Hybrid continuous and periodic barrier strategies in the dual model: optimality and fluctuation identities

José Luis Pérez, Kazutoshi YamazakiarXiv 2016 · DOI · arXiv

Abstract

In the study of optimal dividend problems, companies often face a choice between continuous dividend payments and periodic payouts, each associated with distinct transaction costs. While previous research by Avanzi et al. (2016) established the optimality of hybrid strategies within a Brownian motion framework, this paper generalizes those findings to the dual model—specifically, spectrally positive Lévy processes. By utilizing the mathematical framework of scale functions, we demonstrate that a hybrid-barrier strategy is optimal for maximizing the expected net present value (NPV) of dividends. Our analytical results are supported by numerical experiments using phase-type Lévy processes.

Introduction

De Finetti’s optimal dividend problem is a cornerstone of actuarial science, seeking to determine the payout strategy that maximizes the expected discounted dividends until the point of ruin. Traditionally, this problem has been analyzed through two lenses: continuous strategies, where dividends are paid whenever the surplus exceeds a certain threshold, and periodic strategies, where payments occur only at discrete time intervals (often modeled by Poisson arrival times).

In practice, these two methods are not mutually exclusive but are governed by different transaction costs. This paper bridges the gap by considering a hybrid model where both continuous and periodic dividends are possible. Furthermore, we move beyond the limitations of the Brownian motion model to incorporate spectrally positive Lévy processes. These "dual models" are particularly relevant for insurance and investment portfolios as they account for random upward jumps (gains) while maintaining continuous downward movements (costs).

The Dual Model and Hybrid Strategy

We model the surplus process as a spectrally positive Lévy process. The hybrid strategy is defined by two distinct thresholds: a periodic barrier $a$ and a continuous barrier $b$, where $0 \le a \le b$.

  1. Periodic Payouts: At the arrival times of an independent Poisson process with intensity $r$, any surplus exceeding level $a$ is paid out as a dividend, bringing the surplus back down to $a$.
  2. Continuous Payouts: The surplus is never allowed to exceed level $b$. Any upward pressure above $b$ is immediately paid out as a continuous dividend.

To solve this stochastic control problem, we employ a "guess and verify" approach. The core of our analysis relies on scale functions ($W^{(q)}$ and $Z^{(q)}$), which allow us to express the fluctuation identities of the Lévy process and the resulting value function in a unified manner.

Optimality and the Cost Ratio

A critical finding of this research is that the optimal strategy is dictated by the cost ratio $\beta$, representing the ratio of net proportions of continuous versus periodic dividends. The optimal regime is identified as follows:

  • Pure Periodic Strategy: Optimal when $\beta \le r/(q+r)$, where $q$ is the discount rate. In this scenario, continuous dividends are too expensive relative to periodic ones.
  • Pure Continuous Strategy: Optimal when $\beta \ge 1$. Here, the efficiency of continuous payments outweighs the benefits of waiting for periodic intervals.
  • Hybrid Strategy: Optimal when $r/(q+r) < \beta < 1$. In this "Goldilocks" zone, the most efficient path involves a combination of both payout types.

We prove that the value function satisfies the necessary variational inequalities and exhibits the required smoothness (differentiability) at the barriers, which varies depending on whether the underlying process has bounded or unbounded variation.

Numerical Confirmation

To validate our theoretical findings, we conducted numerical experiments using phase-type Lévy processes. These processes are computationally advantageous because their scale functions can be expressed as sums of exponentials. Since phase-type distributions are dense in the field of all distributions, these results serve as a robust approximation for general Lévy processes. The experiments confirmed that the hybrid strategy strictly dominates sub-optimal single-barrier strategies within the predicted parameter ranges, and the value function correctly aligns with the analytical "smooth fit" conditions.

Conclusion and Future Work

This paper rigorously establishes the optimality of hybrid-barrier dividend strategies in the dual model. By providing a concise expression for the value function using scale functions, we offer a framework for calculating exact dividend thresholds based on transaction costs.

Future research may expand this model in several directions. While this study assumes a positive drift (where the surplus tends toward infinity), exploring cases with negative drift could provide insights into riskier ventures. Additionally, incorporating fixed transaction costs—which might shift the optimal strategy toward impulse control ($S, s$ type)—and addressing the "standard" insurance model (spectrally negative Lévy processes) remain significant challenges due to the complexities of downward jumps and barrier overshooting.

Summary updated

Abstract

Avanzi et al. (2016) recently studied an optimal dividend problem where dividends are paid both periodically and continuously with different transaction costs. In the Brownian model with Poissonian periodic dividend payment opportunities, they showed that the optimal strategy is either of the pure-continuous, pure-periodic, or hybrid-barrier type. In this paper, we generalize the results of their previous study to the dual (spectrally positive Lévy) model. The optimal strategy is again of the hybrid-barrier type and can be concisely expressed using the scale function. These results are confirmed through a sequence of numerical experiments.
BibTeX
@misc{perez2016hybrid,
  author = {José Luis Pérez and Kazutoshi Yamazaki},
  title = {{Hybrid continuous and periodic barrier strategies in the dual model: optimality and fluctuation identities}},
  year = {2016},
  eprint = {1612.02444},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1612.02444},
}
Figure