The owl of Minerva

The American National Election Studies 2016 data has been released.  There is a lot which can and will be done with these data, but I'll start with some basic things.
     1.  The ANES asked people how they voted in 2012, and I originally intended to write about the relationship between 2012 and 2016 vote.  However, about half the the non-black, non-Hispanic people* who said that they voted in 2012 reported voting for Obama.  That's too high--the actual figure would be a little over 40%--suggesting that a significant fraction of the people who recalled voting for Obama actually voted for Romney.  So I'll leave that aside.
   2.  A lot has already been said about the relationship between education and vote, but here is the table:

          Didn't Vote  Clinton  Trump   Johnson  Stein Other
HS or less     37%        21%      38%     1.6%  1.3%  1.3%
Some college   22%        24%      47%     3.8%  1.0%  1.1%
College grad   11%        46%      36%     4.3%  0.7%  2.0%

Although most accounts focused on Trump's support among people with a high school diploma or less (which journalists like to call the "working class"), he did even better among people with middle levels of education.  The major contrast is between college graduates and everyone else.  Most people with no post-secondary education don't vote at all (the figures for non-voting in the ANES are too low, partly because some people say that they voted when they actually didn't, and partly because the sort of people who don't participate in surveys are less likely to vote).
  3.  Income is more complicated.  On the average, Trump voters had lower incomes than Clinton voters.  After controlling for education, there was no clear difference, but after adding a control for being married, it was back again.  (Married people were substantially more likely to vote for Trump and have higher family incomes).  However, the effect of income seems to differ by education.  Here is a figure showing estimated support for Clinton by education, income, and marital status.  (Income is measured by 28 categories, going from under $5,000 to over $250,000--9 is about $25,000; 15 is about $50,000, and 23 is about $100,000).

I was surprised by how much difference marital status made--I knew it had become a fairly important factor but didn't think it would be that big.  Income has more effect among college graduates.  Another way to put it is that education makes only a little difference among people with low incomes, but a lot of difference among people with high incomes.
4.  The ANES also includes a variable for occupation, but this is a preliminary release, and doesn't include the occupation variable yet.  The occupation variable might help to illuminate the educational differences among people with high incomes.

*all of the analyses are limited to people who are neither black nor Hispanic