TLDR: Don't trust this paper as meaningful, it's meant only as a tentative step towards something which may be meaningful.
They put Alabama and New Mexico in the same category? Even when the latter has "red or green [chile]" as the official state question, and has a very different environment due to being in the high desert?
And all of South Africa into a single category? The supplementary materials say "For South Africa in the B&S dataset1, the cookbooks from which the recipes were derived are written by Afrikaans-speaking authors8, so we have used the Afrikaans language to represent this cuisine." but that is only one of many cultures in South Africa, including a large Indian population with spicy food.
This makes me regard the paper with a wary eye.
Ohhh! All of the peer reviews are published! See "Peer Review Information" at https://www.nature.com/articles/s41562-020-01039-8 . It took a while to get this paper through, with lots of comments about issues.
Reviewer #1, after giving the okay, points out:
"I am aware this was a limitation of the sources upon which the study relied, and that similar biases were present in previous work, and this does not undermine the study overall because it is a “tenative step” towards better understanding, but please make this a very clear and strong caveat in the discussion and conclusions. It may also be leading to the rather odd correlation between GDP and spiciness which I refer to above."
where that correlation above is:
"the rather limited global sampling (see below) may also be playing a role here. For example, off the cuff, adding in say Singapore (which I imagine might have spicy food and be rich) and Nicaragua (which has rather bland food but is poor) could make a difference…I would not suggest the authors try to expand their sample (unless very straightfoward), but please at least discuss this possibility."