The much-leaked report of the Intergovernmental Panel on Climate Change will be released today (this evening our time). As everyone knows, it will conclude that it is very likely (more than 90 per cent) that human activity is the main cause of observed global warming. The IPCC best estimate of the impact of business as usual is an increase of around 3 degrees C relative to preindustrial levels by 2100.

There’s still room for debate over the central estimate, particularly regarding the projections of CO2 emissions. For example, the population growth projections used in the estimate are probably too high. On the other hand, I doubt that some feedback effects like bushfires have been fully taken into account. And the climate models themselves are still being refined. Still, it doesn’t seem likely that the estimates are going to be changed much by another 5 years of data and improved modelling, so we can probably take this estimate as settled for the moment.

In many ways, though, the real interest now is in the tails of the distribution. Most of the attention so far has been focused on the lower tail, representing the possibility that global warming will turn out to be modest or non-existent. Denialist arguments that the whole idea of anthropogenic global warming is wrong have received a lot of attention but have been thoroughly refuted by now. On the other hand, if all the questions now in doubt turned out the right way for us (forcings at the low end of estimates, feedbacks from water vapour and so on less positive, historical trends at the low end of the margin of error) and some factors we haven’t yet considered turned out to reduce warming, we could see lower numbers, small enough to make adaption rather than large-scale mitigation the best response (in hindsight).

But all these arguments are symmetrical. If warming could be slower than the best projections, it could also be faster. According to the ABC, the report says gains of up to 6.3 degrees can’t be ruled out, though this figure is from only one model, and does’t fit well with other data.

Although the probability distribution of possible outcomes is essentially symmetrical, however, this doesn’t mean that we can ignore the tails. This is because the costs of climate change grow much more than linearly with the rate of change. Let’s say, for illustrative purposes, that 3 degrees of warming would impose costs equivalent to a 5 per cent reduction in income. Taking account of a small probability (less than 10 per cent) that warming will be very modest, so that costs are zero, doesn’t change the analysis much. By contrast, 6 degrees of warming would be catastrophic – it’s equal to the change since the last Ice Age. Even a small probability of 6 degrees of warming (say 5 per cent)** would greatly increase the expected cost of warming. For plausible levels of risk aversion, the expected cost could be as much as 10 per cent of GDP. So the cost of warming is greatly affected by the probability of these extreme events, and this is an issue that is still unresolved.

The general point here has made a number of times already,notably by Tyler Cowen and Brad DeLong . Uncertainty about the impact of climate change strengthens the case for action to mitigate it.

** This number isn’t really crucial to the argument, but the lower the number the greater the relative importance of the low-probability tail events. A lot of published estimates, like those of Nordhaus and Tol are lower than this, but (as noted in the previous posts) these estimates impute trivial costs to the severe ecological damage that would inevitably arise with 3 degrees of warming.