The paper presents the deep uncertainty that stems from the fact that climate science has not been able to give a precise number for climate sensitivity. Simply put, this is the amount of warming we would get through a doubling of CO2 concentrations in the atmosphere. This amount has been estimated to be around 3 degrees, ranging from 1.5 to 4.5 deg C.
That range has persisted almost all of the past 35 years. That consistently wide range is the first and one of the two most significant uncertainties linked to climate sensitivity. Charney et al. (1979) first established the ’likely’ range of climate sensitivity that remains in use today, despite dozens of efforts to refine it.
The problem with the wide range is the vastly different social and economic consequences. Wagner and Zeckhauser are economists and focus on economic aspects. I will have something to say from a sociological perspective later on.
First the economic argument. Basically, the higher the temperature rise, the more costly the consequences: 'Marginal damages from rising temperatures increase rapidly.' However, to achieve a higher temperature rise will take longer, so during that time society will grow richer. An additional problem are 'tipping points' which could occur but their timing and size which are uncertain. As the authors put it,
Perfectly constructed expected damages curves take account of tipping points. Thus, any jumps in damages get accounted for by the slope of the certainty equivalent curve. In practice, computing such a curve is exceedingly difficult to do.
The IPCC has always stuck to the 1.5-4.5 range, with one exception. Between Assessment Reports Four and Five the lower bound of the ‘likely’ range was lowered.
In particular, in 2013, the IPCC lowered the lower bound from 2°C in 2007 back down to 1.5°C, where it had been since 1979. Superficially, this looks like unambiguous good news; part of the distribution had shifted downward. That indeed would be unambiguous good news if the distribution outside the likely range remained the same. However, this expansion of the likely region reflected greater uncertainty.And:
Had global warming turned out to be less severe than previously thought, say if the whole distribution or even some portion of the distribution shifted downward, what would be cause for celebration. But merely reducing the bottom value of the likely distribution in the IPCC report hardly represents such a shift. Rather, it also spreads the overall distribution. Thus, it tells us about the current capabilities of climate science—notably the current understanding of the climate sensitivity parameter—and indicates that the uncertainties are greater than we thought. That alone is disturbing, since greater variability indicates greater expected cost. The news is, thus, bittersweet, a probable reduction in the mean in exchange for greater variability. And the surprise on uncertainty is even more disturbing, since the relationship between carbon dioxide emissions and global temperatures is perhaps the most studied relationship in the climate debate. Uncertainties elsewhere may be even greater.
The authors then refer to Knight's definition of risk and uncertainty and argue that in climate change we are dealing with uncertainty (where we do not know probabilities, unlike in risk), even with ignorance (we don't know the likely future system states): 'It is the realm of unknown unknowns. That is the realm where we are with climate sensitivity and, hence, long-run climate change projections in general.'
I find this paper remarkable for two reasons. First, it does not pretend that climate policy must be based on 'virtually certain' science. It explores the deep uncertainties that exist in our knowledge, despite the fact that climate science has established some basics about detection and attribution. Second, the deep uncertainties resemble problems in social policy that are familiar to social scientists, to policy makers, and to the public at large. The analogy I made in my Nature comment is thus worth re-examining.
Remember, my argument was that climate change resembles economic policy, health policy or education policy much more than it resembles ozone policy. Take unemployment as an example. This is seen as undesirable but there is deep uncertainty about what rates of unemployment will have hugely undesirable consequences. Social scientists have studied historical examples (like: Hitler's rise to power was based on surging unemployment) and therefore we need to stay below a certain unemployment rate to avoid a re-run of history (a more nuanced version of this can be found in a recent paper that argued a link between financial crises and the rise of right wing parties). You could develop scenarios for other political surprises, like the outbreak of left-wing insurgency, or of a descent into chaotic anarchy.
Likewise, climate scientists use historical analogues to establish 'tipping points' (of the sort: 'the last time we saw CO2 concentrations at these levels the planet was in such and such a state...Therefore we must do something about it urgently'). These are all more or less plausible conjectures, resting on an array of assumptions (most prominently the ceteris paribus -- but has the Earth/Society not changed in important aspects since 'then'? Can we be sure it has not? Can we be sure it has?)
Wagner and Zeckhauser argue that more uncertainty calls for more prudence and I agree with them. But the problem is that many other issues than climate change compete for the attention of decision makers, and the wider public. Most burning problems of today have to do with political violence, war, poverty, ill health. Climate activists like to proclaim climate change as the greatest problem facing mankind. But how do we know, and how could we measure and compare this?