Why can't you get this.
Atheists have no shared beliefs - they simply had a shared disbelief that a god or gods exist.
Some Atheists may say that religion causes war, some may not. For myself I think the answer is obvious that it has.
That is a testable hypothesis: Using Pew 2014 relgious data we can isolate at least 3 religion groups about the same size as atheists & throw in agnostics to compare. Then we look at the chance you make an acurate guess based on the combined averages closest to the stereotype[top result]. The five comparison groups I pulled being Atheists, Agnostics, Methodists, Pentecostals & Nondenominational Evangelicals as they are around the same size.
Atheists
Guess: Male, White, Non-immigrant, Non-Parent
~36% chance of being right
Guess: Liberal leaning Democrat
39% chance of being right
Ratio of agreement of 5 contraversial questions:
8.11 : 1
Agnostics
Guess: Male, White, Non-immigrant, Non-Parent
~35% chance of being right
Guess: Liberal leaning Democrat
30% chance of being right
Ratio of agreement of 5 controversial questions:
7.73 : 1
Methodists
Guess: female, White, Non-immigrant, Non-Parent
~38% chance of being right
Guess: Conservative leaning Republican
23%
Ratio of agreement of 5 controversial questions:
1.68 : 1
Pentecostals
Guess: female, White, Non-immigrant, Non-Parent
~38% chance of being right
Guess: Conservative leaning Republican
24% chance of being right
Ratio of agreement of 5 controversial questions:
1.60 : 1
Nondenominational Evangelical
Guess: female, White, Non-immigrant, Non-Parent
~23% chance of being right
Guess: Conservative leaning Republican
34% chance of being right
Ratio of agreement of 5 controversial questions:
1.87 : 1
In other words, "atheists"/"agnostics" at least in America are slightly more homogenous and predicable than similar religious groups. Does that change when the question is something less political like universal wonder, well-being, relative morality & sources of guidance on right and wrong? Not really, you still find a slightly higher to simlar homogeneity. Does that change if we use lower probabaility with more specific generalizations like education, income or age? No.
So, sorry theoretically that may be true but in practice you can pretty accurately guess.
PS: I would note knowing gender or age + religion you could be even more accurate. As all that is based on religion alone.