The cumulant process and Esscher's change of measure (Q1424691): Difference between revisions

An exponential semimartingale \(S=\exp{(X)}\) can be represented as stochastic exponential \(S={\mathcal E}(\tilde{X})\) (here \(X\) and \(\tilde{X}\) are some semimartingales). The process \(\tilde{X}\) is called stochastic logarithm or exponential transform of \(X\). If \(\tilde{X}\) is a special semimartingale, then its predictable part of finite variation \(\tilde{K}\) is denoted as the Laplace cumulant process of \(X\). This process has close relations with the exponential compensator of \(X\) (the unique predictable process \(K\) of finite variation such that \(\exp{(X-K)}\) is a local martingale). \(\tilde{K}\) is the exponential transform of \(K\). These notions and their properties are considered in the first part of the article. In the second part it is shown that the cumulant process naturally leads to a hierarchy of sufficient criterions for uniform integrability of \(\exp{(X-K)}\), which includes in particular the Novikov and Kazamaki type conditions in the sense of \textit{D. Revuz} and \textit{M. Yor} [``Continuous martingales and Brownian motion'' (1999; Zbl 0917.60006)]. Measure transformations with the density process \(\frac{dP_{\theta}}{dP}|_{{\mathcal F}_t}=\exp{(\theta X_t-k(\theta)t)}\) for a given Lévy process \(X\) and real numbers \(\theta, k(\theta)\) are considered as the Esscher transformation for contingent claim pricing. This concept can be generalized to integrals \(\int\theta_sdX_s\) of a class of semimartingales if the cumulant \(k(\theta)t\) is replaced by the cumulant process of \(\int\theta_sdX_s\). This approach is discussed in the article.