The Paris Climate Agreement was agreed by most of the world’s nations in 2015. It aimed to keep global average temperature increase over pre-industrial levels to 1.5°C by 2100, or in any event well below 2°C. The 2°C threshold, viewed as a tipping point, was chosen because of scientific opinion that climate-related risks would increase dramatically beyond this point. In an article just published with Jitong Jiang and Skylar Shi, we analyzed progress since then.
The agreement was to be implemented by international cooperation. Individual countries made promises to cut their emissions, in most cases over the ensuing 15 years, to 2030, by amounts they specified themselves. The amounts varied but the global average was on the order of 30% over the 15 years. These promises, called Nationally Defined Contributions (NDCs), were voluntary with no legal enforcement mechanism beyond international peer pressure.
In an article with Peiran Liu in 2021, we analyzed climate change prospects to 2100 based on data up to 2015, just before the Paris Agreement started, using a fully statistical Bayesian probabilistic approach. The method was based on the IPAT equation in the form of the Kaya identity, which breaks down carbon emissions from each country as the product of three factors: Carbon Emissions = Population x GDP per capita x Carbon Intensity, where Carbon Intensity is the amount of carbon emitted per unit of GDP. We then used the IPCC’s linear relationship between cumulative carbon emissions and temperature change to infer the distribution of temperature increase in 2100.
Based on pre-Paris data, we found that the predicted temperature increase was 2.8°C, with very likely range 2.1-3.9°C, while the chance of staying below 2°C was just 5%. Of the three components of the IPAT equation, population is sticky, while reducing GDP per capita would likely be unpopular. This leaves carbon intensity as the main lever under potential policy control.
Now, with a decade of post-Paris data, our research group revisited this analysis based on what has happened since the Paris Agreement started in 2015. We updated the methods and the data. Over the period 2015-2024 for which we have data, we found that carbon intensity declined (i.e. improved) by about 3% per year, as shown in the left panel of Figure 1. This is much faster than the roughly 1% per year over the previous period, 1960-2015. The NDCs implied an annual reduction of about 4% per year, so in terms of carbon intensity, countries came closer to meeting their promises than they would have previously. We can’t claim for sure that this improvement has been caused by the Paris Agreement, but the data are consistent with it.
But, in spite of this good news, we found that total carbon emissions increased, by almost 1% per year, as shown in the center panel of Figure 1. How could carbon emissions actually increase, when carbon intensity was improving so fast? The answer is that GDP per capita increased fast from 2015 to 2024, by 4% per year, as shown in the right panel of Figure 1. This can be compared with only 2% previously, from 1960 to 2015. This rapid rise in GDP more than cancelled out the progress in carbon intensity.
To understand this, it might help to consider a simple analogy. Think of a toaster factory, which in 1960 was emitting substantial amounts of carbon, and so had high carbon intensity. Over the years, as the manufacturing process became more energy efficient, the amount of carbon per toaster would go down, and so would carbon intensity. Also, if a worker who previously drove to work, started taking transit, or driving an electric car, the factory's carbon intensity would go down. This would also happen if the factory put solar panels on its roof to produce some of the electricity it uses. Overall technological improvements would also reduce carbon intensity. For example, in the 1990s, standard light bulbs were largely replaced by LED bulbs, which use 90% less energy; this would also have reduced our factory’s carbon intensity. Finally, public measures to encourage these developments, such as legislation, regulations or tax incentives, could have helped.
But at the same time as the carbon intensity from our toaster factory was improving over the decades, millions of people around the world, including many in developing countries, were becoming more affluent and buying toasters, including from our factory. So overall our toaster factory could have been emitting more carbon, even though each toaster was being more efficiently produced. This illustrates one reason why climate change is sometimes described as a “super wicked problem.”
What are the long-term climate implications of this? Based on pre-Paris data, we had projected that annual carbon emissions would stay relatively flat until 2100, while now we project that they would decline by 64%. The projected global average temperature increase to 2100 is now 2.4°C, compared with the previous 2.8°C, a notable but still inadequate improvement. The very likely range is now 1.7-3.2°C. The probability of staying below the 2°C tipping point is still only 17%. One bright spot is that the chance of the very worst warming, above 3°C, has gone down from an alarming 26% pre-Paris, to 9% now.
These projections are based on current trends. But if all countries keep their promises or NDCs, most of which are for 2030, and then continue to improve at the same rate for the following 70 years, the projection is lower, at 2.1°C.
In 2025, the newly elected US President Donald Trump pulled the US out of the Paris Agreement, with effect from 2026. If the US continues on current trends, rather than continuing to improve at the rate of its current NDC, then our projection for global average temperature increase goes up by 0.1°C, to 2.2°C. In that case, the chance of keeping the increase to below 2°C would go down from 34% to 27%. However, this is optimistic, since President Trump has expressed hostility to climate mitigation measures, and has indeed called for actual increases in fossil fuel consumption and hence in carbon emissions, and not just for a slowing of efforts to reduce them. If this were to happen, the projected additional increase, even assuming that all other countries keep their promises, could well be higher than 0.1°C.
I have given results for the world as a whole, but carbon intensity and carbon emissions vary widely between countries. By some measures, the top three countries by GDP are the US, China and Germany. All three improved carbon intensity substantially and at similar rates over the past decade, but their absolute levels are very different. The US’s carbon intensity is 50% higher than that of Germany, which is striking since the two countries have many cultural and economic similarities. The difference is due in part to lifestyle differences, with Germans using transit more than Americans and so driving less, and living in smaller homes, that require less heating and emit less carbon in other ways.
China’s carbon intensity is a whopping three times that of Germany, partly because of its heavier use of coal. The US and China are the world’s two biggest carbon emitters, with about 34% and 12% of the world’s total, respectively. So there is a lot of scope for them to become more carbon efficient, including by becoming more like Germany in their energy use.
Our main result suggests that just reducing carbon intensity may not be enough, and that we may need to consider reducing unnecessary consumption as well. Some of China’s emissions arise in the production of consumer goods, or "stuff", that are then exported to the US, with the transportation generating further emissions. In 2025, President Trump imposed substantial tariffs on Chinese imports. It seems plausible that such tariffs could reduce American consumption of Chinese imports, thus potentially reducing Chinese carbon emissions, and perhaps US ones also. So, ironically, while President Trump is hostile to action to combat climate change, his signature policy of tariffs may actually help to do so.