How hot is too hot? The Economics Nobel Prize winning DICE model suggests 3.5oC of global warming this century is optimal. The UN Paris agreement has 2oC warming as the maximum. While there appears to be a large discrepancy here, by updating DICE with the latest climate and social science, we find support for targets at or below 2 oC.

For at least a decade, there has been a clear divide between the UNFCCC negotiations that aim to restrict the increase in the global average mean surface temperature to less than 2°C above the pre-industrial level and the main message from the most influential Cost-Benefit Analysis (CBA) on climate change.  In this analysis, the economic estimates of the damages caused by climate change are weighed against the costs of reducing emissions. One of the two recipients of the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel in 2018, William Nordhaus, has developed a model to structure the debate and make it possible to analyse what is an optimal climate policy and the associated temperature response. In his own calibrations of the model, he has argued that the world community must do something about climate change as, if nothing is done, temperatures will eventually rise by more than 6°C. In Nordhaus’ calibration, the optimum instead implies a maximum warming of 3.5°C in 2100 and 4°C around 2150. On the basis of this, Nordhaus argues for sizeable global carbon taxes (or tradable permits giving a price of around 50 $/tonne CO₂). Yet most governments have failed to take this advice; carbon taxes are pitifully small and many countries even have carbon subsidies instead. While we agree with Nordhaus that higher carbon prices are needed, we have been puzzled that he finds such a high optimal global temperature. In this paper, we have found multiple causes behind this. When we update the DICE model with recent evidence from the climate and social sciences, we find optimal temperatures that are actually in line with the UNFCCC and Paris goals.

The Nobel Prize and the Paris Agreement

On the 8^th October 2018, the IPCC released a special report cataloguing the strategies needed in order to limit the global average surface temperature increase to the 1.5^oC above pre-industrial levels that signatories committed to in the UN Paris Agreement of 2015. Two hours later, the Sveriges Riksbank’s Prize in Economic Sciences in honour of Alfred Nobel, the so-called “Economics Nobel Prize”, was awarded to Professor William Nordhaus for his work on climate change economics, as embodied in the DICE (Dynamic Integrated Climate Economy) model. Together these were two important global statements about the seriousness of climate change as a global problem and the need for global action. Ostensibly, at least, the world seemed to be pulling in the same direction.

Closer inspection, however, revealed an important rift between these two global announcements. While the signatories of the Paris agreement and the IPCC report were organising around the 1.5-2^oC limit, the central case in the Nobel Prize winning DICE model proposed an optimal path, taking into account the costs and benefits over time, that leads to a temperature rise of 3.5^oC in 2100 and 4^oC in the following century. Most climate scientists would agree that this is an important rift since life on earth with 4^oC of warming is expected to involve serious risks to nature and human societies around the world.  

The fact that a single economic model disagrees with an international political decision might seem unsurprising, even trivial, but this would be to underestimate the impact of Nordhaus’ work on international climate policy. The DICE model is central to, inter alia, the US Environmental Protection Agency’s calculations on the Social Cost of Carbon, it is one of the key economic models used in the IPCC assessment reports, and when Lord Stern released his milestone Review in 2007, its conclusions were tested against the DICE model. Indeed, among climate change economists, DICE represents a testing ground for policy advice and theoretical advances. In short DICE has been hugely influential in policy and practice, hence the award of the Nobel Prize.

All the more reason, then, to understand why the ‘DICE-optimal’ climate response leads to a 3.5-4^oC temperature rise, implying that the costs of meeting the Paris targets of 1.5-2^oC are not justified by the future benefits.

Our team, both economists and climate modellers, have been working on various aspects of climate modelling using DICE for many years. In 2019 Thomas Sterner, who knew of the different strands of work that we were all doing, decided that it would be constructive to try to gather everyone together to take a deep dive into these issues. 

An interdisciplinary workshop on climate economics in Gothenburg

Marty Weitzman was visiting Thomas Sterner in Gothenburg during May 2019, in part to give the yearly Tore Browaldh Lecture. Christian Gollier was also in Gothenburg at the time and Thomas realized that this was a good opportunity to attract other climate and economic scientists for a workshop that would concentrate on these issues.

Marty was very enthusiastic and participated in the workshop as a natural authority. Even though he was not directly involved in any DICE modelling, nor in the writing of our paper, he gave feedback and provided his perspective on numerous issues. 

The workshop was the starting point for a very productive collaboration that benefited from each team member’s willingness to learn from each other. During the subsequent months, we jointly implemented updates of both the climate and the economic science that is built into DICE.

Updating the climate science

First, we considered the climate science. It turns out that the 2018 DICE model has some outdated assumptions in relation to how much carbon is absorbed (the carbon cycle) and how much increased concentrations of greenhouse gases in the atmosphere affect temperature (the energy balance). Concerning the carbon cycle, the rate at which carbon is absorbed from the atmosphere, either by the oceans or vegetation, depends on the atmospheric concentration. The share of a pulse emission into the atmosphere that remains at a certain point in time is higher the higher the background concentration. The DICE model is calibrated around the high concentrations anticipated in the future in a business-as-usual world with no interventions to reduce emissions. It is convenient in a model like DICE to have a simple linearization of the full-scale climate models and Nordhaus uses a linearization that is calibrated to be consistent with the 3.5-4 ^oC optimum he was working with. The problem occurs when this simplified model is used to discuss the possibility of reaching a more ambitious target. Nordhaus’ calibration understates how quickly CO₂ is absorbed in a world that tries to meet a 2^oC target, making it somewhat easier to reach such a target in practice than his work implies. For that reason, we introduced a non-linear carbon cycle that better reflects the carbon uptake at a much broader range of CO₂ concentrations.  Concerning the energy balance, in the current DICE model the temperature responds more slowly at first, and then, after some fifty years, more rapidly to higher atmospheric CO₂ concentrations than is implied by the latest evidence from climate science. These two updates alone reduce the optimal temperature by around 0.5^oC.

Updating the social science

In relation to the social science, the key updates concern the magnitude of climate damages and the weight placed on future generations’ well-being: the social discount rate.

We use Howard and Sterner (2017) to update the damage function. This study undertakes a detailed meta-analysis of the whole literature, starting by setting up criteria for inclusion or exclusion of studies. The studies are systematically weighted to reflect a number of factors including multiple publication bias; the fact that many of them do not provide independent information, and hence studies by the same teams were weighed too heavily before. Howard and Sterner also address heteroskedasticity, methodological controls, and a number of other issues. This leads to a damage function that predicts around 6.7% loss in GDP as a result of 3^oC of warming, rather than the 2.1% GDP loss assumed in DICE. We have had extensive discussions with William Nordhaus on the damage function and he used a version of Howard and Sterner’s damage function in his Nobel lecture.  He agrees that using this damage function lowers the optimal temperature by around 1^oC.

On intergenerational fairness, a recent study by Drupp et al. (2018) revealed the extent of disagreement on the social discount rate for long-term welfare analysis. We update the discounting assumptions in DICE by using the median opinion among 173 experts surveyed, rather than relying on Nordhaus’ position alone, which, it turns out, is not a centrally held view.

The updates to damages and the social discount rate lower the DICE-optimal temperature increase by an additional 1^oC, to below 2^oC. Further adjustments to DICE are also made, including bringing forward the time at which Negative Emissions Technologies (NETS) are available from 2150 to 2050. Such adjustments, when introduced after the changes already made, do not reduce the optimal temperature greatly, but rather, in the case of NETs, relax the speed at which emissions reductions need to happen. This increases welfare. We also adjust the exogenous assumptions in DICE concerning non-CO₂ greenhouse gasses and other radiative forcers.

Singing from the same hymn sheet on climate change

In our paper, entitled simply “Climate economics support for the UN climate targets”, we find that the discrepancy between the DICE model’s findings and the UN climate targets can be reconciled by changing both the climate and social science parameterizations of the DICE model. Some of our updates can be described as corrections to better represent the current state of knowledge (the carbon cycle and temperature response), while others are ethical choices (the social discount rate) and/or newer results from meta-analysis. With these modifications, the optimal balance of costs and benefits in DICE puts us on a path that limits warming to between 1.5^oC and 2^oC. The findings hold in the simplest incarnation of DICE, without consideration of many other factors, including uncertainty and climate tipping points, which would typically strengthen the arguments for immediate and strong action to limit climate change.

What our paper does not say, however, is that this new optimal will be easy to achieve. The fact that aggregate welfare would increase will not reduce the resistance from vested interests including some major coal and oil lobbyists. There are still enormous difficulties when it comes to political implementation and coordination between countries. With higher damages and more weight placed on future generations, the Social Cost of Carbon, and hence the optimal carbon tax, rises quickly to 100 US dollars per ton of CO₂ and more.   The first steps – and here we are in complete agreement with William Nordhaus - will be to raise the price of carbon emissions and other relevant gasses in all contexts.

Article: Martin C. Hänsel, Moritz A. Drupp, Daniel J.A. Johansson, Frikk Nesje, Christian Azar, Mark C. Freeman, Ben Groom, Thomas Sterner: "Climate economics support for the UN Climate targets". Nature Climate Change (2020). DOI: [10.1038/s41558-020-0833-x]

Link to the article: https://rdcu.be/b5AvD