Costs of delaying action on climate change

This post was prompted by a silly twitter argument, about which probably the least said the better. Someone who has set themselves up as some sort of “climate communicator” had asserted that if we don’t halve our emissions in 12 years then the world as we know it will end. Moreover, anyone who even thought this assertion was controversial was, in their eyes, a denier. Well, I thought it was not so much controversial as simply false. But I did wonder, what is the actual effect of delaying decarbonisation of the global economy? In the sense of, let’s hypothesise that we actually can take policy action that decreases carbon emissions, what difference does it make when we start?

I’m sure people must have done (and published) these sort of calcs but to be honest I don’t recall seeing them. Most of the research I’ve seen seems to be more along the lines of: if we delay action then how much more stringent will it have to be, in order to meet a particular target? This pic below shows that sort of thing:

screenshot 2019-01-27 09.42.12

I don’t think this sort of thing is really all that helpful as it gives no clue as to how realistic any of the pathways are. It seems that this sort of graph is basically motivated by a political assertion (“let’s not let warming exceed X degrees|”) rather than any plausible understanding of the world we live in. I also don’t think it is very realistic to think that the world will design and implement carbon emissions policies that credibly aim at a particular max temperature change, at least not within my lifetime. So, here’s an alternative question that although still rather simplistic is (IMO) more directly relevant to the real world. Let’s assume we are able decarbonise at some given rate. How much difference does it make how soon we start?

To answer this, we have to model (a) CO2 emissions and how they vary with policy delays (b) how atmospheric CO2 concentrations vary with emissions (c) how climate change depends on CO2 concentrations, and finally perhaps (d) the economic impacts of climate change.

For (a), I assume an exponential growth rate for historical and future emissions up to the initiation of decarbonisation, followed by an exponential decline. I use a historical growth rate of 1.9% in these calcs. For decarbonisation, I assume a rate of 2% which would halve our emissions in 35 years. This is less than half the rate that would be required to halve emissions in 12 years as hypothesised earlier. Atmospheric CO2 concentrations are then provided by the equation of Myhrvold and Caldeira (2012). I could have used real historical emissions for the historical period of my simulation, but actually I get a marginally closer fit to historical CO2 concs when just using the exponential growth with my chosen rate. Three decarbonisation dates tested are 2020, 2030 and 2070. Ie starting now(ish), or alternatively after a delay of 10 or 50 years.

screenshot 2019-01-27 13.31.02

Current CO2 concentration is about 410ppm, increasing by 2.5ppm per year. I didn’t bother distinguishing or labelling the three lines on each graph as it’s obvious which relates to which scenario. I have marked the date at which decarbonisation starts, so you can see how the concentration increases for quite a while after we start to cut emissions.

The resulting climate change is modelled by the two-layer model of Winton, Takahashi and Held (2010) discussed in several papers by Held, Winton and others (2010 ish). Parameter values can be changed in this model, but the only one that really matters here is the equilibrium climate sensitivity (ECS) to a doubling of CO2. I use historical non-CO2 forcings for the historical era and just hold these fixed at their current values indefinitely into the future. The model simulation matches historical data pretty reasonably as shown. The max temp rises (up to the year 2350) for the three scenarios are shown on the graph, ie you get a 0.25C increase in max temp for a 10 year delay, and 1.6C for 50 years. In other words, each year of delay initially leads to an increase in ultimate warming of about 0.025C , and this number rises steadily to around 0.04C per year in the middle of the century. The differences in temperature seen by 2100 are a little less than this, eg just under 0.2C difference for the 10 year delay.

screenshot 2019-01-27 13.31.36

Raising the sensitivity of the model increases the ultimate temperature rise of course, and also increases the difference between the scenarios. For a sensitivity of 5C (hard to reconcile with what we believe) the 10y delay leads to an additional ultimate warming of almost 0.4C, though in this case significant warming is continuing beyond the end of the simulations in 2350 and the long-term differences will also grow gradually beyond this time. For sensitivity of 2C, the decadal delay leads to an ultimate difference of just under 0.2C, and is only 0.15C at 2100.

So this is the cost, in climate terms, of delaying decarbonisation. I don’t think the underlying assumptions are unreasonable, though no doubt some could be changed. The growth rate of emissions at 1.9% per year is probably debatable but (when fed through the Myhrvold and Caldeira equation) gives reasonable historical results. My decarbonisation rate is a guess, but results are not very sensitive to this. Eg if we can achieve 5% decarbonisation rate, then the cost of delay is reduced slightly to just under 0.2C over a decade. Note that the starting point for this post was an assumption (assertion?) that we can decarbonise at 5% per year, otherwise the world is going to end anyway.

Evaluating the economic impact of the warming may be the most contentious part. Here I’ve just used an estimate based on a version of the (Nobel-winning) Nordhaus DICE model, which I also used in this paper. Other estimates are available. I’ve also used a simple 2% per annum growth rate which some may disagree with, especially when extrapolated out to 2350. But what else should I have done?

screenshot 2019-01-27 13.31.19

There are indeed three lines on this graph, but they aren’t very clearly separated. The 2%(ish) cost of modest climate change just isn’t very visible against the background of several orders of magnitude of economic growth. And to be clear, I don’t think that everything can be readily boiled down to money – recent events show, many millions are apparently willing to squander untold billions (of other peoples’ money, of course) on the hypothetical benefits of “sovereignty”. Yes, I’m talking about brexit, the costs of which will undoubtedly dwarf any plausible impact of climate change on the UK, for many decades to come. But even if we aren’t trying to maximise economic benefit, it’s still an interesting context for the impact of climate change policies.

At this point I will refrain from making any more rhetorical flourishes but will instead leave the reader to decide whether this analysis indicates an end to the world.

8 thoughts on “Costs of delaying action on climate change

  1. Have you assumed that decarbonisation is free? I think so. If so, it would be interesting to add it; perhaps at an exponentially decreasing cost, reflecting the cost of solar panels. If you do that, and the reverse (exponentially increasing into the past) how far back do you have to start decarbonisation so that it bankrupts the world?

    • Yes I’m not making any assumptions about the cost of decarbonisation, and wouldn’t really know where to start with that anyway. It was really the climate impact of delayed action I was interested in, the economic impact is a trivial after-thought. I bet solar panels were expensive back in 1850 though 🙂

  2. I think your analysis is fine James until you get to the economics. You’ve got the GDP growth rate as exogenous. For a good overview of why the damage functions that have gone into these IAMs to date are pretty useless see Pindyck 2013 (another economist in line for a Nobel Prize):

    https://www.nber.org/papers/w19244

    And take a look at the new work emerging on empirical-based damage functions such as that by Marshall Burke at Stanford:

    https://web.stanford.edu/~mburke/climate/BurkeHsiangMiguel2015.pdf

    And a general overview by Thomas Stoerk et al on what the IPCC should do to improve the economics side of its analysis

    https://academic.oup.com/reep/article/12/2/371/5025082?searchresult=1

    And have you noticed how William Nordhaus himself has got distinctly more gloomy? This 2018 paper pushing the need to go faster mitigation-wise when faced with uncertainty. You didn’t see that in his writing 10 years ago (and that’s even using the barking-mad damage function that he uses in that paper).

    https://www.aeaweb.org/articles?id=10.1257/pol.20170046

  3. Well……I don’t really want to get into a fight between different groups of economists, the financial bit was only intended as a bit of an afterthought. I didn’t even try to put in a discount rate 🙂

    • Well, the title is “Costs of delaying action on climate change”. I would have thought the use of the word “costs” suggests that economics is front and centre? 🙂

      Anyway, without being facetious, how can you make statements comparing the cost of Brexit and the cost of climate change for decades to come if you don’t have a view on the damage function? Have can you have any opinion about climate change policy if you don’t have an opinion on costs and therefore the damage function?

      Following from that, I also presume that you view Greta and the kids and Extinction Rebellion as misguided? But if you take a tail argument (it doesn’t have to be particularly fat tail one), their action appears justified purely on a risk of ruin argument (from insurance) or value at risk argument (from finance). Here is Oliver Bettis from the Institute and Faculty of Actuaries laying out the risk of ruin approach from the insurance industry (building on Marty Weitzman’s work) :

      http://www.lse.ac.uk/GranthamInstitute/wp-content/uploads/2014/01/Oliver-Bettis-Risk-Management-and-Climate-Change-Risk-of-Ruin.pdf

      I know from our talks back in Japan that one of your main themes was that you saw a not particularly fat tail at the upper end of the climate sensitivity distribution (5% chance of over 4.5 degrees of warming as opposed to a lot higher?). But even if you combine that well-behaved tail assumption with the fact that economists haven’t got a clue what will happen to GDP over 2 degrees of warming (but with an emerging consensus that whatever will happen could be really bad with even Nordhaus coming around to that view – note that he is perfectly aware of the limitations of his own damage function when we get above 2 degrees of warming), then how can any sensible person not be out on the streets with Greta, the kids, and XR?

      I’m not being sarky. Just can’t see any logic that would combine a 5% change of a 4 degree-plus sensitivity tail plus an economic impact tail of completely unknown fatness and not view this as a shocking risk combo that dwarfs Brexit by orders of magnitude.

      Yup, I know you used the term “many decades to come”, but a current Greta-aged 16-year old will likely have children still alive in the year 2150, so the risk of catastrophic-climate change impacting their future kids will, I think, be just as pressing a concern to them as Brexit, whose downside risk is 6-9 percentage points of GDP (speaking as a Remainer who has been on all the People’s Vote marches, those ‘level’ numbers are 15 years out, Treasury and BOE).

      Anyway, hope you and Jules are doing well and have settled back into UK life well after Japan. All the best (but still curious why you appear unconcerned about this horrible tail risk profile).

  4. Thanks for the link to the interesting paper. I certainly have some sympathy for the arguments presented though I’m not really qualified to judge. However in their “destruction testing” of the model they are using emissions (and thus temp changes) way in excess of anything I’ve written here as well as using parameter values that seem pretty excessive. Regardless of arguments about rate of warming and adaptation, I can’t see any plausible argument as to why the long-term (century and longer) growth rate should happen to be optimised at the pre-industrial temperature.

    Regarding Greta et al, coincidentally I was in Stockholm recently and took the opportunity to talk a little about her in my presentation (which was mostly “pure” science). It would be fair to say that both her and Prof Rockstrom’s more exaggerated claims didn’t attract much support from the audience. Tempting to turn a blind eye to activists pushing the Overton Window but…well…it’s a difficult judgement to make as to how much overt support they should be given. When we haven’t decarbonised in 5 years or 10 years or whatever….then what?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s