THE PRICING OF BASKET OPTIONS BY EDGEWORTH EXPANSION

                   Magnus Blix


We consider the problem of finding the option price of
a sum of correlated log-normal assets. The main
difficulty is that the option price, a discounted
conditional expectation of a sum of log-normal random
variables, does not seem to have a convenient
implementable representation. To solve this problem, the
crucial point is to use an Edgeworth expansion, which
means that the underlying density function is
approximated by a log-normal density plus an appropriate
number of correction terms. Two different choices of the
approximating log-normal density are made. The Edgeworth
expansion approach is compared to Monte-Carlo simulations
of the conditional expectation. Unknown volatilities are
estimated. The theoretical computations are supplemented
by accesible programs.