The desire for robustness is not necessarily a time-invariant feature of economic behavior. It is reasonable to suggest human beings can become more or less pessimistic depending on the events that they have just witnessed or experienced. In this paper we develop a framework to analyze an environmental policy maker that uses emission taxes to regulate a stock pollutant and his degree of uncertainty aversion can vary as he observes pollutant emission shocks of more or less persistence. In order to keep the model tractable we consider two regimes for the emission shocks persistence: low and high. The policy maker’s model works well in the low persistence regime. However, the policy maker mistrusts his model in the high-persistence regime as it is reasonable to believe that the estimated model might misrepresent the firms’ abatement response to the tax, particularly after a period of relative emissions stability. Thus, when the policy maker observes emissions shocks, the degree of uncertainty aversion about the accuracy of his own model can change. We model the fear of model misspecification in the high-persistence regime by introducing robust control. Hence, the policy maker plays a fictitious game with an “evil” nature which distorts the policy maker’s model only in the high-persistence regime. In addition, we use a Markov chain to capture the time-varying degree of uncertainty aversion about the policy maker’s model. This bound, known as the “free” parameter in robust control, measures the uncertainty aversion (or pessimism) of the policy maker and can take on two values. In one regime such parameter is bounded between zero and infinity (the policy maker is pessimistic about his model), while in the other regime the bound approaches infinity (the policy maker is confident in his model). That is, unlike the standard application of Hansen-Sargent’s robust control, we consider a time-varying nature of the bound on the “evil” nature distortion.

The framework developed in this study is applied to the regulation of CO2 emissions leading to climate change. In the evaluation and analysis of different combinations of chosen and true transition probabilities from the low to the high-persistence regime, we compare welfare losses of two types of errors: pessimistic and optimistic. The optimistic error occurs when the policy maker underestimates the probability of switching to the regime with emissions shocks of high-persistence. The pessimistic error takes place when the policy maker overestimates the aforementioned probability. We find that the policy maker obtains lower welfare losses when it overestimates the probability of switching to the regime with high-persistence emissions shocks. That is, our results suggest that the environmental authority should err on the caution side in the presence of time-varying uncertainty about the persistence of emissions shocks.