A philosophy for a new old welfare economics, Part 1.

This is part one of an outline of my planned PhD thesis on the philosophy & practice of welfare economics. I’ll link part II here when I write it. I’m about three months into the research for my thesis, so forgive me if it’s a little loose. I wanted to put it out there to seek feedback, reading suggestions etc. I’ve had to condense months of thinking into very little space, so I apologise if it’s a little abdrupt, or requires you to fill in blanks yourself that should be spelt out in detail.

Welfare economics is a branch of economics that considers how economic policy and resource allocation affect human welfare. Historically welfare economics was often conducted in an explicitly utilitarian spirit, with policies evaluated by their effects on average welfare. In the 1930’s, that began to change.

The New Welfare Economics was an approach to welfare economics developed in the 1930’s, influenced by the global wave of behaviourism in the social sciences. You might think that welfare economics necessarily involves asking questions like “is Bob happier than Sarah” -interpersonal comparison-, as well as other mental concepts. The New Welfare Economics argued it was possible to do away with such concepts. Long after behaviourism receded elsewhere, the New Welfare Economics has remained strong, even hegemonic, in welfare economics.

One of the flagship approaches of New-Welfare economics is the Kaldor-Hicks compensation test. A policy or proposal is said to pass the Kaldor-Hicks compensation test if and only if:

Those who win from the policy could compensate those who lost from the policy if they so chose

Note that it is not essential that the compensation actually is paid for a policy to pass the test. So suppose I had a block of land and you stole it, allowing you to complete a major development worth millions. You could easily compensate me but unsurprisingly do not. This is a Kaldor-Hicks improvement! Obviously, this won’t do as a measure of welfare and numerous economists and philosophers have said as much. Yet it has remained in part because it’s a lot easier to test than other claims related to the question of whether a policy has improved human welfare-it contains no reference to mental states.

Especially in the context of Cost-Benefit analysis which we will discuss more extensively in the second blog-post, the Kaldor-Hicks philosophy is taken to mean that a dollar is a dollar and has the same social value, whether it goes to a rich person or a poor person. If you gain five hundred dollars and I lose four hundred dollars, even if that was the last four hundred dollars I had, you could compensate me. Thus this is a Kaldor-Hicks improvement. The fact that you won’t actually compensate me is disregarded.

This isn’t a niche academic issue. Kaldor-Hicks compensation is the spiritual foundation of cost-benefit analysis, and governments around the world use CBA every day. In the United States, regulations are required by law to pass a cost-benefit analysis and have been since Reagan. It is even possible that one cause for the spike in inequality post-’70s has been the coming to power of cost-benefit analysis. Cost-benefit analysis can be weighted to account for inequality (don’t worry about the details yet) but generally isn’t.

We will explain cost-benefit analysis in greater detail in part 2 of this blog series, but for the moment you can safely think of cost-benefit analysis, as the applied form of Kaldor-Hicks efficiency. In reality they are not equivalent for technical reasons, but Kaldor-Hicks can certainly be thought of as the “spiritual foundation” for cost-benefit analysis.

Tangent: but is the Kaldor-Hicks approach strictly wrong?

If I were to try to defend Kaldor-Hicks against the standard criticisms I’ve raised so far I’d say something like: “While it is true that the Kaldor-Hicks criterion can’t tell us whether a policy is desirable on its own, nonetheless it defines a concept of efficiency. This concept of efficiency can be studied independently of ethical concerns and is of inherent interest”.

This is true. Studying the economy with a Kaldor-Hicks framework is not by any means “wrong” from a factual point of view. What is unclear is that KHE tells us anything meaningful about welfare. Even to the extent that it does tell us something about welfare, it is hard to see that it’s absolutely central place in “welfare economics” is justified. Constantly analysing whether policies are Kaldor-Hicks efficient and calling that practice “welfare economics” places an inherent axiological weight on Kaldor-Hicks that I would argue it can’t bear. There’s also a motte & bailey fallacy where KHE is just an abstractly interesting form of efficiency when it suits, but when the criticism goes away it becomes a yardstick for policy- after all, it must be a good thing to be “efficient”.

(Aside: Speaking about highly technical terms as if they had the same moral weight as superficially similar, but ethically much richer concepts is not a new problem in economics. One of the most frustrating examples is the way many textbooks butcher the concept of “allocative efficiency”, equivocating between meanings to make it sound as if undistorted markets maximise aggregate welfare.)

Another defence of the Kaldor-Hicks approach is to suggest that we can make Kaldor-Hicks improvements and then sort out the distributive issues at tax time. We’ll talk more about this when we talk about CBA in Pt 2 of this blog post. For now I’ll briefly say that there are a number of issues with this approach. One of the main issues is transfer costs. Another is the implicit assumption that non-tax redistribution is as salient as tax-based redistribution- non-tax transfers may be less salient and thus have a smaller negative effect on the labour supply. Also, there are political economy effects insomuch as giving people more stuff equates to giving them more power, which makes it harder to then redistribute the stuff you gave them.

2. Value, science & epistemic risk

Our aim is to outline an alternative approach to welfare economics, and for that we have a few guiding claims. The first point has often been made. Nevertheless, it bears repeating:

Thesis 1: It is both permissible and desirable for scientific research programs to be guided by human needs

The second is an attitude towards epistemic risk

Thesis 2: Doing science involves taking epistemic risks, and a major factor in deciding which epistemic risks we take- what assumptions we are willing to make- should be human needs.

This is a little more complicated so let me explain with an example. There is a philosophical problem around the interpersonal comparison of welfare. What does it mean to say Bob is happier than Sally? How can there be a common unit of measurement of their happiness- is Bob’s happiness really commensurable with Sally’s happiness in the same way that two lengths are commensurable?

There have been various attempts to address this problem, and personally, I find them persuasive. So do most psychologists in practice- indeed many psychologists have never even thought about the problem- happily assuming that your “strongly agree” on a personality test is equivalent to my “strongly agree”- at least roughly. However many economists remain unconvinced. Believing in interpersonal comparisons constitutes an epistemic risk, and this has led many economists to retreat to the New Welfare Economics, ordinalism and calmer epistemic waters generally.

This approach reminds me of the parable of the streetlight. A woman is walking down the road and sees a drunk man searching for something under a streetlight, keeps walking and thinks nothing of it. Ten minutes later she walks past the same spot and sees him again looking in the narrow radius of the streetlight for something. “What are you looking for?” she asks “For my keys” replies the drunk “I lost them around here somewhere”, “Why are you only looking there? You must have searched those few meters a dozen times!” “Well, that’s where the light is!” Much in the same way, the new welfare economics, by refusing brave the difficulties and study actual human welfare, is searching for the keys where the lights are.

While it has special difficulties, we do know how to study welfare. There are well-developed practices for measuring human welfare in psychology, i.e. these OECD guidelines. Research programs that assume the interpersonal comparability of psychological states and traits (including happiness, which is sometimes seen as synonymous with welfare) have been ongoing for decades. While such studies are not always robust to further research, this is a general condition in the social sciences. Indeed it could be argued that happiness research is in a better position to comparable fields in psychology like social psychology.

Economics is more than happy to take its own favoured epistemic risks, the extremely strong assumptions made by research programs like the dynamic stochastic general equilibrium have frequently been remarked. These aren’t small risks. So clearly, whatever is going on in this division between psychologists and economists on the question of interpersonal comparison it’s not just that economists are more averse to potentially risky assumptions.

So welfare economics relentlessly focused on actual human wellbeing studied empirically is possible, yet approaches like the Kaldor-Hicks test retains considerable strength. In the next section, we will consider why.

3. Values & political savvy

Given the flaws of the Kaldor-Hicks compensation test, one has to wonder why it still retains its canonical role, part of the answer, I think is to do with the politics of alternatives.

The Kaldor-Hicks compensation test advertises itself as “value-neutral”. This claim is somewhat ridiculous, KH, and cost-benefit analysis understood as an applied form of KH encodes a particular value- function. It’s a quite strong variant of anti-prioritarianism that no one really holds.

Let me explain. Roughly, prioritarianism is the view that it is more important to improve the welfare of the worst off than the best off. Antiprioritarianism is the view that it is more important to improve the welfare of the best off than the worst off. Almost no one in the world is an antiprioritarian, but the Kaldor-Hicks procedure implicitly is. Since marginal utility is declining in income, to get the result that dollar means the same whether it goes to the rich or poor you’ve got to value the utility of the rich a lot more than the utility of the poor, hence Kaldor-Hicks is antiprioritarian in practice. Moreover, given that marginal utility declines fairly rapidly in income, Kaldor-Hicks is quite strongly antiprioritarian.

Nonetheless, not being tied to any popular existing school of ethics has given KH an advantage. Suppose you wanted to put forward an alternative approach to measuring welfare changes- say hedonic utilitarianism. At this point Amartya Sen is going to pop up and tell you you’re doing it wrong, what you actually need is his form of objective list theory. Not so, says the Rawlsian, what we actually need is preference-based maxmin. There’s a big N-way squabble, and then the committee grumpily decides to revert to Kaldor-Hicks because at least that’s got the institutional inertia.

My solution to this mess is what I call measure utilitarianism or sometimes strategic utilitarianism. Measure utilitarianism is the view that we should unite around a demand for the inclusion of utilitarian measures as a supplemented to bureaucratic exercises like Cost-Benefit analysis even if we are not utilitarians. To arrive at measure utilitarianism consider:

Thesis 3: The social welfare function(s) with which we do our cost-benefit analysis, or similar is not necessarily intended to be a kind of automated social decision-maker. It’s just additional information that may supplement other morally relevant information. Our debate, therefore, should be about whether the information is worth including, not about whether it represents by itself an adequate decision criterion.

Thesis 4: Utilitarianism is much closer to more or less every popular extant view on social welfare functions than the strange anti-prioritarianism that Kaldor-Hicks represents.

Thesis 5: Utilitarian metrics are, in an important sense, especially tractable- utilitarian metrics are easily understood and have no free further parameters that can’t be filled in by empirical research. Hedonic utilitarianism has an enormous research base on how to measure its variables. It is practically feasible to provide reasonable estimates of the effects of many policies on happiness.

These jointly lead to thesis 6

Thesis 6: Measure utilitarianism- from many points of view there is value in taking utilitarian measures to supplement New Welfare Economics based measures like unweighted CBA.

This does not exclude also taking, for example, priortarian measures, in fact, I would support it. We’re focusing on utilitarianism because here we are just trying to establish a narrow case for a particular demand intended break the hegemony of unweighted cost-benefit analysis. That narrow case is that it is hard to object to adding utilitarian measures if the alternative is to use something like unweighted CBA alone.

We use things that could be taken as social welfare functions as mere indicators all the time. For example (mean income*(1-Gini index)) defines a perfectly valid SWF- indeed not even a particularly bad one. Despite this, and despite the existence of multiple inequality indexes, we are happy to use the Gini index as an indicator of inequality, and would certainly never use something obviously inadequate on its own simply because we couldn’t agree whether to use the Gini or Theil index of inequality. I am suggesting a similarly pragmatic approach to utilitarian metrics- as indicators of the broad effects of policies on happiness.

The reader might object that in our argument for measure utilitarianism we’ve mostly had in mind comparison with other welfarist views like prioritarianism. What about non-welfarist views? Non-welfarist views are views which judge the goodness of policy not only through the welfare outcomes of policy but also through other outcomes (say, the preservation of the environment) or whether the policy doesn’t violate a variety of side-constraints. Well, even here many theorists who believe these alternative approaches will place at least some weight on the welfare implications of a policy, and it seems we gain more information about such consequences by including a utilitarian measure than a Kaldor-Hicks analysis alone. Even in the worst-case scenario, when considering a view which is entirely indifferent to the welfare-related consequences of a policy, it’s hard to see that such an approach would have a greater objection to including utilitarian analysis than unweighted CBA.

Ethical debates are interminable. At the current rate, they will never be resolved, and we have no reason to think that will change. Yet in the meantime kludges like Kaldor-Hicks continue to prevail. An ethical Levithan is needed to overcome this (un)ethical state of nature.

What we have tried (and hopefully succeeded) to do in the above argument is sketch out a reason to supplement vanilla welfare-economic analysis with a utilitarian analysis that does not depend on accepting utilitarianism as the one true social welfare function. Our criteria have been primarily political rather than ethical in any direct sense, arguing that such a view is likely to be broadly acceptable to representatives of other ethical viewpoints viz their mutual interest in defeating the hegemony of unsupplemented and unweighted cost-benefit analysis.

I am aware that this argument blurs the line between political strategy and philosophical analysis but I don’t necessarily see this as a bad thing. In fact, I hope to explore the possibilities of such a blurring in my thesis and the ways in which my account of how they might be blurred is different than other accounts (for example, my account doesn’t depend on an overall anti-realist outlook).

I suppose the approach I’ve outlined in section 3 might seem to some to undermine that in section 2, but the pragmatic adoption of utilitarianism as a standard is made possible precisely because a huge range of ethical views are closer to each other in the ways that matter than the Kaldor-Hicks approach. Values led social-scientific inquiry doesn’t require you to be an ethical sectarian.