Scale Functions for Spectrally One-Sided Lévy Processes

44 concepts31 papers2 proposals

Traditional financial and actuarial models rely heavily on Brownian motion, where paths are continuous and change smoothly. However, real-world systems experience sudden, catastrophic shocks—insurance companies face massive clusters of claims, and corporate asset values can plummet instantly during a market crash.

To capture these dynamics, researchers use Lévy processes, which incorporate random jumps. Yet, general jump-diffusion models are notoriously difficult to analyze. Calculating the probability that a process will cross a specific boundary (such as a bank hitting a regulatory capital minimum) typically requires solving complex integro-differential equations or running computationally expensive path-by-path simulations.

A powerful simplification emerges when we restrict the direction of the jumps. Spectrally one-sided Lévy processes can move continuously in one direction but are restricted to one-sided jumps in the opposite direction.

  • Spectrally negative processes drift upward continuously but only jump downward (ideal for modeling an insurance surplus facing sudden claims).
  • Spectrally positive processes (often called the dual model) drift downward continuously but experience random upward shocks (ideal for modeling a company driven by sporadic technological breakthroughs).

The core breakthrough of this research direction is that the entire fluctuation behavior of these asymmetric jump processes can be completely characterized by a family of deterministic functions known as scale functions.

The Intuition of Scale Functions

Why does restricting the jump direction make the mathematics tractable?

Consider a process starting at a value $x$ inside an interval $[0, b]$. We often want to find the "exit probability": what is the chance the process reaches the upper barrier $b$ before falling below zero? If a process can jump both up and down, it could skip right over either boundary. This "overshoot" depends heavily on the exact distribution of jump sizes at the boundary, making the problem highly path-dependent.

However, if the process is spectrally negative, it cannot jump upward. To exit the upper boundary $b$, it must creep across it continuously. It can only overshoot the lower boundary zero via a downward jump.

Because the process cannot jump over the upper barrier, the probability of exiting at $b$ before hitting zero can be mapped directly to a single, one-dimensional deterministic function, $W(x)$, called the scale function. The exit probability simplifies to a clean ratio:

$$\mathbb{P}_x(\tau_b < \tau_0) = \frac{W(x)}{W(b)}$$

As shown by Pistorius in the context of the Skorokhod Embedding Problem, scale functions serve as a fundamental representational strategy. They encapsulate complex path-dependent behaviors—such as the joint distribution of maximums and minimums during an excursion—into deterministic equations, bypassing the need to track every possible path.

Modeling Control Strategies: Reflection and Refraction

In practical applications like corporate finance or inventory management, systems are actively managed rather than left to fluctuate freely. Scale functions are highly effective because they naturally incorporate controlled boundaries.

Reflection

Imagine an insurance company that must inject capital whenever its surplus hits zero to avoid bankruptcy, or a firm that instantly pays out any surplus above a threshold $b$ as dividends. These boundaries "reflect" the process back into the allowed domain. Pérez et al. and Surya et al. demonstrated that scale functions can cleanly express the expected net present value of these capital injections and dividend payouts over an infinite horizon, turning complex stochastic control problems into solvable algebraic equations.

Refraction

Instead of a hard reflection that instantly removes excess capital, a corporate strategy might change the rate of growth. For example, when a firm's assets cross a threshold, it might begin paying out dividends at a bounded, continuous rate, which reduces its upward drift. This change in drift is called "refraction."

The authors in Pérez et al. showed that processes constrained by both lower-boundary reflection and upper-boundary refraction can be fully characterized using scale functions. This provides closed-form expressions for exit times, resolvent measures, and occupation times without requiring path simulations.

Reflected Boundary (Hard Stop)
   ^
   |   /|  <- Excess instantly removed as a lump sum
---|--/-|---------------- Threshold b
   | /  |
   |/   |

Refracted Boundary (Drift Adjust)
   ^
   |   /   <- Upward drift flattens out (bounded dividend stream)
---|--/------------------ Threshold b
   | /
   |/

Bridging Continuous Theory and Discrete Reality

While continuous monitoring is mathematically elegant, real-world managers only observe systems or make decisions at discrete intervals. Scale functions adapt well to these realistic constraints through fluctuation identities.

Parisian Reflection

What if a bank's capital is only audited at random, discrete times? If the surplus dips below zero, it does not trigger instant bankruptcy; ruin only occurs if the process remains below zero at the next observation time (modeled via a Poisson process). Palmowski et al. and Avram et al. utilized scale functions to derive exact mathematical identities for these "Parisian" horizons, calculating the exact probability densities for the timing and state of the system when a boundary is crossed under discrete observation.

Periodic Barriers

In inventory management or dividend distribution, decisions often happen periodically rather than continuously. Pérez et al. and Pérez et al. proved the optimality of periodic barrier strategies for inventory replenishment and dividend payouts. By leveraging scale functions, the authors derived closed-form expressions for expected outcomes even when the underlying demand or supply processes exhibit infinite activity.

Simplifying Optimization via Smooth Fit

How do we find the optimal threshold or barrier to maximize a firm's value or minimize a bank's risk? Traditionally, solving optimal stopping problems requires verifying complex variational inequalities across a free boundary.

Scale functions simplify this process via the continuous and smooth fit principle. As Yamazaki et al. demonstrated, the intuitive requirement that a value function must transition smoothly across a stopping boundary is mathematically equivalent to the first-order condition for maximizing an expected payoff.

Because the value functions are written directly in terms of scale functions, finding the optimal threshold reduces to analyzing the differentiability and smoothness properties of the scale function itself. This transforms a functional optimization problem into a straightforward root-finding exercise.

Computational Tractability: The Phase-Type Shortcut

If scale functions are so powerful, what is the catch? For arbitrary jump distributions, finding an explicit analytical form for the scale function $W(x)$ can be difficult. Practitioners often have to rely on numerical Laplace inversions, which can be unstable near boundaries.

To overcome this computational bottleneck, researchers developed phase-type fitting. Phase-type distributions—formed by a system of interrelated structural exponential stages—can approximate any positive probability distribution with arbitrary precision.

[Start] --> (Stage 1: Exp) --> (Stage 2: Exp) --> [Exit/Jump]
                 |                  ^
                 +------------------+

As Egami et al. and Yamazaki showed, when the jump distributions of a Lévy process are fitted into a phase-type framework, the Wiener-Hopf factorization simplifies significantly. This allows the scale functions to be expressed as elegant, closed-form sums of exponentials.

This phase-type shortcut allows practitioners to compute complex risk metrics—such as the Gerber-Shiu risk function or optimal dividend thresholds—with high accuracy and speed, making scale function analysis a practical tool for real-world financial engineering and risk management.

Key Foundational Work

On an Explicit Skorokhod Embedding for Spectrally Negative Lévy Processes (2008) — contributed 2 concept elements to this direction.
Processes of class Sigma, last passage times and drawdowns (2009) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_scale_function_analysis_b6e4ff0f,
  title = {scale function analysis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/scale-function-analysis/rdr_b6e4ff0fe06c45b8bb7562a9ab6f3e09},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-22},
}

This exact version (version 1, frozen text):

@misc{lacuna_scale_function_analysis_b6e4ff0f_v1,
  title = {scale function analysis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/scale-function-analysis/rdr_b6e4ff0fe06c45b8bb7562a9ab6f3e09/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake