# How can an outcome be Pareto efficient but not Kaldor-Hicks efficient?

An policy is a Pareto improvement if it makes some people better off and no one worse off. And a policy is a Kaldor-Hicks improvement if it can be turned into a Pareto improvement by redistributing money. Every Pareto improvement is a Kaldor-Hicks improvement, but not every Kaldor-Hicks improvement is a Pareto improvement.

Now an outcome is called Pareto efficient if there are no possible Pareto improvements of it, and Kaldor-Hicks efficient if there are no Kaldor-Hicks improvement of it. Now clearly every Kaldor-Hicks efficient outcome is Pareto-efficient, because if a Pareto improvement were possible it would also be a Kaldor-Hicks improvement. But my question is, why is the converse not true?

Here is my argument for why the converse is true. Suppose that X is a Pareto-efficient outcome which is not Kaldor-Hicks efficient. Then there exists a Kaldor-Hicks improvement which can be turned into a Pareto improvement by a redistribution of money. But then that Kaldor-Hicks improvement together with that redistribution is a Pareto improvement, so there exists a possible Pareto improvement of X, contradicting the fact that X was Pareto efficient. Where am I going wrong?

Example. Each month, God gives Adam $$60$$ apples and Eve $$40$$ (for a total of $$100$$ apples). Let's write this allocation as $$X=(60,40)$$.

The Devil now comes along and offers to increase their total monthly allotment of apples to $$101$$, but on the condition that the allocation must be $$Y=(59,42)$$.

Observe:

1. $$X$$ is Pareto efficient but not Kaldor-Hicks efficient.
2. $$Y$$ is both Pareto efficient and Kaldor-Hicks efficient.
3. $$Y$$ is a Kaldor-Hicks, but not a Pareto improvement over $$X$$.
4. Through imaginary redistribution, we could get from $$Y=(59,42)$$ to $$Z=(60,41)$$. And $$Z$$ would then be a Pareto improvement over $$X$$. But the problem is that the option of making the redistribution $$Y\rightarrow Z$$ doesn't actually exist. And so, while $$X \rightarrow Z$$ would indeed be a Pareto improvement, the problem is that it does not exist and is not a possibility. It is merely something we or Adam and Eve may have imagined.

So let's see where you went wrong in your argument:

But then that Kaldor-Hicks improvement together with that redistribution is a Pareto improvement

This is correct. Using the above example, $$X \rightarrow Y$$ is the Kaldor-Hicks improvement, $$Y \rightarrow Z$$ is the imaginary redistribution, and $$X \rightarrow Z$$ is the Pareto improvement.

so there exists a possible Pareto improvement of X

The error is here. The redistribution $$Y\rightarrow Z$$ is imaginary and does not exist. So, the reallocation $$X \rightarrow Z$$ is also imaginary and does not exist.

(In this case the redistribution $$Y\rightarrow Z$$ isn't possible due to arbitrary constraints set by the Devil. But in the real world there will be other constraints.)

Edit: Another example

Suppose you and I are neighbors. I really like practicing the drums at 9 am on a Saturday. You really, really, like sleeping until 10 am on Saturdays, which is, of course, significantly disrupted by my drumming. Suppose I value drumming at \$10, and you value sleeping at \$20. What are the efficient outcomes?

The clear Kaldor-Hicks efficient outcome is for me to not drum. Sure, I lose (the equivalent utility of \$10 to be precise), but I lose less than you gain. If desired, you could compensate my loss, and still be better off than if I drum. The set of Pareto efficient outcomes is bigger. It's Pareto efficient for me to either not drum (letting you sleep), or drum away at 9 am! Of course, we could come to a mutually agreeable resolution where I don't drum and you pay me some amount between \$10 and \\$20. However this adds another layer of complexity that we didn't include: that enforceable agreements and transfers are possible (and costless). In the real world, where there are many other pragmatic issues (such as enforcement, positive transaction costs, lack of information about who the "gainers" and "losers" are, etc.), these payments might not be possible, and hence, we're reduced to considering the initial problem.

• What are the other constraints in the real world? Jan 8, 2020 at 8:57
• Good question but you should ask it separately
– user18
Jan 8, 2020 at 9:13
• @KeshavSrinivasan This answer is right on the money. The key is that the distribution is itself imaginary. I'll edit Kenny's answer to give a more "real world" (though somewhat more convoluted) example (if that's alright with Kenny LJ!) Jan 8, 2020 at 15:06

Youre erring in the definition of Kaldor-Hicks efficiency. K-H improvement is a change where those who lose from the change, lose less than the gain obtained by those who benefit from it. Pareto improvement is a change from which no-one loses and some gain. Thus all Pareto improvements are also K-H improvements but many K-H improvements are not Pareto improvements. A classical example of this is international trade. If a country opens its market to the world, some domestic producers will inevitably loose, but the gain of the domestic consumers from opening the market will be much larger than the loss of the domestic producers. I.e. opening the market for foreign producers is a K-H improvement but it is not a Pareto improvement. Hence autarky is Pareto efficient but it is not K-H efficient.

EDIT

The possibility of side payments you are talking about is a consequence of defining K-H improvements properly, not a part of the definition. This actually matters here. These side payments are not a part of the choice set when youre choosing between opening the market for foreign suppliers or keeping it closed.

I.e. if the market is currently autark and your choice set consists of exactly two options: X = {keep market closed, open market for foreigners}, there is no possibility for side payments and no Pareto improvement but there is a K-H improvement.

In this case both choices in X are Pareto efficient but only open market for foreigners is K-H efficient.

Talking about possible side payments is only a rationalization of the concept of K-H efficiency. The idea of side payments is based on the same logic as the Coase theorem on externalities. It simply is a nice property to have.

• What you said and what I said about Kaldor-Hicks improvements are equivalent - if the winners gain more than the losers lose, then you can redistribute money from the winners to the losers so that no one is worse off than they were before. In the case of trade, free trade combined with a redistribution from domestic consumers to domestic producers is a Pareto improvement. Dec 9, 2019 at 5:28
• @KeshavSrinivasan the possibility of side payments you are talking about is a consequence of defining K-H improvements properly, not a part of the definition. This actually matters here. These side payments are not a part of the choice set when youre deciding whether to open the market for foreign suppliers or keep it closed. Dec 9, 2019 at 6:00
• OK, but whether it’s part of the definition or a consequence of the definition, is it or is it not true that if you start Pareto efficient outcome that is not Kaldor-Hicks efficient, then you can implement a Kaldor-Hicks improvement followed by the associated side payment, and the combination will be a Pareto improvement of the outcome you started initially? And if it is true, isn’t that a contradiction? Dec 9, 2019 at 6:09
• @GradaGukovic I’m not talking about Pareto improvements coinciding with Kaldor-Hicks improvements, I’m talking about Pareto efficient outcomes coinciding with Kaldor-Hicks efficient outcomes. You can have the latter without the former. Dec 9, 2019 at 7:25
• OK, but how is that an example of Pareto efficient outcomes not coinciding with Kaldor-Hicks efficient outcomes. Dec 9, 2019 at 7:29