Pablo Martinez Monsivais, AP
President Donald Trump gestures while speaking about the U.S. role in the Paris climate change accord, Thursday, June 1, 2017, in the Rose Garden of the White House in Washington. (AP Photo/Pablo Martinez Monsivais)

Hillary Clinton this week threw fresh fuel on the fire of post-election speculation about why Trump won the 2016 election, pointing to an array of causes ranging from Russian interference and former FBI director Jim Comey's statements about her email server to Democratic National Committee fecklessness and the editorial decisions of the New York Times.

Clinton's litany of reasons struck some as sour grapes. But Washington continues to buzz with fresh speculation of Russian contact with Trump allies, and Comey is set to testify before the Senate Intelligence Committee on June 8 about his own role in that investigation.

With all the turmoil, its easy to forget that a distinctive group of real people in battleground states actually voted for Trump. And in the months since the election, we've gained a complex and often counterintuitive picture of who they are and what makes them tick.

Here are seven charts that highlight intriguing and sometimes surprising data about who actually voted for Donald Trump.

1. Trump voters have poor health.

Trump voters may pay their bills, but they certainly aren't taking care of their bodies, according to this graphic by The Economist. The Economist was responding to a challenge by pollster Patrick Ruffini. “Find the variable that can beat % of non-college whites in the electorate as a predictor of county swing to Trump," Ruffinni said. So they did.

This graph shows the results. It’s a lot of colored dots, coded by region of the country. But the upshot is that Trump counties (top of graph) had worse health (left side of graph) and were clustered in the south and west (red and yellow).

Atlas of US Presidential Elections; Census Bureau; IPUMS; Institute for Health Metrics and Evaluation; The Economist | Aaron Thorup, Deseret News

There is a tight connection between areas with “non-college educated whites” and poor health measurements, the authors acknowledge. But holding all other variables constant, including race and education, the health measures do perform slightly better in predicting the Trump vote.

According to The Economist, this cluster of four health indicators does a better job of explaining who voted for Trump than the more widely cited category of "non-college whites." winning by a 43 to 41 percent edge. Technically, the authors admit, they didn't find one single variable that met Raffini's challenge, but rather a tight cluster of four. But then, Raffini's challenge itself was two variables (race and education) not one, so who's counting?

2. Trump voters pay their bills.

In December, the Urban Institute checked on the average credit scores in the counties won by Clinton and Trump. The results were striking. Of 55 counties nationwide that had exceptional average credit scores (720 or higher), 51 tipped to Trump. Of the 11 counties that averaged below 600, or subprime, all 11 went to Clinton.

Counties with good credit scores went with Trump over Clinton. | Urban Institute

The result was so counterintuitive, says Diana Elliott, the study's lead author, that many journalists had no idea what to do with it. One prominent journalist, Elliot said, refused to cover it because it did not fit into the dominant narrative.

The Urban Institute also put together a nifty interactive map of U.S. counties allowing the user to hover and filter to see which areas supported Trump and how secure their finances were.

A lot of the talk in the immediate aftermath of Trump's victory focused the economic struggles of the Trump voters, particularly in rural counties and in coal country and the rustbelt regions of the country. This led some to conclude that economic struggles defined the Trump voter. But as a large survey by the research firm PRRI shows, economic fears and skepticism are not the same as economic struggles.

To be strictly accurate, though, the authors of the Urban Institute report do acknowledge a trick here. Credit scores overwhelmingly correlate to race, they note, and was an very racially polarized election. In fact, credit scores had no predictive value at all after controlling for "median income, unemployment rate, percentage white residents, percentage with a bachelor’s degree or higher, and homeownership rate."

3. Trump voters are economic fatalists.

White working class Trump supporters were nearly twice as likely as white working class Clinton voters to question the value of a college degree, a recent poll by the research firm PRRI finds, arguably the most intriguing result in the survey of over 3,000 Americans.

White working -class Trump voters view college education as a gamble. | Heather Tuttle, Deseret News

Voters were asked if going to college was a "risky gamble" or a "smart investment" to measure pessimism or optimism about economic mobility. What the survey was trying to indirectly get at, its authors noted, was whether the respondents see themselves as having options? Or do they feel stuck? On the basis of this question, PRRI defined those who see college as a risky gamble as "economic fatalists."

Skepticism about college lagged just behind fears about immigration and cultural displacement in predicting that a white working class voter — defined as a voter without a four-year college degree who works in a non-salaried job — would vote for Trump over Clinton, PRRI found. White working class voters who said college was a "gamble" were almost twice as likely to vote for Trump over Clinton.

4. Poor counties went to Clinton

Mark Lippman at the hard-left website The Daily Kos took umbrage at post-election buzz about economic struggles undergirding the Trump upset. To attack this narrative, Lippman ran the numbers on the poorest 130 counties, correlating poverty levels and the Trump and Clinton vote. The results were not even close. The poorer the county, the stronger the Clinton support. Clinton easily won all counties with over 25 percent poverty levels.

County-level vote totals reported by board of elections officials population. Below poverty level from U.S. Census Bureau 2015 estimates. | County-level vote totals reported by board of elections officials population. Below poverty level from U.S. Census Bureau 2015 estimates.

The results here are not all that surprising, given what else we know about Trump voters, including the Urban Institute graph above. We know that Trump voters and Trump counties are not the poorest of the poor. As the Deseret News outlined prior to the election, Trump counties were the remnants of a once comfortable blue collar middle class, marked more by growing insecurity and fear of the new economy than by deep, intergenerational poverty.

5. The GOP (and Trump) made huge gains among rural voters.

This NPR analysis shows a significant shift in the rural vote over the past three election cycles. There was little movement in the urban or suburban vote, in either direction. But the GOP gained nine full points in the rural vote. Significantly, this effect does not begin with Trump. Already in 2012, NPR notes, Romney made substantial rural gains over McCain's 2008 performance.

Rural vote shifts to GOPJoseph Tolman, Deseret News

6. Religious affiliation showed no effect.

This graph from the PRRI survey shows a surprising nonresult, a non-barking dog. Attitudes toward religion are shifting rapidly in the rising generation, among working class whites as elsewhere. As this graph shows, nearly half of young working class whites expressed no religious affiliation, compared to 10 percent from their grandparents' generation. But somewhat surprisingly, that shift tells us nothing about the Trump vote. That is, religious affiliation or its absence did not make a working class white voter any more or less likely to support Trump over Clinton.

The young white working class turning away from religion

| Joseph Tolman

7. Trump voters have low education levels.

"I love the poorly educated," Trump famously said after winning the Nevada Caucuses in February, 2016. And the feeling is mutual. No real surprise here, but the data are still intriguing, and it does confirm that education levels mattered more than income levels.

This data, collected by FiveThirtyEight, sifted counties by the percentage of residents with a four-year college degree. Clinton gained over 8 percentage points on Obama's 2012 performance in the 50 best educated counties, but lost over 11 points in the 50 least educated counties.

Clinton surged in educated counties, collapsed in less-educated counties | Joseph Tolman