How Trump Won -- Conclusions
As a few observant readers have noted, one thing that has been missing entirely from this series is race. This is a bit odd, given that virtually all election analysis this cycle has focused on that aspect of Trump’s win.
There are three reasons for this, the last of which leads to our conclusion.
First, it’s been done. A lot. In fact, the lead writer here has probably written well over 100,000 words on demographic analysis. It seemed useful to offer a different perspective.
Second, the impact of demographics is often overstated. For example, Democrats lost Colorado by 100,000 votes in 2004, but won it by 130,000 in 2016. Based upon the exit polls and statewide vote totals, Democrats picked up about 110,000 votes from Hispanics – enough to flip the state – but Republicans picked up 50,000 of their own. What really flipped the state was that Democrats picked up 284,000 votes from whites, while Republicans picked up just 53,000. Hispanics are still just over 10 percent of the electorate in Colorado, and Republicans still win a significant number of their votes.
Also in this series: Part 1: The South
Finally, we come to our overall conclusion for the series: These demographic changes are meaningless without a geographic filter. The reason is simple. Our nation’s politics, at every level, have a geographic basis. Presidents are elected via the Electoral College, not the popular vote. Senators are elected by state. Representatives are elected through geographic districts. Statehouses choose their members in the same manner. There are good normative arguments for and against this, but we place these to the side and simply take the facts as they are, not as they perhaps ought to be.
As the lead author noted two years ago in his chapter for Larry Sabato’s “The Surge,” Hispanics exceed their national share of the population in just nine states, only three of which are swing states (Colorado, Florida and Nevada). Likewise, African-Americans exceed their national share of the population (14 percent) in just 16 states, only a handful of which (Florida, Michigan, North Carolina and Virginia) are swing states.
In other words, the Democrats’ coalition of the ascendant is very inefficiently distributed. We therefore opted to utilize a demographic (urbanicity) that is easily filtered through a geographic component (CBSAs) and that people intuitively think of in geographic terms. What we discovered is a different dimension to the Democrats’ demographic inefficiency problem: They are becoming far too clustered in urban centers to be effective, even when they win the popular vote.
Today, we look at this from a national perspective. We begin with a map:
The large metropolitan areas stick out, and there are some interesting oddities: Teton County, Wyoming, is the home of Jackson Hole, and it is a blue island in a sea of red; the blue counties in southeastern Kansas are home to a number of meat-packing plants, where the Hispanic population has turned once lily-white counties into majority-minority areas.
The general trend here, though, is that most of the land area is red. Land area doesn’t vote, but rural and small-town America casts tens of millions of votes. Shifts of this magnitude add up. The one big exception is California, which has shifted almost uniformly toward Democrats. But this is inefficient; California was already quite blue by 1996; all of these votes are “wasted.” On the other hand, the dark red areas – West Virginia, Kentucky, Tennessee, Missouri, Arkansas, Louisiana – are all places where Democrats performed well in 1996, but no longer do.
We can see this again by using our familiar CBSA divisions, to see how Clinton’s performance compared to previous Democratic performances:
As you can see, Hillary Clinton had the best performance of any Democrat in recent memory in the so-called “Mega-Cities,” winning almost two of every three votes cast. In large cities (those with metro areas of between 1 million and 5 million people), Clinton performed a bit worse than Barack Obama in 2008 and Bill Clinton in 1996. In other words, looking only at large urban areas, she was headed for a big victory.
The problem is that the Democratic coalition fell apart below that. Her performance in small cities was closer to Al Gore’s 2000 performance than either Obama’s or Bill’s landslide wins. Beneath that, her performance was a disaster; she ran behind Michael Dukakis in large towns, about 10 points behind him in small towns, and about 15 points behind him in rural areas. She ran over 20 points behind Bill in small towns and rural areas.
Winning mega-cities by 30 points is great, but her margin there was mostly (though not entirely) neutralized by her poor performance in large rural areas and small towns alone. Again, her vote in these mega-cities was also inefficiently distributed in already-blue states; the swing states with mega-cities tend to have large amounts of rural land, which is why she lost Florida, Georgia and Pennsylvania. Finally, we note that while rural and small-town America are disappearing, that disappearance is happening much more gradually than people appreciate.
What all of this adds up to is a Democratic coalition that has become increasingly unbalanced over time. We believe the next chart is striking. Note that, for all its faults, Dukakis’ coalition in 1988 was fairly balanced. This provided the groundwork for Bill Clinton to achieve a wide-ranging victory in 1992, without changing the coalition that much. He simply moved all of the bars up by similar amounts, with a little extra bump in the mega-cities (reflecting the shift of northern suburbanites to the Democratic Party). But from there on, each election shows the gap between rural areas and urban areas expanding:
We close with three observations:
1. This didn’t happen overnight.
We’ve talked a lot about what “Clinton did wrong” and “how Trump won.” But we hope what you see is that Clinton’s loss is, in many ways, the result of a process that unfolded over the course of multiple decades. The slide from Bill Clinton ’92 (the populist; Paul Tsongas was the “emerging majority” candidate that year) to Bill Clinton ’96 (the progressive centrist) to Al Gore to John Kerry to Barack Obama to Hillary Clinton ’16 is a progression of candidates that are increasingly urban and urbane. This played well in the mega-cities and enabled Democrats to win large cities. But it probably hurt them everywhere else.
The fact that this didn’t happen overnight means that the solution is probably more complicated than simply identifying the right candidate to appeal to rural voters or tweaking the platform.
2. This hurts the Democrats’ chances in the Electoral College, and kills them in the House and Senate.
Of course, elections are determined by fundamental factors like the economy, scandals and incumbency. Why, then, do we focus so much on coalitional factors?
First, it’s simply interesting; changing party coalitions tell us about the parties, why they are pursuing certain agenda items, and what type of candidate they are likely to nominate. You cannot, for example, understand Donald Trump’s nomination without also understanding the influx of Jacksonian Democrats into the Republican Party.
But we get back to our initial point: In our system of government, popular vote metrics are only sensible when put through a geographic filter. This causes problems in the Electoral College, which we’ve recounted before. There are only nine “mega-cities” in America: New York, Los Angeles, Chicago, Washington, D.C., Philadelphia, Miami, Atlanta, Houston, and Dallas. These, in turn, affect 11 states: New York, New Jersey, Connecticut, California, Illinois, Virginia, Maryland, Pennsylvania, Florida, Georgia and Texas.
In other words, in seven of these states, further growth in this area does no good for Democrats, as they are already blue. In three others (Pennsylvania, Florida and Georgia), the rural areas, towns, and small cities cast enough votes to outvote the mega-city. The final one – Texas – may be the key to a Democratic majority down the road, but Hillary Clinton still lost it by nine points, with a lot of Romney’s votes going to third party candidates. Put differently, the place where the Democratic coalition is growing the most does them the least good, electorally speaking.
But if it causes problems in the Electoral College, it wreaks havoc in the Senate, House, and state legislatures. While only 11 states have mega-cities, 18 states have neither mega-cities nor large cities. To put this in perspective, a party that sweeps the rural and town-dominated states starts out with 36 Senate seats. This won’t happen, of course – Vermont isn’t going Republican any time soon – but Republicans also have a solid foundation in states with large cities, like Oklahoma and Kansas. Because of the Democrats’ concentration in cities, and because of the concentration of the urban vote in relatively few states, the Senate is now a natural Republican gerrymander.
In the House, it is largely the same story. To see how this plays out, consider Iowa during the 1990s redistricting:
As you can see, it’s rather difficult not to draw districts where Democrats can win. Republicans ran well here, in part because the coalitions shifted over time, in part because of the GOP wave in 1994, and in part because a pair of unusually strong incumbents stuck around in the eastern portion of the state. But even as late as 2010, Democrats were able to win three of the four districts in the state. But how do Democrats win congressional seats when Clinton’s 2016 coalition appears – a few urban-ish counties and college towns:
You have a map where even an established incumbent Democrat like Dave Loebsack struggles; suddenly Republican Rod Blum’s 2014 win in the formerly Democratic northeastern district doesn’t look like a fluke anymore. Losing some of these seats is a direct consequence of Democratic weakness in rural areas.
Or consider Wisconsin in 2000. Given a Democratic coalition that looked like this, it isn’t surprising that they were able to win in five districts over the course of the decade.
But given this? How much can Democrats accomplish?
Finally, we’ll leave you with an image from Thursday’s Part 4. Obviously there was some gerrymandering in Ohio in 2012, but it’s also relatively easy to see why it was so easy for Republicans to win, and why Democrats would have a hard time drawing even an 8-8 map if they were to control redistricting in 2020 (without resorting to some truly ugly lines).
It should also become apparent why Republicans were able to win the Electoral College while losing the popular vote twice in the past five elections, and why they were able to win a healthy House majority while (narrowly) losing the popular vote in 2012. It flows directly from weaknesses in the Democrats’ coalition.
3. This is reversible.
We close with the following thought. This happened for a reason, and since it happened for a reason, it is probably reversible. But it won’t happen overnight. Democrats were once able to win rural areas, and send large numbers of members of Congress from these places. That was in part because they focused their message on these areas, and tolerated culturally conservative Democrats like Harold Volkmer in Missouri and Sonny Montgomery in Mississippi.
But for much of the Obama administration, these members were forced to take a series of tough votes that rendered the Blue Dog Democrat a near-extinct species. Hillary Clinton’s campaign in 2016 was almost entirely directed toward the coalition of the ascendant in mega-cities, which is a decent enough coalition for the popular vote, but is highly problematic for the Electoral College.
We very much doubt that liberal Democrats will like some of policy compromises that winning back these areas probably entails. We suspect, however, that they will find that preferable to the policies that will be enacted over the next four years.