The main part of the project analyzed the impact of growth uncertainty on optimal mitigation policy. Ceteris paribus, faster growth implies richer future generations and, at the same time, more emissions and higher damages in the climate dependent sectors of the economy.

Earlier Monte Carlo studies suggested that optimal mitigation decreases under long-term economic growth uncertainty. We showed that these suggestive results do not hold in a proper stochastic dynamic programming framework. Under standard preferences, optimal mitigation slightly increases under growth uncertainty.

In a more sophisticated evaluation framework, we first account for observed risk premia by increasing risk aversion, while keep the propensity to smooth consumption over time fixed at its widespread value. From the intergenerational perspective, this propensity to smooth consumption over time can be interpreted as a (relatively high) aversion to intergenerational inequity. In this scenario, long term growth uncertainty more substantially increases optimal CO2 mitigation. Second, we reduce the propensity to smooth consumption over time to better reflect the observed low risk-free rate. Then, growth uncertainty decreases optimal mitigation. However, the reduced propensity to smooth consumption over time, simultaneously makes present generations more willing to invest into a future that is richer in expectation, increasing optimal mitigation. We show that the net effect of using the more comprehensive evaluation framework is always an increase in optimal mitigation policy.

The project funding also enabled us to take a first step towards developing a model that addresses distributional aspects within generations under uncertainty. Hereto, we related our DICE based numerical integrated assessment structure to a recent analytic integrated assessment model by a group of macroeconomists (Krusell, Golosov & Hassler). The models of the latter have the advantage of an explicit energy and resources sector and a direct implementation of the decentralized equilibrium, while our DICE based model has the advantage of overcoming the very restrictive functional forms and much simplified climate dynamics. So far, we have elaborated different model layouts that combine important characteristics of both models while keeping numerical tractability. The transition of Mexican workers away from agriculture will have profound impacts for the U.S. farm sector, which historically has depended on an elastic supply of Mexican farm labor and will now have to compete with Mexican farms for a dwindling supply of labor. To adjust to a smaller supply of labor, in the future U.S. farming will need to become more mechanized, relying on fewer and more productive workers. This will impact both the U.S. farm industry and rural communities; increasing skills and wages of farm workers may help break a vicious circle of farm employment, immigration, and poverty.

We have analyzed the impact of damage uncertainty on optimal mitigation policies in the integrated assessment of climate change. The common approach to analyze uncertainty had been a Monte Carlo simulation averaging deterministic paths. A proper treatment of stochasticity, however, requires a model where decision makers take decisions under uncertainty. For this purpose, we constructed a close relative of the integrated assessment model DICE in a recursive dynamic programming framework. Our recursive approach allows us to model persistent uncertainty and to disentangle effects of risk, risk aversion, and aversion to intertemporal substitution. We show that the common Monte Carlo approach underestimates the effects of damage uncertainty on optimal policies. Moreover, a comprehensive risk evaluation has to distinguish between risk attitude and the propensity to smooth consumption over time. The asset pricing literature has shown that this more comprehensive risk analysis eliminates a multitude of asset pricing puzzles. We show that, in the climate context, the same disentanglement implies a major increase in the optimal abatement of greenhouse gases. The project also served as a starting point for a follow up project that analyzes ambiguity and tipping points in the integrated assessment of climate change.