17.3 The Black-Scholes-Merton Model
Key Takeaways
BSM assumes geometric Brownian motion: terminal stock prices are lognormal and continuously compounded returns are normal with variance σ²T.
Historical volatility is the sample standard deviation of log returns (often annualized by √frequency); realized return over [0,T] is ln(S_T/S_0).
European calls and puts on non-dividend stocks use closed-form N(d1)/N(d2) formulas; puts follow from put-call parity.
Implied volatility is the σ that equates a BSM price to the market quote; dividends and early exercise break plain BSM for Americans.
Yield, FX, and futures Europeans use Black–Scholes–Merton with carry adjustments (q, r_f, or Black ’76 on forwards); warrants require dilution-adjusted share counts.
The Black-Scholes-Merton Model
The Black–Scholes–Merton (BSM) model is the continuous-time limit of the binomial model for European options when the underlying follows geometric Brownian motion. FRM Part I expects you to know the distributional story, the assumptions, how to use the call/put formulas conceptually, what implied volatility means, and how dividends, futures, FX, and warrants change the setup.
Lognormal Prices and Normal Returns
Under GBM, dS = μ S dt + σ S dW. Integrating gives
ln(S_T / S_0) ~ Normal( (μ − σ²/2) T , σ² T )
so S_T is lognormal. Equivalently, the continuously compounded return over horizon T is normal. The −σ²/2 adjustment is the Itô correction: E[S_T] = S_0 e^(μ T), not S_0 e^((μ−σ²/2)T).
| Object | Distribution under GBM |
|---|---|
| S_T | Lognormal |
| ln(S_T / S_0) | Normal |
| Arithmetic return (S_T − S_0)/S_0 | Not normal (skewed) |
Worked lognormal probability
S0 = 100, μ = 0.10, σ = 0.20, T = 1. Then ln(S_T/100) ~ N(0.10 − 0.5×0.04, 0.04) = N(0.08, 0.04). P(S_T > 100) = P(ln(S_T/100) > 0) = P( Z > (0 − 0.08)/0.20 ) = P(Z > −0.40) ≈ 0.655 where Z is standard normal. Median of S_T is 100 e^0.08 ≈ 108.3; mean is 100 e^0.10 ≈ 110.5.
Realized Return and Historical Volatility
Realized (continuously compounded) return over a sample interval is r_i = ln(S_i / S_(i−1)).
Historical volatility: compute the sample standard deviation s of the r_i series, then annualize. If observations are daily and there are 252 trading days,
σ_annual ≈ s × √252
For weekly data, multiply by √52. Use n − 1 in the sample variance denominator unless the question specifies otherwise.
Worked historical vol
Five daily log returns: 0.012, −0.008, 0.005, 0.010, −0.004. Mean m = 0.003. Squared deviations ≈ 0.000081, 0.000121, 0.000004, 0.000049, 0.000049; sum = 0.000304. Sample variance = 0.000304 / 4 = 0.000076; s ≈ 0.00872. Annualized σ ≈ 0.00872 × √252 ≈ 0.00872 × 15.87 ≈ 13.8%.
BSM Assumptions
Classic assumptions (exam checklist):
- Underlying follows GBM with constant μ and constant σ.
- No dividends (original Black–Scholes) or continuous yield q (Merton extension).
- Constant risk-free rate r; continuous compounding.
- No transaction costs or taxes; unlimited short selling; continuous trading.
- No arbitrage; European exercise only (for the closed form).
- Lognormal terminal prices / normal log returns.
Violations (stochastic vol, jumps, discrete dividends, early exercise) push you to trees, local/stochastic-vol models, or numerical PDE methods.
European Non-Dividend Call and Put
With continuous rates and no dividends,
c = S0 N(d1) − K e^(−rT) N(d2)
p = K e^(−rT) N(−d2) − S0 N(−d1)
where
d1 = [ln(S0/K) + (r + σ²/2) T] / (σ √T)
d2 = d1 − σ √T
N(·) is the standard normal CDF. Interpretation: N(d2) is itself the risk-neutral probability that the call finishes in-the-money (P*(S_T > K)); multiplying by e^(−rT) turns it into the price of a cash-or-nothing digital paying $1 if S_T > K. Do not conflate the two—N(d2) is the probability, e^(−rT)N(d2) is the discounted value of that bet. N(d1) is the corresponding probability under the stock numeraire and equals the call's delta for a non-dividend stock.
Worked conceptual calculation
S0 = 49, K = 50, r = 5%, σ = 20%, T = 0.3846 (≈ 20 weeks).
σ √T = 0.20 × √0.3846 ≈ 0.20 × 0.620 ≈ 0.124.
ln(S0/K) = ln(0.98) ≈ −0.0202.
(r + σ²/2)T = (0.05 + 0.02) × 0.3846 ≈ 0.0269.
d1 = (−0.0202 + 0.0269) / 0.124 ≈ 0.054; d2 ≈ 0.054 − 0.124 = −0.070.
N(d1) ≈ 0.5216; N(d2) ≈ 0.4721.
c ≈ 49 × 0.5216 − 50 e^(−0.05×0.3846) × 0.4721 ≈ 25.56 − 50 × 0.981 × 0.4721 ≈ 25.56 − 23.16 ≈ 2.40.
Put from parity: p = c + K e^(−rT) − S0 ≈ 2.40 + 49.05 − 49 ≈ 2.45.
You are not expected to memorize CDF tables beyond rough N(0) = 0.5, but you must know the formula structure and how inputs move price.
Implied Volatility
Implied volatility (IV) is the constant σ plugged into BSM that reproduces the observed market price. It is a quote convention: options are spoken in vol points, not dollars. Because one σ cannot fit all strikes, markets show a smile/skew: IV(K, T) varies with strike and maturity. Historical vol looks backward; IV looks forward through prices.
| Volatility type | Source | Use |
|---|---|---|
| Historical / realized | Past returns | Risk, backtests |
| Implied | Invert BSM from market price | Pricing relative value, forecasting |
| Forward / local | Model surface | Exotic pricing |
If two otherwise identical options imply different σ, at least one assumption (constant vol, diffusion-only) is wrong—or quotes are stale/misaligned.
Dividends and Early Exercise
Continuous yield q (Merton): replace S0 with S0 e^(−qT) in the formulas (or use d1, d2 with r − q):
c = S0 e^(−qT) N(d1) − K e^(−rT) N(d2)
Discrete known dividends: replace S0 with S0 − PV(dividends) for Europeans, then apply non-dividend BSM on the escrowed stock. This fails for Americans when early exercise around ex-dividend dates matters—use a tree.
Early exercise: BSM is for Europeans. American calls on non-dividend stocks equal Europeans (never exercise early). American puts (and American calls with large dividends) need numerical methods; BSM understates American put value.
Dividend, Futures, and FX Europeans
| Contract | Formula family | Key substitution |
|---|---|---|
| Index option | Merton BSM | Dividend yield q |
| FX option (Garman–Kohlhagen) | Merton BSM | q → foreign rate r_f |
| Futures option (Black ’76) | Black on forward/futures | Discount F0 e^(−rT) N(d1) − K e^(−rT) N(d2); d1 uses ln(F0/K) and σ²/2 |
Black ’76 is the market standard for options on futures and many interest-rate options quoted on a forward underlying. Carry is embedded in F0; you do not separately estimate q.
Worked FX mapping
USD call on EUR (right to buy EUR for K dollars). Domestic rate = USD rate r, foreign rate = EUR rate r_f. Same algebra as an index with q = r_f. Higher foreign rates lower the FX call (EUR is “costlier to carry”), analogous to higher stock dividends.
Warrants and Dilution
A warrant is a long-dated call issued by the company on its own stock. When exercised, the firm issues new shares, diluting existing equity. If a firm has N shares outstanding and M warrants with strike K, upon exercise the firm receives M·K cash and the equity claim is split among N+M shares. A common dilution adjustment prices a claim on the firm’s assets with
effective underlying proportional to N/(N+M)
and scales the warrant value accordingly (Black–Scholes on firm value, multiply by N/(N+M) per warrant, under stylized assumptions). Exam cue: ordinary exchange-traded calls do not dilute; warrants do—so warrant value < undiluted BSM call on the stock, all else equal.
Worked dilution sketch
N = 10 million shares, M = 1 million warrants, stock price 50 as if no warrants. Rough dilution factor N/(N+M) = 10/11 ≈ 0.909. If an undiluted European call would be 8.00, a stylized diluted warrant might be worth about 8.00 × 0.909 ≈ 7.27 (actual formulas use asset value and simultaneous equity/warrant valuation—direction matters more than the exact 7.27).
Synthesis
BSM = lognormal S_T + European exercise + constant σ,r (+ optional continuous yield). Use N(d1), N(d2) formulas for non-dividend Europeans; invert for IV; switch to trees for Americans/discrete dividends; map q, r_f, or Black ’76 for indices/FX/futures; dilute for warrants.
Under the BSM geometric Brownian motion assumption, which statement is correct?
Terminal stock price S_T is lognormal and continuously compounded returns are normal
Terminal stock price S_T is normally distributed and ln(S_T/S_0) is lognormal
Both S_T and arithmetic returns are normally distributed
Volatility must be stochastic for S_T to be lognormal
Daily log-return sample standard deviation is 1.2%. Using 252 trading days, approximate annualized historical volatility is:
1.2%
19.0%
30.2%
7.6%
Implied volatility is best defined as:
The forecast from a GARCH model fitted to past returns
The standard deviation of implied dividends
The σ that makes the BSM (or Black) formula match the observed option market price
Always equal to historical volatility at the same horizon
Why is the plain BSM European formula generally inappropriate for a warrant on the issuer’s own stock?
Warrants never expire
Warrants are always American puts
BSM cannot handle continuous dividend yields
Exercise of warrants typically dilutes existing shareholders, so valuation must adjust for new shares issued
Sections you finish are checked off in the contents.