The COP16 meeting, about to take place Cancun, Mexico, faces no such high expectations. It is likely to focus on climate financial aid and deforestation, and on preparing the ground for a broader agreement either in South Africa in 2011 or at the Rio de Janeiro Earth Summit in 2012.

But what sort of agreement should we seek? This question remains most urgent, as global warming continues apace and the developing countries have returned, post-GFC, to rapid rates of growth of greenhouse emissions.

In my paper, ‘Containing Global Warming after Copenhagen‘, two contrasting ways of achieving the objective enshrined in the Copenhagen Accord, to hold global warming to <2°C are examined. The first, which I call the one-shot approach, involves a universal, legally binding agreement on emissions paths that will stabilise the atmospheric concentration of GHGs at an appropriate level. The second, the learning-by-doing approach, involves the major nations entering into an evolving process to contain warming to the target level, by adopting and acting on emissions reduction targets that in some cases are non-binding and are reviewed over time in the light of new information.

A learning-by-doing agreement may well be a legally binding agreement covering many matters, such as climate financing and emissions targets for many countries. But for some countries the agreement will include targets for emissions reductions that they will be able revise if necessary in the light of experience. The learning-by-doing approach sits between top-down and bottom-up: it involves a coordinated attempt to achieve a top-down target (say <2°C) but allows for bottom-up initiatives and learning.

The negotiations at Copenhagen dramatised the nature of the problem: it involves strategic interaction between many countries, pursuing both their individual national and global interests, with differing objective incidence of various types of risk across countries and with different perceptions of risk, in the context of substantial future learning. In both analytical and practical terms this is a deeply complex challenge. Three conclusions were reached about the relative merits of the two approaches.

First, there is no assurance in either theory or practice that a one-shot agreement can be achieved at a level consistent with the warming target, or that this approach is to be preferred to the learning-by-doing approach. The putative benefits of the one-shot, binding approach – avoidance of free-riding, encouragement of country commitments and certainty of outcomes – might be equally well achieved under a learning-by-doing approach. Indeed, there are grounds for the view that a reflexive, learning-by-doing approach might be more effective in dealing with such complex strategic interaction.

Second, good but still insufficient progress towards reducing future emissions and global warming has been made in recent years by using a learning-by-doing, bottom-up approach. Comparing a range of emissions paths associated with different policy settings, from the policies in place in 2006 to those implicit in the Copenhagen Accord, we found that policy changes over that time have reduced the estimated atmospheric concentration of Kyoto gases and mean global warming at 2100 by about 15–18per cent. This contrasts with the impact of the GFC, which was to reduce the 2100 value of these variables by about 4per cent. While the Copenhagen Accord path is still well above a <2°C path, it is about half way between our 2006 unchanged policy path and a range of <2°C paths.

Third, the attempt to meet the <2°C target will involve fundamental changes in technology and industrial structures, to be pursued in the face of rising temperatures and severe impacts in many communities. This will imply many forms of learning: about the actual impacts of rising global mean temperatures, about the pace of development of many different technologies and about their competitive position, about the scale of investment involved in moving to a low carbon economy and the impact of these factors on the competitive position of different countries and regions. There seems to be a strong case for allowing these forms of learning to influence the response of individual countries, through a dynamic, multi-stage process of adjustment to the risks posed by climate change.

Many view the climate policy process over the last five years or so as a failure, because we have failed to achieve the single definitive agreement to control climate change. But much has been achieved, in terms of changes in the policies of governments and in the action of firms and individuals. From a learning-by-doing perspective we have come perhaps half the way that we need to travel to constrain warming to <2°C. What is needed now is an intensive effort to define, understand and negotiate a learning-by-doing agreement that can be put into place in South Africa in 2011, so that this critical target can be achieved.

Peter Sheehan Professor of Economics and Director of the Centre for Strategic Economic Studies at Victoria University, Melbourne