Catastrophes, delays, and learning

Abstract

How to plan for catastrophes that may be under way? In a simple but general model of experimentation, a decision-maker chooses a ow variable contributing to a stock that may trigger a catastrophe at each untried level. Once triggered, the catastrophe itself occurs only after a stochastic delay. Consequently, the rhythm of past experimentations determines the arrival of information. This has strong implications for policies in situations where the planner inherits a history of experiments, like climate change and pandemic crisis. The structure encompasses canonical approaches in the literature.