Multi-Asset Options: When One Stock Is Not Enough

So far in this series we have been looking at options on a single stock. One underlying, one random walk, one volatility. Life was simple. But the real world is messier. Many popular contracts depend on two, five, or even twenty different assets at the same time. Welcome to the world of multi-asset options.

Chapter 11 of Wilmott’s book extends the Black-Scholes framework to handle multiple assets. The math stays mostly the same, but the new ingredient is correlation between the assets. And as we will see, that one ingredient causes a lot of headaches.

Correlated Random Walks

Remember the lognormal random walk for a single stock? Each asset still follows its own random walk with its own drift and volatility. The difference is that the random shocks driving each asset are no longer independent. They are correlated.

If you have d assets, each one follows:

dS_i / S_i = mu_i * dt + sigma_i * dX_i

The key new piece is that the random increments dX_i and dX_j are correlated with correlation coefficient rho_ij. All these correlations together form a correlation matrix, a symmetric table where every diagonal entry equals 1 (each asset is perfectly correlated with itself) and the off-diagonal entries capture how pairs of assets move together.

In plain terms: if two stocks have a correlation of 0.8, they tend to move in the same direction most of the time. If it is -0.5, they often move in opposite directions. If it is zero, their movements are unrelated.

We also need a multidimensional version of Ito’s lemma to handle functions of many random variables. It looks more intimidating with all the summations, but the idea is exactly the same as before: Taylor expand, apply the rules of thumb for Wiener processes, keep the important terms.

Measuring Correlations Is Hard

Wilmott is blunt about this. Correlations measured from historical data are “notoriously unstable.” If you split your data in half and compute correlations separately for each half, the numbers can differ quite a lot. A 60-day rolling correlation between two stocks can swing wildly over time, much more than volatility does.

You can also try to back out an implied correlation from the market price of a multi-asset option, similar to how we back out implied volatility from a single-asset option. The idea is the same: let the market tell you what it thinks the correlation is. But there are not enough traded multi-asset instruments to pin down all the correlations precisely.

This is important to keep in mind. The whole pricing framework relies on knowing correlations, but they are the least stable parameter we have.

Basket Options and Rainbow Options

Options on multiple underlyings are called basket options or rainbow options. The pricing approach is a direct extension of Black-Scholes:

1. Build a portfolio with the basket option and short some amount of each underlying asset
2. Choose the hedge ratios (one delta per asset) to eliminate all the random terms
3. Set the return equal to the risk-free rate

This gives you the multi-dimensional Black-Scholes equation. The delta for each asset is simply the partial derivative of the option value with respect to that asset’s price, just like in the one-dimensional case.

For European contracts with no path dependency, there is even a pricing formula that looks like the single-asset one but with a multi-dimensional integral. You integrate the payoff against a multivariate normal distribution.

Exchange Options: A Nice Special Case

One elegant example is the exchange option, which gives you the right to swap one asset for another. The payoff at expiry is max(q1 * S1 - q2 * S2, 0).

What makes this special is that you can reduce the two-dimensional problem to a one-dimensional one. By working with the ratio of the two asset prices as your variable, the equation turns into a standard Black-Scholes equation with an adjusted volatility:

sigma’ = sqrt(sigma_1^2 + sigma_2^2 - 2 * rho * sigma_1 * sigma_2)

This is the volatility of the ratio, and it neatly captures the effect of correlation. Higher correlation means lower effective volatility for the ratio, which makes sense: if two assets move together, their ratio does not move much.

One fascinating detail: the hedged portfolio for an exchange option has zero value. You are exactly long one and short the other in just the right amounts.

Quantos: Cross-Currency Contracts

A quanto is a contract where the payoff depends on an asset in one country but is paid in another currency. For example, a call option on the Nikkei Dow index, but paid in US dollars.

You might think you need to worry about the exchange rate a lot. But it turns out that for the simplest quantos, the only effect of the cross-currency feature is an adjustment to the dividend yield. The effective dividend yield becomes:

D_effective = D_original + rho * sigma_N * sigma_$

Where rho is the correlation between the asset and the exchange rate, sigma_N is the volatility of the Nikkei, and sigma_$ is the volatility of the exchange rate.

So pricing a quanto is basically the same as pricing a regular option with a tweaked dividend yield. The correlation between the underlying asset and the exchange rate does matter, but it enters in a surprisingly simple way.

The Reality: Easy, Medium, and Hard Problems

Wilmott is practical about what you can actually compute. He breaks multi-asset problems into three categories:

Easy problems have closed-form solutions or can be reduced to integrals that are straightforward to compute numerically. European non-path-dependent contracts usually fall here.

Medium problems have low dimensionality (three or four assets) and can be solved with finite-difference methods. For higher dimensions, Monte Carlo simulation works well. Both handle path-dependent features reasonably.

Hard problems combine high dimensionality (needing Monte Carlo) with early exercise features (needing finite differences). There is no numerical method that handles both well at the same time. This is still an active area of research.

Hedging Is Even Harder Than Pricing

Even if you can compute accurate prices and Greeks, the hedge ratios are very sensitive to the correlation values. And as we discussed, correlations are hard to measure reliably. So your deltas are probably not accurate, which makes delta hedging basket options a risky business.

You might try to hedge correlation risk using other derivatives, similar to vega hedging for volatility risk. But there simply are not enough traded contracts that depend on the specific correlations you need.

Correlation vs. Cointegration

Wilmott ends the chapter with an interesting aside. Correlation measures how two time series move together in the short term. Cointegration is a different statistical concept: two series are cointegrated if some linear combination of them stays close to a constant mean. In simple terms, the two assets never stray too far from each other, even if they wander around a lot individually.

Cointegration is probably a more robust measure of how two financial quantities are linked. But so far, there is not much derivatives theory built on it. It is an idea worth watching.

Key Takeaways

Multi-asset options are a natural extension of everything we have learned so far. The theory is clean. The new ingredient is correlation. But in practice, correlation is the shakiest parameter in the whole framework, which makes pricing uncertain and hedging unreliable. The mathematical elegance of the exchange option and the simplicity of the quanto formula are nice, but for more complex baskets, you are often stuck with numerical methods and hoping your correlation estimates are not too far off.

The practical lesson here: if someone shows you a beautifully precise price for a complex basket option, ask them where they got their correlations. That is usually where the real risk lives.

Previous post: Probability Density Functions

Next post: Delta Hedging in Practice: Implied vs Actual Volatility