Convergence rates for rank-based models with applications to portfolio theory (Q1955832): Difference between revisions

The authors consider a rank-based model that is a multidimensional Markov process whose instantaneous dynamics is a function of the order in which the coordinates can be ranked. Some examples of rank-based models are described, in particular, diffusion model and Atlas model. The main goal of the paper is to determine exponential rates of convergence to the equilibrium for ranked-based models and for the derived reflecting Brownian motions. Sharp Gaussian fluctuations around the equilibrium mean for additive functionals of these processes are obtained. The main difficulty one encounters with these processes is the lack of smoothness in its drift and diffusion parameters. The new approach developed in order to overcome this difficulty is based on recent advances in transportation cost-information inequalities as well as classical Poincare inequalities satisfied by the associated Dirichlet forms. The motivation to study the rank-based models is to solve certain open problems related to the area of stochastic portfolio theory. For example, quantitative bounds are produced for fluctuations of market weights and occupation times of various ranks for individual coordinates. Some comparison of portfolios is also considered as an example.