#

Why Brownian motion doesn't have derivatives?

Why $\frac{d~B_t}{dt}$ does not exist?

Difference quitient of B over the interval $[t,t_h]$: $$D_{t,h} = \frac{B_{t+h} - B_t}{h}$$

Now let's find expectaion and variance of that.

$$\mathbf{E}[D_{t,h}] = \frac{0-0}{h} = 0$$$$\mathbf{Var}[D_{t,h}] = \frac{h}{h^2} = \frac{1}{h}$$

So when $h\to 0$ then $\mathbf{Var}[D_{t,h}]\to \infty$

So since if $D_{t,h}$ had a limit, it would have to be Gaussian, thus, $D_{t,h}$ does not converge. So $\frac{d~B_t}{dt}$ does not exist.

However, check this out:

  • In differential form, we write $d~B_t = B_{t+dt} - B_t$ and $d~B_t$ should be Guassian. So we ought to have
    • $\mathbf{E}[dB_t] = 0$ and
    • $\mathbf{Var}[dB_t] = dt$ (indeed, this means $\mathbf{E}[(dB_t)^2] = dt$)

So, in particular, $\displaystyle \sqrt{\mathbf{Var}[dB_t]} = \sqrt{dt}$.

To understand this, we will need "stochastic calculus".

One more leap of faith,

$$\left(dB_t\right)^2 = dt \ \ \text{Ito's rule}$$

So, we say that $dB_t \sim \mathcal{N}(0, dt)$

Black-Scholes-Merton Model

$$S_t = S_0 ~ \displaystyle e^{\sigma B_t -\frac{1}{2} \sigma^2t}$$

where $\sigma=\text{a constant called volatility}$

This is a good model for stock proces (but not for interest rates, ...)

To see why the BSM is good for stock prices, let's first prove the Ito's formula.

Recall: Taylor's formula

Let $f:\mathbb{R}\to\mathbb{R}$ and $f$ is differentiable.

  • Order 1: $f(b) \approx f(a) + (b-a)f'(a)$

  • Order 2: $f(b) \approx f(a) + (b-a)f'(a) + \frac{(b-a)^2}{2}f''(a)$

  • Differntial motivation: $$f(x+dx) \approx f(x) + f'(x) dx + \frac{1}{2}f''(x)(dx)^2$$

    Also, if $f$ also depends on another parameter $t$, then,

    $$f(t+dt, x+dx) \approx f(t,x) + \frac{\partial~f}{\partial t}dt + \frac{\partial~f}{\partial x}dx + \frac{\partial^2~f}{\partial x^2} (dx)^2$$

    In the above formula, we ignored the $\frac{\partial^2~f}{\partial t \partial x}dtdx$ because in pur stochastic calculus ...

Theorem: Ito's formula

For the function $f(t,x)$ on $\mathbb{R}_t\times\mathbb{R}$ ($f$ has $t-$derivatives of order 1, and $x-$derivatives of order 2).

then, $$f(t+dt,B_{t+dt}) = f(t,B_t) + \frac{\partial~f}{\partial x}(t,B_t)dB_t + \frac{1}{2}\frac{\partial^2~f}{\partial x^2}(t,B_t)dt + \frac{\partial~f}{\partial t}(t,B_t)dt$$

Proof:

Apply the $(t,x)-$ dependent version of Taylor's formula we just saw with $x=B_t$ and $x+dx = B_{t+dt}$.

Simply, notice the notation:

$$dB_t := B_{t+dt} - B_t\ \ \ \ \ \ \ \text{ note }:=\text{ means definition}$$

We see that $dx = dB_t$ we just defined.

Indeed, $\displaystyle (dx)^2 = \left(B_{t+dt} - B_t\right)^2 = (dB_t)^2 = dt$ where the last equality is Ito's rule.

Recall Black-Scholes Model

$$S_t = S_0 ~ \displaystyle e^{\sigma B_t -\frac{1}{2} \sigma^2t}$$

say $S_0=1$ just to simplify.

Thus,

$S_t = f(t,B_t)$ where $$f(t,x) =\displaystyle e^{\sigma x -\frac{1}{2} \sigma^2t}$$

Now, we apply Ito's formula:

  • $\frac{\partial~f}{\partial x}(t,x) = \sigma e^{\sigma x - \frac{1}{2}\sigma^2t}$

  • $\frac{\partial^2~f}{\partial x^2}(t,x) = \sigma^2 e^{\sigma x - \frac{1}{2}\sigma^2t}$

  • $\frac{\partial~f}{\partial t}(t,x) = -\frac{1}{2}\sigma^2 e^{\sigma x - \frac{1}{2}\sigma^2t}$

Let's subsititue $$f(t+dt,B_{t+dt}) = f(t,B_t) + \frac{\partial~f}{\partial x}(t,B_t)dB_t + \frac{1}{2}\frac{\partial^2~f}{\partial x^2}(t,B_t)dt + \frac{\partial~f}{\partial t}(t,B_t)dt$$

Ito's formula:

$$\begin{array}{lc}S_{t+dt} - S_t &= \left(\sigma e^{\sigma B_t - \frac{1}{2}\sigma^2t}\right)dB_t + \\ &\frac{1}{2}\left(\sigma^2 e^{\sigma B_t - \frac{1}{2}\sigma^2t}\right)dt + \\ &\left(-\frac{1}{2}\sigma^2 e^{\sigma B_t - \frac{1}{2}\sigma^2t}\right)dt \\ &= \sigma e^{\sigma x - \frac{1}{2}\sigma^2t} = \sigma S_t dB_t\end{array}$$
Conclusion:

We have just [proved $S_{t+dt} - S_t = \sigma S_t dB_t$

BSM differential equation: $dS_t = \sigma S_t dB_t$

Interpretation: We cannot say which direction the price will go (decrease or increase) but we can say how fast it will go.

Insurance on Financial Risk

Call options: you buy the right at time zero to purchase stock x at time $T$ (pre-detemined) inf future at a pre-determined price $K$.

  • If $S_T<K$ then just buy

  • If $S_T\ge K$ use the option and buy at price $K$ abd sell at price $S_t$ so you make profit $S_T-K$

The insurance contract gives you

$$\left\{\begin{array}{lrr}0&\text{if}&S_T< K\\ S_T - K &\text{if}&S_T \ge K \end{array}\right.$$

The price seeling this insurance contract must find the right proce for it. It turns out the fair price is the expected value of the payoff. In fact, $PAYOFF = \Phi(S_T)$ and fair price is $\mathbf{E}[\Phi(S_T)] = \mathbf{E}[(S_T-K)_t]$

We know everything about the distrinution of $S_T$:

$$S_T = e^{\sigma B_T - \frac{1}{2}\sigma^2 t}$$

$B_T \sim \mathcal{N}(0,T)$ so we write $S_T = e^{\sigma \sqrt{T} Z - \frac{1}{2}\sigma^2 T}$

So, $$\mathbf{E}[\Phi(S_T)] = \int_0^\infty \frac{1}{\sqrt{2\pi}} e^{-Z^2/2}\left(e^{\sigma \sqrt{T} Z - \frac{1}{2}\sigma^2 T} - K\right)_t dZ$$


In [ ]: