On optimal dividend and capital injection strategies for Markov additive processes

Abstract: We study de Finetti's optimal dividend problem, including capital injections, under the assumption that uncontrolled assets behave as the additive component of a Markov additive process. We assume that capital injections can be received at any time. Meanwhile, we primarily address cases in which dividends can only be paid at specific, discrete times, and we establish the necessary and sufficient conditions for the optimality of the strategies. Additionally, we prove the optimality of certain Markov-modulated periodic-classical barrier strategies and cover cases where dividends can be paid at any time via approximation. A key feature of this research is our use of a significantly more general Markov additive process than in prior studies, which leads to different proof approaches.