Geometry of vectorial martingale optimal transport and robust option pricing

Abstract: This paper addresses robust finance, which is concerned with the development of models and approaches that account for market uncertainties. Specifically, we investigate the Vectorial Martingale Optimal Transport (VMOT) problem, the geometry of its solutions, and its application with robust option pricing problems in finance. To this end, we consider two-period market models and show that when the spatial dimension $d$ (the number of underlying assets) is 2, the extremal model for the cap option with a sub- or super-modular payout reduces to a single factor model in the first period, but not in general when $d > 2$. The result demonstrates a subtle relationship between spatial dimension, cost function supermodularity, and their effect on the geometry of solutions to the VMOT problem. We investigate applications of the model to financial problems and demonstrate how the dimensional reduction caused by monotonicity can be used to improve existing computational methods.