Avoiding disastrous climate change is possible but not inevitable.

W hat the world does or does not do about climate change will have profound implications. If the world fails to act, we may cross a climate tipping point, with disastrous consequences all around. No country by itself can prevent this, but a substantial number of countries working together can avert disaster. Collectively, we know what we need to do. However, can we agree on what each of us should do? Can we be assured that, if we act, others will fulfill their pledges? An experimental paper in PNAS (1) provides reasons for optimism and despair, optimism because, with communication, outcomes improve; despair because intransigence by a few countries can block progress. A key impediment to success, this paper (1) shows, is overcoming historical inequities. This paper by Tavoni et al. (1) is best appreciated by considering a previous paper by Milinski et al. (2), also published in PNAS. An experimental game is played by groups of six students. Each student is given an endowment of €40 and must decide whether to contribute €0, €2, or €4 in each of 10 periods to a collective effort to avert dangerous climate change. (You might think of these values as representing the costs of reducing emissions by various levels.) The identities of the players are private, but everyone’s contribution is made public after each round. If all six students collectively contribute €120 or more by the end of the game, each student gets to keep the amount of money he or she has left. If, however, contributions add up to less than €120, each student loses all the money he or she has left with a probability of 90%, 50%, or 10%, depending on the treatment. Notice that, collectively, the students will want to contribute either €0 or €120. Any other total contribution will be wasted. For example, if €119 is contributed, the prospect of disaster will be unaffected and the money contributed will be lost. If €121 is contributed, disaster will be prevented but the same outcome would have been assured by contributing €1 less. If students contribute €120, they are sure to get a collective payoff of €120. If they contribute €0, they will get an expected payoff of €24 if the probability of disaster is 90%, €120 if this probability is 50%, and €216 if this probability is 10%. Their collective-best contribution thus depends on the probability of disaster and the students’ aversion to risk. The result that Milinski et al. (2) obtained for the 90% treatment was especially startling. Disaster was avoided only half of the time, even though the collective incentive to avoid disaster was powerful. Reading this paper, one cannot help but ask, “Are people really this foolish?” To understand what is going on, it helps to examine the private and not only the collective incentives to contribute. If no one else contributes, each player is better off not contributing. However, if the others contribute, say, €20 each, every student will want to contribute €20 under the 90% treatment. In this game, there is strong incentive for each student to contribute only if assured that the others, or at least enough others, will contribute. It is possible for people to be individually rational and collectively foolish. The paper by Tavoni et al. (1) explores the robustness of these previous results while at the same time casting the experiment in a more realistic light. In particular, Tavoni et al. (1) allow the students to make pledges for how they intend to contribute in the future. They also construct treatments in which inequities appear attributable to different histories. Unfortunately, they limit their analysis to the 50% probability of disaster. Are these experiments plausible representations of the game countries are playing now? Some differences stand out. The real climate change game is being played by 192 countries rather than 6. Catastrophe will cost us more than €40 each. Less obvious differences are also important. For example, both papers assume that we know that a threshold exists and that we know precisely where it is and how to avoid it. This assumption substantially strengthens the incentives for countries to act collectively. It also gives communication its bite. The real challenge, however, is riddled with uncertainties. Although the Copenhagen and Cancun agreements say that a temperature change greater than 2 °C should be avoided, other thresholds have been identified (3), and the uncertain relationship between concentrations and temperature makes it difficult to know how to meet a given temperature threshold. As well, both papers assume that efforts short of and in excess of the threshold are wasted, whereas every ton of emissions prevented reduces “gradual” climate change in addition to the probability of triggering “abrupt and catastrophic” climate change. These matters aside, what can we learn from the Tavoni et al. (1) paper? Start by considering the game that is played without communication. In game of Milinski et al. (2), with 50% treatment, only 1 of 10 groups cooperated. In the game of Tavoni et al. (1), by contrast, 5 of 10 groups avoided catastrophe. Why the difference? There are two reasons. First, in the game of Tavoni et al. (1), the computer chose a contribution of €2 per round for every player in each of the first three rounds. Hence, if each player contributed €0 in each of the remaining €20