This paper examines the intergenerational welfare economics of risk. It characterizes a family of criteria that avoid some serious drawbacks of expected utilitarianism, such as the inability to disentangle risk aversion and inequality aversion, the lack of ethical considerations for learning, and Weitzman’s “dismal theorem”. Risk and learning are modeled as a decision tree: in each period, the outcome assigned to the current one-period living generation is to be traded-off against uncertain benefits of future generations; as time passes, the planner observes the realized shocks and becomes more informed about the true state of the world.

According to the characterized family of fair intergenerational utilitarian criteria, each generation’s welfare should be measured by a CES aggregation of the outcome at each history relative to the fair prospect; total welfare is the discounted sum of a CRRA transform of each generation’s welfare. The discount rate is time-varying: depending on the magnitude of risk, the learning process, and the planner’s risk attitude, specific discounting formulas obtain, including exponential and quasi-hyperbolic.