I base this assertion on the IBD/Tipp poll, supposedly one of the most highly respected pollsters and rated as most accurate over the last 3 Presidential cycles (rated A- by 538). They are also very generous with their table data.
I present to you the smoking gun, religious affiliation:
Based on 2012 exit polling voters were 53% Protestant, 25% Catholic, 7% Other, and 12% None. Keep in mind exit polling, while not perfect, is fairly accurate.
However in the IBD/Tipp poll here are the results for religious affiliation:
Protestants: 55% Trump, 36% Clinton (+3% over Romney's margin)
Catholics: 54% Trump, 38% Clinton (Obama won this in 2012 by 2%!)
Other: 40% Trump, 44% Clinton (Obama won this group by 51%!!!)
None: 22% Trump, 67% Clinton (Trump +4%, Clinton -3% from 2012)
So what are the results when you do the simple math, multiplying the share of each religious group in exit polling times the percentage vote in this poll?
TRUMP = (53)(.55)+(25)(.54)+(7)(.4)+(12)(.22) = 48.09%
Clinton = (53)(.36)+(25)(.38)+(7)(.44)+(12)(.67) = 39.7%
TRUMP LEADS BY 8%, HE IS NOT TIED!
If you do a similar analysis with non-college educated, some college, and college degrees, the same problem pops up, a gross under-sampling of Trump's biggest vote demographic (non-college and some college)! This is clear evidence of an attempt to over-sample atheistic college graduates.
