Thread by @Jabaluck

It's time for normative economics Saturday wherein we discuss whether economic analysis reaches insane conclusions about racism. (Indirectly inspired by:

https://twitter.com/Simon_Mongey/status/1269059038018330624

)

Suppose society is 75% white people who like being racist and are rich(er), so they are willing to pay $50,000 to preserve the racist status quo. It's 10% black people who are each willing to pay $50,000 to end the racist status quo. Does pareto efficiency say: let's be racist?

After all, we might say, instead of ending racism, keep being racist, but tax rich white people, subsidize black people, and then everyone is better off?

One tempting (for economists) response is to say, "There is a frontier of pareto-efficient outcomes corresponding to different initial endowments. The racist world might be 'a' pareto-efficient outcome but it is not 'the' pareto-efficient outcome."

This response misses the main issue. It may well be the case that if endowments were different, the pareto-efficient outcome would change. But it still seems wrong to say: if white people are rich and numerous, racism makes everyone better off with appropriate transfers.

The real problem is with the idea of taking preferences as given because "there is no arguing about tastes". This is a *terrible normative assumption.*

A quick aside: Becker and Stigler put forward "De Gustibus Non Est Disputandum" as a program for positive economics wherein instead of *explaining* things with taste differences, we try to model taste differences as arising from more primitive factors.

I'm not talking about the usefulness of that program here--I'm talking about the absurd normative idea shared by some economists that we shouldn't try to evaluate whether preferences are reasonable.

If you are willing to pay $50K to be racist, we shouldn't regard you as $50K worse off if we end racism, your preference is worthless trash. If you're willing to pay $100 million to see someone you dislike die, welfare isn't improved if you pay them to play Russian Roulette.

Is the problem here only with "Other Regarding Preferences"? (see: pdfs.semanticscholar.org/d16c/442d67fd7…). No! Self-regarding preferences are often wrong (uninformed) and arguably unethical (e.g. unfair to your future self).

Any call to say, "Well, let's weigh the claims of the protestors against the willingness to pay of police to exert power over them" would be rightly met with derision -- this isn't a shortcoming of conventional economic analysis but it highlights the need to be normative.

John Stuart Kill (

https://twitter.com/Sophie_Jallais/status/1263944532971982850

) is right that we should generally give deference to people about their own preferences, and more deference as we consider more onerous forms of paternalism.

But this doesn't excuse us from having to evaluate preferences, and it doesn't even make it sensible to take preferences as given in most settings. It means that we should be very cautious to avoid the common error of thinking we know better when we don't.

Every time we take preferences as given in welfare analyses we are making a substantive assumption that preferences are informed, consistent and deserving of ethical weight. These assumptions should be made explicit and defended.