[Overview] Trump Voters II – Geography and Culture

[Part I; Part II; Part III; Part IV; Part V; Part VI; Part VII; Part VIII; Single page]

by David Severa





As you can see in these county-level maps from the New York Times, while Trump’s support rarely dips below 20% outside of the Mormon West, there is nevertheless significant regional variation in his support. (Ignore Wyoming, which has an odd caucus system. West Virginia and Nebraska voted after everyone else had dropped out.) Some states, Texas and Ohio, he lost to a favorite son. He also has notable pockets of weakness in the Great Plains and Midwest. He does well in the Deep South, mid-Atlantic, and parts of the West. For the most part, his support goes from 20-something percent to 50-something percent, a more significant spread than for the other factors I’ve looked at so far.

However, this regional map is in some ways distorting, as some states are much more Republican than others and some states much more populous than others. (A map distorted by population density would be useful.) This is a graph of Trump voters relative to voters for non-Trump Republican candidates, and to the rest of a state’s population:

image (1)

(Data from 1, 2, 3; graph created by me; Colorado, Wyoming, North Dakota, states that voted after every other candidate had withdrawn, and all territories are excluded)

The first striking thing is how small many of these primaries are, even in Republican states in a high-turnout year. (And the restrictive and time-consuming caucuses are even tinier.) But Trump’s support is clearly smaller in the Northeast than elsewhere; he’s dominating a smaller party. Many of these primaries come down to a few thousand ballots out of only a few tens of thousands cast.

But what drives this regional variation among Republicans? 

The places where Trump has done well cut across many of the usual fault lines of American politics — North and South, liberal and conservative, rural and suburban. One element common to a significant share of his supporters is that they have largely missed the generation-long transition of the United States away from manufacturing and into a diverse, information-driven economy deeply intertwined with the rest of the world.

They “compared hundreds of demographic and economic variables from census data, along with results from past elections, with this year’s results in the 23 states that have held primaries and caucuses.” What they found:

Capture (Source)

I’ll cover this more in the next section, but what “American” ancestry means is that, when asked to describe their background, some people simply say “American” rather than naming a country in Europe or wherever. In practice, the people who do so are white Americans mainly from Greater Appalachia and the Deep South. Also, I’m pretty sure that the authors are implicitly dividing whites into three groups: those with Catholic ancestry, those who report”American” ancestry (who are mostly actually White Anglo-Saxon Protestants), and others. They call this last group WASPs, but I believe it also includes people with German and Scandinavian heritage. Of these three groups, those with Catholic and “American” ancestry are the ones supporting Trump.

Brian Arbour and Jeremy M. Teigen did their own, similar analysis. (Note that their term for “American” ancestry is “unhyphenated American”.) They found similar correlations for ancestry, less education, and worse economic conditions:


Notably, these two studies differ, at least in interpretation, on the impact of race and immigration. Irwin and Katz:

Despite evidence that some individual Trump voters are driven by racial hostility, this analysis didn’t show a particularly powerful relationship between the racial breakdown of a county and its likelihood of voting for Trump. There are Trump-supporting counties with both very high and very low proportions of African-Americans, for example.

One of the strongest predictors of Trump support is the proportion of the population that is native-born. Relatively few people in the places where Trump is strong are immigrants — and, as their answers on their ancestry reveal, they very much wear Americanness on their sleeve.

Versus Arbour and Teigen:

Interestingly, Trump’s support increases as the percentage of African Americans and immigrants in a county increases. Trump and his campaign have been criticized for bashing immigrants and for being slow to disavow the Ku Klux Klan’s support. We doubt, then, that blacks and immigrants are the ones voting for him. Rather, apparently those unhyphenated white GOP primary voters are pulling the lever for Trump’s anti-immigrant and anti-minority platform and attitudes specifically because they feel threatened by those “others.” Some scholars refer to it as the “racial threat” hypothesis.

Preliminary research by one of us (Arbour) shows that regions with concentrations of unhyphenated Americans also voice more racial resentment than the rest of the country — and even more than in the rest of the South.

Unfortunately, neither article provides enough information to resolve this discrepancy. Perhaps the relationship between local diversity and Trump support is different in some parts of the country in a way that isn’t being captured by these studies, but that’s pure speculation on my part. I’m also not sure why the second study finds that Latino and immigrant populations have opposite effects on Trump support.

The Washington Post has an interesting interactive graphic that lets you see how well the candidates did in each county relative to various measures like race, education, income, and so on, with more states included than in the above two analyses. It’s a bit confusing, but if you can draw a line that fits through the red dots from bottom left to upper right, that indicates a positive correlation with Trump’s vote share. Here’s a screenshot of Trump’s support relative to the black population in a county:


The author says:

Trump’s success in heavily black counties isn’t a function of black support; this is the total population of the county that’s depicted. That strength in black counties is also why he seems to do obviously better in poorer counties — until you limit that to only the white population. Then, the apparent link drops significantly.

So this agrees with the other Washington Post analysis: (white) Republicans vote Trump in counties that are more black. However, it doesn’t seem to be the main factor. “This isn’t like the Democratic contest, where knowing the percentage of the population that’s black can give you a good sense of who will win.”

Trump country is also relatively poor and uneducated. (But again, we’re just talking about relative concentration of support!) You can find profiles of these ultra-Trumpist areas. As the Times says, these regions have been suffering for a long while:

There were only weak correlations between Trump support and various measures of economic performance from 2007 to 2014, including the lingering damage from the 2008 global financial crisis. Rather, the economic problems that line up with strong Trump support have long been in the making, and defy simple fixes….

In places where Trump does well, relatively high proportions of workers are in fields that involve working with one’s hands, especially manufacturing. The decline in manufacturing employment is not a story of merely a rough few years for the economy; nationwide factory employment peaked in 1979, and as a proportion of total jobs has been declining almost continually since 1943. Forces including mechanization and trade have put employment prospects in the sector in an ever-worsening position.

Note that this analysis only shows the social conditions correlated with Trump support, it doesn’t show which voters in these counties are the ones voting for him. For instance, it could be the relatively well-to-do, while the poorest aren’t voting at all. Yes, Trump voters are poorer and less educated than other Republicans, but that doesn’t mean that they’re poor themselves. At any rate, it’s a caveat worth keeping in mind.

And as I quoted above, the relationship between support for Trump and poverty is attenuated when you only look at the white poverty rate, which weakens the case that Trump voters are themselves necessarily disproportionately poor:


At least some of the relationship between Trump support and poverty is actually the relationship between Trump support and race. Further research into how race and the economy tie together at the county level would likely be very informative.

White ethnicity

Understandably, there has been much discussion about Trump and race. But there has been relatively little discussion of Trump and ethnicity. Specifically, whether support varies by ancestry among whites. Let’s return to the gap between unhyphenated Americans/those with “American” ancestry and everyone else.


According to the Washington Post:

The people who identify this way are concentrated in Appalachia and rural areas in and around the South. They are heavily Protestant and less likely than average to have graduated college. They cluster especially along the migration path taken by highland Southerners before the Civil War — those farmers from the hardscrabble Southern hills who never had enough money to buy land in the fertile parts of the Deep South and who moved from the Appalachian highlands across the Upper South and into the near Southwest….

But are whites’ feelings of “American” ancestry why they support Trump, or is it just that Trump does well generally among downscale and rural voters? To find out, we ran a model that tests them together, using election results and county-level census data on American ethnic identity, socioeconomic status, race, religion, immigrant status and age.

Here’s what we found: When we control for those other factors, places with more unhyphenated Americans do indeed vote more strongly for Trump. For every 10 percent increase in a county’s share of unhyphenated voters, we found about 3 percent more support for Trump.

In other words, this region and these voters break for Trump above and beyond what would be predicted by other, especially economic, variables.

Outside of Appalachia, whites with Catholic ancestry also support Trump disproportionately. Discussing the New York primary, Nate Cohn says:

His best district might be New York’s 11th, which includes Staten Island and which has many voters of Italian ancestry. The model estimates he’ll receive 58 percent of the vote there.

It might seem strange to focus on something like Italian ancestry, but the number of voters with ancestry from predominantly Roman Catholic countries in Europe (like Italy, Ireland or Spain) is one of the strongest drivers of Mr. Trump’s support. Conversely, a high share of voters with ancestry from Protestant northwest Europe (like England, Germany or the Netherlands) has predicted lack of support. Unsurprisingly, Mr. Trump’s support in the area includes two prominent Catholic politicians, Rudy Giuliani and Chris Christie.

Of note, both of these groups historically favored Democrats until the realignment that began in the 1960s, while WASPs, who have supported Republicans longer, aren’t voting for the Republican insurgent.

Cohn also discusses this split further west:

Iowa, Utah, Kansas — and Wisconsin — have something else in common: a large population who report their ancestry from predominantly Protestant countries in Northern Europe.

These voters represented the base of the Republican Party for the century after the Civil War, whether it’s the old-stock “Yankees” who spread west from New England, or the German, Scandinavian and Dutch immigrants who generally settled over the same stretch of the northern part of the United States later in the 19th century.

These voters are probably the biggest problem for Mr. Trump that you haven’t heard of: He would fare about 30 points worse in counties where all of the white residents reported their ancestry from Protestant countries in Northern Europe than he would in a place where none did, according to our model. It’s the type of thing that helps separate Northern Virginia — where Mr. Trump struggled greatly — from the Boston area, where he excelled.

He further speculates that cultural “niceness”, deeper ties to the Republican party, and less racism might be the cause. Conservative pundit Michael Barone has wondered if the cause is higher social connectedness:

My first clues came from the Dutch. Heavily Dutch-American counties in northwest and central Iowa and western Michigan, around Grand Rapids, were Huckabee and Santorum territory in past years.

This year, unlike surrounding territory, they voted for Ted Cruz, with Trump a poor third. Dutch-Americans have dense networks of churches and civic groups — unusually high social connectedness….

[Robert] Putnam reports that social connectedness is highest in states with large Scandinavian- and German-American populations and in Utah. It’s lowest in — no surprise — Nevada, one of Trump’s best states.

In the 13 states highest in social connectedness, Trump has gotten just 21 to 35 percent in primaries and caucuses. In the 11 states lowest in social connectedness (except for Cruz’s Texas), his percentages ranged from 33 to 47 percent.

As we’ve seen and will see more fully, much of Trump’s base is in those parts of white America that aren’t doing well in a variety of ways. So if ethnic background is correlated with, or if there are cultural causes underlying these social problems then we should expect to see a correlation with support for Trump. It’s an issue that should be investigated more fully.


Death has been in the news lately. Last year Princeton economists Anne Case and Angus Deaton released findings that, unlike for any other group in America or in any other developed country, “The mortality rate for white men and women ages 45-54 with less than a college education increased markedly between 1999 and 2013, most likely because of problems with legal and illegal drugs, alcohol and suicide”. There has since been some debate whether the mortality rate is actually increasing or just stagnant, but everyone agrees that something is seriously wrong.

The death rate of whites is another predictor of Trump’s support. Even after controlling for factors related to income, employment, education, and urbanization, “the middle-aged white death rate in a county was still a significant predictor of the share of votes that went to Trump.”  (Note that this is measuring higher mortality in general, not the recent possible increase)


Yet again, a negative social indicator predicts Trump support.


There is a great deal of regional variation in support for Trump, but some of his highest support comes from Greater Appalachia and the Northeast. Relatedly, white Americans with Catholic ancestry and self-reported “American” ancestry vote for Trump. The areas that support Trump seem to suffer from various social ills, from higher mortality to low education to declining economies. Race and ethnicity also play a role. Living in areas with a larger black population also seems to be correlated with support.

Next: Trump, religion, and economic disaffection

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s