We study, using the St. Petersburg paradox, the risk attitudes expressed by
utility functions without the normal continuity and differentiability assu
mptions. We model the repeated St. Petersburg lottery with two parameters,
starting wealth and return ratio, as a stochastic process with a simple con
trol mechanism and the objective of maximizing utility of the outcome. A st
opping function expresses the earliest stopping stage; a finite value resol
ves the paradox. This new approach can measure the risk attitude of discont
inuous utilities in only a finite horizon repetition of the lottery, openin
g the door to new usefulness of the utility concept in modeling. As an Exam
ple we give an application to behavior of persons receiving entitlements to
wards additional income from working. (C) 1999 Elsevier Science Inc. All ri
ghts reserved.