Monday, 25 May 2015

Is Pareto optimality the most misleading definition in economics?

John Quiggan thinks so (HT: Tim Haford). I often disagree with John but understand the different perspective he is coming from. But this post has me really perplexed.

John's post starts with a discussion of Pareto's political views, suggesting that his construction of the concept of Pareto optimality and his other main intellectual contribution, the Pareto income distribution, were motivated by an antipathy to government redistribution policy. This may be right, but that fact in itself doesn't make the the concept of Pareto optimality misleading or dangerous. 

On the term itself, John's first point is that the use of the word "optimality" is misleading. On this, I agree to an extent. Optimality suggests the result of an optimisation problem with a well-defined objective, but one of the first things we teach our students when invoking Pareto optimality is that there are many different allocations that are Pareto optimal, and the Pareto welfare criterion (the value judgement that making someone better off without making anyone else worse off is a good thing), is silent on how to compare these different optimal allocations. Indeed, typically the first thing undergraduate economics students see after encountering Pareto optimality is the two welfare theorems, the second of which specifically refers to different Pareto optimal allocations, and how all can be sustained by a competitive market given an particular allocation of property rights. Yes, optimal is a poor choice of word, but the context in which it is used does make it clear that it doesn't imply the single-best (by some notion) outcome. 

But then John says: 
Recognising the inappropriateness of describing radically unfair allocations as “optimal”, some economists have used the description “Pareto efficient” instead, but this is not much better. It corresponds neither to the ordinary meaning of “efficient” nor to the meaning with which the term is commonly used in economics, which is also misleading, but in a different way.
A few comments are in order here: First, the "some economists", surprised me. I thought that "Pareto efficiency" was overwhelmingly the preferred term for the concept. I quick google of "Pareto optimality" and "Pareto efficiency", yielded 263,000 hits for the former and 377,00 for the latter, which is a way smaller ratio for the latter than I was expecting, but still a majority. Second, I thought that Pareto efficiency corresponds exactly to the way the term "efficiency" is used in economics. In fact, I always teach my students that if the noun "efficiency" is used without an adjective, it is always Pareto efficiency that is implied. I would be interested to know how our commenters perceive the term. 

John suggests instead using "opportunity cost". 
The concept of opportunity cost gives us a better way to think about the possibility of making some people better off while no one is worse off. If such possibilities exist, then there are potential benefits that have no opportunity costs. Conversely, if there is a positive opportunity cost for any benefit, then we can’t make anyone better off without making someone else worse off. So, a “Pareto optimal” situation may be described, more simply as one where all opportunity costs are positive.
This doesn't seem right to me. Opportunity cost is a tautological statement about scarcity, not about welfare. If I take resources away from some use (say private consumption) and use it for another (say the provision of a public good), and do it in such a way that it is Pareto improving, there has still been an opportunity cost. TINSTAAFL still applies: we didn't get the public consumption for free; it's just that we all preferred what we got to what we had to give up to get it. 

So while "Pareto optimality" may be a bit misleading, I don't think "Pareto efficient" is, and I'm not sure what is a viable alternative to what is a very useful concept. So let's get back to whether the concept itself is dangerous. 

The nicest thing about the concept of Pareto efficiency is that it expresses outcomes purely in terms of the values of the individuals on society, not the values of the particular economist or government, employing the concept. There are more things that we might value than just Pareto efficiency, but how we weight them is a personal value judgement. Best to leave them as separate concepts and report them all, rather than obscure the weighting value judgement in the construction of a single well-being metric. (That is the main reason I am nervous about moves to move beyond GDP to "Gross National Happiness", but that is the subject for another post.)

Where Pareto efficiency becomes dangerous is when it is confused with the Kaldor-Hicks notion of potential Pareto improvements and then implicitly used as a welfare measure. Recall that according to this notion, a policy is desirable if in principle it would be possible to implement it and then make side payments where the winners compensate the losers even if those side payments are not made. Of course, it would be rare for an economics course to put that principle in front of students and suggest that it is a reasonable one, but sometimes, in taking the view "economists can talk about efficiency, but have nothing to say about equity", economists make it all to easy to confuse "we are only qualified to talk about efficiency, not equity" with "only efficiency matters, not equity". This does not mean in any way that we should abandon the concept of Pareto efficiency or change its name; we just need to be more careful to be clear that it is not a single measure or well-being. 


  1. Thanks for this Seamus, I also thought his post was odd when I saw it.

    Two quick things. You said: " Optimality suggests the result of an optimisation problem with a well-defined objective, but one of the first things we teach our
    students when invoking Pareto optimality is that there are many
    different allocations that are Pareto optimal". It might be worth noting that Negishi's Theorem shows that every Pareto efficient allocation can be described as the solution to a social planner's problem with appropriate weights on the utility functions of each agent in society.

    Second, on the popularity of "pareto efficient" vs. "pareto optimal": when I search for those terms exactly, the latter far outweighs the former. Also, a search for those terms in books published over the last 200 years confirms the more common use of "optimality":

  2. I discussed reasons why good effects of alcohol were likely to be minimized in public discussion in a blog post some years back:

  3. Nice post.

    The anti- piece Slate Star Codex linked was pretty terrible - it failed to include any of the serious J-curve evidence (Di Castelnuovo, Rimm & Moats, etc), as expected from that set of authors. But I do need to find time to give the new gene-based study a closer look.

  4. I too had read JQ's comments; and was happy enough to go along with them until he got to the "opportunity cost" part. Like you - it seemed a bit wrong, or unusual at least. But I think the whole section I was reading was from a draft, and I did wonder if that section would remain in later editions.

  5. Re: the opportunity cost thing, I think he's talking about the difference of the opportunity costs of the two situations.

    So if I have mutually exclusive options A and B, preferring B, then the opportunity cost of A is that I can't have B, and the opportunity cost of B is that I can't have A. Since I prefer B, then I will consider the one cost to be greater than the other, so the net opportunity cost of moving from A to B is negative, and from B to A it is positive. This seems to capture all the information contained in the concept of "Pareto efficiency", I think? Which makes it a coherent alternative.

    No comment on whether the one conceptualisation is superior to the other.

  6. Your search terms were slightly different; you searched for the noun form while James for the adjective.

    If I hadn't any prior exposure to either of the phrases, I probably would have guessed that "Pareto optimality" was used less frequently than "Pareto efficiency", simply because "optimality" on its own isn't that common a term. I suggest that explains your results.

  7. Ah I see, that actually makes things a lot clearer for me. And yes, that's what I understand Negishi to be doing, I just misunderstood in thinking that you were trying to say that pareto optima cannot be derived from *any* optimization problem.

    I've definitely more often seen "efficiency" over "optimality" (and likewise, "efficient" over "optimal"). And like you, I read "optimal" and think "unique maximum", so I prefer "efficiency". But then I'm also conscious of something you said in class once (something to the effect of): "to the layman, 'efficiency' conjures up notions of cold, soulless, number-crunching, corporate-raiders. When talking to the public then, rather than to economists, it might pay not to use that particular word".

  8. Yes, "efficiency" would still be in my top-4 list of misleading terms in Economics. But as long as we are keeping discussions within the profession, I am happy with "Pareto efficiency".

  9. What are your other three, pray tell?

  10. Ugh. I hadn't noticed that James had gone for the adjective.

  11. It's a post in draft.

  12. "Is it the noble lie: that folks who've heard of the J-curve would use it to rationalize far greater drinking, so it's best to pretend it doesn't exist?"

    Probably. One of the stated purposes of the NZ Drug Foundation (for example) is "how to minimise the harm of using drugs", and one could make a good argument for the net utility of a white lie in this case.

    As a former heavy drinker, it's very tempting to try and confirm my bias that maybe it wasn't that bad after all.