A new Stanford-led analysis of corporate carbon disclosures finds companies undercount emissions from their supply chains by billions of tons.
The shortfall arises from use of a popular statistical model that assumes all suppliers are located in the United States. "Supply chains are global, though, so a model that assumes everything is made domestically is going to give us a wrong answer," said Steve Davis , a professor of Earth system science in the Stanford Doerr School of Sustainability and lead author of the Dec. 20 study in Nature Communications .
Davis and co-authors set out to find out how far off estimates are under the common approach compared to results from an alternative model that considers emissions data from the regions where suppliers actually operate. They found the widely used single-region U.S. model , maintained until recently by the U.S. Environmental Protection Agency, missed about 2 billion tons, or 10% of emissions tied to more than 400 companies' supply chains in 2023. That's roughly equivalent to the annual emissions of Russia or India.
The additional 2023 emissions included approximately 973 million tons of emissions attributed to suppliers in China, which relies heavily on coal. The biggest gaps emerged in energy-intensive manufacturing sectors such as steel and concrete, construction machinery, fabricated metals used for cars and infrastructure, and electronic components.
Companies relying on estimates from single-region U.S. models may overlook the emissions impact of importing energy-intensive manufactured goods from China and Russia, according to the study. They may also miss opportunities to reduce emissions and associated costs by sourcing those goods from countries with cleaner grids, such as the U.S., France, and Brazil.
At the start of 2026, as an expanded carbon border tariff takes effect in Europe, the cost of upstream emissions is increasing for EU importers of manufactured goods, including steel, aluminum, and cement. "A company that's interested in reducing the emissions in their supply chain needs to know not just how big the number is, but also where these emissions are coming from," Davis said.
Expanding access to better data
Together with collaborators including lab member Wesley Ingwersen , who created the U.S. EPA model and led work on it until the agency shuttered it in August 2025, Davis is now working to make a global model freely available and easy to use through an effort called Cornerstone . "The reason companies haven't been using the global models is they're not as easy to come by. They are a lot more involved to build, and there hasn't been an easy, open-source version," said Davis.
The group is integrating the former government database with the multi-region model analyzed in the study, which was developed by a private company called Watershed, where Davis chairs the science advisory board and previously served as head of climate science.
"When available tools neglect international sources of emissions, companies' sustainability decisions suffer," said study co-author Michael Steffen, Watershed's head of climate analytics.
The team aims to release the merged model in late 2026, with ambitions to account for emissions from land-use changes and deforestation in later research and iterations. "If you're getting soybeans from Iowa, it has a very different footprint than if you've cut down some of the Amazon to grow those soybeans," Davis said.
Scientists from the World Wildlife Fund and CDP, formerly known as the Carbon Disclosure Project, co-authored the paper. "These are NGOs that are really interested in minimizing greenwashing and making sure that corporate climate actions are as beneficial as possible," Davis said.
Some critics question whether models based on sector-wide averages and spending like those analyzed in the study are the right approach to estimate upstream emissions at all – regardless of whether global or single-region models are used. "I think we could all agree that if you had perfect data about exactly what was going on in your supply chain, you could make even more accurate estimates and leapfrog all of these models," Davis said. "The reality is, though, that it's still really difficult to get a lot of that data, and there's little prospect of getting it without much more stringent regulations than are even on the table." As a result, he said, there's value in improving the modeling approach even if the long-term strategy is to get better data.
Corporations seeking to track and reduce emissions in their supply chains have the potential to make a meaningful difference in global carbon pollution, Davis said. "They are making sizable investments. If those dollars are directed to the right places, it could meaningfully reduce global emissions," he said.
Co-authors of the study not mentioned above include Andrew Dumit, Mo Li, and Yohanna Maldonado of Watershed; Martha Stevenson of the World Wildlife Fund; Tatiana Boldyreva of CDP; and Sangwon Suh, who is affiliated with both Watershed and the University of California, Santa Barbara.
