Worst-case climate scenarios may be off the table, but that’s not because climate action succeeded


Global warming is usually defined as an increase or deviation from a “preindustrial” baseline temperature, normally centered around 1880 or so. Currently the temperature anomaly for air at the Earth’s surface, which is to say the increase in said temperature since the baseline period, is about 1.1ºC.

While no one knows how much warming will take place from now until the end of the century, there is a growing realization (you might even call it a consensus) that it will be less than the 4ºC or 5ºC that many scenarios projected until recently. And some climate researchers are claiming that this averted warming stems, essentially, from successful climate action. Put it other way, they’re saying that emissions now and in the recent past are and have been significantly lower than they would have been if the world had done nothing about climate change, and these lower emissions are the cause of the scaled-back warming forecasts. This article uses figures of global economic output and CO2 emissions to show that claim is inaccurate.

The article is organized as follows:

1) Definitions of some of the terms used, as well as links to the data used in the article

2) Links to the specific statements by researchers. I’ve focused on some scientists who have made this claim explicitly and recently, but there are surely many more (to say nothing of the non-researchers making similar claims).

3) A demonstration that this claim is wrong so far: the evolution of CO2 emissions since ‘climate action’ started, around year 2000, does not deviate from what one would have expected under a ‘do nothing’ scenario. This does not mean that all climate policies are ineffective, but it does rule out any significant net effect.

4) Some reflections on what climate policies may achieve until the end of this century, since that is the timeframe usually referred to in those scenarios.

5) An appendix illustrating Simpson’s paradox. This is a statistical phenomenon that could have hidden a successful climate action at the global scale; in other words, it’s possible for countries to independently succeed at avoiding emissions without this effect showing up in the world stats. Though this didn’t actually happen, it’s an issue to consider when looking at global trends.

1: Data and definitions

The measure of greenhouse gas emissions I’m using is CO2 emissions from combustion (i.e. fossil fuel burning), with figures from the BP Statistical Review of World Energy 2019. The reasons for excluding other greenhouse gas emissions are:

· Fossil fuel use is very well tracked globally. Different sources agree approximately as to how much coal and hydrocarbons we’re burning. Since the carbon content of each of those fuels is more or less known, it’s easy to derive CO2 emissions (BP already does that for us). There is far less certainty as to how much methane is released from rice paddies and landfills, how much nitrous oxide comes from fertilized soils or thawing permafrost, etc.

· In any case, CO2 accounts for the bulk of radiative forcing caused by man-made greenhouse gases, which is to say these gases’ warming effect. In recent years its share of said radiative forcing has been about 80% (see Table 2 of NOAA’s greenhouse gas index page — though note that it doesn’t include ozone, so CO2’s actual share is a bit lower than it states).

There are also some CO2 emissions from land use (agriculture, forestry, etc), which are more difficult to quantify, and minor emissions from chemical processes such as cement manufacturing. The article would be better if I had a better grasp on how to account for those, but I believe the conclusions are clear enough with the numbers available.

In any case, precisely because combustion is the best-quantified source of greenhouse gases, it’s also the most actionable. If you want to reduce your CO2 emissions, you burn less fossil fuel, and this translates directly into reduced concentrations of said gas (versus the concentrations that would have taken place otherwise). It’s by contrast far more difficult to track the exact emissions impact from changes in land use, livestock rearing and so on. If there is one metric which can track the success or failure of global climate action, it’s fossil fuel consumption.

Now, in this article I don’t look only at fossil fuel use. I also check global GDP figures, in constant (i.e. inflation-adjusted) dollars, from the World Bank. The BP and World Bank spreadsheets, along with an Excel file containing my calculations and charts, are available in this Google Drive folder.

In the article I also use the terms ‘decarbonization’ and ‘mitigation’. I’ll define those in section 3.

2: The claim: extreme warming is now unlikely and climate policies are to thank for that avoided warming

I’ll point first to this article by Zeke Hausfather and Justin Ritchie. It makes the argument that the new ‘business as usual’ scenario (i.e. continuing existing climate policies) would most likely lead to a temperature anomaly of 3ºC by the end of this century, rather than the 4º or 5ºC envisioned in some previous scenarios. The article does not say that this change in forecasts is exclusively the result of successful climate action, but all the reasons it cites point to that cause. Here’s a paragraph near the beginning of the article:

“Our business-as-usual projection of 3C of warming — rather than 4 or 5C — is a testament to the progress in global decarbonization over the last few decades. It also reflects the fact that rapid growth in coal use during the 2000s was not necessarily characteristic of longer-term energy use trends. The world has taken concrete steps to move away from coal in the past decade, and this progress should be reflected in our assessment of likely emissions pathways — and their resulting climate impacts — going forward.”

The article also states that this progress could slow down, resulting in 4ºC rather than 3ºC warming by the end of the century.

“For example, a SSP3-style world of resurgent nationalism, isolationism, and conflict with high population growth, inequality, and low rates of technological development could result in a modest return to coal and warming of around 4C by 2100.”

The message is clear. A warming of ‘only’ 3ºC depends on countries continuing their current policies and trends; if these were weakened or reversed, we might see 4ºC instead.

Hausfather is also the lead author of a recent paper that evaluates the performance of old climate models. The major argument of the article is that past projections of global warming often assumed concentrations of greenhouse gases different from what actually happened, and one has to take this into account when assessing how accurate or inaccurate those projections were.

And how does the paper explain that concentrations turned out to be lower than expected by the modellers? It chalks this up to CO2 emissions being different than forecast. From page 2:

“The best physics-based model will still be inaccurate if it is driven by future changes in emissions that differ from reality.”

And page 3:

“While climate models should be evaluated based on the accuracy of model physics formulations, climate modelers cannot be expected to accurately project future emissions and associated changes in external forcings, which depend on human behavior, technological change, and economic and population growth.”

The problem is that the figures in the Supplementary Information of the paper, S3 and especially S4, refer only to concentrations of greenhouse gases (in S4 only CO2); no figure either in the paper or the Supplementary Information actually shows emissions.

This isn’t just nitpicking: concentrations depend on emissions and the fraction of these that remains in the atmosphere (called — you guessed it — airborn fraction). And emissions depend on human behavior, but the airborn fraction does not. If the airborn fraction is lower in reality than was expected by climate modellers, then concentrations will also be lower than forecast — even if emissions are in line with or even higher than expectations.

Delving deeply into the airborne fraction issue would require a major detour for this article, so I won’t belabor this point except to note that the authors provide no evidence on CO2 emissions, and do not mention the airborn fraction at all.

One of the co-authors of the aforementioned paper, Gavin Schmidt, claimed something similar on Twitter:

Okay, he doesn’t say that this ‘successful campaign to avoid worst case scenarios’ involved climate change and CO2 emissions. But what else could he be talking about?

3: The data: climate policies have had no effect on global emissions so far

Here’s when we have to define two terms: decarbonization and mitigation. To some, decarbonization implies reducing CO2 emissions. I use a different definition: reducing CO2 emissions per unit of GDP. This is the same as increasing GDP per unit of CO2.

The reason for using this definition is that, if we’re looking at the effect of climate policies (or changes in human behavior more broadly) on CO2 emissions, then there are two things that we don’t want to mix up:

· Emissions that were avoided because people bought more efficient cars, switched to LED lighting, etc.

· Emissions that were ‘avoided’ simply because the economy happened to grow less than expected.

Looking at the CO2 intensity of GDP means that we only count the first kind of emission ‘reductions’. And I use quotation marks around that word because these avoided emissions need not result in a reduction in absolute emissions. Just like collapsing emissions do not imply that climate action was successful (e.g. Syria, Venezuela), emissions that happened to increase over time do not imply that climate action was a failure. It may simply be that economic growth was faster than expected.

Think of the economy as a motorbike. A motorbike’s economic ‘output’, or the utility it provides to consumers, is the distance it covers; the input is fuel consumed. By definition, going a greater distance with a given quantity of fuel means the motorbike has become more efficient. Likewise, the world economy outputs a series of products and services, for which it uses inputs (fossil fuels among them). Since fossil fuel combustion necessarily results in CO2 emissions (and only a rounding error of those are captured and stored artificially), then you can consider those CO2 emissions as an input to the global economy. The amount of fuel consumed per unit of distance is the motorbike’s fuel efficiency, and the amount of CO2 released per unit of GDP is the economy’s CO2 efficiency.

So for each ton of CO2 emitted the world economy ‘produces’ a given quantity of GDP. If said quantity increases, as it normally does, then by this article’s definition the economy is decarbonizing. Thus, for the article, ‘decarbonization’ does not mean or imply a reduction in carbon emissions. It simply means the economy is getting more GDP out of each unit (ton) of CO2 emitted. The decarbonization rate is the rate of said improvement, per year.

What about mitigation? This term is typically used in the climate debate to mean something like ‘reduced emissions’. I dislike ‘reduced’ because, as mentioned before, it implies cuts in absolute emissions — which may not happen and may not actually indicate a successful climate policy even if they occur. Sometimes ‘mitigation’ is used as a synonym of ‘decarbonization’, but that’s problematic too because, well, decarbonization has been happening pretty much since stats became available! Increased CO2 efficiency over time is the rule, not the exception.

So what is mitigation? To me, it’s a conscious effort to accelerate decarbonization. To increase the decarbonization rate. To make emissions lower than they would have been under a normal decarbonization rate, for a given GDP. Mitigation is a more formal term for what, in previous sections of this article as well as the headline, I called called climate action.

Mitigation is supposed to governments’ policy, but conceivably it could also happen due to the action of individuals, corporations, and so on. There are plenty of social and economic trends that are not caused by any policy in particular — they arise because of decisions taken by, well, everybody.

Now let’s get to the math. BP offers numbers on CO2 emissions since 1965, but since we’re interested in the change in emissions per unit of GDP, the first year for which we can see a decarbonization rate is 1966; the last is 2018. As you can see, the decarbonization rate bounces up and down, normally within the 1–2% range, but shows no long-term trend.

If you’re wondering, the 1966–1997 average is 1.28%, while for 1998–2018 it was 0.98%. I don’t consider this decline meaningful, especially because of the elephant in the chart: China. Removing this country changes the numbers quite a bit, and about half the decline in the rate of decarbonization disappears (1.51% yearly decarbonization until 1997, 1.34% thereafter). I chose 1997 as it’s the year the Kyoto Protocol was signed, but the results are similar if you instead look at 2000, 2005, etc.

Now, why exclude China? In the first decade of this century, the country increased coal consumption very quickly; this large increase in CO2 emissions pulled down the global decarbonization rate.

One could argue that part of China’s increased emissions happened because the country became a manufacturing hub for the rest of the world. From that point of view, excluding China overstates the rate of decarbonization for the rest of the world — the ‘decarbonization’ of some countries simply consisted, to a degree, of moving factories somewhere else. Furhermore, since electricity generation in China is far more CO2-intensive than in the countries where this manufacturing formerly took place, this could be considered a ‘recarbonization’ of the world economy.

On the other hand, it’s unclear how much of China’s increased emissions were really ‘caused’ by the offshoring of manufacturing. The country’s production of energy-intensive materials like steel, cement, etc. is also huge, and those materials are overwhelmingly consumed inside China. So it may be ‘unfair’ to criticize the decarbonization efforts of other countries just because they look bad when lumped in with China.

The issue of what is the ‘correct’ form to calculate the decarbonization rate is too complex to be settled in this article, so I prefer simply to display both results. In any case, the rate of decarbonization has continued much as you’d expect if no mitigation at all had happened.

Now, does this mean that all climate policies everywhere are useless? No. There isn’t enough space in this article, but there is tentative evidence that some countries have been successful in raising their decarbonization rate. If these results don’t show up in the global stats it may be because:

· The emissions in said countries are simply too small to make a difference at the global level

· These successful policies are offset by counter-productive policies somewhere else. The most obvious example is Germany, where decarbonization has been slower since Energiewende started (around year 2000) than before.

· Part of this ‘decarbonization’ is simply the offshoring of energy-intensive sectors. Manufacturing is an obvious case, but for instance electricity generation can be also be partly offshored. This was an issue when looking at the whole world minus China, and it’s a much bigger problem if we look at a single country.

And there’s also the relatively quick decarbonization of the electric sector in countries that adopted nuclear energy, like France and Belgium; it really shows up when looking at emissions vs GDP growth for these particular countries. The transition to nuclear mostly happened for reasons unrelated to climate change and before Kyoto, so you may consider it ‘accidental mitigation’.

4: Okay, climate policies have achieved essentially nothing so far. But what about the future?

The article by Hausfather and Ritchie does not clarify exactly how much warming they believe would happen under a do-nothing scenario. They mention that even a ‘do nothing’ scenario that started today would keep current policies in place but, as we’ve seen, policies adopted so far have had virtually no effect on emissions.

Any estimate of how much the world will warm is inherently dependent on a number of assumptions; the problem is not when assumptions are made, but when they go unmentioned. So in this article I’ll lay out my assumptions. The first thing to say is that, rather than how much the world warms, the variable of interest is how much warming can be avoided by mitigation (i.e. by increasing the decarbonization rate).

The warming caused by a given increase in CO2 concentration depends on a metric known as the transient climate response, or TCR; temperature changes of course also depend on other natural factors, but since those are by definition outside humanity’s control I won’t consider them in the calculation. Hausfather’s paper (the peer-reviewed one) precisely deals with this issue: calculating TCR. One weakness is that Hausfather and the other authors estimate TCR using only the years for which a specific model made forecasts, which in my view is wrong; there is evidence that TCR is close to constant over time, so there is no good reason for throwing away most of the years (data) available.

In Hausfather’s paper (see Figure 2) the estimated TCR is about 1.8ºC per doubling of CO2. This is about the same as in the last generation of climate models, CMIP5. However, the correct value of TCR is still somewhat in doubt; a paper from earlier in 2019 calculated instead 1.57ºC, and other papers find values of 1.3ºC or 1.4ºC. On the other hand, by cutting fossil fuel consumption we will also curtail part of the non-CO2 forcing, mostly methane (natural gas is essentially methane — reduce natural gas consumption and you also reduce methane leaks).

So let’s say that, when including the associated methane forcing, a doubling of CO2 concentrations leads to warming of 1.8ºC by the end of the century. Then how much warming can we avoid? The following numbers are just a made-up example so that the reader may have an idea of how to arrive at an estimate. Assume that:

· The do-nothing CO2 concentration by the end of the century is 700 parts per million (ppm)

· The do-something, mitigation scenario leads to 600 ppm

This may seem like a poor mitigation effort, but considering that we’re already at 410 ppm, and that the airborne fraction of CO2 has been rather stable over time even though emissions have multiplied, it means that under one scenario we would emit about 53% more than under the other (because concentrations would increase by 290ppm, rather than 190ppm). For instance, let’s say mitigation raises the rate of decarbonization by 1% versus whatever rate there would have been otherwise. If under the mitigation scenario emissions decline by 0.5% a year, while under the do-nothing scenario they increase by 0.5% a year, then over the following 80-year period cumulative emissions under the do-nothing scenario would indeed be about 50% higher than under mitigation. Which is pretty much as the example above.

So what happens if end-of-the-century concentrations are 600 rather than 700 ppm? That difference is 16.7% in arithmetic terms, but since the forcing of CO2 is approximately logarithmic, in forcing terms it’s equivalent to 22.2% of a doubling (because 1.167 ^ 4.5 = 2 and 1 / 4.5 = 0.222). So, going back to the 1.8ºC we had approximately calculated as the immediate warming that results from a doubling of CO2 levels (and this is a rough calculation because we’re including the effect of methane which does not follow the same rules), we get 0.222 * 1.8º C = 0.4ºC.

To emphasize, this is just an example, but it would indeed be a case of somewhat successful mitigation.

Now the reader may be wondering: isn’t that very little? 0.4ºC? Surely the people talking about climate change discuss 2ºC vs 4ºC and so on? What is wrong the official scenarios?

The answer is that it’s very difficult to tell what’s wrong with the ‘official’ scenarios because they don’t focus on the decarbonization rate as a separate variable. Two scenarios that project different levels of warming do so by assuming different rates of population growth, different rates of economic growth, different technologies and so on. In fact there is evidence that the warming projected by models is inconsistent with climate models’ own TCR (see page 49 of the Lewis & Crok report); different climate scenarios may in fact involve different TCRs! The scenarios differ between them in lots of things, but in the real world there is only one TCR; and, while future population and GDP are unknown, presumably we don’t want to curtail them. So all the things that we do in order to avoid emissions will show up in the decarbonization rate. It’s the only lever we can push.

Pretty much nobody talking about climate change scenarios can tell you, for a given projection:

· What rate of GDP growth is expected as the ‘default’ or ‘business as usual’

· What’s supposed to be the ‘normal’ rate of decarbonization and how much of an improvement over that the scenario represents

· Even finding out what’s the forecasted concentration of CO2 by the end of the century is a challenge

Indeed, if you read the Hausfather and Ritchie article you’ll find lots of references to different scenarios but no references to GDP growth rates, decarbonization rates, or CO2 concentrations. These are the key elements to calculate how much warming can be averted, yet there’s no info on them. I hope the authors do not to take this as a personal attack; they’re doing the best they can with very opaque tools.

My view is that, if you want to know how effective climate policies can be, you have to do the math yourself. Believing in the ‘scenarios’ is an act of faith.

Appendix: An illustration of Simpson’s paradox

Suppose the world is divided into two countries: Westland and Eastland. The fomer has a GDP in Year 1 of $80 billion, versus $20 billion in the latter. CO2 emissions are 60 tons in Westland and 40 tons in Eastland. Since global emissions are 100 tons of CO2 and global GDP is $100 billion, the CO2 efficiency of the world economy is $1 billion per ton of CO2, but this differs between Westland (which generates $1.33 billion per ton) and Eastland ($0.5 billion per ton).

Now suppose both countries increase their GDP by 30% in one year — unrealistic, yes, but exaggerated numbers make the issue clearer. Both countries also decarbonize their economies at a rate of 20% a year. For Westland, this means GDP jumps to $104 billion; since the new CO2 efficiency of GDP is $1.33 * 1.2 = $1.6 billion per ton of CO2, the emissions only increase to 104 billion / 1.6 = 65 tons of CO2. For Eastland, the new numbers are: $26 billion GDP, $0.6 billion per ton of CO2, and 43.33 tons of CO2 emitted.

(Another way to calculate the new emissions, using Eastland as an example: 130 / 120 = 1.083, meaning that with GDP growth of 30% and decarbonization of 20% emissions must grow at a rate of 8.3%. And indeed, 60 * 1.083 = 65)

For the global economy, the Year 2 numbers re:

· GDP: $84 billion + $26 billion = $130 billion

· Emissions: 65 tons + 43.33 tons = 108.33 tons of CO2

· CO2 efficiency: $130 billion / 108.33 tons = $1.2 billion per ton of CO2

This CO2 efficiency is indeed 20% higher than in Year 1. This is what you might expect, because both countries that make up this fictional world had a decarbonization of 20%.

Here comes the problem: the global decarbonization rate will not remain 20% if one of the countries grows faster than the other. The different levels of absolute CO2 efficiency between countries are to blame for that. This is what’s known as Simpson’s paradox: a trend that affects sections of a group may not be evident when looking at the group as a whole.

Suppose that Eastland starts growing much faster than Westland — by 50% a year, rather than 30%. Then what would happen in Year 3? Even though both countries retain the same decarbonization rate of 20%, the global rate falls to about 16%. Furthermore, unless Eastland converges to Westland levels of CO2 efficiency, the global decarbonization rate will always remain lower than 20%; by Year 20 of the simulation Eastland dwarfs Westland in terms of GDP, but decarbonization is still well below the Year 2 level.

In the real world, over the 2000s some coal-heavy Asian economies, using more CO2 per unit of GDP than the rest of the world, significantly increased their share of global GDP. This means that there could have been an increase in the decarbonization rate of each individual country, and yet a decline in the global decarbonization rate. In this article I showed that’s not really what happened, at least when considering China. But one cannot know that until one runs the numbers.