Introduction

Greenhouse gas emissions from emerging economies are a primary driver of the future course of climate change, with China standing out as it generates one-third of global emissions. Appendix Figure A.1 shows that its CO2 emissions are more than double those of the US and more than three times those of the EU or India. In July 2021, China launched a nationwide carbon market, starting in the power sector which accounts for 15% of global emissions. Unlike traditional cap-and-trade programmes, China’s national Emissions Trading Scheme (ETS) is rate-based, meaning compliance depends on emission intensity—the ratio of emissions to current output. Because allowances increase with output, this design implicitly subsidises production.

China’s national ETS

The national ETS regulates 2,257 power plants that emitted 26,000 tonnes or more of CO2 in any year between 2013 and 2018, covering approximately 40% of China’s total carbon emissions. Regulated plants receive a free allocation of emission allowances based on assigned benchmarks: 0.877 tCO2/MWh for conventional coal-fired units with a rated power of at least 400 MW, and 0.979 tCO2/MWh for those below. More efficient firms with initial intensities below the benchmark receive surplus allowances, providing incentives to increase production. Conversely, less efficient firms must buy allowances or reduce emission intensity, as reducing output is implicitly taxed through a reduction in allowances received.

Data and methods for power generation and carbon emissions

To overcome challenges regarding monitoring and transparency, the study utilises satellite imagery and machine learning to estimate plant-level output for 413 coal-fired plants. In collaboration with Climate Trace, electricity generation is predicted from visible water vapour plumes, which are robust indicators of activity. These predictions are then combined with plant-specific emission factors derived from fuel and technology types to estimate total CO2 emissions. This satellite-based measure provides an objective benchmark against annual data reported by firms. The study focuses on “specialised” plants that use a single technology: 54% are sub-critical and 35% are super-critical or ultra-super-critical.

Data and methods for ambient air pollution

The research incorporates ambient air pollution data from approximately 2,000 ground monitors across China and high-resolution satellite-derived estimates. To track the transport of pollutants, the study employs the Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model. This atmospheric dispersal model generates particle trajectories from emitting power plants, matching them to specific pollution monitors within a certain radius. This allows for an assessment of how the ETS alters the spatial allocation of air pollution relative to plant efficiency and location. Factors such as stack height, date, and hour are used to generate forward trajectories.

Empirical strategy

The primary analysis compares coal-fired power plants within China by their initial generation efficiency before and after the ETS launch. Plants are classified as high-efficiency (super-critical or ultra-super-critical technology) or low-efficiency (sub-critical technology). The difference-in-differences approach estimates the differential response between these groups, controlling for plant and year fixed effects, plant-by-year capacity, and province-by-year GDP. A further cross-country comparison is conducted using coal plants in Indonesia and the Philippines to establish a counterfactual for aggregate emission implications.

Results

Within China, high-efficiency plants increased their annual utilisation rate by 2.6 percentage points (approximately 6.3%) and CO2 emissions by 186 kilotonnes (about 5.2%) relative to low-efficiency plants. There is no evidence of a differential reduction in emission intensity, which was the margin the market intended to move. Ambient air pollution increased downwind of high-efficiency plants: SO2 concentrations rose by 4.4%, PM2.5 by 3.2%, and PM10 by 2.8%. The cross-country comparison suggests that Chinese high-efficiency plants increased emissions relative to foreign counterparts, while low-efficiency plants moved similarly to foreign plants. A bounding exercise suggests the net change among observed Chinese plants was an increase of approximately 14 million tonnes of CO2 per year.

Summary and discussion

The evidence suggests that China’s rate-based ETS encouraged net expansion of emissions rather than reallocation. Generous benchmarks allowed inefficient plants to comply at low cost, depressing carbon prices and weakening incentives to cut intensity. To effectively reduce carbon emissions, benchmarks must be tightened within the existing system, or an alternative system that does not incentivise electricity production, such as an absolute cap, must be adopted. The findings carry cautionary implications for other emerging economies, such as Indonesia and India, which are considering similar rate-based designs that may encourage output expansion by efficient producers and suppress carbon prices.