Welfare economics has been historically built on a more or less loosely defined principle of consumer sovereignty. This principle is closely articulated to the definition of welfare as preference-satisfaction that has dominated the “new” welfare economics since the pioneering work of Vilfredo Pareto, John Hicks and Nicholas Kaldor. Relying on several experimental results coming from behavioral economics, an increasing number of economists are however subscribing to paternalistic social evaluations and policies. Most of them are endorsing a “light” or “libertarian” form of paternalism in an attempt to reconcile the libertarian foundations of normative economics with the empirical evidence that people often do not behave in their own interest.

An interesting feature of these normative extensions of behavioral economics is that most of them have not given up the welfarist criterion of preference-satisfaction. This is explicit in Douglas Bernheim and Antonio Rangel's significant attempt to build a “behavioral welfare economics” [(Bernheim and Rangel 2007); (Bernheim and Rangel 2009)]. This is also the case – though more implicitly – in the oft-cited writings of Thaler and Sunstein on libertarian paternalism. This article precisely deals with the philosophical and theoretical implications of this rather unnatural association between the welfarist criterion of preference-satisfaction and the rising paternalistic stance of a part of normative economics. I shall argue that the perspectives of this encounter between welfare economics and paternalism are not promising. The main reason lies in the fact that all the plausible conceptual and theoretical strategies to deal with the fact that preference are endogenous and more particularly context-dependent are ill-qualified to account for the issue of autonomy that is at the core of all discussions about paternalism – which is even more true when one is endorsing a “libertarian” or “soft” version of paternalism.

The fact that welfarism is unable to deal adequately with autonomy-related issues should not be a surprise. Amartya Sen's famous Paretian-liberal paradox (Sen 1970a) is an early result stating that a weak form of welfarism, Paretianism, is inconsistent with the satisfaction of a condition of minimal liberty. However, the normative turn of behavioral economics threatens the foundations of welfare economics even more deeply because it casts doubts on the very definition of welfare and on how we should measure it. The main point I shall make is that rescuing the preference-satisfaction welfare criterion against behavioral anomalies almost inevitably implies to give up the notion that a person is an economic agent and the main locus for welfare considerations. Since autonomy and freedom are values that necessarily pertain to persons, behaviorally informed welfare economics totally disconnects welfare and autonomy. As a result, welfare economics cannot be of any help to discuss paternalistic social evaluations and policies. Instead, following Gerald Dworkin [(1972); (2010)] and John Rawls (1971), I suggest that a justification of paternalism compatible with liberal principles depends on the ability for reasonable persons to voluntarily consent to a collective choice rule with paternalistic tendencies. This argument relies on a distinction between preferences (which can be attached to other entities than persons) and values unknown to welfare economics.