Just a very fundamental point: People use quantitative data in order to debunk qualitative impressions. But when the data we are discussing is putative welfare measures (e.g. real income or wealth, GDP per capita, etc), the *qualitative* measures are foundational, the quantitative mere proxies. 1/

GDP per capita is widely used as a welfare measure not because it conceptually maps well to welfare (for all kinds of reasons it does not!), but because from the mid-20th to early 21st Century it mapped pretty well to our qualitative intuitions about relative welfare. 2/

The consensus that there is a good correlation between GDP per capita and qualitative welfare has broken down more recently. 3/

We can have arguments about why (inequality, differences in how medical and social insurance are accounted in GDP, market power, treatment of leisure). 4/

But fundamentally, it was only ever a good measure because there was a widespread consensus that it tracked qualitative outcomes. Once that consensus has broken down, there is no reason to think it *should* be a welfare measure. 5/

(The inventor of GDP, Simon Kuznets, explicitly argued that it should not be!) 6/

The same is true of "real" wealth or purchasing power measures! They are not inherently welfare measures. (I belabored this in a recent post.) 7/

If people are making qualitative claims that some group's welfare is poor, and you debunk them with quantitative data, whether GDP-per-capita or real purchasing power measures, you are engaged in a kind of circular reasoning. 8/

The only reason we think these *should* be welfare measures is because they sometimes seem to work well at capturing our qualitative experience and intuition about relative welfare. 9/

If qualitatively they seem to cease to work well as welfare measures, then there is no reason to think they are good welfare measures! When you debunk widespread qualitative expressions of welfare with this "data", you are really debunking the quality of your measures! 10/

That's not to say all unevidenced claims about qualitative welfare must be taken as gospel, at face value. The claims could still be wrong! 11/

Welfare is unobservable, hard to measure. This is economics' foundational demon as a "science". 12/

The moments when there is a strong consensus that any quantitative measure maps well to welfare are fleeting and precious. During those exceptional moments, it seems plausible that we might maximize welfare "scientifically". 13/

The rest of the time, like now, we have to cop to the fact that human welfare is not a scientific observable, but something we construct normatively and strive to achieve politically. /fin

( i've turned this thread into a blog post: drafts.interfluidity.com/2025/11/19/t... )