Max-plus decomposition of supermartingales and convex order. Application to American options and portfolio insurance (Q2482283)

The authors establish the existence of the new type of supermartingale decomposition where the operations \((+,x)\) are changed for \((\max,+)\). It means that for supermartingale \(Z\) they find a martingale \(M\) and adapted increasing process \(\Lambda\) with max-plus density \(L\) such that \(M=\max(Z,\Lambda)\) and \(\Lambda_t=\sup_{s\leq t}L_s\). So, the result consists in expressing any supermartingale of class \((\mathcal D)\) as a conditional expectation of some running supremum process. The existence is proved via convex analysis argument. The martingale \(M\) is also characterized as the optimal solution of some martingale problem. An optimization problem consists in finding of the best martingale dominating a given floor process (on every intermediate date), w.r.t. the convex order on terminal values. As an application, it is demonstrated how the Max-Plus supermartingale decomposition allows, in particular, to solve the American optimal stopping problem without having to compute the option price.