This paper considers a model of reference-dependent utility in which the individual makes a conscious choice of her reference point for future consumption. The model incorporates the combination of loss aversion and anticipatory utility as competing forces in the determination of the optimal reference point: anticipating better outcomes boosts current utility but also raises the reference level for future consumption, making the individual more susceptible to losses. A central focus of the paper is on the implications of this model of Optimal Anticipation for attitudes toward risk in dynamic environments. The main representation is formulated in an infinite-horizon framework, and axiomatic foundations are provided. I also describe special cases and show in particular that recursive expected utility in the sense of Epstein and Zin (1989) and Kreps and Porteus (1978) can be reinterpreted in terms of optimal anticipation and loss aversion. Finally, I describe a homogeneous version of the model and apply it to a portfolio choice problem. I show that asset pricing for the Optimal Anticipation model is based on simple modifications of standard Euler equations. While maintaining tractability, this model is rich enough to permit first-order risk aversion and can overcome several deficits of standard expected utility, such as the equity premium puzzle and Rabin's paradox. JEL Classification: D03, D81, G12
