1980) are all prominent models satisfying LS condition. The implication is that it is possible
to translate many EU-based models into the MS framework with no loss of accuracy and that
these transformed MS models can be analyzed on the basis of EU-based behavioral
hypotheses. Thus, the MS approach is not only practical but is also flexible enough to
provide full explanations for the LS class of EU-based economic models. Attempts have
been made to take advantage of the practical and flexible framework of MS approach in
developing joint analysis models from economic theories that satisfy the LS condition (e.g.,
Saha 1997; Abdulkadri, Langemeier and Featherstone 2003; Nakashima 2006). There is no
doubt that employing MS approach instead of EU theory simplifies the models. Besides,
theoretically, the models are free from ex ante assumptions on vNM utility function and
distribution of random factor. For example, Hawawini (1978), Meyer and Robison (1988)
and Leather and Quiggin (1991) enjoyed in their applied theoretical studies the MS
framework that is flexible enough as well as tractable. However, when it comes to empirical
applications, the theoretical flexibility of the MS-based empirical models is reduced, because
MS functions have to be specified. Therefore, examining how the specification of MS
functions restricts the theoretical flexibility is an important research area to empirically
exploit the potential of MS approach under the LS condition.
It is well known that if some restrictions are imposed on vNM utility function and/or
distribution of random payoff so that EU theory derives MS function, the derived MS
function also needs to be restricted properly. For example, if vNM utility function is
restricted to negative exponential utility function and random payoff is assumed to follow
normal distribution, then the EU theory yields MS function that is specified as the linear
mean-variance (LMV) model, supposedly one of the most frequently applied MS functions
in the field of agricultural economics.2 Likewise, if random payoffs are restricted to the
distribution class that satisfies LS condition, then the derived MS function has to be properly