Jump Diffusion vs Variance Gamma

Side-by-Side

When to Use Jump Diffusion

• Event-driven jump pricing. Earnings, FDA approvals, M&A close decisions, and macro prints all produce discrete priced jumps in single-name options. Jump diffusion explicitly models the jump as a discrete Poisson event with calibrated frequency and magnitude. This matches how option markets actually price these events: with a binary-like jump component layered on top of normal-vol diffusion.
• Short-tenor smile capture. At 1-7 DTE, smile is dominated by jump-risk pricing because the diffusion vol has insufficient time to produce smile curvature on its own. Jump-diffusion models capture short-tenor smiles where pure stochastic-vol models fail.
• Bates / SVCJ for full-surface. Bates (1996) combines Heston stochastic vol with Merton jumps. SVCJ (Eraker-Johannes-Polson 2003) adds correlated jumps in spot and variance. Both are practitioner standards for full-surface fitting that captures both diffusion-vol skew and jump-tail premium.
• Decomposing risk premia. By separately calibrating diffusion-vol parameters (sigma) and jump parameters (lambda, mu_J, sigma_J), you can decompose the variance risk premium into its diffusion and jump-tail components. This is operationally useful for risk-management.
• Asymmetric jump structure. Kou's double-exponential jump-diffusion specifically models asymmetric jumps (different probabilities and magnitudes for up-jumps vs down-jumps). This captures the empirical fact that equity downside-jumps are more frequent and larger in magnitude than upside-jumps.

When to Use Variance Gamma

• Long-tenor fat-tail fits. Variance Gamma's pure-jump structure produces excess kurtosis at every horizon, including long-tenor. Pure stochastic-vol models tend to under-capture kurtosis at long-tenor; VG handles this naturally through its variance-time mechanism.
• High-frequency microstructure. The infinite-activity property of VG means it produces continuous-looking sample paths at high resolution while still being a pure jump process. This matches the empirical microstructure of liquid options.
• Closed-form characteristic function. VG has a particularly tractable characteristic function that integrates cleanly via FFT. For batch pricing across thousands of strikes, VG is the fastest fat-tail-aware model.
• CGMY extension for asymmetric tails. The CGMY/CMY extension (Carr-Geman-Madan-Yor 2002) generalizes VG to four parameters that independently control left-tail thickness, right-tail thickness, and tail decay rate. This is the most flexible pure-jump model for asymmetric fat tails.
• Pricing without the diffusion component. VG explicitly removes the Brownian-motion component. This is a strength for pure jump-based research applications where the question is "how much of returns is due to jumps?"

Where They Agree

Both produce fat-tailed return distributions and capture skew. Both calibrate to listed option prices and produce fitted IV surfaces with measurable smile and skew. Both have characteristic-function representations that price options via Fourier-domain methods (FFT, COS, Lewis). Both are arbitrage-free Levy-process specifications.

For ATM short-tenor pricing in calm regimes, the two models often produce nearly identical prices. The differences emerge at OTM strikes (where pure-jump fat tails diverge from diffusion-plus-jump fat tails) and at long-tenor (where pure-jump kurtosis decays at a different rate than diffusion-dominated fat tails).

Where They Diverge

• Path discontinuity. Jump diffusion has continuous diffusion paths interrupted by discrete jumps. VG paths are pure jumps everywhere, with no continuous component. For modeling stop-loss execution risk and barrier-option pricing, this matters: jump-diffusion stops can fire on continuous diffusion moves; VG stops always fire on discrete jumps.
• Variance swap pricing. Both models compute variance swap fair values, but the structure differs. Jump diffusion produces variance swap rates with both diffusion-vol and jump-variance components. VG produces variance swap rates from a pure-jump variance accumulation.
• Smile shape at short-tenor. Jump diffusion can produce sharply asymmetric short-tenor smiles tied to specific jump events (earnings dates). VG produces smoother short-tenor smiles without event-localized features.
• Stochastic-vol coupling. Jump diffusion couples cleanly with stochastic-vol models (Bates, SVCJ). VG couples less cleanly because it lacks a continuous-vol component to be modulated.

Why They're Often Confused

Both are categorized as "fat-tail" or "Levy" models, and both produce skew and smile via jump-based mechanisms. Practitioners sometimes blur the distinction or use the terms interchangeably. The right framing: jump diffusion is "GBM plus jumps" - additive structure. Variance Gamma is "subordinated Brownian motion" - multiplicative structure where time itself is randomized.

The choice between them is rarely either-or in production systems. Bates (Heston + jump-diffusion) is the standard for full-surface stochastic-vol-plus-jumps. CGMY (extended Variance Gamma) is the standard for pure-jump fat-tail research. They serve different niches.

Further Reading

• Jump Diffusion Model Documentation
• Variance Gamma Model Documentation
• Heston Model (couples with jump-diffusion via Bates)
• Tail Risk in Options
• Risk-Neutral Density
• Pricing Model Landscape

References

• Merton, R. C. (1976). "Option pricing when underlying stock returns are discontinuous." Journal of Financial Economics, 3, 125-144. The seminal jump-diffusion paper.
• Kou, S. G. (2002). "A jump-diffusion model for option pricing." Management Science, 48(8), 1086-1101. The asymmetric double-exponential extension.
• Bates, D. S. (1996). "Jumps and Stochastic Volatility: Exchange Rate Processes Implicit in Deutsche Mark Options." Review of Financial Studies, 9(1), 69-107. The Bates stochastic-volatility jump-diffusion model.
• Eraker, B., Johannes, M., and Polson, N. (2003). "The Impact of Jumps in Volatility and Returns." Journal of Finance, 58(3), 1269-1300. The SVCJ reference for correlated jumps in returns and volatility.
• Madan, D. B., Carr, P., and Chang, E. C. (1998). "The Variance Gamma Process and Option Pricing." European Finance Review, 2, 79-105. The canonical VG paper.
• Carr, P., Geman, H., Madan, D., and Yor, M. (2002). "The Fine Structure of Asset Returns." Journal of Business, 75(2), 305-332. The CGMY extension.
• Cont, R. and Tankov, P. (2003). Financial Modelling with Jump Processes. Chapman & Hall. Comprehensive Levy-process treatment.

This is one of the model-vs-model comparison pages. For the full landscape of pricing models and their relationships, see the Pricing Model Landscape.