When it comes to normative decision-making in risky intertemporal situations, for instance optimal climate policy, economists face two main challenges. The first challenge is to adequately represent climate and economic risks. This means using the best available scientific information to inform t he quantification and adequately integrate stochastic risk into social choice rather than averaging over realizations of that risk in a deterministic (risk-less) way. The second challenge is to select a social choice criterion that is adapted to these risks. Climate change is a risk bearing on all generations, i.e. a risk on intertemporal utility: considering absolute aversion with respect to this aggregate risk, also called temporal risk aversion, might matter for optimal policy. The social choice criterion should also have suitable normative properties, for instance monotonicity with respect to first-order stochastic dominance which ensures that dominated social strategies are never chosen when facing risk. A substantial body of literature has developed to drive this research agenda forward, drawing on advances in the theoretical literature to propose quantifications in full-fledged climate-economy models and bring theoretical results on intertemporal social choice closer to applied circles and decision-makers.
Quantifying tipping risk
Lemoine and Traeger (2014) offer a quantification of the impact of one specific risk under expected utility. They represent a stylized risk of climate tipping point, i.e. a critical threshold in the climate system at which global or regional climate changes from one stable state to another stable state (e.g. West Antarctic ice sheet disintegration or Amazon rainforest dieback). By break- ing down the value function, they highlight and quantify the channels through which a tipping risk can affect optimal climate policy. The decision maker can either change the welfare impact of tipping (self-insurance) or the likelihood of tipping (self-protection). In their setting, optimal policy is more stringent under tipping risk mostly because of the second channel: the social planner is ready to give up some wealth at initial time to decrease the probability of a catastrophic event. Compared to a model without tipping risk, the resulting policy paths lower optimal peak warming by up to 0.5°C.
Stochastic vs deterministic climate-economy models
Cai and Lontzek (2019) develop a state-of-the-art stochastic climate-economy model. In comparison with the deterministic models often used to define optimal climate policy in the past, this model offers a quantification of a large number of economic and climate uncertainties under Epstein-Zin-Weil preferences. Unlike standard discounted expected utility, Epstein-Zin-Weil preferences allow to dis- entangle preference over risk and preference over time. Under a climate tipping risk, the authors show that it is optimal to more than double mitigation efforts throughout this century.
Monotone recursive preferences
Bommier, Kochov and Le Grand (2017) explore preferences that are both recursive and monotone which avoid some of the shortcomings of the Epstein-Zin-Weil form. They show the interest of this monotonicity property for modelling dynamic choices: in intertemporal normative settings, monotonicity ensures that a more risk-averse planner consistently prioritizes risk reduction.
Multiplicatively separable preferences
Bommier, Lanz and Zuber (2015) compare optimal policy under the standard expected utility model to multiplicatively separable preferences which display absolute risk aversion with respect to intertemporal risk, i.e. temporal risk aversion, but no pure time preferences in a setting with a simple stylized catastrophic risk. Under catastrophic risk, multiplicatively separable preferences are associated with a much higher value of catastrophic risk reduction and a more stringent policy response.
Tipping and temporal risk aversion
Fillon, Guivarch and Taconet (2023) explore risk-sensitive preferences that are discounted, recursive, monotone and that display temporal risk aversion. They quantify how much temporal risk aversion matters for optimal policy under a stylized tipping risk, and analyse the channels through which the tipping risk affects the risk-sensitive planner. Optimal climate policy is more stringent under temporal risk aversion, in order to reduce all present and future probabilities of crossing the tipping point and avoid a situation where all generations are badly off. Temporal risk aversion implies a 30% increase in the social cost of carbon (SCC) under their benchmark calibration and for a 10% irreversible increase in the level of economic damage from climate change.