Merton Jump Diffusion: Pricing Gap Risk in Options

Key Takeaways

• Merton jump diffusion adds a compound Poisson process to Brownian motion; the European price is an infinite weighted sum of Black-Scholes prices by jump count.
• A biotech OTM put at 30-vol under Black-Scholes can be many times more expensive under Merton once jump intensity and size are calibrated to FDA event risk.
• A common mistake: using realized historical volatility as the diffusion parameter, historical vol includes past jumps and double-counts them if used as σ alone.
• Jump-diffusion explains short-dated OTM option skew that stochastic volatility alone cannot capture; combining the two is the SVJ model.

What It Is

Robert Merton published "Option Pricing When Underlying Stock Returns Are Discontinuous" in the Journal of Financial Economics in 1976. The model superimposes a compound Poisson jump process on the geometric Brownian motion of Black-Scholes. Between jumps, the underlying drifts and diffuses as usual. At Poisson arrival times, it takes a lognormal-sized jump.

Merton derived a semi-closed-form solution for European options as an infinite weighted sum of Black-Scholes prices, each corresponding to a specific number of jumps before expiration, weighted by the Poisson probability of that jump count.

The Intuition

Black-Scholes models prices as if they never skip. A real stock chart is full of gaps: earnings misses, FDA rulings, takeover announcements, macro surprises. These are discrete shocks, not small continuous moves. The lognormal distribution that Black-Scholes assumes has thin tails, so it assigns near-zero probability to large single-day moves that actually happen a few times per decade.

Jump-diffusion fixes that by separating the return process into two pieces. Small day-to-day noise is captured by the diffusion component. Rare, discrete shocks are captured by the jump component. Tune the jump intensity and size, and you get leptokurtic returns with fatter tails that match what data shows. The model prices the "gap risk" that Black-Scholes ignores, which is especially important for short-dated out-of-the-money options.

How It Works

Merton's stochastic differential equation for the underlying is:

dS_t / S_t = (mu - lambda * k) * dt + sigma * dW_t + (J - 1) * dN_t

Where:

• sigma * dW_t is the Black-Scholes diffusion term
• dN_t is a Poisson process with intensity lambda (expected jumps per year)
• J is the random jump size multiplier, typically lognormal with parameters mu_J and sigma_J
• k = E[J - 1] is the expected relative jump size, subtracted from the drift to keep the martingale property

The parameters of interest are:

lambda   = expected number of jumps per year
mu_J     = mean of log-jump size
sigma_J  = standard deviation of log-jump size
sigma    = diffusion volatility (not total volatility)

The European call price is an infinite sum:

C_Merton = sum over n of [ (exp(-lambda' * T) * (lambda' * T)^n / n!) * BS_Call(S, K, T, r_n, sigma_n) ]

Where lambda' = lambda * (1 + k), and r_n and sigma_n are adjusted for the average effect of exactly n jumps before expiration. In practice the sum is truncated at around 30 to 50 terms; beyond that the Poisson weights are negligible.

Worked Example

Suppose a biotech trades at 50 before an FDA decision that is two weeks away. Historical vol is 40 percent. The market expects a rare but large move on the announcement. Calibrate Merton with sigma = 0.30 (diffusion), lambda = 2 (two expected jumps per year on news events), mu_J = -0.05, and sigma_J = 0.20 (log-jumps roughly 5 percent down on average with 20 percent dispersion).

An OTM put with strike 40 expiring in one month has almost zero Black-Scholes value at 30 percent vol because the required move is many standard deviations. Under Merton, the jump component assigns meaningful probability to a single 20 percent down-gap that would put the strike in the money. The Merton price for that put is therefore many multiples of the Black-Scholes price. Back-solving for the Black-Scholes IV that would match the Merton price produces a much higher number than 30, which is precisely the volatility smile that appears in real event-driven names.

Common Mistakes

1. Using historical volatility as the diffusion parameter. Historical vol captures the combined effect of diffusion plus past jumps. Plugging it in as sigma alone double-counts jumps and inflates the model. Calibrate sigma against a filtered series that removes jump days, or calibrate all parameters jointly to the vanilla smile.

2. Over-fitting lambda to one event. A single FDA approval or earnings miss is not a statistically reliable estimate of long-run jump intensity. Traders who calibrate lambda off one sample size overestimate expected jump frequency and overprice OTM options going forward.

3. Ignoring jump risk premium. The risk-neutral jump intensity is usually higher than the realized physical intensity because investors pay a premium to insure against jumps. Using historical jump counts to price options underestimates the risk-neutral lambda and understates OTM premiums.

4. Assuming jumps are symmetric. Equity jumps are mostly to the downside. A symmetric jump distribution produces too little skew. Most calibrations set mu_J < 0 with moderate sigma_J to match observed put skew.

5. Treating Merton as a substitute for stochastic volatility. Merton handles sudden shocks. It does not handle slow-moving vol clustering. Combining jump-diffusion with a Heston-style variance process (the SVJ model, Bates 1996) tends to fit real markets better than either component alone.

Frequently Asked Questions

Q: What is the Merton jump-diffusion model in simple terms? Merton's model adds sudden, random jumps to the usual smooth stock-price diffusion. The underlying drifts and bounces continuously most of the time, but occasionally gaps by a large amount on news. This prices the fat tails that Black-Scholes ignores.

Q: How does the jump-diffusion model affect investment decisions? It correctly prices short-dated out-of-the-money options that Black-Scholes dramatically undervalues. For biotech stocks before FDA rulings, or any name with known jump events, Black-Scholes premiums are too cheap and Merton provides a more realistic floor.

Q: What is a real-world example of jump-diffusion pricing? A biotech at 50 before an FDA decision. Black-Scholes at 30% vol gives near-zero value to a 40-strike put. Merton with jump intensity 2 and mean log-jump -0.05 assigns meaningful probability to a 20-percent gap down, producing a put premium that matches the market.

Q: How can investors use the jump-diffusion framework practically? Treat the Merton price as a lower bound for any option near a known event date. If the market premium exceeds even the Merton price, implied vol is genuinely rich. If the market is below Merton, the option may be underpriced for its true jump risk.

Q: How is the Merton jump-diffusion model different from the Heston model? Merton handles sudden discrete shocks; Heston handles slowly varying stochastic volatility. Real markets have both, that is why practitioners often combine them in the SVJ (stochastic volatility plus jumps) framework attributed to Bates (1996).

Sources

1. Merton, R.C. (1976). "Option Pricing When Underlying Stock Returns Are Discontinuous." Journal of Financial Economics, 3(1-2), 125-144. https://www.sciencedirect.com/science/article/abs/pii/0304405X76900222
2. Bates, D.S. "Pricing Options Under Jump-Diffusion Processes." University of Iowa. https://www.biz.uiowa.edu/faculty/dbates/papers/chapter3.pdf
3. Kou, S.G. "A Jump-Diffusion Model for Option Pricing." Columbia University. http://www.columbia.edu/~sk75/MagSci02.pdf
4. QuantNext. "The Merton Jump Diffusion Model." https://quant-next.com/the-merton-jump-diffusion-model/

Disclaimer

This article is educational content only and is not financial advice. Nothing here is a recommendation to buy, sell, or hold any security. Consult a licensed advisor before making investment decisions.