The curious role of "learning" in climate policy : should we wait for more data?

Abstract

Given the large uncertainties regarding potential damages from climate change and the significant but also uncertain costs of reducing greenhouse emissions, the debate over a policy response is often framed as a choice of either acting now or waiting until the uncertainty is reduced. Implicit behind the "wait to learn" argument is the notion that the ability to learn in the future necessarily implies that less restrictive policies should be chosen in the near-term. I demonstrate in the general case that the ability to learn in the future can lead to either less restrictive or more restrictive policies today. I also show that the initial decision made under uncertainty will be affected by future learning only if the actions taken today change the marginal costs or marginal damages in the future. Without this interaction, learning has no effect on what we do today, regardless of what we learn in the future. Results from an intermediate-scale integrated model of climate and economics indicate that the choice of current emissions restrictions is independent of whether or not uncertainty is resolved before future decisions, because the cross-period interactions in the model are minimal. Indeed, most climate and economic models fail to capture potentially important cross-period interaction effects. I construct a simple example to show that with stronger interactions, the effect of learning on initial period decisions can be more important.