefficient β is the vector of coefficients for predictors. Regarding the other components, ui is the province-specific intercept; wt is the year-specific intercept; and ɛi,t is the unique residual.
Pretests are performed following the approaches of previous studies (Hao, 2020, 2023). First, the Levin–Lin–Chu unit-root test results show that panels are stationary instead of containing unit roots. Second, the Pesaran cross-sectional dependence tests reject the null hypothesis of cross-sectional independence. To tackle these challenges, province-specific and year-specific intercepts are incorporated to account for heterogeneity, converting the model into a two-way fixed-effects framework (Wooldridge, 2016). First-order autocorrelation (AR(1) disturbances) is adjusted and assumed to be uniform across all panels. This approach enhances model robustness against omitted time-invariant variable bias and yields estimates that more closely approximate experimental conditions than alternative methods (Hsiao, 2003).
Results
The analysis yields several principal findings regarding the propositions derived from the aforementioned theoretical arguments. First, a province's GDP is significantly related to its carbon emissions, and provinces with higher GDP tend to emit more CO2. According to Model 1, a 1% increase in GDP is accompanied by a 0.565% growth in CO2 emissions. The pattern remains in Model 2 after adding the population measure. GDP per capita maintains a positive and statistically significant association with carbon emissions per capita (β = 0.758). Thus, the consistent findings from these two models signal that the TOP argument is partially supported, insofar as there is a strong association between economic development and emissions.
Second, Model 3 suggests all interactions are significant, and most are negative. In this model, the coefficient for GDP represents the change in CO2 emissions in 1997 for every point increase in GDP in the same year. Regarding GDP's effects for other time points (1998–2021), they equal the combination of the GDP's coefficient (the coefficient in 1997) and the interaction term. As we can see, GDP's effect on carbon emissions increased from 1997 to 2002, with interaction terms becoming positive for the period 2000–2002. Thereafter, however, the interaction terms are consistently negative and become increasingly so over time (2003–2021). Based on these results, two separate regression models are estimated for two distinct periods. One focuses on the period from 1999 to 2002. The coefficient for GDP is statistically insignificant, likely due to the inadequate number of observations (n = 120) and insufficient variability over these 4 years. The second model, covering data from 2002 to 2021, shows a positive and significant GDP coefficient consistent with Model 1's results, demonstrating that economic growth is associated with increased emissions over time. The decoupling effect becomes evident only when the interaction terms are incorporated into the model for analysis. The results of the two additional models are displayed in Table A1 in the Appendix.
The variation in the regression coefficient in Model 3 is presented in Figure 6. To sum up, a 1% growth in GDP led to a 0.586% growth in CO2 emissions in 1997. The percentage grew to 0.992% in 2002, declined to 0.400% in 2010, and then to 0.326% in 2021. There is a noticeably sharp decline in the coefficient between 2002 and 2003, indicating that China's economy became substantially less carbon-intensive over this short period. While many reasons might account for the dramatic shift, and identifying the exact cause is beyond the scope of this study, one possible speculation is that China's ratification of the Kyoto Protocol in August 2002 led to climate initiatives focusing on carbon emissions reduction.
Regression coefficients of GDP on CO2 emissions, 1997–2021.
The analyses provide empirical evidence regarding the TOP and EM theories. While GDP positively correlates with CO2 emissions during the study period, an in-depth investigation of the long-term trend shows a decoupled pattern instead of an intensified effect. In other words, even though GDP growth drives carbon emissions, the magnitude of the force has lessened over the past two decades, which is contrary to Hypothesis 1 but consistent with the EM argument and Hypothesis 2. For the control variable, the population has an insignificant effect, as shown in Model 2. An alternative model estimating the impact of GDP per capita on CO2 emissions per capita with time interactions yields analogous results. GDP per capita has a positive but decoupled influence on CO2 emissions per capita between 2000 and 2021 when data are available.
Several diagnostic tests were conducted to ensure the validity of the modeling assumptions. First, alternative models, including a fixed-effects model with robust standard errors and a first-difference baseline model, were estimated to determine whether the Prais–Winsten model produced spurious results. The findings from these models regarding the direction and statistical significance of the coefficients closely align with the reported results. Thus, the analyses are not biased due to issues related to unit roots and cross-sectional dependence. Second, a multicollinearity test within panels revealed no significant issues, as the variance inflation factor for each independent variable remained below 2, indicating that collinearity does not meaningfully impact the estimations (Belsley et al., 1980). Third, the model was re-estimated by systematically excluding each province and each year to assess the influence of individual cases, with results remaining consistent across models, suggesting no issues with influential observations. Finally, further analyses confirmed that key assumptions, including constant variance and normality, were met. As noted in a previous cross-national study that tests the TOP/EM arguments and utilizes the same modeling approach, the techniques applied are well-suited for evaluating the temporal stability of associations between covariates and outcomes (Jorgenson and Clark, 2012). The high R-squared values result from unreported province-specific and year-specific intercepts. The rho values indicate that approximately 80% of the variance is attributed to differences across years within provinces.
Discussion
The rigorous examination reveals two primary findings. First, China's economy continues to correlate strongly with climate outcomes. In the past 25 years, its GDP grew by 15 times, and CO2 emissions jumped 3.5-fold. Model 1 suggests GDP growth leads to carbon emissions. Drawing arguments from the TOP theory (Gould et al., 2004), the growth imperative and the pursuit of maximizing profit drive economic expansion, which requires constant resource inputs, unescapably generating pollution and carbon emissions that impact the overall state of environmental conditions.
Second, the results in Model 3 and Figure 6 show a decoupling of the impact of the economy from the environment, as expected by Hypothesis 2, but in contradiction to Hypothesis 1. In particular, the extent to which GDP drives carbon emissions has weakened in recent years compared to the beginning of the twenty-first century, and the decoupling has become more evident from 2002 on. The coefficient size in 2021 (0.326) is only around one-third of the peak coefficient size in 2002 (0.992), suggesting that for the same level of economic growth, carbon emissions in 2021 were only 30% of what they were 20 years ago. In other words, while GDP still leads to carbon emissions, the amount of emissions generated as a result of a given unit of GDP has declined. A closer look at the growth pattern of carbon emissions shows that emissions increased by just 14% in the past 10 years (from around 9 billion tons to 10 billion tons) compared to a 250% jump during the first decade of the twenty-first century (from 3 billion tons to over 7 billion tons). The EM theory articulates that the modernization process helps transform society to better manage the climate crisis and reduce the economy's damaging impact over the course of development (Grossman and Krueger, 1995).
Multiple factors might account for the decoupling relationship, and one underlying cause could be the deploying of renewable energy to replace fossil fuels such as coal, oil, and natural gas. China is the world's largest producer of solar panels and wind turbines, most of which are connected to its electricity grid. Almost two-thirds of the large solar and wind plants under construction globally are in China. A significant amount of electricity is also generated via hydropower (IEA, 2024a). The established infrastructure will condition the expansion of renewables and transmit green electricity nationwide. The latest World Bank data (2025) show that renewable energy sources accounted for around 11% of China's total energy consumption in 2011, and this number rose to over 15% in 2021. This significant rise in merely 10 years highlights a promising pattern, suggesting that renewables may become an even more vital component of the energy mix moving forward.
An alternative perspective to illustrate the identified relationship is examining the ratio of CO2 emissions to GDP. Figure 7 shows that China emitted 367 million tons of CO2 per trillion yuan of GDP in 1997. The number fell below 200 million tons by 2010 and further dropped to under 100 million tons by 2021. The figure for 2021 (90 million tons) was just one-quarter of that for 1997. This trend suggests that while both CO2 emissions and GDP grew dramatically, China emitted a considerably reduced amount of CO2 over time for a given level of economic output, which might imply that the manufacturing and transportation sectors have become more energy-efficient. As a result, economic growth is less reliant on fossil fuels, and its effect on carbon pollution has decoupled.
CO2 emissions (million tons) per trillion yuan of GDP, 1997–2021.
Overall, the energy used to power the economy has become less carbon-intensive and more sustainable. The ratification of the Kyoto Protocol mentioned earlier could be an underlying catalyst for the momentum, marking a major turning point in national energy/climate policy. The transition of energy production from fossil fuels to renewables and the promotion of efficient production can only occur when a supportive national framework is in place. Meanwhile, the clean energy industry could generate diverse socio-economic benefits, including employment opportunities and tax revenues. Additionally, recent technological advancements have helped promote electric vehicles as a viable alternative to gas-powered cars. The Chinese manufacturer BYD produced more battery-only powered cars and hybrids in 2024 than Tesla, becoming the largest electric vehicle company worldwide. However, despite the rapid increase in electric vehicle adoption in China, it is important to note that 49% of these vehicles are classified as SUVs and an additional 16% fall into the large car category, while small cars account for only 9%, according to estimates from the International Energy Agency (2024b). The substantial size of these vehicles, although electric, raises environmental concerns due to their greater resource and energy demands throughout the production and usage phases.
Despite the broader decoupling trend observed since 2002, it is worth highlighting that the relationship between GDP and CO₂ emissions remained relatively stable from 2014 to 2021. From a theoretical standpoint, such a stabilized association provides weaker support for the EM theory compared to the more pronounced decoupling pattern in earlier years and does not help address the debate over the TOP theory. On the policy front, this plateau may indicate that China's progress in decarbonizing its economy has encountered a bottleneck, presenting significant challenges to continued advancement. Rejuvenated momentum through policy reorientation or industrial restructuring will be essential to maintain progress. Since the data in this study cover a limited period and cannot fully address the uncertainties, it is necessary to continue monitoring the ongoing trend by employing up-to-date indicators to assess the trajectory more accurately. Subsequent analyses incorporating measures of the next decade will be key to addressing the TOP versus EM debate. If a stronger decoupling association is identified, it would support the EM theory and suggest that China has overcome existing obstacles. However, a weaker decoupling or even a reversed coupling pattern might suggest the opposite scenario, wherein the economy becomes increasingly carbon-intensive.
From a theoretical standpoint, neither the TOP theory nor the EM theory considers the role of globalization when examining economy–environment associations. Both theories need to refine the framing and would benefit from macro-sociological perspectives, particularly those that focus explicitly on how the structure of the world economy yields an uneven distribution of environmental destruction across nations and regions. The ecologically unequal exchange theory can fulfill this function by describing the asymmetrical exchange and the displacement of environmental harm via international trade (Bunker, 1984; Hornborg, 1998). The argument aligns with post-Keynesian economic thought that holds a critical view of the mainstream neoclassical models and emphasizes the impact of the global political economy on national development.
The principal thesis of the ecologically unequal exchange theory examines ways that the hierarchical world system produces and perpetuates unequal development, which has led to contrasting environmental outcomes between high-income and low-income countries. The former group occupies core positions in the world system and some are hegemons, while the latter is in a peripheral position (Freudenburg, 2005). In this system, the hegemons have excessive access to resources and the ability to outsource pollution while the less-developed countries serve as a tap for supply and a sink for waste disposal (Roberts et al., 2003). Specifically, the “imperial mode of living” prevalent in Global North countries and emerging economies requires a disproportionate thirst for global resources to satisfy production and consumption demands, especially amid rapid industrialization (Brand and Wissen, 2012, 2021). For instance, the production of electric vehicle batteries needs critical minerals such as lithium and aluminum, often extracted and imported from other countries using environmentally harmful methods. In this light, the decoupled environmental impact observed in some countries might not stem from the factors projected by EM theory but rather from the global economic process that enables offshoring carbon emissions (Hao, 2020; Jorgenson, 2012). Extending this logic, the decoupling pattern found in China might partly be attributed to the relocation of pollution-intensive manufacturing and mining industries to less-developed countries, which externalizes environmental costs but simultaneously contribute to pollution in those regions. In other words, China's reduced carbon emissions could be offset by concomitant emissions growth in other countries due to transnational production shifts. Such assumptions need to be examined in subsequent research when relevant data are available.
Furthermore, work on the “Great Acceleration” highlights the parallels among economic growth, intensified resource use, and severe environmental degradation. The conclusion is based on the observation of global development after World War II, specifically from the 1950s onward. The unprecedented acceleration in human activities, including economic expansion, population growth, and technological advancements, has profoundly impacted the climate, atmosphere, and human well-being, placing the planet under pressure (Steffen et al., 2005). During this period, Earth system indicators, such as CO2 and nitrous oxide emissions, continued their long-term post-industrial rise, with clear evidence of fundamental shifts to the Anthropocene (Steffen et al., 2015). In addition, the Great Acceleration is not homogeneous but consists of multiple phases of different trajectories and distinct regions. The concept of the “Second Great Acceleration” has been introduced to characterize a subsequent period (post-2000), marked by intensified resource consumption, primarily driven by emerging economies (Görg et al., 2020). When applying this perspective to China, where GDP grew 15-fold between 1997 and 2021 alongside numerous accelerated socio-economic trends, environmental degradation could be an inevitable consequence. Therefore, similar to the prior argument derived from the ecologically unequal exchange theory, it is prudent to remain cautious about attributing the observed decoupling patterns identified in the analysis solely to the effects of modernization.
Conclusion
Climate change is an overwhelming task facing the world that has far-reaching impacts on the ecological system and socio-economic well-being (Carleton and Hsiang, 2016). China is one of the world's largest sources of planet-warming greenhouse gases. Economic growth is a primary driver of the climate crisis, and the existing literature has made the economy–environment nexus a focal point in theoretical debates and policy considerations (Fisher and Jorgenson, 2019). The TOP theory argues for an ironclad association between ecological harm and economic growth (Schnaiberg, 1980). On the contrary, the EM theory envisions that the economy and environment can be synergistic and that environmental protection is in tandem with economic growth (Mol, 2003). This study thoroughly evaluates these propositions in order to understand the landscape of how China's economy shapes its climate outcomes, and its findings generally back the EM argument.
Investigating human activities that lead to climate change has been a major research area in environmental sociology (Bohr and Dunlap, 2018; Dietz, 2023; Rosa et al., 2015). Regarding the environment–economy interaction, current studies have performed numerous analyses using country-level data (Jorgenson and Clark, 2012; Knight and Schor, 2014; Thombs, 2018). By analyzing China's longitudinal, provincial-level data, this survey contributes to the literature dominated by cross-national design and presents findings of geographical and temporal variations. Patterns of within-country variations complement the between-country variations identified in existing research, provide novel insights, and help account for the causes behind CO2 emissions.
Methodologically, using panel data and including time interactions in regression models generates robust results and enables tracking of the nuanced dynamics of how the economy impacts the environment over time. More importantly, examining provincial-level data exposes subnational disparities typically omitted by conventional national metrics. Figure 2 illustrates significant variations in carbon emissions across provinces, with coastal provinces emitting substantially greater CO2 than their inland counterparts. An analogous pattern is observed for GDP, as coastal provinces are considerably wealthier than those inland. The regression results for 30 provinces over 25 years show a shifting association between GDP and CO2 emissions, which, despite remaining positive throughout this period, intensified at the beginning before showing signs of decoupling over most of the last two decades. Thus, considering China's wide domestic disparities in economic structure and environmental governance, investigating provincial-level data is indispensable to uncover the evolving economy–environment association under diverse local contexts within a single national setting. Relying solely on country-level indicators would conceal these dynamics. Meanwhile, this distinctive analytical leverage also provides valuable insights into evaluating the theoretical debate between the TOP and EM frameworks. These arguments have been tested in current literature primarily using country-level measures, making this study's subnational design an important addition.
In addition to its scholarly contributions, this study has nontrivial policy implications, particularly given China's vast economy and carbon emissions. The Paris Agreement signed in 2015 (UNFCCC, 2015) and the Glasgow Climate Pact approved in 2021 (UNFCCC, 2021) recognize the indispensability of substantially cutting greenhouse gas emissions by 42% by 2030 and 57% by 2035 to ensure the global temperature is below 1.5°C above preindustrial levels. However, the 2023 United Nations Climate Change Conference pointed out that countries are “not yet collectively on track towards achieving the purpose of the Paris Agreement and its long-term goals” (UNFCCC, 2023).
According to the findings of this study, despite the mounting carbon emissions, China is likely on a path of low-carbon development, with its GDP gradually exerting less impact on CO2 emissions. Several reasons contribute to this promising trend. At the country level, the growing prominence of the service industry and the digital sector in the national economy helps reduce energy intensity and offset the heavy energy usage of traditional industries. Meanwhile, even in carbon-intensive sectors such as manufacturing and energy production, technological progress has enhanced efficiency, enabling the same output level with reduced energy consumption. Therefore, the declining energy intensity plus the growing deployment of renewable energy mentioned earlier might be the primary mechanisms driving China's transition toward a low-carbon economy. China pledged to peak carbon emissions before 2030 and reduce emissions by 7% to 10% from peak levels by 2035, and the data show that the country's emissions are seemingly reaching a plateau. The growth rate has decelerated in recent years, even exhibiting a pattern of declining emissions between 2013 (9.5 billion tons) and 2016 (9.25 billion tons), and a potential long-term decline could be on the horizon.
China's government unveiled new strategies to reconstitute the organization of the economy along ecologically rational lines during the United Nations-sponsored climate summit in Baku, Azerbaijan, in 2024. Meanwhile, there is a strong push for renewable energy. The Chinese government released a new plan in October 2024 to boost renewables and replace fossil fuels that sets ambitious targets: increasing annual renewable energy consumption to 1.1 billion tons by 2025, a 30% rise from 2023 levels, and reaching 1.5 billion tons by 2030, a further 36% increase from the 2025 levels. Through efforts to upgrade infrastructure, electrify key sectors, and advance green technologies, China is laying the groundwork for its next phase of economic transformation. Formerly a peripheral country in the world system, China was once a primary recipient of carbon emissions outsourced by wealthier core nations. However, this trend has slowed and even started to reverse. The Global Carbon Atlas (2025) data showed that China absorbed roughly 1.4 billion tons of CO2 in 2011, which fell to about 1 billion tons in 2021, representing a 25% reduction. Altogether, China's shift toward renewable energy and its diminishing role as a repository of carbon pollution for the Global North countries might offer valuable insights for other Global South nations pursuing a low-carbon, sustainable path to economic growth. These countries resemble China in their historical dependence on carbon-intensive manufacturing and in being major recipients of emissions offloaded by high-income nations.
Building on the momentum, China needs to reduce its dependency on a carbon-intensive economy and transition from fossil fuels to renewable energy (Hao, 2025; Hao and Shao, 2021). Policies that aim to mitigate carbon emissions might focus more on the coastal provinces since they share a dominant proportion of China's total emissions and have a larger economy and population. High-tech companies are concentrated in the Yangtze River and Pearl River Delta regions, which will facilitate the application of new technologies, such as building renewable energy facilities and manufacturing electric vehicles.
Despite the contribution, this study has limitations that future research can resolve. First, this study estimates the impact of two predictors on carbon emissions. The modeling approach would be improved by the inclusion of additional variables for estimation. The measure of urban population was initially included in the model but dropped due to its high correlation with the population measure. Predictors such as manufacturing GDP, income inequality, and non-dependent population could condition a province's CO2 emissions, and it is worth examining their effects when provincial-level, cross-year data are available (Fitzgerald et al., 2018; Thombs, 2025). Second, the outcome measure of climate change is constrained to a single indicator, CO2 emissions, which might be complemented by alternative indicators such as methane emissions, a more potent greenhouse gas (Jorgenson, 2006). Meanwhile, following the examples of cross-national analyses, it is helpful to compare territorial and consumption-based CO2 emissions (Knight and Schor, 2014), or to estimate the potentially varying effects of economic development on emissions related to domestic production, international trade, or household consumption (Huang, 2024). Third, building on the mechanism established in this study, a follow-up project might continue tracking the associations for a more extended period by using data from before 1997 and the latest measures from after 2021. Employing data from more recent years will capture recent changes in the economic structure and in energy consumption patterns. Finally, future research on this subject should situate the case of China within a broader global context and identify the international forces that have shaped China's economic growth and environmental consequences.
Climate change is becoming increasingly disastrous and warrants immediate attention. It requires extensive government efforts to promote clean energy and push high-polluting industries to decarbonize. The growing proportion of less-carbon-intensive sectors in China's economy and greater deployment of renewable energy are examples that could indicate a scenario in which the decoupled impact remains or becomes more significant. Policy implications derived from energy transition and technological advancement could cover a broader scope of ways that the government might propose to mitigate the impact of the climate crisis. Given China's pivotal role in global mitigation efforts, subsequent studies on this topic carry significant value and continue to be needed.
Footnotes
Acknowledgments
Previous versions of the article were presented during invited talks at the Sociology Department of Nanjing University, Hohai University, and Xi’an Jiaotong University in the summer of 2024. I am grateful for the thoughtful feedback provided by faculty and students.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Notes
Appendix
Table A1. Regression of CO2 emissions on GDP for two periods.