image: It displays the feature importance graphs derived from the Random Forest and the Extra Trees models. Feature importance is measured by each feature’s percentage of total predictive power; a higher feature importance indicates stronger predictive power for that variable. The black solid lines within each bar represent the standard deviation of the importance value for each variable of interest. The figure indicates that the amount of fixed assets (i.e. property, plant, and equipment), industry, and the number of employees are the three most important variables for the ET model. Conversely, total assets is the least important variable, likely because the model already includes the amount of fixed assets and non-current assets.
Credit: Lucas S. Li (Shanghai American School, China) Yan Zhao (City College of the City University of New York, USA)
Background and Motivation
Climate change, driven by carbon emissions, has become a core concern for policymakers, investors, and corporate leaders worldwide. In recent years, global efforts such as the Paris Agreement and divestment strategies by major asset managers have aimed to accelerate the transition to a low-carbon economy. Yet, the question remains: Do these initiatives influence investor behaviour in capital markets? Are investors becoming more attentive to the carbon profiles of firms? China Finance Review International (CFRI) brings you a new article titled ‘Do investors care about carbon emissions? Evidence based on stock return co-movement with machine learning-augmented data’, which examines whether and how the carbon emissions of US-listed firms shape patterns in stock return co-movement over time.
Methodology and Scope
This study leverages a unique, two-pronged approach.
- It constructs a comprehensive database of US public companies’ carbon emissions (scope 1) from 2004 to 2020, using both publicly disclosed data and machine learning models to predict emissions for firms that do not report them. This data augmentation, powered by advanced algorithms such as Extreme Trees, increases sample size and diversity, reducing bias.
- The research calculates yearly pairwise stock return correlations and emission similarities for firm pairs, controlling for firm fundamentals, industry, and other factors. The analysis compares co-movement patterns before and after 2012, and further tests robustness through double-sorting and analysis of institutional investor flows.
Key Findings and Contributions
- Reveal increased co-movement among stocks with similar carbon emissions: The study finds that, since 2012, the stock returns of firms with similar carbon emission levels have moved more closely together, suggesting that investors are increasingly attentive to carbon-related information.
- Demonstrate that investor carbon awareness has strengthened over time: The increase in return co-movement post-2012 is significant, and the effect is most pronounced among larger, higher-emitting firms, reflecting a shift in market sentiment toward sustainability.
- Show that exposure to environmental news magnifies the carbon effect: The link between emission similarity and stock return co-movement becomes stronger in periods when investors face more frequent environmental news and attention shocks.
- Establish a robust methodology for carbon-related asset pricing research: By applying machine learning to expand carbon emission datasets and focusing on co-movement rather than average returns, the study offers new tools and perspectives for examining ESG factors in capital markets.
- Highlight the impact of regulatory action: State-level regulatory emission targets significantly influence both investor behaviour and firm stock price dynamics, confirming the real effects of policy on market activity.
Why It Matters
These findings demonstrate a clear evolution in investor behaviour: market participants are increasingly pricing carbon risk, especially in response to regulatory signals and growing environmental awareness. As responsible investment practices gain traction, companies’ carbon footprints are becoming a more material factor in their valuation and risk assessment. This has significant implications for corporate disclosure, capital allocation, and the broader transition toward sustainable finance.
Practical Applications
- For Researchers: This study provides a robust empirical framework—using both machine learning and novel asset pricing tests—for analysing how environmental factors influence capital markets, opening new avenues for ESG and carbon finance research.
- For Investors: The results underscore the importance of factoring carbon risk and emission profiles into portfolio construction, as the market increasingly rewards or penalises firms based on their environmental impact.
- For Policymakers: Evidence that regulatory actions and investor attention shape market pricing of carbon risk supports the continued development of emission targets, disclosure standards, and environmental regulation as levers to guide market behaviour toward sustainability.
- For Listed Companies: Firms should be aware that both their own carbon disclosures and the evolving carbon consciousness of investors can affect their stock price dynamics and capital market perception, emphasising the strategic value of transparent and proactive sustainability practices.
Discover high-quality academic insights in finance from this article published in China Finance Review International. Click the DOI below to read the full-text original! Open access for a limited time!
Journal
China Finance Review International
Method of Research
News article
Article Title
Do investors care about carbon emissions Evidence based on stock return co-movement with machine learning-augmented data
Article Publication Date
5-Jun-2025