Developing Long-run Agricultural R&D Policy in the Face of Uncertain Economic Growth

Projecting economic growth, population and climate over the 21st century is challenging. One approach to this problem has been the develop of "Shared Socio-economic Pathways" designed to provide a consistent characterization of alternative evolutions of population, per capita income and climate. However, recent analysis has shown that the true extent of future growth uncertainty is likely far greater than that embodied in the SSPs. We build on the innovative work of Christensen et al., in order to construct 13 independent probability distributions of economic growth in the 21st century. For each of these distributions, we use a stochastic dynamic partial equilibrium model of global land use to compute the optimal rate of R&D investment as well as the ensuing path of Total Factor Productivity (TFP) growth to 2100. When there is a significant probability of non-positive growth, the optimal response is to invest a lot in R&D today, and maintain a fairly flat trajectory over the entire century. This is in sharp contrast to the optimal path when growth rates are strictly positive. In this case, R&D spending starts out slow, and accelerates over time. Since we do not know which expert, if any, is correct, we propose a novel approach to dealing with this ambiguity by minimizing the maximum regret across all 13 optimal growth paths. This results in 50% higher R&D spending early in the century than that dictated by a mean growth rate deterministic model. However, by mid-century, optimal R&D spending levels off, and the resulting TFP plateaus at a level which is about 75% higher than at the start of the century.