A share price grows $1+\epsilon$ times with prob. $p$ or falls $1-\epsilon$ times with prob. $1-p$ each day, what's its expected value after $n$ days?

67 Views Asked by At

Task:

Today's share price is $Q$. Every day, independently of other days, it grows to the value of $q(1+\epsilon)$ with probability $p$ or falls to the value of $q(1-\epsilon)$ with probability $1-p$, where $q$ is the price of the previous day and $0\leq\epsilon<1$ is a constant. Compute the expected value and variance of the share price after $n$ days.

How to solve this task? My attempt:

Let $X$ denote how many times the price grew. Then $X$ is clearly given by the binomial distribution, so we have $\operatorname{P}(X=k)={n\choose k}p^k(1-p)^{n-k}$ and $\operatorname{E}X=np$. Now let $f(X)$ compute the price after $n$ days, then $f(k)=Q(1+\epsilon)^k(1-\epsilon)^{n-k}$. Now I'd like to be able to say that $\operatorname{E}f(X)=f(\operatorname{E}X)=f(np)=Q(1+\epsilon)^{np}(1-\epsilon)^{n(1-p)}$, BUT...

I don't think if this is correct and even if it is correct I can't prove this. And unless I can prove why I can do something like that the only thing I have is the general formula for computing the expected value of a function of a random variable, and that is... $\operatorname{E}f(X)=\sum_k\operatorname{P}(X=k)f(k)$.

And this formula leads to a horrible sum...: $\operatorname{E}f(X)=\sum_{k=0}^n{n\choose k}p^k(1-p)^kQ(1+\epsilon)^k(1-\epsilon)^{n-k}={Q\sum^n_{k=0}{n\choose k}\bigl(p(1+\epsilon)\bigr)^k\bigl((1-p)(1-\epsilon)\bigr)^{n-k}}$

Whatever should I do with this sum? I have no idea. Or is there any other way to solve this task that wouldn't lead to this sum?

1

There are 1 best solutions below

2
On BEST ANSWER

Using the law of total expectation we can avoid having to compute any complicated series for the expectation; that is we use the fact that

$$ \mathbf{E}[X_{n+1}] \, = \, \mathbf{E}[ \mathbf{E}[X_{n+1}\, | \, X_n ]]$$

If we denote $X_n$ to be the value on day $n$, then conditioned on $X_n$ the expectation of $X_{n+1}$ is

\begin{align*} \mathbf{E}[X_{n+1} \, | \, X_n] &= p(1 + \epsilon)X_n + (1-p)(1 - \epsilon)X_n \\ & = (1 - \epsilon + 2p\epsilon)X_n \end{align*}

Then the law of total expectation gives $$ \mathbf{E}[X_{n+1}] \, = \, \mathbf{E}[ \mathbf{E}[X_{n+1}\, | \, X_n ] ] \, = \, (1 - \epsilon + 2p \epsilon) \mathbf{E}[X_n],$$

And so iterating the above (i.e doing the same for $X_n$, etc.) we obtain

$$\mathbf{E}[X_{n+1}] = (1 - \epsilon + 2p \epsilon) \mathbf{E}[X_n] \, = \, \cdots \, = \, (1 - \epsilon + 2p\epsilon)^{n+1} \mathbf{E}[X_0].$$

Assuming that $X_0= Q$ then this simplifies to

$$\mathbf{E}[X_n] = Q(1- \epsilon + 2p \epsilon)^n.$$

Similarly one can apply the law of total variance to get an expression for the variance. That is

$$ \text{Var}(X_{n+1}) = \mathbf{E}[ \text{Var}(X_{n+1}\, | \, X_n)] + \text{Var}( \mathbf{E} [ X_{n+1} | X_{n} ] ).$$