On Skorokhod embeddings and Poisson equations

Leif Döring, Lukas Gonon, David J. Prömel, Oleg ReichmannOxford 2017

TL;DR

This paper generalizes the classic Skorokhod embedding problem (SEP) to arbitrary Markov processes, not just Brownian motion. They establish necessary and sufficient conditions for the existence of a finite mean solution, linking it to the solvability of a Poisson equation involving the process's generator and the initial/target distributions. For Lévy processes, they provide an explicit construction of the stopping time.

What Problem Does This Paper Address?

The Skorokhod embedding problem (SEP) asks for a stopping time $\tau$ such that a stochastic process $W_t$ (typically Brownian motion) stopped at $\tau$ has a specific target distribution $\mu_1$, given an initial distribution $\mu_0$. This problem has a rich history and numerous applications, particularly in finance for pricing exotic options and in robust mathematical finance. While solutions exist for Brownian motion, generalizing this to other Markov processes, especially those without local times, and providing explicit conditions for existence and construction of solutions has been a significant challenge. This paper tackles this by formulating a generalized SEP for any Markov process and deriving conditions for the existence of a finite mean solution, offering an explicit construction for Lévy processes.

Methods

The core of the paper's methodology lies in connecting the Skorokhod embedding problem to the solvability of a Poisson equation. For a Markov process $X$ with generator $A$ and initial/target densities $h_0, h_1$, they propose that the existence of a finite mean embedding is equivalent to the existence of a non-negative, integrable solution $H$ to the Poisson equation $A^*H = h_1 - h_0$, where $A^*$ is the formal adjoint of the generator.

For the specific case of Lévy processes, they leverage Fourier transform techniques. The Poisson equation can be solved explicitly in Fourier space, leading to an explicit form for $H$. This explicit form then provides the necessary and sufficient conditions for the existence of the embedding. The construction of the stopping time $\tau$ for Lévy processes is achieved through a time-change mechanism, inspired by Bass's work on Brownian motion, but adapted to the general Lévy process setting. This involves constructing a local speed function $\sigma$ and then defining $\tau$ as a solution to a random ordinary differential equation.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Poisson Equation as a Bridge: The paper's central insight is the direct link between the existence of a finite mean Skorokhod embedding and the solvability of a Poisson equation involving the process's generator and the initial/target densities. This provides a powerful analytical tool for studying the SEP.
  • Generalization Beyond Brownian Motion: The work extends the SEP framework to a much broader class of Markov processes, including those without local times, which are crucial for many real-world applications.
  • Explicit Conditions for Lévy Processes: For Lévy processes, the paper derives concrete, checkable conditions for the existence of an embedding, expressed in terms of the process's characteristic exponent and the densities of the initial and target distributions.
  • Constructive Approach: Beyond existence, the paper provides an explicit construction of the stopping time for Lévy processes, making the theoretical results practically applicable.
  • Regularity Assumptions are Key: The paper highlights how different types of Lévy processes (classified by their characteristic exponents) require different regularity assumptions on the initial and target densities for the embedding to exist.

Key Results

  • Necessary and Sufficient Conditions: For a Markov process with densities $h_0, h_1$ and generator $A$ with adjoint $A^*$, the existence of a finite mean Skorokhod embedding is equivalent to the existence of a non-negative, integrable solution $H$ to $A^*H = h_1 - h_0$.
  • Explicit Conditions for Lévy Processes: For Lévy processes, the conditions for a finite mean embedding are given by:
  • $(\mu_1 - \mu_0)/\eta \in L^1(\mathbb{R})$

    1. $H \ge 0$
    2. $H \in L^1(\mathbb{R})$ where $\eta$ is the characteristic exponent and $H$ is an explicit function derived from the Fourier transform of $(\mu_1 - \mu_0)/\eta$.
  • Explicit Construction for Lévy Processes: A stopping time $\tau$ is explicitly constructed using the solution $H$ of the Poisson equation and the densities $h_0, h_1$.

Limitations and Future Work

  • Regularity Assumptions: The paper relies on certain "sufficiently smooth" regularity assumptions for the initial and target densities ($h_0, h_1$). While these are necessary for the explicit construction and analysis, extending the results to less regular densities might be a future direction.
  • General Markov Processes: While the paper establishes the connection to the Poisson equation for general Markov processes, the explicit construction and conditions are primarily detailed for Lévy processes. Extending the explicit construction to more general Markov processes remains an open challenge.
  • Computational Aspects: While the conditions are explicit for Lévy processes, their practical computation might still involve numerical integration or Fourier transform inversion, which could be computationally intensive.

The authors suggest that their findings can be extended to more general Markov processes, hinting at future research in that direction. The connection to the Poisson equation is presented as a fundamental insight that could be leveraged for broader classes of processes.

Summary updated

Abstract

The classical Skorokhod embedding problem for a Brownian motion W asks to find a stopping time τ so that Wτ is distributed according to a prescribed probability distribution μ. Many solutions have been proposed during the past 50 years and applications in different fields emerged. This article deals with a generalized Skorokhod embedding problem (SEP): Let X be a Markov process with initial marginal distribution μ0 and let μ1 be a probability measure. The task is to find a stopping time τ such that Xτ is distributed according to μ1. More precisely, we study the question of deciding if a finite mean solution to the SEP can exist for given μ0,μ1 and the task of giving a solution which is as explicit as possible. If μ0 and μ1 have positive densities h0 and h1 and the generator A of X has a formal adjoint operator A∗, then we propose necessary and sufficient conditions for the existence of an embedding in terms of the Poisson equation A∗H=h1−h0 and give a fairly explicit construction of the stopping time using the solution of the Poisson equation. For the class of Lévy processes we carry out the procedure and extend a result of Bertoin and Le Jan to Lévy processes without local times.
BibTeX
@inproceedings{doring2017skorokhod,
  author = {Leif Döring and Lukas Gonon and David J. Prömel and Oleg Reichmann},
  title = {{On Skorokhod embeddings and Poisson equations}},
  booktitle = {Oxford},
  year = {2017},
}
Figure