Risk-Neutral Measure for Quant Interviews: Pricing, Girsanov, and Why μ Disappears
The risk-neutral measure is one of the most subtle concepts in quant finance and one of the most misunderstood by candidates. The textbook line — “under the risk-neutral measure, all assets earn the risk-free rate” — is correct but misleading without context. Understanding why this transformation works, where it comes from, and what it implies for derivatives pricing is what separates strong quant-research candidates from those who can compute Black-Scholes mechanically without understanding the foundation. This guide covers what gets tested and what’s just deep theory.
The Setup: Why We Need Two Measures
In financial markets, asset prices have two perspectives:
- Physical (or “real-world”) measure P: the actual probability distribution of future prices, reflecting investors’ beliefs and risk preferences. Stocks have expected return μ (typically greater than the risk-free rate r) because investors demand compensation for risk.
- Risk-neutral (or “pricing”) measure Q: a transformed probability distribution under which every traded asset has expected return equal to the risk-free rate r. This isn’t a “true” probability; it’s a mathematical construct used for pricing.
Why do we need Q? Because pricing derivatives by replication doesn’t depend on individual investors’ risk preferences. The no-arbitrage requirement implies the existence of an “equivalent martingale measure” Q under which discounted prices are martingales. Pricing under Q gives derivative values that are arbitrage-free regardless of how risk-averse different investors are.
The Core Result
Under the risk-neutral measure Q, the price of any contingent claim with payoff f(S_T) at time T is:
V(0) = e^(-rT) E^Q[f(S_T)]
Discount the expected payoff under Q at the risk-free rate. This is the foundation of all derivatives pricing in continuous-time models.
Under Q, the stock SDE becomes (for a non-dividend-paying stock following GBM under P):
- Under P: dS = μS dt + σS dW
- Under Q: dS = rS dt + σS dW̃
Where W̃ is Brownian motion under Q, related to W by W̃ = W + (μ – r)/σ × t (this is the Girsanov change of variables).
Where Does Risk-Neutrality Come From?
The intuition: in a complete, arbitrage-free market, we can replicate any derivative with a self-financing portfolio of the underlying and cash. The price of the derivative must equal the cost of replication, which doesn’t depend on investors’ risk preferences — it depends only on the dynamics of the underlying and the cost of cash (r).
Mathematically: the no-arbitrage requirement is equivalent to the existence of an equivalent measure under which discounted asset prices are martingales. This measure Q is unique in complete markets and exists by the First Fundamental Theorem of Asset Pricing.
Why μ disappears
Under P, the stock has drift μ — investors demand compensation for risk. Under Q, the drift is r. The difference (μ – r)/σ is the “market price of risk” or “Sharpe ratio of the market”; the change of measure absorbs this into the new Brownian motion W̃.
The economic interpretation: in pricing a derivative by replication, you only care about how the underlying moves relative to the risk-free benchmark. The drift “above” the risk-free rate (the risk premium) doesn’t affect the cost of replication, so it doesn’t affect the derivative price. The risk-neutral transformation strips out this risk premium; what remains is the volatility, which determines how much hedging will cost.
Girsanov’s Theorem (Just Enough)
Girsanov’s theorem formalizes the change of measure. If W is a Brownian motion under P, and we define a new measure Q via the Radon-Nikodym derivative dQ/dP = exp(-∫θ dW – (1/2)∫θ² dt) for some adapted process θ, then W̃(t) = W(t) + ∫θ ds is a Brownian motion under Q.
For Black-Scholes, θ = (μ – r)/σ (the market price of risk). The change of measure shifts the Brownian motion by this amount, transforming the stock drift from μ to r.
For interview purposes: you don’t need to derive Girsanov from measure theory. You need to know that the change of measure exists, is given by Radon-Nikodym derivative, and shifts the drift of Brownian motion by the market price of risk.
Risk-Neutral Pricing of Common Derivatives
European call option
V(0) = e^(-rT) E^Q[max(S_T – K, 0)]
Under Q, S_T follows lognormal: ln(S_T) ~ N(ln S_0 + (r – σ²/2)T, σ²T). The expectation can be evaluated in closed form, yielding the Black-Scholes formula.
Forward contract
V(0) = e^(-rT) E^Q[S_T – K] = e^(-rT) (E^Q[S_T] – K) = e^(-rT) (S_0 e^(rT) – K) = S_0 – K e^(-rT)
The forward value at initiation equals zero when K = S_0 × e^(rT) (the no-arbitrage forward price). This is consistent with the cost-of-carry derivation; risk-neutral pricing recovers it.
Put option (via put-call parity)
V_put = e^(-rT) E^Q[max(K – S_T, 0)] = V_call – S_0 + K e^(-rT). Put-call parity falls directly out of risk-neutral pricing.
Common Interview Questions
Explain risk-neutral pricing
“What’s the risk-neutral measure and why is it useful?” Articulate: it’s a probability measure under which all assets earn the risk-free rate. Useful because pricing derivatives by replication doesn’t depend on investor risk preferences, only on volatility and the risk-free rate. Strong candidates discuss the no-arbitrage foundation: existence of Q is equivalent to absence of arbitrage.
Show E^Q[S_T] = S_0 × e^(rT)
Under Q, S follows GBM with drift r. So S_T = S_0 × exp((r – σ²/2)T + σ W̃_T). Take expectation under Q: E^Q[exp(σ W̃_T)] = exp(σ²T / 2) (lognormal formula). So E^Q[S_T] = S_0 × exp((r – σ²/2)T + σ²T/2) = S_0 × e^(rT). The σ²/2 corrections cancel.
Explain why μ disappears
Under risk-neutral measure, stocks are priced as if they earn r, not μ. The transformation absorbs (μ – r)/σ into a shifted Brownian motion via Girsanov. Economic intuition: pricing by replication only cares about volatility and financing cost, not risk premium.
Compute E^Q[max(S_T – K, 0)] (call option price)
Set up the integral: E^Q[max(S_T – K, 0)] = ∫(S_T – K)+ dQ. Substitute the lognormal density for S_T under Q, change variables, recognize the standard normal cumulative distribution function. The result is the Black-Scholes formula. Strong candidates can outline this derivation; full mechanical execution is rare in interviews.
Discuss incomplete markets
“What if the market isn’t complete?” Multiple equivalent martingale measures exist; Q is not unique. Different measures give different “fair” prices for the same derivative. Practical implication: in incomplete markets, replication doesn’t pin down a unique price, and additional assumptions (utility maximization, equilibrium models) are needed. Strong candidates know this is a non-trivial complication and don’t blithely apply Q in non-traded markets.
Common Misunderstandings
“Risk-neutral measure means investors are risk-neutral”
No. The risk-neutral measure is a mathematical construct; investors are still risk-averse. The measure transforms probabilities for pricing purposes. The actual probability of stock outcomes under the physical measure includes a risk premium reflecting investor risk aversion.
“Q is the ‘true’ probability”
No. Q is a pricing measure derived from no-arbitrage requirements. It’s not what we believe will actually happen; it’s how we should price derivatives consistently. The physical measure P is closer to “what we believe.”
“All assets earn r under Q, so no investor would hold stocks”
This conflates the two measures. Investors care about returns under P (where stocks earn μ > r); pricing happens under Q (where stocks earn r in the formula). Both are simultaneously correct in their respective contexts.
Frequently Asked Questions
How deep does risk-neutral measure understanding need to go?
For derivatives-pricing roles at investment banks (Goldman Strats, JPMorgan Markets Quants, Morgan Stanley), you should understand the change of measure intuitively, know Girsanov’s theorem at the level of “shift Brownian motion by market price of risk,” and price common derivatives via E^Q. Full measure-theoretic foundations rarely come up. For systematic hedge fund roles (Two Sigma, D. E. Shaw, Cubist), the topic is peripheral — statistical signal generation matters more than continuous-time pricing. Match prep to your target firms.
What’s the Radon-Nikodym derivative and do I need to know it?
The Radon-Nikodym derivative dQ/dP measures how Q reweights P. For Black-Scholes, dQ/dP = exp(-θ W_T – (1/2) θ² T) where θ = (μ – r)/σ. You should know that this derivative defines the change of measure and that Girsanov’s theorem uses it to transform Brownian motion. You don’t need to derive properties of the derivative or compute it for novel models; that’s measure-theoretic detail that doesn’t typically come up.
What’s the connection between risk-neutral pricing and replicating portfolios?
They’re equivalent in complete markets. The price of a derivative equals (1) the expected discounted payoff under Q, and (2) the cost of constructing a replicating portfolio. Both give the same answer because no-arbitrage requires it. If they gave different answers, you could arbitrage by buying the cheaper and selling the more expensive. The two perspectives are useful in different ways: risk-neutral pricing gives the formula; replicating portfolios give the hedge.
What books should I use for risk-neutral pricing prep?
Shreve’s Stochastic Calculus for Finance II chapters 5–6 cover risk-neutral pricing rigorously but accessibly. Hull’s Options, Futures, and Other Derivatives covers the same ground less rigorously but more readably. Tomas Björk’s Arbitrage Theory in Continuous Time is a deeper academic treatment. For interviews, Shreve II is the reference everyone studies; Hull is the accessible alternative.
How do real-world quant traders use risk-neutral measure ideas?
Constantly, often without naming the framework. Pricing OTC derivatives, calibrating volatility surfaces, computing implied parameters from market prices — all are risk-neutral pricing operations. The risk-neutral framework is the lingua franca of derivatives. Strong quants speak it fluently; those who only understand it as a textbook concept struggle with calibration, hedging, and inter-derivative pricing relationships.
See also: Stochastic Calculus for Quant Interviews • Options Pricing for Quant Interviews • Monte Carlo Methods for Quant Interviews