Thread by @cremieuxrecueil

How many Americans receive food stamps?

If you go based off of surveys like the ACS, CPS ASEC, or SIPP, you might end up with the wrong answer.

If you use administrative records, you'll find that people usually underreport the welfare they receive🧵

If you look at receipt of Temporary Assistance for Needy Families (TANF) and unemployment insurance, a similar pattern shows up:

What we're seeing here is something that affects not only the estimated prevalence of benefits receipt, but also the magnitude of it. In other words, people underreport the size of their welfare checks.

If we take these facts together, then we'll notice something interesting: America's poverty rates are overestimated.

Poverty is, in fact, somewhat less common, less severe, and considerably less racialized than is generally understood:

An open question is Why do people misestimate? Well, go look back at the first two charts again.

Did you notice that misestimation of benefits receipt was greater for Blacks and Hispanics than for Asians and Whites? That's an important detail here.

When people respond to surveys, errors in their responses are predicted by tons of variables, from sex to race to education to height and so on.

One of my favorite examples is that people with low levels of educations misreport their heights more often.

What surveys show us about welfare usage is different from the reality, and that has unfortunate consequences, like misleading us about the extent to which Americans are poverty-stricken and underserved by their governments.

What this also tells us is that, somewhere, some bureaucrat could have probably assembled this data and contradicted a years-old literature, because all the researchers were using survey data that turned out to be, in some ways, wrong!

If we want to ensure this doesn't keep happening, we'll need the government to be more open with administrative data.

To learn more, check out my latest article: cremieux.xyz/p/why-americas…