[notahacker]

Even the paper itself acknowledges the weaknesses of the sample. It's much easier to make "based on consistent responses in our sample, humans generally behave in a certain way" generalisations from a small, unrepresentative sample than to conclude "this cohort of humans we've unrepresentatively sampled generally behaves differently from that cohort of humans we've unrepresentatively sampled"

Consider the hypothesis that (for the sake of simplicity) half of conservatives support conservative candidates because they're "susceptible to bullshit" and half of conservatives support them because they're hardworking professionals who believe those candidates' brand of fiscal conservatism is most aligned with their rational self interest in not being taxed too highly.

Which of these two categories of Trump and Cruz supporters would be underrepresented on AMT, a place for underemployed people to find menial work?