This overall narrative is correct, but any complex story that’s condensed into a single sentence loses some details. While major cities in general moved toward the Democrats, there was significant variation in how much each individual city moved and which party it moved toward. So I used some statistical tools to try to figure out exactly what makes a city like Pittsburgh move right while Philadelphia moves left.
And I found that, as with a fractal, zooming in on one part of the puzzle revealed the same patterns from the larger picture repeating themselves again and again.
Specifically, I found that both the long-term and short-term trends within the cities mirrored the broader trends taking place across the country.
The Long-Term Trend and How to See It in the Map
The long-term trend in big cities (with some notable exceptions) favors Democrats.
These areas represent the 53 most populous cities in the country (a large or mega city, according to the CBSA Division definitions laid out here). There’s more than one way to define a city, and I used divisions produced by the Office of Management and Budget. The basic idea is that each city is more than a small, dense core of skyscrapers and high-end restaurants. Cities also include the suburbs, exurbs and surrounding areas that, along with the core, make up an economic unit. The OMB lines take this into account and draw large areas that incorporate suburbs and exurbs.
This map shows the change in how each city voted from 1988 to 2016, adjusted for the difference between the national results in those years. The basic idea is that if Hillary Clinton improved on Michael Dukakis’s vote share in a given city by more than she improved on his overall popular vote share, the city will be blue (and red when the comparison is between Trump and Bush 41). The bluer or redder a city is, the more it moved toward Democrats or Republicans, respectively.
Obviously this method of map-drawing, like any method, has its shortcomings. Focusing on the whole metro area can lose some level of granularity, as countervailing trends in different parts of a city may cancel each other out. And correcting for the national popular vote isn’t always the right move. As demonstrated in longer pieces on coalition-building, it can make more sense not to correct for changes in the popular vote. But doing so brings out trends in a helpful way here.
The message of the map is simple: The long-term trends that have shaped national politics for the last 30 years explain the difference between the right-trending and liberalizing cities.
Most cities, especially the largest ones, have moved left over the course of the last few decades. Three quarters of the cities shown on the map became more Democratic after correcting for the difference in the popular vote, often sharply so. This is especially true in big metro areas like Los Angeles, New York, Chicago, Washington, D.C., San Francisco, Philadelphia and Miami. This makes sense. Democrats have moved to the left culturally over this time period (think of the difference in tone and substance between Bill Clinton and Hillary Clinton), and many cities have seen an influx of Democratic-leaning racial minorities. Both of those trends help Democrats in the nation’s largest cities.
The exceptions to this pattern, mostly in the heartland, encapsulate some of the countervailing national-level trends. As Democrats become more culturally cosmopolitan and diverse, they’ve lost some white voters (especially downscale whites). That’s probably part of why cities with a lower average income and a larger share of non-Hispanic white voters were more likely to trend Republican in the long term.
Some of this movement is regional. Republicans have made gains in a number of Southern and semi-Southern states (i.e. Oklahoma, Missouri) for a combination of reasons (increasing racial polarization, suburban growth, etc.), and cities such as Nashville, Birmingham, Louisville, St. Louis and Oklahoma City weren’t immune to those trends. It’s also not a coincidence that the only major Appalachian city, Pittsburgh, has moved right as Democrats lost that region. And some of the cities that buck those trends (e.g. Atlanta, Raleigh-Durham, Washington, D.C.) feature a large number of residents who were born out of state — many of whom originally come from the Northeast or other areas that are less conservative than the South.
In other words, the same trends that shaped national politics for the past 30 years have been acting within cities, pushing and pulling them in different directions with varying levels of strength.
The Short-Term Trend Varies in Different Ways
The short-term trends initially look similar, but actually differ significantly.
This map shows how much each city moved between 2012 and 2016 (again, colors are corrected for the difference between the popular vote results in those years) with the size of each circle representing the city’s population.
The conventional wisdom about Trump’s win — that the most culturally cosmopolitan areas gave more votes to Clinton while the heartland moved toward Trump — holds up reasonably well. Trump made gains, relative to the change in the popular vote, in Cleveland, Detroit, Buffalo, St. Louis, Pittsburgh and other smaller cities in the middle of the country. Clinton made significant gains in Boston, D.C., Los Angeles, Seattle, San Francisco, Miami, Atlanta, and big-city Texas. Most of these areas are commonly seen as centers of economic growth, political power or cultural production.
This narrative doesn’t work perfectly. The New York City area as a whole didn’t move much more than the popular vote, though some subparts of it definitively shifted left. And some heartland cities (e.g. Minneapolis, Oklahoma City, Kansas City, Nashville) moved significantly toward the Democrats. Other cities also moved partially due to idiosyncratic local concerns. Salt Lake City is very liberal, but part of its leftward trend is likely due to typically Republican Mormon voters choosing conservative third-party candidate Evan McMullin over Trump.
Some familiar demographic storylines also pop up when we run regressions on the data. I used multiple demographic variables to predict which cities would move right or left, and three variables — education level among whites, non-Hispanic black percentage of the population and median age — stood out as having significant predictive power. Together, they explained about two-thirds of the data on short-term political trends in these cities.
Lower rates of college education among whites led to improvements of Trump’s vote share compared to Mitt Romney’s, and higher rates led to Clinton improving on Obama’s vote share. Oddly enough, higher concentrations of black voters led to improvements for Trump over Romney. And a higher median age tended to mitigate improvements for Clinton or help Trump increase his vote share.
These patterns roughly mirrored national trends. Trump won over a large number of normally Democratic non-college-educated whites while losing some college-educated white suburbanites. That made education among whites a powerful predictor of his gains over Romney in many parts of the country. Diminished black turnout hurt Hillary Clinton nationwide, and it (along with strong racial polarization in some areas) may have led to her losing ground compared to Barack Obama in cities with a significant number of black residents. Age is slightly more difficult to interpret since different age groups trended in different directions and vary in their base-line political leanings. But these findings cohere with the overall idea that millennials disliked Trump and that he did well with older voters.
In that way, urban America isn’t so different from the rest of the country — the broad trends that moved voters everywhere also pushed and pulled cities in 2016.