The atmosphere is a very complex system. Although climate models have greatly improved over time, our ability to predict exactly how much warming greenhouse gas emissions cause remains highly imperfect. While there is little disagreement on the basic notion of climate change, a wide range of scenarios remain plausible, even if we know how much carbon will be emitted in the coming decades. And that itself remains an open question.

This uncertainty presents a major challenge to climate policy. Even if global climate policy was perfectly rational (you wish), designing an optimal response would be very difficult because the exact benefits of early action cannot be measured with any accuracy. In my view, this basic observation means that standard cost-effectiveness calculations are not terribly useful for climate policy formation. The uncertainty surrounding any cost-benefit analyses is so vast that the results are almost certain to be very crude approximations at best, and highly misleading at worst.

In most optimization problems, a simple solution to uncertainty is to collect more information. If an engineer is designing a new production line, she can continue to experiment with different approaches until she is satisfied with the result.

Unfortunately, this solution is not available for global warming. One of the key features of the energy economy is carbon lock-in: the cost of building energy infrastructure is so high that past investments influence the carbon intensity of the world economy for a long time, so business as usual would greatly complicate future attempts towards decarbonization.

Another problem is that the only way to see exactly how much global temperatures increase is to continue to pollute. Unfortunately, global warming will have virtually irreversible effects from species extinction to desertification.

On the other hand, uncertainty itself is not a compelling reason for decisive action. In reality, there is always some uncertainty, so the idea of precaution should not be applied frivolously to regulate everything in sight. After all, regulation itself will have unforeseen consequences that may cause great harm.

In my view, the uncertainty surrounding global warming is more profound than that surrounding virtually any other policy problem. Perhaps paradoxically, the reason is that we already know enough to construct plausible pathways to a genuine climate catastrophe.

The idea that global temperatures might rise by five or six degrees Celsius, perhaps triggering positive feedback loops that cause further warming, is not a fantasy, as one of my favorite economists, Professor Martin Weitzman at Harvard, explains in his working paper, “The Extreme Uncertainty of Extreme Climate Change.” It is a very real possibility that should be addressed in any credible proposal for global emissions trajectories.

For this simple reason, the uncertainty surrounding climate change should be an impetus for action. Clean energy deployment and efficiency improvements may prove much costlier than expected, as the economy is also a highly complex system, but it is simply not plausible that these public policies would cause a catastrophe. They may raise energy prices or reduce growth, but not on a scale comparable to the damage caused by a climate catastrophe. By acting now, societies can greatly reduce the probability of a future disaster at a cost that is almost certainly not extreme.

To be sure, this does not imply that we should not continue to collect information. In addition to investing in climate science, the need to develop functional methodologies to deal with extreme uncertainties is urgent. So is the need to develop decision-making procedures in circumstances of deep uncertainty. But in the case of climate change, new information cannot substitute for decisive action.