Applying the principle of maximum entropy (PME) to infer an implied probabi
lity density from option prices is appealing from a theoretical standpoint
because the resulting density will be the least prejudiced estimate, as "it
will be maximally noncommittal with respect to missing or unknown informat
ion."(1) Buchen and Kelly (1996) showed that, with a set of well-spread sim
ulated exact-option prices, the maximum-entropy distribution (MED) approxim
ates a risk-neutral distribution to a high degree of accuracy. However, whe
n random noise is added to the simulated option prices, the MED poorly fits
the exact distribution. Motivated by the characteristic that a call price
is a convex function of the option's strike price, this study suggests a si
mple convex-spline procedure to reduce the impact of noise on observed opti
on prices before inferring the MED. Numerical examples show that the convex
-spline smoothing method yields satisfactory empirical results that are con
sistent with prior studies. (C) 2001 John Wiley & Sons, Inc.