Carolyn Kaster/AP

The 2016 election reinforced America’s deepest, long-held divides.

The post-mortem on the 2016 presidential election is still in process, but there are already some clear narratives: Clinton lost because of lower Democratic voter turnout, especially among minorities; Trump mobilized the white working class and was able to breach the ‘blue wall’ in the Midwest; 2016 realigned the electorate along regional lines. Plus, nearly all the polls and data analytics failed to predict the outcome. All of those conclusions are true to a certain extent. But the bigger reality according to my own analysis is that the 2016 election follows the same basic contours of class and location—the same divides of knowledge and density—of the last several election cycles.

Writing here in October, I used polling data to show that “despite Trump” who I called one of “the strangest” candidates in American presidential history, the 2016 election “conforms to America’s underlying basic economic, demographic, and political divides.”

My analysis of the actual state-by-state election results confirms this. Rather than being a significant break with the past, this election reinforces the nation’s existing divides between richer, more highly educated, more knowledge-based states and less advantaged, less diverse, and whiter ones. More than anything else, I would argue, the 2016 election hardened these long-standing divides.

(My colleague Charlotta Mellander conducted the statistical analysis and Taylor Blake of the Martin Prosperity Institute helped to organize and map the data).

Despite the very different outcome—a Republican victory compared to two previous Democratic wins—the first thing that jumps out is how closely aligned the basic state voting patterns are over the past three presidential elections: Obama vs. McCain in 2008, Obama vs. Romney in 2012, and Clinton vs Trump in 2016.

(Taylor Blake)

Across the 50 states and the District of Columbia, Clinton votes were extremely closely correlated with Obama’s in 2012 (.97) and with Obama’s in 2008 (.92). Trump’s were similarly closely correlated with Romney’s votes in 2012 (.91) and McCain’s in 2008 (.87). Across the states, the final vote tallies also line up reasonably well with the polls, with correlations of .95 for Clinton and .97 for Trump against Pollster’s mid-to-late October polls.


Class was and is the central feature of America’s political divide. Clinton support was concentrated in states with higher wages (.82), greater shares of college grads (.77), and greater share of the workforce in knowledge, professional and creative jobs (.72). Trump support was concentrated in states with lower wages (-.81), smaller shares of college grads (-.81), and smaller shares of the workforce in knowledge, professional, and creative jobs (-.73.) These correlations are all greater than for the past two election cycles, as the table below shows.

Clinton 2016 Trump 2016 Obama 2012 Romney 2012 Obama 2008 McCain 2008
Annual Wage 0.82 -0.81 0.76 -0.76 0.61 -0.61
College Grads 0.77 -0.81 0.72 -0.73 0.62 -0.63
Creative Class 0.72 -0.73 0.66 -0.66 0.46 -0.46
Working Class -0.77 0.79 -0.75 0.75 -0.63 0.64
Service Class 0.32 -0.33 0.38 -0.39 0.40 -0.4

Trump support was highly concentrated in states with greater concentrations of the working class (.79), similar to Romney in 2012 but up somewhat from McCain in 2008. Clinton support was highly negatively correlated with a larger blue-collar share of the workforce (-.77). Again, this was roughly on par with Obama in 2012, but greater than Obama in 2008.

An even more interesting set of results turn up for unionization. While the working class states went for Trump and against Clinton, the effect of unionization was the opposite. States with more union members favored Clinton over Trump. Clinton support was positively associated with the level of unionization across states (.46), whole Trump support was negatively associated with it (-.44). These correlations are slightly weaker than the past two election cycles.

Most analyses of America’s class divide juxtapose these two classes, the new knowledge class and the older working class. But few look at the largest class, the service class, which is made up of nearly 70 million American workers, or 45 percent of the workforce who earn low wages in retail shops, office and clerical work, and food service. Clinton support was stronger in states where the service class is larger (.32), while Trump’s support was weaker (-.33).

While Democrats may be considering strategies to win back the white working class, they might do well to remember that the 70 million member, multi-ethnic service class is more than double the size, and makes a fraction of the income.

Density and Urbanization

Density and urbanization remain key factors in America’s political divide.

Clinton support was positively associated with both density (.71) and the share of the state that is urbanized (.63), while Trump support was negatively associated with both (-.61 with density and -.54 with the urbanized share). These correlations are roughly similar to 2012 but somewhat greater than in 2008.

Clinton 2016 Trump 2016 Obama 2012 Romney 2012 Obama 2008 McCain 2008
Density 0.71 -0.61 0.67 -0.64 0.57 -0.53
Urbanization 0.63 -0.54 0.60 -0.56 0.52 -0.47
Drive Alone to Work -0.46 0.53 -0.41 0.43 -0.39 0.41

On the flip side, Trump support was positively associated with states where a larger share of people drive to work alone, a proxy indicator for sprawl (.53), while Clinton support was negatively associated with it (-.46).

Housing is another feature of America’s political divide. Trump support was positively associated with states where the share of residents who own their own homes is higher (.61), while Clinton support was negatively associated with it (-.63). These correlations are up substantially from 2012 and even more so from 2008. Housing prices appear to have played an even bigger role. Clinton support was higher in states with more expensive housing (with a correlation of .75 to median housing values), while Trump’s was negative (-.80).

Clinton 2016 Trump 2016 Obama 2012 Romney 2012 Obama 2008 McCain 2008
Home-Ownership Share -0.63 0.61 -0.55 0.56 -0.30 0.30
Median Housing Value 0.75 -0.80 0.71 -0.72 0.60 -0.61


Much has been made of the role of race in the election. But at the state level, race played a somewhat less significant role in determining whether a state went Democratic or Republican.

White states went against Clinton, with a negative correlation of -.36 to the white share of the population. The white share of the population was positive but statistically insignificant for Trump. This is similar to 2012 but greater than in 2008. Given the rhetoric of the campaign, it is less surprising that more Hispanic states went for Clinton (.31) and even more so against Trump (-.36). This is significantly greater than in 2012 or 2008.

Clinton 2016 Trump 2016 Obama 2012 Romney 2012 Obama 2008 McCain 2008
White Share of Population -0.36 0.28 -0.30 0.28 -0.22 0.19
Hispanic Share of Population 0.31 -0.36 0.17 -0.19 0.18 -0.17
Black Share of Population 0.13 0.02 0.05 -0.02 -0.03 0.07

The share of immigrants or foreign-born people played a bigger role.  The share of foreign-born people in a state was positively associated with Clinton votes (.67) and negatively associated with Trump votes (-.67).

A surprising finding I pointed to in October is the result for the share of the population that is black. While black voters voted overwhelmingly for Clinton and remain a core element of the Democratic coalition, the correlation for the black share of the population is insignificant for both Clinton and Trump at the state level.

The Culture Wars

Americans’ differences on hot-button cultural and social issues continue to matter as well, in ways that reflect the underlying divides of class and geography.

Religion remains a serious fault-line in America, but its salience appears to have diminished slightly in 2016. The level of religiosity in a state—measured via Gallup surveys of the share of the state population who say they are very religious, was positively correlated with Trump support (.54) and negatively associated with Clinton (-.55). These correlations are down from both 2012 and 2008.

Abortion is another key line of division. Clinton support was positively correlated with access to abortion services, measured as the number of abortion providers normalized for population (.66), while Trump support was negatively correlated with it (-.69). These correlations are similar to the past two election cycles.

Gays and gay rights are yet another perhaps even more powerful axis of division. Clinton support was positively associated with the share of LGBT people in a state (.75), while Trump support was similarly negatively correlated with it (-.75). These correlations are substantially greater than for 2012 or 2008.

That said, on the policy front, taxes remain a bread-and-butter issue separating Democrats from Republicans. Clinton support is 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 (.44), while Trump support is negatively correlated (-.42) with such state-by-state tax burdens.

Clinton 2016 Trump 2016 Obama 2012 Romney 2012 Obama 2008 McCain 2008
Religiosity -0.55 0.54 -0.61 0.63 -0.65 0.67
Abortion Providers Per Capita 0.66 -0.70 0.66 -0.68 0.65 -0.66
Gay Index 0.75 -0.75 0.62 -0.63 0.57 -0.57

For many of America’s urban dwellers, the Trump victory is a huge, depressing blow. But the evidence suggests it is not as strange of an event as it seems. America is a deeply divided nation along the lines of class and geography. Those divides are not going anywhere.

The much bigger question is: Can we—or how can we—survive as such a divided nation? I for one find it hard to fathom experiencing another similar election. What happens if we continue to swing back and forth between red and blue? Is it time to think about some sort of mutual co-existence strategy for these two very different Americas? If so, that could very well mean devolving power from the federal to the state and local levels, in ways that more explicitly recognize and respect the great differences between these two nations.

About the Author

Most Popular

  1. Illustration: two roommates share a couch with a Covid-19 virus.

    For Roommates Under Coronavirus Lockdown, There Are a Lot of New Rules

    Renters in apartments and houses share more than just germs with their roommates: Life under coronavirus lockdown means negotiating new social rules.

  2. Equity

    The Problem With a Coronavirus Rent Strike

    Because of coronavirus, millions of tenants won’t be able to write rent checks. But calls for a rent holiday often ignore the longer-term economic effects.

  3. photo: a For Rent sign in a window in San Francisco.

    Do Landlords Deserve a Coronavirus Bailout, Too?

    Some renters and homeowners are getting financial assistance during the economic disruption from the coronavirus pandemic. What about landlords?

  4. Equity

    We'll Need To Reopen Our Cities. But Not Without Making Changes First.

    We must prepare for a protracted battle with coronavirus. But there are changes we can make now to prepare locked-down cities for what’s next.

  5. photo: South Korean soldiers attempt to disinfect the sidewalks of Seoul's Gagnam district in response to the spread of COVID-19.

    Pandemics Are Also an Urban Planning Problem

    Will COVID-19 change how cities are designed? Michele Acuto of the Connected Cities Lab talks about density, urbanization and pandemic preparation.