Reasonable Homunculi Can Disagree: The Impossibility of Welfare Economics

I’ve just uploaded a new working paper, which is a slightly edited version of a chapter from my thesis, to SSRN. Here’s the abstract:

This paper draws on the “preference reversal” literature in psychology and behavioural economics to argue for the impossibility of welfare economics. The effect of normatively-irrelevant contextual factors shows that humans do not have a coherent preference function which pre-exists and informs choice. Every choice is a constructive act which forces us to choose among incommensurable values: choice creates preference. This rules out the possibility of a value-free welfare economics and forces social scientists wishing to make normative conclusions to engage in indeterminate moral reasoning.

Quote of the Day: Bigotry and Utilitarianism Edition

There are many who consider as an injury to themselves any conduct which they have a distaste for, and resent it as an outrage to their feelings; as a religious bigot when charged with disregarding the religious feelings of others, has been known to retort that they disregard his feelings, by persisting in their abominable worship or creed. But there is no parity between the feeling of a person for his own opinion, and the feeling of another who is offended at his holding it; no more than between the desire of a thief to take a purse, and the desire of the right owner to keep it. And a person’s taste is as much his own peculiar concern as his opinion or his purse.

John Stuart Mill, On Liberty, Ch. 4, paragraph 12.

Liberals and libertarians need to think very carefully about the kinds of preferences/harms should be considered valid policy concerns. The are obvious cases: I wrong you when kick you in the shin, but not when I wear clothes you find distasteful. It seems that this is so even when you have a very high tolerance for shin pain and a low tolerance for fashion crimes, and the harm/disutility is equal in each case. Most people find it reasonable that people have a presumptive right not to be physically attacked, but no such right not to be visually offended by poor taste. There is a large grey area in between these two cases.  Utilitarianism as a moral theory is incapable of considering this question, or even admitting that it is a problem. This, more than anything else, is why I am not a utilitarian. The Mill of On Liberty was not a utilitarian in this respect either. On some readings, not even the Mill of Utilitarianism was really a utilitarian.

Update: Eric disagrees. Update 2: Not really. See the comments on Eric’s post.

Against ‘Efficiency’

Sandy Ikeda shares the abstract of an essay he is contributing to a Festschrift for Jane Jacobs. The Death and Life of Great American Cities is one of my favourite books, and I look forward to reading the Festschrift, particularly Ikeda’s essay. My favourite two paragraphs from the abstract:

The modern demand to rationalize the city and to make it “more efficient” is misplaced.  A living city cannot be efficient.  Efficiency, in the economic sense, presupposes an overarching plan against which measured outcomes can be evaluated.  A living city, however, follows no such plan.  It is itself the unplanned, collective result of the countless individual plans executed continuously in it, day after day.

Neither can it be inefficient, because that too presupposes a system-wide plan.  Both efficiency and inefficiency presume that we know how things ought to be, what success and failure look like, and that’s impossible in the urban dynamic.  Instead, borrowing from ecology (and certain heterodox schools of economic thought), we might say that a living city is a “dynamically stable” process, in which the forces of positive and negative feedback, as well as sudden mutation and diversity, combine under the right conditions to generate order through time.  It embodies trial and error, surpluses and shortages, apparently useless duplication, conflict and disappointment, trust and opportunism, and discovery and radical change.  These are in the nature of the living city.

This point applies to any form of social organization. I have recently become very suspicious of anyone positing a social welfare function as the basis of policy (something I try to get at here and here). I think Arrow’s Impossibility Theorem and the impossibility of interpersonal utility comparison make it fairly obvious that no social welfare function could hope to be neutral. Most often, advocacy of utility maximization is a way of dressing up implicit moral judgements as value-free economic science.

Pluralism, Subjectivism, and Welfare Economics

Some recent posts by Rauparaha at The Visible Hand in Economics have got me thinking about the problems of subjectivism and value pluralism in welfare economics. Rauparaha is using widely-accepted methodology and taking the conventional view on this, so my criticisms are not directed at him particularly, but at the general approach taken by most welfare economists.  He says, for example:

It’s great to see the government taking economic incentives seriously. Their latest initiative considers imposing a 5c/bag tariff on plastic bags in supermarkets. The idea is that the market price for the bags doesn’t take into account the full environmental cost of non-biodegradable bags. By taxing the bags the government can adjust the market price of the bags to match their social cost.

It seems to me that the idea that government can use Pigouvian taxes and subsidies to internalise externalities stems from a combination of technocratic hubris and a fundamental misunderstanding of the concept of welfare.

First off, and least interestingly, I don’t think the messy workings of government are particularly likely to produce those policies welfare economists would find optimal. Public Choice Theory tells us to treat markets and governments on equal terms: neither businessmen nor politicians are angels and both markets and governments fail. The benevolent despot model of government implicit in much of economics is unjustified, since government is neither entirely benevolent nor under the control of a single decision-maker: it is neither benevolent nor a despot. This, by itself, doesn’t make conventional welfare economics utterly worthless, since there may be some value in specifying what it would be good to do, even if we recognise that it’s not likely to happen that way in practice.

A more fundamental problem with the idea that government can provide incentives through taxes and subsidies to induce efficient levels of consumption of some good is in the very idea that there exists a definite amount of utility, either within or across individuals, which the government can attempt to maximise.  

We value many things and these are not easily commensurable. Everyone has surely experienced the difficulty of trading off different preferences against each other: higher pay or more job satisfaction, night of pure hedonism or productive morning, new TV or new computer. Of course, we do manage to trade these things off and on average our decisions can reasonably be said to make us better off. It is only through facing the trade-offs and thus being forced to weigh them, however, that we discover what we really value. As Mises puts it:

Economic action is always in accord only with the importance that acting man attaches to the limited quantities among which he must directly choose. It does not refer to the importance that the total supply at his disposal has for him nor to the altogether impractical judgment of the social philosopher concerning the importance for humanity of the total supply that men can obtain.

I see no reason to suppose that there is something analogous to a substance that we are maximising here. When I make myself richer but have less fun in the process, I am improving my life in one sense and worsening it in another. I have demonstrated through action my preference, but why should there be any basis for supposing that I have increased the quantity of some common thing within myself called ‘utility’ to which both short-term fun and wealth can be reduced? I like Sen’s idea of utility as a vector, but I don’t think it goes far enough. It is only through trade or other decision-making that we come can come to say that good X is more valuable than good Y to agent A at time t. We cannot come to know this by simply asking them, since this does not approximate a real decision and will not put them in a genuine trading-off frame of mind. I suspect that expressive preferences are likely to dominate in such situations.

I’m tempted to say that there is no fact of the matter whether A prefers X over Y until he actually faces the choice. I have no idea whether I’d prefer world peace to a twenty percent increase in global economic growth, because this is not a decision I have ever had to make. I can make a number of arguments either way, but these are biased by how I would like to see myself. If I ever were in the position to push the button, I’m sure my decision-making would become less biased and more consequentialist. If I cannot predict in advance what I would prefer given the choice, can I really be said to have a true preference? Rather than maximising utility from some pre-defined function, human action itself in some sense defines our preferences: exchange is performative.  

All of this becomes far more problematic once we begin to consider interpersonal utility comparisons. We clearly can’t aggregate something we can’t quantify, and there is no neutral way to make tradeoffs between the preferences of different individuals. How do we trade off satisfying the preferences of many people versus leaving a few utterly miserable? How do we weigh the value of poetry and pushpin? Without individual decision-making in the real world, it’s hard to see how this is even possible in principle.

Bringing this back to taxing plastic bags: it seems likely that discarded plastic bags cause people some displeasure, and they may hypothetically be willing to give up some other resources to avoid this displeasure. What they are willing to give up, however, is fundamentally unknowable unless we create some sort of market for bag-reduction or observe the compensating differentials in bag versus non-bag areas: i.e. for practical purposes, we can’t. It is not just that we don’t know the precise value of the disutility caused by discarded plastic bags, if this were the case it might be reasonable to set a conservative Pigouvian tax and feel confident that we are improving things somewhat while not making things perfect. The real issue is that we have absolutely no idea of how people are affected, or even if they would be willing to give up anything at all to see less plastic bags, and have no reliable and realistic way of finding out.