When Prices Gap: Merton Puts Jumps Into Option Pricing

The Black-Scholes model has a hidden assumption so natural that most people never notice it: the stock price never gaps. It moves continuously, wiggling in infinitesimally small steps, so that if you watch it closely enough it never skips a value on its way from 100 to 95. It passes through 99.99, then 99.98, and so on.

That assumption is what makes perfect hedging possible. If the price moves in tiny increments, you can adjust your hedge in tiny increments, and you never get caught out.

Real prices do not behave this way. A company announces a failed drug trial and the stock opens 40 percent lower. A central bank surprises the market and a currency moves 3 percent in a second. Nobody traded through the intermediate prices. The stock jumped.

In 1976, Robert Merton asked the obvious follow-up question: what does option pricing look like if we admit that prices jump? The answer changed the shape of the whole field, and it introduced an uncomfortable idea that the industry has been living with ever since.

The problem: the hedge that fails exactly when you need it

Delta hedging is the beating heart of the Black-Scholes world. You sell an option, you hold some stock against it, and as the stock drifts you continuously rebalance. Your hedge is calibrated for small moves, because in the model, small moves are the only kind there are.

Now insert a jump. The stock is at 100, you are hedged for a 100-ish world, and overnight it opens at 65. Your hedge, tuned for a gentle drift, is catastrophically wrong. You did not get the chance to rebalance on the way down, because there was no way down, just a gap. The loss is real, it is large, and it is not a rounding error in a theory that promised you a riskless position.

This is not an academic quibble. It is the mechanism behind most of the genuinely painful losses ever suffered on hedged option books. The theory says your risk is zero. The jump says otherwise.

The key idea via analogy: two kinds of weather

Think about the price of a stock as being driven by two different kinds of news.

The first kind is ordinary news: a broker upgrade, a mild change in sentiment, a slight shift in the sector. It arrives constantly, in small doses, and it nudges the price up and down in a continuous shimmer. This is the drizzle. Black-Scholes models the drizzle perfectly.

The second kind is extraordinary news: an earnings miss, a lawsuit, a takeover, a bankruptcy. It does not drizzle. It arrives at random, unpredictable moments, and when it arrives it moves the price by a large, discrete amount. This is the lightning strike.

Merton's model is simply: drizzle plus lightning. The continuous Brownian shimmer of Black-Scholes, plus a separate process that fires off at random times (a Poisson process, the standard mathematical way of describing "rare events arriving at some average rate") and, when it fires, multiplies the stock price by a random factor.

The mathematics of this is elegant. Merton derived a pricing formula that expresses the option's value as a weighted average of Black-Scholes prices: one for the scenario where no jumps occur before expiry, one for exactly one jump, one for two jumps, and so on, each weighted by the probability of that many jumps happening. Each of those Black-Scholes prices uses a volatility and rate adjusted for the jumps already accounted for. The answer is an infinite sum, but it converges fast, and it is entirely computable.

The uncomfortable bit: the hedge is gone

Here is where Merton had to be intellectually honest, and where the paper gets interesting.

The whole Black-Scholes argument depends on being able to build a perfect copy of the option out of stock and cash. With jumps, you cannot. You have one hedging instrument, the stock, and now two sources of risk, the drizzle and the lightning. One dial, two problems. You can neutralise the small continuous moves, but nothing you hold in the stock will protect you against a random-sized gap. The market is incomplete, and no-arbitrage alone no longer pins down a unique option price.

Merton's response was a pragmatic assumption: he supposed that jump risk is diversifiable, that is, that jumps in individual stocks are idiosyncratic company-specific events, uncorrelated with the market as a whole, so that a well-diversified investor can wash them out and would therefore demand no extra premium for bearing them. Under that assumption you can proceed as if jump risk carries no risk premium, and the pricing formula follows.

It is a reasonable assumption for a biotech blowing up on trial results. It is a terrible assumption for the stock market as a whole, where the jumps everybody actually fears, crashes, are precisely the ones that hit every asset simultaneously and cannot be diversified away. Merton knew this. He was clear about what he was assuming. Later researchers spent decades trying to price jump risk properly, precisely because the assumption is the model's weak point.

Why it mattered

• It produced the volatility smile, before the smile existed. Jumps make the distribution of returns fat-tailed and skewed: big moves are far more likely than a bell curve allows. Options far from the money are therefore worth more than Black-Scholes says. Plot the implied volatilities of a jump-diffusion model against strike and you get a smile or a skew. After the 1987 crash, when the real options market started showing exactly that shape, Merton's model was sitting there waiting with an explanation.
• It legitimised discontinuity. It is the direct ancestor of an entire family of models, Kou's double-exponential jumps, Bates's jumps plus stochastic volatility, the whole Levy-process literature. All of them are variations on the drizzle-plus-lightning theme.
• It made incompleteness visible. This is the paper where the industry had to confront the fact that the perfect hedge is a modelling convenience, not a fact about markets. That realisation has arguably done more for risk management than any formula.
• It gave crash risk a language. Traders talk about "gap risk" and "jump-to-default" and price it explicitly. Merton is why those are quantifiable ideas rather than vague anxieties.

The honest limitations

• The diversifiable-jump assumption is the model's soft underbelly. For index options, the jumps that matter are systematic, they hit everything at once, and investors demand a large premium to bear them. Assuming that premium is zero underprices exactly the crash protection that the market cares most about. Empirically, index options price in far more jump premium than Merton's assumption allows.
• You cannot hedge your way out, so the price is not unique. Different reasonable views on how jump risk should be compensated give different prices. The model gives you a number, but that number rests on a judgment call, not on an arbitrage.
• It is hard to calibrate. Jump frequency, average jump size and jump volatility are all parameters you must estimate from data in which, by definition, jumps are rare. Small samples, huge uncertainty. Two analysts can fit the same data and disagree substantially.
• Real jumps do not arrive at a constant average rate. Merton's Poisson process fires at a steady long-run rate. Real markets cluster: crashes come in bunches, and calm periods can go years without one. Later work on self-exciting processes exists because of this gap.
• It still assumes constant volatility between jumps. Real volatility is itself random and persistent, which is why Bates later grafted Heston's stochastic volatility onto Merton's jumps.

The one-line takeaway

Merton added lightning to Black-Scholes' drizzle, showing that once prices are allowed to gap, the perfect hedge that justifies the whole theory quietly stops existing, and along the way he produced the first model that naturally generates fat tails and a volatility smile, at the cost of an assumption about crash risk that nobody quite believes.