Markov Models of Policy Support for Technology Transitions

Abstract

Abstract In this paper, we develop and analyze two Markov models to obtain generalizable insights into technology policy decision making under uncertainty. The first model is a Markov reward process (MRP) that represents policy interventions with one-time, upfront costs, while the second is a Markov decision process (MDP) that represents interventions with recurring costs. For each model, we derive analytical expressions for the policymaker’s willingness to pay (WTP) to raise the probabilities of advancing a technology development or diffusion process at various stages and compare and contrast the behaviors of the MRP and MDP models. Then, we conduct numerical sensitivity analysis to explore how the optimal technology policy portfolio varies with certain parameters, and present a case study on lithium-ion batteries for electric vehicles. We find that the MRP and MDP models share some key similarities. Most notably, the possibility of regressing from a more advanced development or diffusion stage back to an earlier one reduces the value of earlier policy intervention and enhances the value of later intervention. This effect is stronger in the MDP because moving the process forward helps avoid incurring policy costs repeatedly by lingering in stages affected by the policy. Another consequence of the models’ different cost accounting schemes is that increases in WTP with respect to policy-enhanced transition probabilities exhibit diminishing returns in the MRP, but constant returns in the MDP. Lastly, our case study on the development of lithium-ion batteries demonstrates the practical application of our model to technology policy decision making.