The welfare effects of an intervention, g, that affects utility directly as well as behaviour, is, generally,
but since ∂U/∂ai = λpi, and p1a1 + p2a2 + p3a3 is constant, the marginal welfare effect is given by the partial derivative ∂U/∂g. It is thus clear that exogenous information - a favourable intervention affecting only γ - will be beneficial only for basically pessimistic individuals
while prevention - a favourable intervention affecting only F - is always welfare improving,
Financial support from the Swedish Medical Research Council (grant no. 27P-10737) and the Bank of Sweden Tercentenary Foundation is gratefully acknowledged. The paper was presented at the 19th Arne Ryde symposium on Individual Decisions for Health in Lund, August 1999, and we are most grateful for insightful comments and suggestions from our discussant, Louis Eeckhoudt. We also appreciate comments from Tomas Philipson and the audience at the seminar. In addition, we have received helpful comments from Dan Anderberg, John Hey, Håkan J. Holm, Bengt Liljas, Björn Lindgren, and two anonymous referees. Any remaining errors are the sole responsibility of the authors.
Examples include Chang (1996), Cropper (1977), Dardanoni and Wagstaff (1987, 1990), Liljas (1998), Picone et al. (1998), and Selden (1993).
Significant contributions with deterministic models are still appearing in the demand-for-health tradition (Grossman, 1998; Ried, 1998), as well as in other areas.
In fact, Eichberger and Kelsey (1999) have provided axiomatic foundations for a utility function which is similar to (3.1); i.e. which is a convex combination of an expected utility and a generalized expected utility.
See, e.g. Schmeidler (1989). The notion of genuine uncertainty was introduced by Knight (1921); it is sometimes referred to as ‘Knightian uncertainty’.
The generalized probabilities generate weights, vs, by means of which the modified expectation is computed: u0 (a) = Σvsu(hs, a).
A framework employing a utility function similar to ours is employed by Mukerji (1998) in an exploration of the effect of uncertainty on contractual incompleteness.
Note that it may be the case that the outcomes (h) of a firm state and an uncertain state coincide; increasing γ then leads to a change in the expectation only due to better knowledge of the probability of this outcome.