Modeling Discontinuous Jumps in Continuous Stochastic Systems

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 2176 papers249 conceptsfrozen 2026-10-04

Most classical models of the world assume a certain level of smoothness. In physics, finance, and biology, we often represent the evolution of a system using Stochastic Differential Equations (SDEs) driven by Brownian motion. These models produce continuous paths—trajectories that never "teleport" from one state to another without passing through every point in between.

However, reality is frequently discontinuous. A stock price doesn't just slide down during a market crash; it gaps downward. An insurance fund doesn't experience a smooth trickle of claims; it faces discrete, large-scale shocks. A social media post doesn't gradually gain traction; it spikes following a specific event. This research direction focuses on integrating these abrupt, non-linear "jumps" into our continuous-time mathematical frameworks.

The Failure of the Continuity Assumption

The assumption of continuous price paths is the bedrock of the Black-Scholes era of finance. But as Mark H. Davis notes in Market completion using options (Banach 2008), this fails to account for sudden market shocks. When a model assumes continuity, it assumes that if a price moves from $100 to $90, it must have spent some infinitesimal amount of time at $95.

This creates a "barrier problem." In Robust hedging of double touch barrier options, Alexander M. G. Cox and Jan Obloj explain that if an asset price jumps, it can bypass a "barrier" (a price level that triggers a trade) without ever activating the necessary hedge. This "overshoot" makes traditional delta-hedging strategies fail, as the system state leaps over the very thresholds designed to protect it.

From Brownian Motion to Lévy Processes

To fix this, researchers move from simple diffusion to Lévy processes—mathematical models that allow for both continuous drift and discrete leaps. Marcel Nutz demonstrates in Power Utility Maximization in Constrained Exponential Lévy Models that while these processes better reflect real-world asset prices, they introduce significant complexity. You can no longer rely on simple calculus; you must account for the "jump intensity" and the distribution of jump sizes.

In credit risk, this distinction is a matter of solvency. Kazutoshi Yamazaki and Masahiko Egami argue in Precautionary Measures for Credit Risk Management in Jump Models that modeling a bank's net worth with jumps is essential. A continuous model might suggest a bank has time to raise capital as its value nears a regulatory limit, but a jump model reveals the risk of an "abrupt default" where the value drops below the limit instantly, leaving no window for intervention.

Modeling Endogenous and Exogenous Shocks

Modern research has extended these ideas into machine learning and high-frequency trading. The challenge is determining when a jump will occur.

  1. Temporal Point Processes: In Bellman Meets Hawkes, Hongyuan Mei and colleagues use Neural Hawkes Processes to model systems where events are asynchronous and self-exciting. Unlike smooth SDEs, this framework explicitly models the timing of discrete events—like a sudden spike in activity after a single viral post.
  2. Neural SDEs with Jumps: For trading, Mengyuan Yang introduced SpotlightTrader in Spotlight News Driven Quantitative Trading Based on Trajectory Optimization. This model treats market reactions to news not as a trend, but as a discontinuous jump within a stochastic differential equation. This allows the agent to "plan" for the non-linear volatility that follows a major news release.

The Estimation Problem

If jumps are part of the system, how do we distinguish a "jump" from mere high-frequency noise? Axel Munk and Johannes Schmidt-Hieber address this in Nonparametric estimation of the volatility function in a high-frequency model corrupted by noise. They show that in the presence of "microstructure noise" (the tiny, random fluctuations in tick data), standard estimators fail. We need spectral decomposition methods to separate the smooth volatility of the market from the genuine, discrete jumps that signal a change in fundamental value.

Similarly, Liang Peng and Cindy Yu suggest in Parameter estimation and model testing for Markov processes via conditional characteristic functions that "infinite-activity" jump models—where a system experiences a continuous stream of tiny shocks—often provide a better fit for interest rate data than models assuming smooth transitions.

Why It Works: Capturing the "Overshoot"

The transition from continuous to jump-diffusion modeling works because it acknowledges the information asymmetry of the real world. In a continuous world, information is revealed perfectly and incrementally. In a jump-prone world, information (like a court ruling or a default) arrives in discrete packets that revalue an entire system instantly.

By using tools like the "continuous and smooth fit" principles for Lévy models, as explored by Kazutoshi Yamazaki in On the continuous and smooth fit principle for optimal stopping problems in spectrally negative Levy models, researchers can finally calculate the exact moment to act in a world that doesn't always move in a straight line.

Papers in this version

Showing 60 of 176 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 149 papers
  2. Version 1 · 2026-10-04 · 176 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_jump_diffusion_modeling_4132f03c,
  title = {jump-diffusion modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/jump-diffusion-modeling/rdr_4132f03c88984376aa949506beaf4e6a},
  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_modeling_discontinuous_jumps_in_continuo_4132f03c_v1,
  title = {Modeling Discontinuous Jumps in Continuous Stochastic Systems},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/modeling-discontinuous-jumps-in-continuous-stochastic-systems/rdr_4132f03c88984376aa949506beaf4e6a/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-22},
}