The Black-Scholes Option pricing model – Quant Finance
5 March 2022In the early 1970s, Fischer Black, Myron Scholes, and Robert Merton achieved a major breakthrough in the pricing of European stock Options. This was the development of what has become known as the Black–Scholes–Merton model (or simply Black–Scholes model). The model became very important and in 1997, Scholes and Merton were awarded the Nobel prize for economics (Black died in 1995).
The model has had a huge influence on the way that traders price and hedge derivatives. In this post we will firstly see how to derive the Black-Scholes (BS) differential equation satisfied by the price of vanilla stock Options.
Then we will also explicitly derive the BS formulas for both European Call and Put Options.
Assumptions
Let us first set up the principal assumptions.
- The stock price (underlying), $S_t$, follows the Geometric Brownian Motion with $\mu$ and $\sigma$ constant: $$ \label{eq:geom:brow:mot} dS_{t} = \mu S_{t}\,dt + \sigma S_t \, dB_{t}. \qquad\qquad (1) $$ See here for further details. In turn, this implies that the underlying price $S_t$ follows a log-normal distribution. See here.
- The short selling of securities with full use of proceeds is allowed.
- There are no transaction costs or taxes and all securities are perfectly divisible.
- There are no dividends during the life of the derivative.
- Security trading is continuous.
- The risk-free rate of interest, $r$, is constant and the same for all maturities.
The idea: Building a riskless portfolio
The idea behind the derivation of the BS differential equation satisfied by Option prices involves setting up a riskless portfolio consisting of one position in the derivative and one position in the underlying. In case of no arbitrage opportunities, this means that the total return from such a portfolio must be the risk-free interest rate $r$.
A riskless portfolio can be set up since the stock price and the Option price are both affected by the same source of uncertainty: the stock price movements.
In any short period of time, the price of the derivative is in fact perfectly correlated with the price of the underlying stock. Hence, if an appropriate strategy involving the stock and the derivative is set up, the gain (resp. loss) from the stock position will be offset by the loss (resp. gain) from the derivative position. In this way, the value of the portfolio at the end of the short period of time is known with certainty.
For example, suppose that at certain time, the relationship between the small change $\Delta S$ in the underlying price and the corresponding small change $\Delta C$ in the price of, e.g., a European Call Option is given by: $$ \Delta C = 0.7\,\Delta S. $$ Then, a riskless portfolio is constructed as follows:
- One long position in 70 shares of the underlying stock.
- A short position in 100 Call Options.
For instance, suppose that the underlying price increases by $5$ cents, that is, $\Delta S = +\$0.05$. We have that \[ \Delta C = 0.7 \cdot (+0.05) = +\$0.035, \] that is, the Call price is supposed to increase by $3.5$ cents.
Therefore, on the one hand, the short position in $100$ Calls will give us a loss of $0.035 \cdot 100 = \$3.5$ (the Call price increases and we are short). On the other hand, the long position in $70$ shares will provide a gain of $0.05 \cdot 70 = \$3.5$. Hence, such a portfolio is riskless and perfectly balanced.
However, it is clear that the position in the stock and the derivative is riskless only for a very short period of time. Theoretically, it remains riskless only for an instantaneously short period of time. To be riskless, it must be adjusted or rebalanced frequently.
Derivation of the Black-Scholes equation
Let the stock price $S_t$ follow the stochastic differential equation $(1)$ introduced above.
Also, let $F=F(S_t,t)$ be the price of an Option (either a Call or a Put) with maturity date $T$, and some strike price $K$ (which is, however, not relevant at this point). For a quick introduction about options and their properties see this post and this one too.
Using Itô's Lemma, with $a=\mu S_{t}$ and $b=\sigma S_t$, we have: $$ \begin{split} dF_t = \left( \frac{\partial F}{\partial S_t}\mu S_t + \frac{\partial F}{\partial t} + \frac{1}{2}\frac{\partial^2 F}{\partial S_t^2}\sigma^2 S_t^2\right)dt + \frac{\partial F}{\partial S_t}\sigma S_t \,dB_t. \end{split} \qquad\qquad (2) $$ Itô's Lemma was introduced here. It can be considered as the stochastic extension of Taylor expansion. It is a fundamental result in finance and we are going to use it to derive the BS differential equation linking the price of the option with the price of the underlying (along with other things).
The corresponding discrete versions of equations $(1)$ and $(2)$ are $$ \Delta S_{t} = \mu S_{t}\,\Delta t + \sigma S_t \, \Delta B_{t}, \qquad\qquad (3) $$ and $$ \Delta F_t = \left( \frac{\partial F}{\partial S_t}\mu S_t + \frac{\partial F}{\partial t} + \frac{1}{2}\frac{\partial^2 F}{\partial S_t^2}\sigma^2 S_t^2\right)\Delta t + \frac{\partial F}{\partial S_t}\sigma S_t \, \Delta B_t, \qquad\qquad (4) $$ where $\Delta F$ and $\Delta S$ are the corresponding changes in the price $F_t$ of the derivative and of the underlying $S_t$ in a small interval of time, respectively.
As the Brownian Motion $B_t$ in the processes for both $F_t$ and $S_t$ is the same, (i.e., the term $\Delta B_t$ in equations $(3)$ and $(4)$), we can construct a strategy involving both the underlying and the derivative in such a way that the Brownian motion is eliminated.
Consider the following portfolio of two positions:
- $-1$ in the Option,
- +$\partial F_t/\partial S_t$ in shares of the stock,
If we let $\Pi_t$ denote the value of the portfolio at current time $t$, then by construction, the value of the portfolio is $$ \Pi_t = -F_t + \frac{\partial F_t}{\partial S_t} S_t. \qquad\qquad (5) $$ Hence, the change in value of the portfolio, $\Delta \Pi$ is, $$ \Delta \Pi_t = -\Delta F_t + \frac{\partial F_t}{\partial S_t} \Delta S_t. \qquad\qquad (6) $$
If we now substitute equations $(3)$ and $(4)$ into $(6)$, we get: $$ \begin{split} \Delta \Pi_t &= - \left(\left( \frac{\partial F_t}{\partial S_t}\mu S_t + \frac{\partial F}{\partial t} + \frac{1}{2}\frac{\partial^2 F_t}{\partial S_t^2}\sigma^2 S_t^2\right)\Delta t + \frac{\partial F_t}{\partial S_t}\sigma S_t \, \Delta B_t\right) + \frac{\partial F_t}{\partial S_t} \left(\mu S_{t}\,\Delta t + \sigma S_t \, \Delta B_{t}\right) \\ &= -\frac{\partial F_t}{\partial S_t}\mu S_t \Delta t -\frac{\partial F_t}{\partial t} \Delta t -\frac{1}{2}\frac{\partial^2 F_t}{\partial S_t^2}\sigma^2 S_t^2 \Delta t -\frac{\partial F_t}{\partial S_t}\sigma S_t \, \Delta B_t +\frac{\partial F_t}{\partial S_t}\mu S_{t}\,\Delta t +\frac{\partial F_t}{\partial S_t}\sigma S_t \, \Delta B_{t} \\ &= \left( -\frac{\partial F_t}{\partial t}-\frac{1}{2}\frac{\partial^2 F}{\partial S_t^2}\sigma^2 S_t^2 \right)\Delta t. \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad (7) \end{split} $$ This shows that the value of the portfolio does not involve $\Delta B_t$, that is, the portfolio is deterministic and riskless during the time $\Delta t$.
Now, similarly to what happened to the risk-free portfolio we built in previous section, our assumptions here imply that the portfolio must instantaneously earn the same rate of return, $r$, as other short-term risk-free securities.
If it earned more than $r$, one could make a riskless profit by borrowing money at rate $r$ and buy such portfolio which will provide a larger rate (likewise for the other way round).
Therefore, at time $t$, in the short term $\Delta t$, we should have a return equal to $r \Pi_t$, i.e., it holds $$ \frac{\Delta \Pi_t}{\Delta t} = r\Pi_t \quad\Longrightarrow\quad \Delta \Pi_t = r\Pi_t \Delta t, \qquad\qquad (8) $$ where $r$ is the risk-free interest rate. If we now substitute equations $(5)$ and $(7)$ into $(8)$, we get: $$ \left( -\frac{\partial F_t}{\partial t}-\frac{1}{2}\frac{\partial^2 F_t}{\partial S_t^2}\sigma^2 S_t^2 \right)\Delta t = r \left(-F_t + \frac{\partial F_t}{\partial S_t} S_t\right) \Delta t, $$ so that $$ \frac{\partial F_t}{\partial t} + r S_t \frac{\partial F_t}{\partial S_t} + \sigma^2 S_t^2 \frac{1}{2}\frac{\partial^2 F_t}{\partial S_t^2} = rF_t, \qquad\qquad (9) $$ which is the Black-Scholes equation.
This equation has many solutions and one must set the boundary conditions. For example, we know that in the case of a European Call Option on non dividend-paying stocks, the boundary condition is the payoff paid by the Call at the maturity $T$, $$ F_T = C_T := \max(S_T - K,0). \qquad\qquad (10) $$ If the derivative is a European Put, then the boundary condition is the payoff paid by the Put at the maturity, $$ F_T = P_T := \max(K - S_T,0), \qquad\qquad (11) $$ Here, in both cases, $K$ is the strike price of the option.
Black-Scholes pricing formulas for Calls and Puts
Risk-Neutral valuation
One way of deriving the Black-Scholes pricing formulas $(12)$ and $(13)$ is to use the so called risk-neutral valuation.
The risk-neutral valuation states that, when valuing a derivative, one can assume that investors are risk-neutral. This means that they don't increase the expected return they require from an investment to compensate for increease in risk. We used to read the expression risk-neutral world to indicate the assumption of a worlds of risk-neutral investors.
A risk-neutral world has two particular features which helps the pricing of derivatives:
- The expected return on an investment is the risk-free rate.
- The discount rate used for the expected payoff on an option (or another derivative) is the risk-free rate.
- Assume that the expected return from the underlying is the risk-free interest rate, $r$.
- Recover the expected payoff at maturity (we know the payoff of both Calls and Puts).
- Discount the expected payoff at the risk-free interest rate to get the price of the option today.
Let us firstly consider the case of a European Call option on a non dividend-paying stock.
The expected value of the Call at maturity $T$ in a risk-neutral world is $$ \mathbb{E}\left(\max(S_T - K,0) \right), $$ where $\mathbb{E}$ is the expected value in the risk-neutral world and $\max(S_T - K,0)$ is the payoff of the Call paid at maturity. Here, $K$ stays fixed, hence the expected values is relative to the possible values of the undelrying $S$ at maturity $T$.
The risk-neutral valuation argument shows that the Call price, $C = C_0$, at time $0$, is given by this expected value discounted at the risk-free $r$, that is, $$ C = C_0 = e^{-rT}\,\mathbb{E}(\max(S_T - K,0)). \qquad\qquad (16) $$ We will see how this would lead to equation $(12)$ (and then to equation $(13)$ in case of a Put option). To this end, we first prove a key result that is needed for the proof.
A key result about lognormal random variables
Proof of the Lemma: Let $f_X$ be the Probability Density Function (PDF) of $X$. Then, $$ \label{lem:res1} \mathbb{E}(\max(X-K,0)) = \int_{-\infty}^{\infty} \max(x - K , 0) f_X(x) dx = \int_{K}^{\infty} ( x - K ) f_X(x) dx. \qquad\qquad (20) $$ Since $X$ is lognormal, then the variable $\ln(X)$ is by definition normally distributed. Also, by hypothesis, its standard deviation is $v$. From properties of the lognormal distribution, the mean of $\ln(X)$ is given by $$ \label{lem:res2} m := \mathbb{E}( \ln(X) ) = \ln( \mathbb{E}(X) ) - v^2/2. \qquad\qquad (21) $$ Now, define the following standardized variable $$ \label{lem:res3} Y := \frac{\ln(X)-m}{v}, \qquad\qquad (22) $$ which is then normally distributed with mean equal to $0$ and variance equal to $1$, i.e., $Y\sim\mathcal{N}(0,1)$. Let $f_Y$ be the probability density function of $Y$, $$ \label{lem:fy:normal} f_Y(y)= \frac{1}{\sqrt{2\pi}}\exp\left(-\frac{y^2}{2}\right) = N'(y), \qquad\qquad (23) $$ where $N$ is given in $(15)$. Before proceeding, note that due to the symmetry of the normal density function, $$ \label{lem:sym:normal:1} \begin{split} N(y) = \int_{-\infty}^{y}\frac{1}{\sqrt{2\pi}}\exp\left(-\frac{s^2}{2}\right)ds &= 1 - \int_{y}^{\infty} \frac{1}{\sqrt{2\pi}}\exp\left(-\frac{s^2}{2}\right)ds \\ &= 1 - \int_{-\infty}^{-y} \frac{1}{\sqrt{2\pi}}\exp\left(-\frac{s^2}{2}\right)ds = 1 - N(-y). \end{split} \qquad\qquad (24) $$
We have to substitute equation $(22)$ into $(20)$ to transform the integral in the variable $X$ into an integral in the variable $Y$. Note that when $X=K$, then $Y = (\ln(K)-m)/v$, and therefore $$ \label{lem:res4} \begin{split} \mathbb{E}(\max(X-K,0)) &= \int_{K}^{\infty}(x-K)f_X(x)dx = \int_{(\ln(K)-m)/v}^{\infty} (\exp\left(y v + m\right) - K) f_Y(y) dy \\ &= \int_{(\ln(K)-m)/v}^{\infty} \exp\left(y v + m\right) f_Y(y) dy - K \int_{(\ln(K)-m)/v}^{\infty} f_Y(y) dy. \end{split} \qquad\qquad (25) $$ Note that from $(23)$, we have that: \begin{equation*} \begin{split} \exp\left(y v + m\right) f_Y(y) &= \exp(y v + m) \frac{1}{\sqrt{2\pi}} \exp\left(-\frac{y^2}{2}\right) \\ &= \frac{1}{\sqrt{2\pi}} \exp\left(-\frac{y^2}{2} + yv + m\right) \\ &= \frac{1}{\sqrt{2\pi}} \exp\left(\frac{1}{2}\left(-y^2 + 2yv + 2m\right)\right) \\ &= \frac{1}{\sqrt{2\pi}} \exp\left(\frac{1}{2}\left(-(y-v)^2 + 2m + v^2\right)\right) \\ &= \exp\left(m+v^2/2\right) \underbrace{ \frac{1}{\sqrt{2\pi}} \exp\left(-\frac{\left(y-v\right)^2}{2}\right) }_{=f_{Y}(y-v)} = \exp\left(m+v^2/2\right) f_{Y}(y-v), \end{split} \qquad\qquad (26) \end{equation*} and then equation $(25)$ becomes $$ \label{lem:res5} \begin{split} \mathbb{E}(\max(X-K,0)) = \exp\left(m+v^2/2\right) \int_{(\ln(K)-m)/v}^{\infty} f_{Y}(y-v) dy - K \int_{(\ln(K)-m)/v}^{\infty} f_Y(y) dy. \end{split} \qquad\qquad (27) $$ Now, if we use equations $(21)$ and $(24)$, respectively, the first integral in previous equation $(25)$ becomes $$ \label{eq:lem:res6} \begin{split} \int_{(\ln(K)-m)/v}^{\infty} f_{Y}(y-v) dy & = 1 - \int_{-\infty}^{(\ln(K)-m)/v} f_{Y}(y-v) dy \\ & = 1 - N((\ln(K)-m)/v - v) \\ & = N( (-\ln(K)+m)/v + v ) \\ & = N( (-\ln(K)+ \ln(\mathbb{E}(X)) - v^2/2)/v + v ) \\ & = N\left( \underbrace{\frac{\ln(\mathbb{E}(X)/K) + v^2/2}{v}}_{=:h_1}\right) = N(h_1). \end{split} \qquad\qquad (28) $$ Likewise, the second integral in $(25)$ will lead to $N(h_2)$: $$ \label{eq:lem:res7} \begin{split} \int_{(\ln(K)-m)/v}^{\infty} f_Y(y) dy & = 1 - \int_{-\infty}^{(\ln(K)-m)/v} f_{Y}(y) dy \\ & = 1 - N((\ln(K)-m)/v) \\ & = N( (-\ln(K)+m)/v ) \\ & = N( (-\ln(K)+ \ln(\mathbb{E}(X)) - v^2/2)/v ) \\ & = N\left(\underbrace{\frac{\ln(\mathbb{E}(X)/K) - v^2/2}{v}}_{=:h_2}\right) = N(h_2). \end{split} \qquad\qquad (29) $$ Therefore, equation $(27)$ becomes: $$ \label{eq:lem:res8} \mathbb{E}(\max(X-K,0)) = \exp(m+v^2/2)N(h_1) - KN(h_2). \qquad\qquad (30) $$ Substituting for $m$ from equation $(21)$, equation $(30)$ becomes $$ \begin{split} \mathbb{E}(\max(X-K,0)) = \exp(\ln(\mathbb{E}(X)) - v^2/2 +v^2/2)N(h_1) - KN(h_2) &= \mathbb{E}(X)N(h_1) - KN(h_2). \end{split} $$ This proves the lemma.
Proof of the Black-Scholes option pricing formulas: European Calls
Finally, we are in position to prove the Black-Scholes formulas $(12)$.
Let us consider an European Call option on a non dividend-paying stock with maturity $T$. Let $K$ be the strike price, $r$ be the risk-free rate, $S_0$ be the current price of the stock, and $\sigma$ be the (annualized) volatility of the stock.
As already said, under the risk-neutral valuation argument, the initial price of the Call, $C = C_0$, is given by the expected value of the payoff at maturity discounted at the risk-free rate of interest (see equation $(16)$).
Recall that $S_T$ follows a lognormal distribution (see above in the Assumptions) and that $\mathbb{E}(S_T) = S_0e^{rT}$ (in a risk-neutral world, the expected returns of an investments equals the risk-free returns).
Furthermore, the logarithm of the prices, $\ln(S_T)$, are normally distributed with standard deviation equal $v_T := \sigma\sqrt{T}$ (see here).
Therefore, we can now use the key result of the lemma proved in previous section: $$ \label{eq:bs:initialprice:proved:1} \begin{split} C = C_0 = e^{-rT}\, \mathbb{E}(\max(S_T - K,0)) &= e^{-rT}\,\left( \;\mathbb{E}(S_T) N(h_1) - K N(h_2)\; \right) \\ &= e^{-rT}\,\left( S_0e^{r T}N(h_1) - KN(h_2) \right) \\ &= S_0N(h_1) - Ke^{-rT}N(h_2), \end{split} \qquad\qquad (31) $$ where $h_1$ and $h_2$ are defined in $(18)$ and $(19)$, respectively, and then given by $$ \begin{align} h_1 = \frac{\ln(\mathbb{E}(S_T)/K)+v_T^2/2}{v} = \frac{\ln(S_0e^{rT}/K)+\sigma^2 T /2}{\sigma\sqrt{T}} = \frac{\ln(S_0/K)+ (r + \sigma^2/2)T}{\sigma\sqrt{T}}, \qquad\qquad (32)\\ h_2 = \frac{\ln(\mathbb{E}(S_T)/K)-v_T^2/2}{v} = \frac{\ln(S_0e^{rT}/K) - \sigma^2 T /2}{\sigma\sqrt{T}} = \frac{\ln(S_0/K)+ (r - \sigma^2/2)T}{\sigma\sqrt{T}}. \qquad\qquad (33) \end{align} $$ Notice that since $$ \label{eq:h1:h2} h_2 = \underbrace{h_1}_{=:h} - \sigma\sqrt{T} = h - \sigma\sqrt{T}, \qquad\qquad (34) $$ we can rewrite equation $(31)$ as $$ \label{eq:bs:initialprice:proved:2} C = C_0 = S_0N(h) - Ke^{-rT}N(h - \sigma\sqrt{T}). \qquad\qquad (35) $$ This is therefore the price of the Call option at time 0, when the entire time to maturity is $T$.
Notice that the Call price $C_t$, for a time $t>0$, is given nothing but that by equation $(12)$ when we consider the time-to-maturity $\tau := T-t$.
Proof of the Black-Scholes option pricing formulas: European Puts
Let us conclude this quite long post by prooving the result for Put options. In case of a European Put, we use the same risk-neutral argument with the payoff of the Put Option at maturity (see equation $(11)$).
First of all, notice that, for any $x$ and $y$, $$ \label{eq:max:x:y} \max(x-y, 0) = x - y + \max(y-x,0). \qquad\qquad (36) $$ Therefore, by using $(36)$, the result for the Call $(31)$, and the symmetric property $(24)$, we can write: $$ \begin{split} P = P_0 &= e^{-rT}\,\mathbb{E}(\max(K - S_T,0)) \\ &= e^{-rT}\, (\; K - S_T + \mathbb{E}(\max(S_T-K,0)) \;) \\ &= e^{-rT}(K - S_T) + e^{-rT}\,\mathbb{E}(\max(S_T-K,0)) \\ &= Ke^{-rT} - S_Te^{-rT} + S_0 N(h_1) - Ke^{-rT} N(h_2) \\ &= Ke^{-rT} - S_0 + S_0N(h_1) - Ke^{-rT}N(h_2) \\ &= S_0(N(h_1)-1) - Ke^{-rT}(N(h_2) + 1) \\ &= S_0(-N(-h_1)) - Ke^{-rT}(-N(-h_2)) = -S_0 N(-h) + Ke^{-rT}N(-h+\sigma\sqrt{T}), \end{split} $$ where in the sixth equation, we have used the fact that $S_0=S_Te^{-rT}$ and in the last equation we have used $(34)$. This proves the formula for a Put option.
As for the case of the Call Option, notice that the Put price $P_t$, for a time $t>0$, is then given by equation $(13)$ where we consider the time-to-maturity $\tau:=T-t$.