All posts by admin

Shreve Exercise I-1.6

Bank A has a long position in a derivative that pays either $V_1(H)=3$ or $V_1(T)=0$ at at time 1.
The price of the derivative is $V_0=1.2$.
In order to hedge the long position, Bank A proceeds as follows:
First, Bank A borrows 1.2 from Bank B.
Then Bank A invests this amount 1.2 in the money market, by handing it over to Bank C.
Next Bank A sells $-\Delta_0=0.5$ shares of the underlying stock short.
This gives Bank A the amount $-\Delta_0 S_0=2.00$ on hand, which it invests in the money market, giving $(1+r)2.00=2.50$ at time 1.
The debt to Bank B becomes $(1+r)1.2=1.50$ at time 1. Thus the proceeds from the short sale, and the debt to Bank B,
together add up to $2.5-1.5=+1.00$. The short sale of the stock means that Bank A will be liable to pay someone either $-\Delta_0 uS_0=4$ or $-\Delta_0 dS_0=1$ at time 1.
Together with the +1.00 this leaves the bank in a position of $1-4=-3$ or $1-1=0$ at time 1. However, the bank also has that long position in the derivative, balancing the -3 or 0 position out.
In the end, then, the only thing left is the investment in the money market via Bank C. It has grown to $(1+r)1.2=1.5$ at time 1.
Thus Bank A has successfully hedged a long position in the derivative.

Actually this can be simplified: the bank just sells 0.5 shares short.
The bank thus receives 2.00 at time 0, which it places in the money market, leading to 2.50 at time 1. When the bank has to pay the value of the stock at time 1 it has to pay either 4 or 1. At the same time the bank receives 3 or 0 because of the long position in the derivative. This gives a net loss of 1 at time 1, leaving the bank with $2.50-1.00=1.50$ at the end.

The probability distribution as a computational resource for randomness testing

The probability distribution as a computational resource for randomness testing
Journal of Logic and Analysis 2 (2010), no. 10, 1—13.

This paper grew out of my assignment to teach Math 472, Statistical Inference, in Spring 2009 at UH. By analyzing the proof of the law of large numbers and some other work I showed that Hippocratic and Galenic randomness coincide for Bernoulli measures. [There has been follow-up work to this paper which it would make sense to describe here.]

Effective dimension of points visited by Brownian motion

Effective dimension of points visited by Brownian motion
(with Anil Nerode). Theoretical Computer Science 410 (2009), no. 4-5, 347—354.

Preliminary version: The law of the iterated logarithm for algorithmically random paths of Brownian motion.
Logical Foundations of Computer Science, Lecture Notes in Computer Science, vol. 4514, Springer, Berlin, 2007, pp. 310—317.

In this paper we answered a question of Fouché regarding algorithmically random Brownian motion. The key idea was to use Carathéodory’s measure algebra isomorphism theorem.

Nearest walk to Brownian motion

Mathematics 472 (Statistical Inference), University of Hawaii at Manoa, Spring 2009.
This project concerns the distribution of the nearest walk of length \(n\) to a randomly chosen walk of length \(m>n\),
where the walks have step size \(1/n\) on the \(x\)-axis and \(1/\sqrt n\) on the \(y\)-axis.
It is well known that the distribution of such walks converges to Brownian motion.

Two of the groups in the class found that strings of the form \(1^a0^b\) and \(0^a 1^b\) were the least likely nearest walks, whereas strings of the form \(0101\ldots\) or \(1010\ldots\) were the most likely. Strings of the form \(000\ldots\) or \(111\ldots\) were the second least likely.

math overflow questions: