I think that a popular approach to welfare economics requires simultaneously holding two contradictory beliefs.
1) Utility is not comparable across individuals so we can only say whether an outcome is Pareto efficient or not.
2) Models which are not reality but which approximate reality are useful.
Both are really central to many policy recommendations made by many economists. They don’t fit together.
Long argument after the jump.
If we accept principle 1) we ask if a proposed reform will achieve a Pareto improvement help some people while hurting none. If we consider a standard economic model, it is often possible to find situations in which a Pareto improvement, that is to say, in which the current situation is Pareto inefficient.
For example, imagine a standard neoclassical general equilibrium model in which all people are equal with equal wealth and income and tastes and … Then moving from any other tax to a lump sum poll tax will cause a Pareto improvement. It removes distortions to incentives and, with such a tax, everyone is perfectly equal. This gives a Pareto efficient outcome. Also it is the outcome which gives maximum possible total happiness given tastes and technology.
The above paragraph is obviously pointless. We don’t like poll taxes, because they would impose unreasonable suffering on the poor. We can’t talk about them sensibly if we start out by assuming there are no rich and no poor.
I think this is a universal problem. When Pareto wrote his book on political economy some (possibly mythical) reviewer said it was a good book but that it contained this silly idea that some reforms might help someone and not hurt anyone. Now for many economists that idea is all that remains of Pareto’s life work. I think the (possibly mythical) reviewer was right.
Pareto improvements can be found in economic models because they are models. They are simpler than reality. The constraint that the reform hurt no one is much less complicated, because the model is simpler than reality. If the key question is “is there a Pareto improvement” then the assumptions made for clarity and tractability are directly related to the answer.
Think of a real world example. How about the holocaust. That sure seems to have been a suboptimal policy. But if Hitler enjoyed it, then preventing it would not have been a Pareto improvement.
If we consider the real world, a Pareto improvement must satisfy 7 billion constraints, actually many more as the reform can’t hurt the unborn. Simplifying assumptions can’t be innocent if we are interested in Pareto efficiency.
OK, now consider statement 2. Everyone agrees that economic theory is a set of models, that the models are not reality, and that, at best, they are useful approximations to reality. So we hope that, while things won’t be exactly as the model predicts, they will be similar. So, for example, if the model says a reform helps people and harms 0% of people, we hope that in the real world it harms a tiny percent of people, say 0.1%.
That’s enough that it isn’t a Pareto improvement, so if economists really thought that the only thing we can say about welfare is that this or that is or isn’t a Pareto improvement, then they should conclude that basing policy on models is a silly mistake.
Needless to say the approach to welfare analysis based on contradictory assumptions is the Fresh Water School’s approach to welfare analysis. To be frank, I think the post, by itself, shows that everything they have ever said about policy is nonsense.
By the way, I made this argument once in front of Andreu Mas Colell and he nodded (I don’t claim he has any recollection of this event or even that he definitely nodded in agreement). He, uhm, was highly regarded at the University of Minnesota where he was the resident math genius/general equilibrium theorist. Now he has never been a Republican (his political development was communist, having been jailed 3 times by the Fascists, socialist (I think) at the time I made that argument and now Catalan nationalist. But he sure knows economic theory.