One of the central problems of a majority of economic integrated assessment models of climate change is that they mostly ignore or at best underplay all the uncertainty around the problem they try to analyse. Meanwhile, these models are mostly presented in calibrated form, i.e. including concrete values for different variables – just what a cost-benefit analysis is expected to do. Even if there are some confidence intervals specified, they are mostly rather crude and are a kind of sherry-picking of “meaningful” uncertainties. However, the dynamics of the climate itself and the consequences of his changing are profoundly uncertain and full of feedback mechanisms. In the following I would like to present and discuss some major sources of uncertainty in economic climate modelling, arguing that it is even deeper than in “just” climate modelling and that the consequences for standard approaches are profound.
The uncertainty climate economists have to deal with in their modelling efforts starts in climate science. This uncertainty is mainly expressed in the notion of climate sensitivity, i.e. the warming response of the climate due to a certain increase in atmospheric concentrations of greenhouse gases when all feedback mechanisms have been taken into account. Climate sensitivity is mostly specified as the expected rise in the equilibrium global mean temperature after a doubling of atmospheric carbon concentrations. Already in the last IPCC Assessment Report the confidence intervals for this magnitude were rather large (with 2 to 4.5°C considered likely). Since then, even higher values have been found in research – e.g., Hansen et al. (2008) claim that, according to their analysis of paleoclimatic data, climate sensitivity may be as high as 6°C. There are many reasons why this number is so uncertain, even though we know with near-certainty the climate sensitivity without any feedback mechanisms (about 1.5°C). The most important feedback mechanisms are: the interactions between greenhouse gases and (cooling) aerosols; changes in Earth’s albedo; releases of methane from permafrost and oceans; changes in the ability of oceans, soils and biosphere to absorb CO2 due to temperature rise; cloud building etc.
However, there are many more sources of uncertainty economists have to deal with (and should take into account) than only those coming from the narrowly understood climate science. First of all, there is uncertainty about how ecosystems and cycles are likely to respond to particular changes in global mean temperature: how fast are glaciers likely to melt and thus to contribute to sea level rise (and albedo change)? what will be the response of important ocean circulations (above all, the Thermohaline Circulation, of which the Gulf Stream is a part)? how much warming can tropical forests endure without collapsing? how will species adapt and which ones cannot and will go extinct? what about desertification? how exactly will precipitation patterns change? and the intensity and frequency of extreme weather events? All these questions are highly important for climate economics since human livelihoods and the entire human economy is dependent on various ecosystem services, and so they will be forced to adapt to the changing environment.
A very important aspect of the problems mentioned above are the uncertain (and often just unknown) interdependencies between ecosystems (and cycles). Nature is a highly complex system that we don’t understand properly. Ecosystems can interact in various ways and we are not able to predict which interactions are important and even what they are. But it would be a fallacy to believe that, say, a collapse of the Amazon rainforest would have no repercussions for the boreal ecosystems in Northern Canada. The problem is that we just are not able to tell what exactly (or often even approximately) they will be. The consequence of this inherent uncertainty is that the valuation of ecosystems (important as it is for economic climate modelling), when facing such potentially abrupt changes as those climate change may trigger, is likely to be pure blind guessing with a particularly high potential for underestimation.
Given the challenges above it appears to be a big mistake that economists are simplifying all these natural interactions and uncertainties by just specifying an ad hoc “damage function” (the most common one being a quadratic function of the temperature increase) in most integrated assessment models of the economics of climate change.
But we have not reached the end of our list of uncertainties meaningful for economic climate modelling yet. Just as ecosystems’ adaptation is uncertain, so is human communities’ ability to adapt: how well are we going to adapt our agriculture to different precipitation patterns and temperatures? how are we going to respond to the rising probability (and severity) of flooding in densely populated coastal areas across the world? These are also highly important questions that cannot but be answered on a local or regional basis – and that gives rise to the danger of a lack of cooperation and coordination.
A technical, but very important source of uncertainty in economic climate modelling is the choice of a discount rate – a I extensively discussed elsewhere, there is no one “right” or “proper” rate of discount in dealing with such a complex and long-term matter as climate change. Many arguments for quite different rates of discount have been brought into debate, but no consensus has emerged.
Purely technically, economic models of climate change can integrate most of the uncertainty listed above by varying parameters (a kind of enlarged sensitivity analysis) – although only at the cost of relative simplicity and traceability. However, given the vast scope and amount of uncertainties the models would have to take into account (and the interdependencies among them), it is doubtful that they would still be able to keep their explanatory power. More likely than not, the confidence intervals of their output would be as large as to make them futile.
So, is it time to abandon integrated assessment modelling in climate economics? I would say: no, not at all. It still can provide insights, e.g. into the likely consequences of particular policy measures (see, for instance, this discussion paper on the likely effects of various kinds of carbon tax). Modelling of limited scope can indeed be useful. But we have to abandon the idea that there is a need for economic modelling of all aspects of climate change at once – and, indeed, that it is possible to do it. We know enough (or little enough, given the uncertainties) to be able to reach agreement on acting immediately on climate change without having to support this decision with(allegedly) exact numbers provided by economists. There is no more need for analysing questionable cost-benefit models that weigh acting against not acting. In the end, we must accept that most of the sources of uncertainty discussed above do exist, are likely unresolvable and make much of economic climate modelling futile.