Libertarian Paternalism @ CATO Unbound

Glen Whitman has prompted an interesting exchange at Cato Unbound on Libertarian Paternalism. Richard Thaler replies:

Consider the following common problem. Most firms have an open enrollment period in November when employees can elect their benefit package for the following year. At my employer, the University of Chicago, you have a few weeks to log on to the appropriate web site and make your selections. The question is, what should the employer do for those employees who forget to log on? (Professors’ reputations for absent-mindedness are well deserved.) For each of the choices the employee has, the employer needs to select a default option for those who do not log on, and normally the default is either “same choice as last year” or “back to zero” (meaning, decline this option). At Chicago the default option for the health insurance plan is the same as last year.

Of course it is possible to criticize this choice of the default option, but it is essential to understand that the employer must choose something. Some employers use “back to zero,” which minimizes the costs to the employer; somewhat less drastic would be to default employees into the plan that is cheapest for the employer; one could even choose a default plan at random (don’t laugh — this is the strategy used for some participants in the Medicare Part D prescription drug coverage implemented by the Bush administration); or the employer could somehow force employees to make a choice. The Nudge philosophy here is that the person who designs the plan, whom we call the choice architect, should choose the default that she thinks, all things considered, will make the participants best off. Does Professor Whitman have a better suggestion?

It's hard NOT to support this version of Libertarian Paternalism (of course, I'd prefer the choice architect be a private entity). If a default must be set, one should (in my normative opinion--and apparently Thaler's and others') set the default to what would most likely make the participant best off.

It seems, however, that this is a very narrow subset of the LibPat agenda. Typically, the government (not the employer) is the choice architect. And when the employer IS the choice architect (say in that cafeteria/fruit example) the government intervenes as supreme default-setter. Nudging is transformed from an ethical guideline that individuals should employ to a government mandated policy. Yet Thaler and others continue to present the very narrow subset of LibPat as the whole agenda. Is my perception of the debate correct? Or have I missed something?