You are on page 1of 5

4.

3

The Black-Scholes Partial Differential Equation

Let S be the price at time t of a particular asset. After a (short) time interval of length dt,
the asset price changes by dS, to S + dS. Rather than measuring the absolute change dS,
we measure the return on the asset which is defined to be
dS
.
S
Note: This return expresses the change in the asset price as a proportion of the original
asset price.
One common mathematical model of the return has two components. The first is a
predictable, deterministic component (similar to the return on a risk-free investment in a
bank). This is
µdt.
The parameter µ is called the drift. It is a measure of the average rate of growth of the asset
price.
The second contribution to the return dS
S is
σdX.
Here σ is called the volatility and is a measure of the standard deviation of the returns. The
quantity dX is a random variable having a normal distribution with mean 0 and variance
dt :

dX ∼ N (0, ( dt)2 ).
This component is a random contribution to the return. For each interval dt, dX is a sample

drawn from the distribution N (0, ( dt)2 ) - this is multiplied by σ to produce the term σdX.
The value of the parameters σ and µ may be estimated from historical data.
We obtain the following stochastic differential equation (stochastic analysis is the study
of functions of random variables).
dS
= µdt + σdX.
S

(4.1)

Notes
1. If σ = 0 then the behaviour of the asset price is totally deterministic and we have the
ordinary differential equation
dS
= µdt.
S
This can be solved to give
S = S0 eµt
where S0 is the asset price at time t = 0.
2. The equation 4.1 is an example of a random walk. It cannot be solved to give a
deterministic path for the share price but it gives probabilistic information about the
behaviour of S.

39

05)) = −0.4.35. assuming 250 business days in a year.2(−0. = 10 − 0.2(0. = 0.916.17.4/250 + 0.179. dS 10.2(0. dt = (This value of dt is basically one day. The equation 4.4/250 + 0.063.916(0.4/250 + 0. A value for dX is chosen from N (0. Step 1 S0 = 10. Step 2 S1 = 9. = 9.179 = 10.4/250 + 0. 250 µ = 0.17.17(0.1 can be considered to be a scheme for constructing time series that may be realised by share prices.4/250 + 0.3. take dX = 0. 1/250) .12) = 9.08) = 10. σ = 0. dS 9.916.4/250 + 0.08.05) = 10(0.17 dS S1 The following is a graphical representation of this time series : 40 .08)) = 0. = 10. Then dS 10 dS S1 = 0.1 The price S of a particular share today is e10.254 = 10.choose dX = −0.916 + 0.17 + 0. Here dX is drawn (at each step) from a normal distribution with mean 0 and standard √ deviation 1/ 250 ≈ 0.916 dS S2 Step 3 S2 = 10.2(0.) Solution: We have dS − µdt + σdX = 0.4 S  1 250  + 0. Example 4.2(−0.12)) = 0.2dX.254. = 0.2.084.2(0.05. take dX = 0.084 = 9.12.3. Construct a time series for the share price over three intervals if 1 .

...... .. If this time step were used in practice however......... . 10..2 s 10.. ...............S 10.... We need Itˆ o’s Lemma....... .... .... .. .... . ... ........ .... . .. .... .. We finish these lecture notes now with a brief outline of such a model...... ...4 s s . ........ . ........ . ... . ... .. .. ... . .... ......... .8 s s t=0 s s s t = 1/250 s t=2/250 s t = 3/250 In real life asset prices are quoted at discrete intervals of time... .... . ............. .. . The Taylor Series of f at x = a is ∞ X f (n) (a) (x − a)n . .. .......... .. .. Recall: Taylor Series Let f be a function with derivatives of all orders on an interval I containing a point a.... ....... .. which is a version of Taylor’s Theorem for functions of random variables.. .. ...... ... . ..... ... n! n=0 If this series converges to f on I then f (x) = f (a) + =⇒ f (x) − f (a) = ∞ X f (n) (a) (x − a)n n! n=1 ∞ X f (n) (a) (x − a)n ........... .. . .... and so there is a practical lower bound for the basic time step dt of our random walk... n! n=1 Now replace x with x + ∆x and a with x to obtain f 00 (x) (∆x)2 + ... . ..... ..0 s 9. .. .. ... ... .... ........... ... .... ...... One approach is to develop a continuous model by taking a limit as dt −→ 0...... We now return to our consideration of what happens to ∆f = f (x + ∆x) − f (x) = f 0 (x) + dS = µdt + σdX S 41 .. ... .. . 2! This relates the small change ∆f in the function f to the small change ∆x in x. .. ...... the sheer quantity of data involved would be unmanageable...

. The value of the portfolio is Π = V − ∆S. There is a deterministic component dt and a random component dX. We need the following fact which we state without proof : With probability 1 (dX)2 −→ dt as dt −→ 0. dS 2 dS 2 Now dS = S(µdt + σdS) and (dS)2 = S 2 µ2 (dt)2 + 2µσdtdX + σ 2 (dX)2  Now since (dX)2 −→ dt as dt −→ 0. dV = σS ∂S ∂S 2 ∂S ∂t Consider a portfolio containing one option and −∆ units of the underlying stock. t) be the value of an option (this is usually called C(S. In fact we need a version of Itˆ o’s Lemma for a function of more than one variable : if f is a function of two variables S. dX + µS + σ2S 2 2 + df = σS ∂S ∂S 2 ∂S ∂t The Black-Scholes PDE Let V (S. the term S 2 σ 2 (dX)2 dominates the above expression for (dS)2 as dt becomes small. . If we change S by a small amount dS then by Taylor’s Theorem we have df = 1 d2 S df dS + (dS)2 + . dS dS 2 dt This is Itˆ o’s Lemma relating a small change in a function of a random variable to a small change in the variable itself. Retaining only this term we use S 2 σ 2 dt as an approximation for (dS)2 as dt −→ 0. t) for a call and P (S. t) for a put). Let r be the interest rate and let µ and σ be as above.as t −→ 0. We then have df df = df 1 d2 S 2 2 (S σ dt) dS + dS 2 dt2 = df d2 S (Sµdt + σdX) + 2 (S 2 σ 2 dt) dS dt = σS   df 1 d2 S df dX + µS + σ 2 S 2 2 dt. Using Itˆ o’s Lemma we have   ∂V 1 ∂2V ∂V ∂V dX + µS + σ2S 2 2 + dt. t we have   ∂f ∂f 1 ∂2f ∂f dt. Suppose that f (S) is a function of the asset price S. 42 .

∂V 1 ∂2V ∂V + σ 2 S 2 2 + rS − rV = 0. Then for a fair price we should have   1 2 2 ∂2V ∂V rΠdt = dt σ S + 2 ∂S 2 ∂t   ∂V S = =⇒ r V − ∂S ∂V 1 2 2 ∂2V + σ S 2 ∂S 2 ∂t Thus we obtain the Black-Scholes PDE. Now if Π was invested in riskless assets it would see a growth of rΠdt in the interval of length dt. dΠ = σS = σS Choose ∆ =   ∂V ∂V 1 ∂2V ∂V dt − ∆Sµdt − ∆SσdX dX + µS + σ2S 2 2 + ∂S ∂S 2 ∂S ∂t     ∂V ∂V 1 ∂2V ∂V − ∆ dX + µS + σ2S 2 2 + − µ∆S dt ∂S ∂S 2 ∂S ∂t ∂V to get ∂S dΠ =  ∂V 1 2 2 ∂2V + σ S 2 2 ∂S ∂t  dt. ∂t 2 ∂S ∂S 43 .Thus dΠ = dV − ∆dS.