Please click here if you are unable to view this page.
TOPIC:
FROM LOCAL UTILITY TO NEURAL NETWORKS
ABSTRACT
By formalizing a preference-based notion of local linearity in the spirit of Machina (1982), we introduce two utility representations. Both are equivalent to continuous finite piecewise linear functions. In the first, it is as if the decision maker has an optimistic self and a cautious self playing a zero-sum game. In the second, the decision maker evaluates an alternative through a neural network. The representations are easy to apply and estimate, can be used for local utility analyses and analyzing choices under ambiguity, and nest the constant loss aversion model and models with hierarchical subjective product attributes as special cases.