Skorokhod embeddings, minimality and non-centred target distributions (Q2494406): Difference between revisions

The Skorokhod embedding problem may be described thus: Given a Brownian motion \((B_t)\) and a centred target law \(\mu\), can we find a stopping time (s.t.) \(T\) such that \(B_T\) has distribution \(\mu\) (i.e., \(T\) embeds \(\mu)\)? Following a paper by \textit{I. Monroe} [Ann. Math. Stat. 43, 1293--1311 (1972; Zbl 0267.60050)] an s.t. \(T\) for a process \(X\) is called minimal if whenver \(S\leq T\) is an s.t. such that \(X_S\), \(X_T\) have the same distribution, then \(S=T\) a.s. The authors give conditions which are necessary and sufficient for the minimality of an s.t. \(T\) embedding an integrable distribution with negative mean in Brownian motion. These results are applied to the following problem: Let \(M_T:=\sup_{s\leq T}B_s\). Given a target distribution \(\mu\) (not necessarily centred), find a minimal s.t. \(T\) such that \(P(M_T\geq x)\) is maximal over all minimal s.t. \(T\) embedding \(\mu\), uniformly in \(x\). The corresponding problem for regular time-homogeneous diffusions can be reduced to the case of Brownian motion via the scale function, using a key idea due to \textit{J. Azéma} and \textit{N. Yor} [in: Séminaire de probabilités XIII. Lect. Notes Math. 721, 625--633 (1979; Zbl 0414.60056)].