My problem with average utilitarianism

Today’s most influential moral framework is average utilitarianism (AU for short): the idea that whatever increases the average person’s welfare is good. It’s the default way economists use to make welfare comparisons. It is also reflected in the use of GDP per person as a measure of a society’s well-being.

By contrast, total utilitarianism (TU) says that good is whatever increases the sum of welfare. This only makes a difference when the number of people under consideration changes. For example, 1000 people with 2 welfare units each is better than 100 people with 3 welfare units each according to TU, but worse according to AU. (Never mind how welfare units are calculated or compared across people. That’s a problem for both varieties, and for another blog post.)

John Rawls puts the difference nicely. TU is like taking a God’s eye view of humanity and looking at the total good in it. AU is motivated by the thought “what if I were a randomly selected person?” So it chimes well with contractarian, veil-of-ignorance reasoning. If you are not sure who you will be in society, and wish to maximize your own welfare, then maximizing average welfare could be the right thing to do. (Rawls thought you should maximize the minimum – again, that’s another debate.)

Another problem for TU is that it requires a “zero point of welfare” – someone who is just happy enough that her happiness neither adds to the total nor subtracts from it. It’s not obvious what that zero point should be. AU doesn’t need a zero point. You just look at whether the new person is happier than the average.

The difference has consequences in applied work. For example, it is often claimed that the agricultural revolution left people worse off on average. After agriculture, skeletons found in archaeological digs appear shorter, and show that people were less well-nourished (this paper has some references). On the other hand, agriculture let many more people exist on the planet. Similarly, average incomes may have fallen in the early Industrial Revolution, or during the British Raj in India. But at the same time, the population increased greatly. So, did these events make the world better or worse?

AU has a weird property that make me suspicious of it: the evaluation in one area depends on what happens elsewhere. Go back to the example:

State X: 100 people with 3 utility each

State Y: 1000 people with 2 utility each

Here, X is better than Y. (An average of 3 beats an average of 2.) However, now add another country where there are 100 people with 0 average utility. Now state X has average utility of 1.5, while state Y has average utility of about 1.8.

This is a practical problem, because we can now only evaluate welfare at a global level. For example a recent book claims that Soviet socialism was actually good for Soviet citizens: by reducing the birth rate, it increased average welfare more than if the Tsars had stayed in power. But even if the factual argument is correct, we still need to know what was happening in the rest of the world before we can evaluate whether socialism made the world a better place or not.

Even worse, if you are philosophically minded: maybe the “other country” is really another planet in a distant galaxy! Now your evaluation of even global welfare depends on some unknowable facts about the universe. It seems more sensible to let the welfare evaluation of states X and Y depend only on the people affected by X and Y. (TU achieves this. To evaluate X and Y, just see which gives more total welfare to all affected people. Then by definition, that one also adds more to the total welfare of all beings, no matter what the Martians are doing.)

Aside from measurement trouble, it seems that as a matter of principle, evaluation of a state of affairs should depend only on that state. Surely if X is better than Y, once we have taken account of all the people to whom X and Y make a difference, then it must always be so, whatever else is happening in the world. This is a kind of “independence of irrelevant states” principle for welfare evaluation.

I know very little about this topic. I am sure moral philosophers have done it to death, and I would love to read a good discussion of it.