There are LOTS of opinions and narratives out there about white voters.

     Hillary Clinton lost because white women abandoned her.

     White, non-college educated voters are Trump’s base. They are never coming back/will come back to Democrats.

     Donald Trump’s testosterone-laden presidency alienated lots of white, college-educated women who held their noses and voted for him in 2016.

Most of these narratives are built on data supplied by the 2016 exit polls and the "education level" cross-tabs in current polling. However, new data and analysis of the 2016 vote suggest that many of these assumptions are worth reassessing.

First, let’s look at this political aphorism that Hillary Clinton lost because white women abandoned her. That assumption comes directly from the 2016 exit poll that showed Trump winning white women by 9-points (52 percent to 43 percent).

However, a Pew Research assessment of the 2016 electorate, which was highlighted in an early August report, showed Trump did much worse with white women and white college-educated voters than what the exit polls found.

The exits show Trump winning white voters by nine-points. But, Pew’s validated voter survey (they matched voter file with the survey respondents) showed white women narrowly preferring Trump to Clinton by 2 points - 47 percent to 45 percent. Instead of winning decisively among this group (as the narratives and exits have shown), the Pew data suggests white women have always been, at best, ambivalent about Trump.

The exit polls and the Pew surveys both show a sizable gap in voter preference among white college-educated and white, non-college educated voters. In the exit poll, Trump narrowly carried white college-educated voters by three points (48 to 45 percent). But, the Pew survey found Clinton won these voters by 17 points! The exit polls also showed the electorate to be more heavily populated by white, college-educated graduates than the census and other data experts believe to be true.

Now, compare how white voters voted in 2016 to Trump’s current job approval rating and the congressional ballot question from the NBC/Wall Street Journal January-July merged poll (3,995 voters). Looking at the Pew data, you can see that opinions about Trump and vote preference among white women isn’t all that much different from where it was in 2016. The exit poll data, however, shows a significant drop-off in support for Trump among those same voters.

For example, according to exit polls, Trump carried white, college-educated voters with 48 percent of the vote. Today, his approval rating among these voters is just 37 percent. More ominously, after voting narrowly for Trump in 2016, these voters overwhelmingly prefer a Democrat for Congress over a Republican (54 percent to 39 percent).

But, what if you compared Trump’s current standing with the Pew data. Trump took 38 percent of college-educated voters in 2016, and his current standing with these voters is….37 percent. Their vote preference in 2016 (38 percent Trump to 55 percent Clinton), pretty much mirrors their vote preference for 2018 - 39 percent Republican to 54 percent Democrat.

In other words, if you only use the exit polls as your benchmark, it looks like there’s been a lot of movement among white women and white college-educated voters since the election. If you use only the Pew data, it looks like very little has changed since 2016. I am not a pollster (and I’m also terrible at math), but my gut tells me that we were more deeply polarized in 2016 than we appreciated and that polarization hasn’t widened, it’s hardened.

What you do see in both in both Pew and exits, is a drop-off among white, non-college voters. The exit polls found Trump getting 66 percent among this group, while Pew put it at 64 percent. Trump’s approval rating among these voters now stands at 57 percent. So, perhaps this is the group that has soured more on Trump that we appreciate.

But, Mike Podhorzer, AFL-CIO’s political director, suggests that if we want to have a better understanding of white, non-college educated voters, we need to stop lumping them into one, catch-all category. What really distinguishes a Trump-supporting white voter from one who doesn’t isn’t education or even gender, it's whether or not that voter is evangelical.

Using a data set from Public Religion Research Institute, Podhorzer broke out white voters by gender, education and whether they identified as evangelical. The gap between white voters who approve and disapprove of Trump by gender was 25 points. By education (college versus non-college) it was about the same at 26 percent. But the gap in perceptions of the president between white voters who are evangelical and those who aren’t was a whopping 60 percent!

This evangelical support gap transcends education and gender. For example, among white evangelicals, college-educated men and non-college educated men give Trump equally impressive job approval ratings (78 percent and 80 percent respectively). But, among white men who aren’t evangelical, the education gap is significant. Those without a college degree give Trump a 52 percent job approval rating, while just 40 percent of those with a college degree approve of the job he’s doing.

Meanwhile, among women, if you remove evangelicals, white women with and without a college degree have the same (very low) opinion of the president.

White evangelical women without a college degree give Trump a 68 percent job approval rating, while those with a degree give him a much lower, though still positive 51 percent approval rating. Meanwhile, Trump’s approval among white, non-evangelical women without a college degree is 35 percent, just five points higher than the 30 percent approval rating he gets from white, non-evangelical college-educated women.

Podhorzer’s analysis leads to two conclusions. First, stop assuming that all white, non-college voters are core Trump supporters. Trump’s base is evangelical white voters, regardless of education level. Second, white non-evangelical, non-college women are the ultimate swing voters.

More from the Cook Political Report