Welfare Evaluation, Behavioral Agents and Rational Agents

A reader sends me the following comment, further to the several VC posts on behavioral economics (initially occasioned by Andy Ferguson’s Weekly Standard essay):

One basic issue that this whole-“behavioral econ– good-or-bad?”
discussion seems to have neglected the following simple point: welfare
evaluation is much harder with “behavioral agents” than “rational
agents.”

With “rational” agents we know that subject to tons and tons of
asumptions markets are great (first welfare theorem).  And we have a
pretty good idea of what constitutes a market failure (externalities)
and when a “social planner” can help.  Thus, there is a principled
econ case for certain forms of “social planner” intervention that we
know will raise welfare (whether a government can act as an optimal
social planner is another question).

With “behavioral” agents, the basic issue is that people’s preferences
are at some level time-inconsistent.  My self of today wishes that my
self of tomorrow would put money in to a 401(K) but my self of
tomorrow wishes it to be the next period’s self and etc.  Thus, the
person sitting at today does *not* have the same preferences as the
person sitting at tomorrow.   If you make the person of today put
money into a 401(k) *today* you make them worse off (since they wanted
to put money in a 401(k) tomorrow), but you make their yesterday’s
self happy.   As a social planner, whose utility do you maximize?

It’s not obvious how you do this.  There are some attempts to work
this out in the literature (e.g. http://www.nber.org/papers/w13737)
but it’s not settled.

I guess the main take-away is that claiming policy implications from
behavioral research is *much* harder than from other kinds of econ
research, so at some level behavioral people are jumping the gun a bit
in claiming that they have a new way to do policy; and this objection
is totally independent of worries about slippery slopes and whatnot.
That said, much of the actual-existing behavioral influenced policy
moves (e.g. doing stimulus through withholding rather than lump sum)
seems like a good idea since it is formally identical to what would
have been the status quo.

The paper cited at NBER is interesting, but I would add that this seems to me an area in which philosophy does have something to say.  The problems of the self over time have been much discussed, and at least some of the leading arguments about “whose” utility you maximize have been formally adduced, including the proposition that this present-self, future-self, successive-selves way of thinking about things is appealing in part because it matches up to marginality analysis, but is not coherent as an account of identity.  This is not an argument about values or an argument from moral philosophy; it is an argument about the nature of identity and self, and I think the philosophers of mind, identity, and such fields do indeed have something to say as to the conceptually valid and invalid ways of framing the issue of the self.

(In a quite different approach to the time-identity problem, framed as a matter of constitutional law and politics, I recommend highly Jed Rubenfeld’s short, compelling book, Freedom and Time: A Theory of Constitutional Self-Government.  When it first appeared, I found it – with apologies to Jed – very smart but frankly weird.  It has grown on me since then – grown on me a lot.)

Thanks to the VC reader for this thoughtful comment.