John Minchillo/AP

The 2016 presidential campaign is unique in many ways, but it still reinforces the basic divides of American society.

There can be little doubt of the fact that reality television star Donald Trump has made the 2016 presidential campaign season one of the strangest in American history. Yet in many ways, this election conforms to America’s underlying basic economic, demographic, and political divides. In fact, the 2016 election reinforces the nation’s divides between richer, more highly educated, more diverse, and more urban Blue States and less advantaged, working class, less diverse, and whiter Red States.

That’s the big takeaway of an analysis I undertook of the key economic, political, and cultural factors that are associated with support for Trump versus Clinton across America’s states. To get at this, my colleague Charlotta Mellander conducted a basic correlation analysis of Trump and Clinton support based on mid-October state-by-state polls. She ran this analysis on three sets of polls: Pollster, Real Clear Politics, and YouGov.

(Taylor Blake)

Here I report the results for the correlations based on Pollster, the aggregated polling source which has appeared to provide the most systematic coverage. The three polls track consistently with one another, with the Pollster figures being correlated at .97 with YouGov and .95 with Real Clear Politics for Clinton and at correlations of .94 and .91 for Trump. (There are a few factors for which there is discrepancy among the correlations across the different polls. This is usually when a variable is statistically insignificant using one poll but significant in another. In those cases, I report the correlations for both.)

As usual, I point out that correlation does not imply causation, but simply points to associations between variables. Still, this analysis helps illuminate not just the patterns of the current outlier of an election, but on the fundamental divides of the American electorate and society writ large.

The Continuing Significance of Class

Clinton Trump
Income 0.57 -0.60
Hourly Wages 0.67 -0.68
Share of Population in Poverty -0.21 0.39

Class continues to be a basic dividing line in this election. Clinton states are richer (being positively correlated with both average incomes, at .57, and hourly wages, at .67) while Trump states are poorer (being negatively correlated with both income, at -.60 and with hourly wages, at -.68).

Trump also draws more support in states with greater concentrations of poverty, measured as the share of families living below the poverty line (the correlation there is .39). Clinton support has no significant correlation to the share of people living in poverty.

Writing in The New York Times, Nate Cohn points out that this election has seen education become a marker of class, replacing cultural issues and the culture war as the central fault-line in American politics. Our analysis similarly finds education to be a basic dividing line in the 2016 election. Clinton support is higher in more educated states, being positively associated with the share of adults that are college graduates (.60), while Trump support is negatively associated (-.70) with it. The same trend comes through when we look at the number of universities per ten million people in each state: it too is positively associated with state-level Clinton support (.53) and negatively associated with Trump support (-.59).

Clinton Trump
College graduates 0.60 -0.70
Universities 0.53 -0.59
High Tech 0.56 -0.67
Creative Class 0.53 -0.63
Working Class -0.60 0.67

Clinton support is also markedly higher in states with greater concentrations of science, technology, and innovation, being positively associated with the Milken’s Institute’s newly released State Science and Technology Index (.56), while Trump support is even more negatively associated with it (-.67).

The kinds of work people do—whether they are members of the knowledge-based creative class or blue-collar working class—also continues to play a key role in America’s political divide. My Atlantic colleague Ron Brownstein has pointed to this class inversion in American politics, as the Democrats become the party of the more affluent and more educated, while the Republicans become the party of the less advantaged working class—a pattern which has accelerated this election season.

Back in 2012, Brownstein and I found that Obama drew support from counties with large shares of creative class knowledge workers, winning two-thirds of the 2008 vote in the top 100 counties with the highest percentages of creative class workers, while McCain drew his support from working class counties. The same basic pattern remains in 2016.

If Clinton support comes from creative class states, Trump’s support comes from working class states. Clinton support at the state-level is positively correlated with the creative class (.53), while state-level support for Trump is negatively associated with the share of workers in creative class jobs (-.63). Conversely, Trump support is positively associated with the share of workers in blue-collar working class jobs (.67), while Clinton support is negatively associated with the working class (.60).

Interestingly, the results for the level of unionization—measured as the share of workers in a state who are unionized—are the opposite.

Clinton Trump
Unionization 0.49 -0.47
Life expectancy 0.35 -0.56

Gini (Pollster)
Gini (YouGov)



States with more union members favor Clinton over Trump. Clinton support at the state level is positively associated with the level of unionization (.49) while Trump support is negatively correlated (-.47) with the share of workers in a state that are unionized. This may reflect the fact that only a relatively small share of blue-collar workers are unionized and that a large share of union members are more highly paid and highly educated public sector workers.

A more detailed (and as-yet-unpublished) analysis of the role of unions and union density in the 2016 election by my former MPI colleague and current UCLA research fellow Patrick Adler lends support to this view. He finds that the level of public sector unionization overall is associated with greater Clinton support, but that Trump support is more closely correlated with higher levels of unionization in manufacturing and construction jobs. Nonetheless, it shows that Democrats continue to draw considerable support from states where more workers are unionized, even in light of the ongoing class inversion of the electorate.

A detailed analysis by Jonathan Rothwell of the Gallup Organization finds that while Trump draws considerable support from less educated working-class voters, his support is more related to less directly economic aspects of class status, such as health and the ability to move up the economic ladder, as well as racial isolation. Our analysis also finds the most basic measure of health, life expectancy, to be a basic dividing line in American politics. Clinton support is higher in states where life expectancy is longer (.35), while life expectancy is negatively correlated with Trump support (-.56).

Inequality is often said to be an issue that is driving politics and class divisions in the United States. But it seems to be playing a more limited role in the 2016 election. Income inequality, based on the standard measure of the Gini coefficient, is not associated with Trump support and positively associated with Clinton support (.39). That said, income inequality is more highly associated with Clinton in the YouGov poll (-.50) and positively and negatively significantly associated with Trump in it as well (-.39).

Urbanization and density are key issues in defining which states are blue and which are red. According to an analysis by Dave Troy of the 2012 election, places turn from red to blue as density approaches 800 people per square mile. Clinton continues to draw support from more urbanized states.

Clinton Trump
Urbanization 0.58 -0.40
Housing Prices 0.57 -0.64
State Income Taxes 0.48 -0.42

In our correlations, Clinton support is positively associated with the share of a state that is urbanized (.58), while Trump support is negatively correlated (-.40) with more urbanized states.

Housing prices are another feature of America’s political divide. According to an analysis by Jed Kolko, housing costs are almost twice as much in deep-blue markets ($227 per square foot) than in red markets ($119). Clinton support is higher in states with higher median housing values (being positively correlated at .57, while Trump support is negatively associate with it at -.64).

Taxes remain a bread-and-butter issue separating Democrats from Republicans. Clinton support is not surprisingly higher in states with higher state income taxes, while Trump support comes from states with lower tax rates. Clinton support at the state level is positively associated with state income taxes per capita (.48), while Trump support is negatively correlated (-.42) with such state-by-state tax burdens. This likely reflects the fact that higher-tax states are richer and more educated than lower-tax ones. But social policy related to issues such as gays, guns, and abortion are bigger issues than taxes.

The Continuing Role of the Culture Wars

While class is a key axis of division in American politics, the culture wars continue to matter as well.

Clinton Trump
Religiosity -0.50 0.58
Abortion Providers 0.54 -0.68
Teen Birth Rates -0.49 0.61
Gun Deaths -0.47 0.56
LGBT Community 0.61 -0.69

Religion remains a key dividing line in American politics. The level of religiosity in a state—measured via Gallup surveys of the share of the state population who say they are very religious, is positively correlated with Trump support (.58) and negatively associated with Clinton (-.50).

Abortion remains another key fault-line issue in American politics, as we heard in the final presidential debate. Trump support at the state level is substantially negatively correlated with the number of abortion providers per capita (-.68), while Clinton support is positively correlated with them (.54). On the flip side, states with higher teen birth rates are more likely to support Trump (.61) than Clinton (-.49).

Guns are another dividing line. Trump support is higher in states with more gun deaths per capita (with a correlation of .56), while Clinton support is lower (with a negative correlation of -.47).

Gay rights are yet another key fault-line issue in American politics. Clinton support is positively associated with the share of LGBT people in a state (.61), while Trump support is even more highly negatively correlated with it (-.69).

Clinton Trump
Foreign Born* .57 -0.55

White (Pollster)
White (YouGov)



Black 0.24 0.06
Hispanic 0.23 -0.29

The results of our analysis for race are surprising. We hear much about the racial divide in this election, and especially how Trump draws support from white voters. And we do find that states where whites make up a larger share of the population are less likely to support Clinton (with a negative correlation of -.30). But we do not necessarily find that states with larger shares of the white population are more likely to support Trump (the correlation there is statistically insignificant). However, the share of the white population is positively and significantly associated with Trump support (.30) when we use the YouGov poll numbers.

Here it appears the key factor is not the overall share of the white population but their degree of racial isolation, as Rothwell has identified.
Immigration has been one of the biggest issues in the Republican primaries and the current general election campaign. Given his anti-immigrant position, it’s not surprising that Trump support is negatively correlated with the share of foreign-born people in a state (-.55), while Clinton support is positively correlated with the share of immigrants (.57).

Similarly, Trump’s positions on immigration have alienated large numbers of Hispanic voters. That said, out analysis finds that the share of the Hispanic population is having little effect on state-level support for either candidate; our correlations are statistically insignificant for both.

African-Americans have long been a key member of the Democratic coalition. Here again, we find that the share of the African-American population has little effect on state level support for either Clinton or Trump, being statistically insignificant for each of them. These results for African-Americans and Hispanics hold across all three polls.

Clinton Trump
Obama 0.95 -0.86
Romney -0.94 0.87
Obama 0.84 -0.81
McCain -0.82 0.82

Despite Trump’s bizarreness as a candidate, the election is not a break with a past. According to our analysis, Clinton and Trump support this election cycle basically lines up with Obama and Romney support in 2012. Clinton states are correlated at a whopping .95 with Obama’s vote shares by states in 2012; and Trump states are correlated at .87 with the shares of the vote won by Romney.

Ultimately, our analysis suggests that even with the strange, outlier candidacy of Donald Trump in the 2016 presidential election, the United States remains systematically and fundamentally divided along the lines of class, as well as the cultural issues that class informs and shapes. Writing over at The Atlantic, Ron Brownstein and Leah Askarinam show that the bluer map of the 2016 election is a product of a relatively small number of denser, more highly educated counties in swing states that are helping turn red states to blue. But make no mistake about, even if Clinton wins by a large margin, America remains as politically and economically divided as ever.

*CORRECTION: An earlier version of this story had switched the correlation coefficients for Clinton and Trump support in relation to a state's share of foreign-born people.

About the Author

Most Popular

  1. A hawk perches on a tree in the ramble area of Central Park in New York.

    The Toxic Intersection of Racism and Public Space

    For black men like Christian Cooper, the threat of a call to police casts a cloud of fear over parks and public spaces that others associate with safety.

  2. A photo of a police officer in El Paso, Texas.

    What New Research Says About Race and Police Shootings

    Two new studies have revived the long-running debate over how police respond to white criminal suspects versus African Americans.

  3. A woman stares out at crowds from behind a screen, reflecting on a post-pandemic world where exposure with others feels scary.

    What Our Post-Pandemic Behavior Might Look Like

    After each epidemic and disaster, our social norms and behaviors change. As researchers begin to study coronavirus’s impacts, history offers clues.

  4. Maps

    Your Maps of Life Under Lockdown

    Stressful commutes, unexpected routines, and emergent wildlife appear in your homemade maps of life during the coronavirus pandemic.

  5. Maps

    The Three Personalities of America, Mapped

    People in different regions of the U.S. have measurably different psychological profiles.