The biggest flaw is parceling populations by State. US states are great statistical aggregrators, but not good disciminators. You can tell because of the narrow range of numbers. None is beyond one statistical deviation from the norm. All are clustered in the middle.

You need to look at cities. Even there, some have huge "metroplex" suburban satellites; and some of those have been regionalized into county plans. (Los Angeles is the paradigm.)

When we talk about the erroneous learning via public schools, we fail to identify "deep claims" such as the fact that the USA consists of 50 states. Even the Libertarian Party arrays itself alphabetically by state. Rail hubs and airports might tell you more about the lands and the peoples.

And they only checked 500k tweets across the nation. Even if those tweets were from a single person and no one person had multiple tweets counted, that's .1% of the population.

This is a ridiculous study. They didn't measure IQ, they measured the way people talk. Just because people use poor grammar doesn't mean they aren't intelligent.