Reconciling normative and behavioural economics

Abstract

The aim of this thesis is to address from a methodological, philosophical and theoretical perspective the problem of how to reconcile normative and behavioural economics — the “reconciliation problem”. The first part develops a methodological assessment of behavioural welfare economics (and more specifically of libertarian paternalism), which currently constitutes the most accepted approach to deal with the reconciliation problem. We argue that behavioural welfare economics is in the direct continuation of neoclassical welfare economics, and that its conception of individual behaviour is tightly connected to Pareto’s reductionist model of the Homo economicus. It is assumed that an individual can be defined by “true” preferences — distinct from her revealed preferences — and that those preferences are consistent and context-independent. The satisfaction of those true preferences is taken as the normative criterion. The existence of true preferences requires accepting that people have access to a latent mode of reasoning able to generate consistent and context-independent preferences, and to treat individual decision-making as if it was the result of an optimisation problem faced by an inner rational agent trapped in an outer psychological shell. We argue on the contrary that we cannot define unambiguously such true preferences. We then argue that the normative challenge raised by behavioural economics is that individuals may lack of autonomy: rather than focusing on the satisfaction of one’s preferences, we argue that what matters is the ability to choose one’s own preferences. Since we suggest that the difficulties of behavioural welfare economics are related to its commitment to an implausible model of individual preferences, we provide in the second part of the thesis a model of preferences compatible with our methodological analysis: we take Bacharach’s variable frame theory as the primitive of our analysis, and build a model of endogenous preferences, in which individuals can choose to some extent their own preferences. The main contributions of this model are that (i) it does not require the existence of latent true preferences, and (ii) the individuals are allowed to team reason. We assume that collective preferences are strategically chosen, and show that team reasoners tend to be aggressive (cooperative) with other teams in submodular (supermodular) games. We finally show that team reasoning can be empirically justified as an ecologically rational heuristic.