Climate 201: What’s Driving Greenhouse Gas Emissions?

Physical Attraction
13 min readMar 7, 2021


Author’s note: In 2020 I started publishing a series of episodes under the title of “Climate 201”. The idea was to introduce topics in climate change, climate science, and climate policy — which I’ve studied for some years as a student- and explore them — and their implications — in much more detail than the simplistic framings you often see on the news. Ideally in the process I wanted to answer any questions that my audience had about climate change.

You can find all of the episodes in the Climate 201 series here.

What On Earth is a Kaya Decomposition?

Hello, and welcome to Climate 201. This episode is going to be a little bit nerdier and wonkier than usual, so I hope you’ll forgive me for it — but I think it will provide an interesting insight into some broader issues in the field of climate change and it’s nice and analytical.

In 1750, human activities very roughly resulted in approximately 10 million tonnes of carbon dioxide a year being emitted into the atmosphere. By 2017, there were 37 billion tonnes of CO2 being emitted a year — a factor of 3700 more.

In 1750, the world population was perhaps 700 million — by 2017, it was around 7 billion, a factor of ten more.

GDP, as I always say, is a flawed economic measure, but it does very roughly tell you the overall “volume of activity” in the economy — even if that activity is not necessarily good! It wasn’t invented properly until 1934, but there have been plenty of estimates published that attempt to project it back in time. Using one of these, we can see that global GDP was perhaps $800bn in 1750 and rose to over $100trn by 2017 — in other words, the global economy is perhaps 120x bigger today, in terms of raw activity, than it was in 1750.

Since 1750, humans have come up with more and more uses for energy. It’s worth pointing out that energy consumption can be described in different ways. “Primary energy” means the embodied energy in all of the fuels and inputs to a system, while “final energy” is the energy that’s actually used. So, for example, if I burn a lump of coal in a power plant, converting 33% of its energy to electricity, the eventual use of that energy to (say) power a lightbulb will only be a third of the primary energy consumption.

According to Vaclav Smil, in 1800, humanity in total used around 5,600TWh of primary energy every year. Back then, this was mostly in the form of “traditional biofuels” — i.e. a great deal of this is people burning wood to heat their homes, or to power rudimentary devices — and it was fairly static across the years, as this demand was pretty constant. But by 2018, this had grown to 157,000TWh of primary energy consumption, from a whole range of different sources including fossil fuels, nuclear, wind and solar, and hydropower energy. In other words, the primary energy consumption has grown by about a factor of 28x between roughly 1750 and today.

So what is the point of telling you all of this?

The point is that there are several different things that can drive changes in CO2 emissions from a society, and it’s important to be able to disaggregate them to determine what’s driving any changes in emissions. For example, between 1750 and 2018, the population has grown by a factor of 10, the economy has grown by a factor of 120, the energy use has grown by a factor of 28, and all of these things have tended to happen together — as our CO2 emissions have grown by a factor of 3700. GDP growth is driven by population growth — more people means more economic activity, and also more demand for energy.

So if you’re emitting more CO2, is it because your population has increased? Because the population is wealthier and therefore demanding more products which require more CO2 to increase? Because your economy is using more energy? Or because the means that you’re using to generate energy have become dirtier or more polluting?

This is the purpose of the so-called “Kaya decomposition” or Kaya Identity, developed by Japanese economist Yoichi Kaya in 1997. Essentially, this tries to split up CO2 emissions into the drivers of those emissions.

So, the CO2 emissions of a country, or a planet, can be expressed as:

CO2 emissions = CO2 per unit energy * energy per unit GDP * GDP per person * population

So, we start with the CO2 per unit energy — a measure of how polluting our procedures are to generate energy, the carbon intensity of energy. Then we multiply by energy per unit GDP — in other words, how energy-intensive is our economy? How much energy are we consuming to generate the wealth that we have? Then we multiply by GDP per person — how wealthy is the economy, once you account for population? And finally, we multiply by the population.

Looking at things in this way allows us to start to analyse what’s driving changing CO2 emissions. And it also illustrates part of why the baseline is for emissions to continue to accelerate in their growth. The OECD recently suggested that global GDP per head would rise 2.6% a year, every year, until 2060. The world’s population will continue to grow by 1–2% a year until the middle of the century, according to many analysts. With these long-term trends pushing emissions ever higher, even if we don’t grow hungrier for energy or choose to burn more fossil fuels to produce that energy, the baseline would be that emissions would rise. If the economy and population continue to grow, then we need to get cleaner in terms of our energy production — or more energy efficient in terms of our economy — every year just to tread water and keep emissions the same.

One of the advantages here is that we can start to analyse what any change in CO2 emissions is down to. For example, we can look at the Soviet Union. Famously, the CO2 emissions from the Soviet Union countries fell dramatically as the USSR collapsed — they fell by around 35% in these countries between 1990 and 2010, even while emissions across the world in general rose by 44%. So we can ask: was this due to people moving away from these countries? Was it due to economic collapse in these countries? Did the countries shift from economies that used a lot of fossil fuels to generate their wealth? Or did they start producing energy a lot more cleanly? What fraction of the change is down to each of these factors? And so on.

When you do this decomposition, although it shifts from country to country, it’s clear that the overall driving factor is a big economic recession associated with the collapse of the Soviet Union which caused large drops in CO2 emissions. There’s a paper, “Drivers of CO2 emissions in the former Soviet Union: A country level IPAT analysis from 1990 to 2010”, which looks at this in much more detail using the IPAT analysis which is similar to the Kaya identity.

It finds that CO2 emissions fell by around 48% between 1991–2000 in the former Soviet Union. The changes to the energy mix and carbon intensity are quite small — around 4% each — and the economy actually becomes more energy intensive, so greater energy production per unit GDP, by about 8%. The population is pretty static, so we can see that a huge fraction of the decline in CO2 emissions was caused by the approximately 40% decline in the economy over this decade.

Let’s apply this to another example — my favourite, my home country of the UK.

The UK has widely been touted as one of the world’s leaders on climate change, due to its record on greenhouse gas emissions. And, indeed, since 1990, the UK’s CO2 emissions have fallen from around 600MtCO2 a year in 1990 to around 370MtCO2 a year in 2018, a decline of about 38%.

But, for example, you could argue about how this is being driven. Perhaps the UK’s economy has slowed down. Maybe it’s just that the economy depends much less on energy production now — for example, as we have switched from manufacturing to a services economy, or as energy efficiency measures have been installed. Or perhaps it’s due to a switch where our energy is being made in a less CO2-intensive way.

Always to be acknowledged in climate geopolitics is that, if economies gradually evolve from high-intensity manufacturing to services economies that import goods from overseas — as has largely been the case in Europe and the US — then there’s a lot of unfairness in us asking developing nations to simultaneously manufacture all of our goods and also to reduce their CO2 emissions rapidly, having already made ourselves rich in part with the unconstrained burning of fossil fuels.

Looking at the specifics of the UK between 1990 and 2018, we can see that the UK GDP has increased from $1.1trn to $2.85trn in 2018 (although even in 2018 it was still lower than its peak in 2007, so this is not some monotonic increase), so the economy is about 2.5x bigger.

In terms of the energy intensity of GDP — a sort of combined measure of how efficiently energy is used, and how much of the wealth production of a country depends on energy use — we see that the UK has declined from around 0.125 “kilos of oil equivalent” (a measure of primary energy use) per $ of GDP in 1990 to around 0.062 “kilos of oil equivalent” per $ of GDP in 2018. In other words, our demand for energy per unit GDP has more than halved in these years.

Finally, we can look at the CO2 emissions per unit of primary energy consumption in the UK — a metric of how well we are removing CO2 emissions from our systems for providing energy in electricity, transport, industry, and agriculture. And this has fallen, according to my calculations from statistics provided by Enerdata — where you can look into this from your own country — that we’ve shifted from 2.68kg of CO2 emissions per kg of oil used equivalent to 2.05kg of CO2 emissions per kilogram of oil used equivalent. i.e. this has fallen by about 24%.

So what can we really say about Britain’s decline in CO2 emissions? The economy has grown quite a bit since 1990, but the effect of this has been offset by a more energy-efficient economy, which in turn will be due to more efficient use of energy but also an economy that focuses more on services and not energy-intensive industrial manufacturing processes. Meanwhile, the UK’s energy production has decarbonised — driven by a shift to wind power and natural gas, and away from coal in the electricity sector — which has contributed to the large reduction in emissions.

You can of course repeat this kind of case study for all sorts of different nations and different time periods, to try and understand what the primary drivers of CO2 emissions are in the world. Eneroutlook from Enerdata, which you can access at, is a fun example of this.

Typically, globally, what we’ve seen in recent years is that primarily economic growth, and growth in energy demand, has offset any deployment of clean technologies or shift towards energy efficiency. The main force pushing emissions down has been economies using less energy than they once did to produce the same $ GDP — even though, as the economy has grown, overall energy demand has increased substantially since 2000. The CO2 emissions per energy use has only gradually begun to decline, and in 2015 was still higher than it was in 2000, so this has not yet been a major contributor to any slowing of emissions growth.

So looking at the implications of this kind of decomposition, we can see what we might ideally want to happen. Assuming we are ruthless economists who always want to maximise GDP and economic growth — which we’ll come back to later — then we want to reduce the CO2/Energy, by deploying more renewables and more carbon-neutral alternatives to existing fuels, and we want to reduce the Energy/GDP, by making our economies more energy-efficient and less dependent on processes that require large consumption of energy.

These are obviously going to be much better than trying to shrink the economy or the population, in my view. People will occasionally talk about “reducing the population” as a solution to climate change. But obviously to actually reduce the population by any meaningful amount, you would need to kill or avoid the birth of billions of people. Which seems like a rather dramatic thing to do when you could achieve similar % reductions in the world’s CO2 emissions with a push to install renewables, to drive more energy-efficient cars, or to phase out coal. Similarly, realistically speaking, given that a global reduction of GDP by just 0.03% in a year (2009, according to Our World In Data) was referred to as “the Great Recession” and the “Global Financial Crash”, you can see that we cannot pull too hard on the GDP lever without radical, radical changes in how the world is run, given that the world is currently addicted to endless GDP growth. If we want to get emissions to net zero, we might want to focus on something else — and, thankfully, I would argue, it’s much easier to become more energy efficient and decarbonise energy supply than it is to convince the world to cut its GDP in half.

But if we accept that our best levers to pull on are energy efficiency and CO2 efficiency of energy production, and we want GDP and population to be able to continue to grow (or at least not shrink too much), we have to reduce these things fast enough to overcome any growth in the economy or population that is going on at the same time if we actually want to reduce emissions. The aim is to decouple economic growth from increased use of fossil fuels and increased greenhouse gas emissions — reversing the broad sweep of human history in nations across the world, where until very recently, the increased use of fossil fuels has driven a lot of economic growth. Luckily, we increasingly have cheaper and cleaner alternatives.

There’s a very real risk that the world, in trying to balance economic growth and our obligation to cut emissions and stop climate change, acts too slowly and lets this stalemate carry on, as we’ve seen in recent years, where clean technologies are deployed too slowly to overcome this economic growth.

And, indeed, this leads to some of the criticisms of the Kaya decomposition as it is — by centring things like GDP, you can argue that our focus is being put in the wrong place.

Because obviously, we are simply choosing to break down the processes that lead to emissions in this way to reveal these “drivers”, but we could have chosen something else. We have tried to choose quantities — like economic growth, population, energy consumption, and the emissions when energy is used — that we reasonably expect to be drivers of changes to CO2. But we could of course have picked some other set of factors.

Maybe a better decomposition would be to focus on something harder to measure, but more fundamental — human happiness, the progress of human society etc — something more aligned with our values than raw, brutal economic growth which misses out a lot of subtlety. Then we would be rewarding ourselves for producing happier societies with less use of CO2, as opposed to larger economies with less use of CO2 — and there’s an argument that this would be substantially better. There are plenty who would argue that the whole effort to drive down CO2 emissions at the rate we need to would be substantially easier if we were willing to contemplate GDP standing still, or even declining (degrowth), and not prioritise growth of the economy at all costs. I have a lot of sympathy for this view, but given we’re in the econometric world of kaya identities and decompositions, I’ll save it for another episode.

You can of course criticise several aspects here. For example — the four “drivers” here aren’t necessarily independent of each other. Predominantly service-based economies, with a lower energy intensity of GDP, also have higher GDP per capita.

And naturally, there are aspects that are concealed within each driver. Take, for example, the energy intensity of GDP. It could be that a society’s overall energy use is declining as its economy changes; it could be that energy-efficiency measures are leading to less wasted energy; or it could simply be that some sectors of the economy are growing more quickly than others. So the drivers of the drivers are of course concealed by a simple Kaya analysis. You can’t determine which sectors are the largest contributors to the emissions, for example.

You also can’t necessarily extrapolate trends in a Kaya decomposition. For example, the UK’s 38% decline in CO2 emissions has been largely driven by a decline in the power-generation sector as we switched from coal to wind and natural gas. Analysis by Carbon Brief suggested that around a third of the decline was due to a cleaner system for generating electricity, a third was reduced fuel consumption by businesses and industries, and 20% was down to reduced electricity consumption and energy efficiency measures. But only 7% was due to changes in the transport sector. So it’s arguable that, with the “easy reductions” in the electricity sector largely achieved, the harder reductions in transport and agriculture remain. In other words, we can’t simply extrapolate the CO2 intensity improvements in the UK into the future — you need to know where they come from, as they are not uniform across sectors.

Some of this type of thinking has been important in terms of climate geopolitics, too. In the Paris Agreement, every party determines its own target — called an NDC. Not all of these are specified in terms of emissions cuts. For example, India’s target is to reduce the “carbon intensity” of its GDP by a third by 2030. In other words, with an economy growing as quickly as India’s, its actual emissions could easily be higher in 2030 and still meet this pledge. But since its economy would be decoupling from fossil fuels, it still counts as progress in decarbonising. And this is perhaps a slightly more equitable way of dividing up responsibilities than insisting on emissions cuts from every country.

Finally, I would point you to one paper that was quite interesting that came out recently — called “Less Than 2 °C Warming by 2100 Unlikely” which you can read for free online. Essentially, this paper looks at countries of the world through a Kaya decomposition — analysing trends in energy efficiency, GDP, population, and carbon intensity etc. to try and determine what might happen if current trends continue. So this is something like an attempt to construct a climate change scenario, but assuming that current trends continue. What they find is that, in general, GDP per capita is expected to climb by 1.8% a year while carbon intensity declines by 1.9% a year. In other words, if current trends continue, these effects can be expected to cancel each other out for the next few decades, leading to emissions that flatline or rise slowly throughout the century. This would easily blast us through the 2C Paris target, which requires emissions to start falling rapidly, and leads — according to the authors — to an average temperature increase of 3.2C by the end of the century. [They do this with some Bayesian modelling to try and estimate various different possibilities so there’s actually a “probability range” of temperatures, based on current trends.]
In other words, the message from this kind of decomposition is the same — to get to anything like the Paris agreement goal of 2C, we need radical change and rapidly accelerating decarbonisation compared to the trends of the last few decades. Thanks for listening etc.