How Trump Won: The South
The first in a four-part series
In 2009, the lead author of this piece co-wrote a series of articles with Jay Cost, who was then with RealClearPolitics. That series (the lead author’s first published articles at RealClearPolitics) focused on a crucial question: How Barack Obama won the 2008 election. It examined census regions and used a combination of maps, election returns, census data, and exit poll data to try to determine what President Obama’s coalition looked like at the sub-national level.
This series will emulate that approach, with a deep dive into election returns and state-level politics. As with the 2009 series, we focus on census regions. These census regions and census divisions are shown on the following map:
The 2009 article examined the South in two parts, one focusing upon the West South Central and East South Central Division, and another focusing on the South Atlantic. We will not do so here. This is because these divisions moved in different directions, and had different electoral significances, in 2008. The West and East South Central divisions had been the heart of the Democratic Party for over 150 years. Only one Democrat (John F. Kennedy in 1960) had won without Kentucky or Tennessee, and he only lost them narrowly. In 2008, however, these regions stayed in the Republican fold while Barack Obama won the presidency comfortably.
The South Atlantic Division, by contrast, had been a core part of the Republican coalition since the 1970s. Virginia, North Carolina, South Carolina, Georgia and Florida had all been thought to be reliably Republican. They combined to give George W. Bush 78 electoral votes in 2004, more than a quarter of the total he needed to win. Yet in 2008, Barack Obama won most of the division’s electoral votes, which was a key component of his victory.
The 2016 cycle was different. The story in the South is actually that all three divisions did basically the same thing.
This is part of an even broader story, however. What happened in the South in 2008 is probably key to understanding what happened in the Midwest in 2016, and the election in general. And it is here that we find the overarching theme of our series, and tie it in with those 2009 articles. In retrospect, the 2009 series identified the problems that would beset the Democratic coalition in the 2010s, and would cost Hillary Clinton the election in 2016. As Cost and Trende observed, when discussing the East South Central and West South Central division:
While [John] McCain's victories here were not surprising, they do carry some significance. After all, for the last 20 years (counting George H.W. Bush as a Texan, given that he began his political career there), the chief executive of the nation has hailed from the West South Central division -- and the overlap between Clinton’s and Bush's voting coalitions here seems to have been pretty significant. Clinton was able to forge a coalition in this part of the country that united African-Americans with lower income, rural/small-town whites. Bush brought many of the latter into a coalition built upon the Republican base. So, while they don't get much press, these voters have, in a strange way, been ruling the roost for 20 years.
But this era is now over. Even as much of the rest of America swung toward Obama and the Democrats -- these voters stuck with the Republican Party. In fact, it appears the GOP actually won a few more of them. For almost 20 years, you could say that the "first choice" of these swing voters would become president, but that's no longer the case. That seems pretty significant.
The oddity that this article noted – that these divisions, which were once the backbone of the Democratic Party, had stayed Republican while the rest of the country shifted leftward – was actually quite a bit more significant than the article suggested.
In retrospect, what the article identified was, in fact, a serious contagion in the Democratic coalition. While President Obama performed very well among certain core Democratic groups, he amputated a portion of the historical Democratic base in the process – rural whites in Appalachia. He had, as the lead author of this piece suggested in his book, “The Lost Majority,” traded the traditional Democratic coalition for a narrower but deeper coalition.
We now view this contagion as spreading throughout the Democratic congressional and statehouse maps in the ’10s, as Democrats who had held on in down-ticket races for decades suddenly found it impossible to do so in 2016. Congressional losses in 2010 were the most pronounced in Appalachia and the Deep South, which lie in the heart of this region. Obama was able to piece together enough of his 2008 coalition to win in 2012, but in 2014 four of the nine Senate seats Democrats lost were in the South, while the party found itself with only two statewide officeholders in the Old Confederacy outside of Virginia and North Carolina, and was locked out of every legislative chamber in the census region (Republicans controlled two-thirds of the seats in over half of the chambers). In 2016, somehow, things got worse.
We look at the data in three ways (exit poll data are not available throughout the region for this election), all of which suggest deepening problems for Democrats: First, we look at the rural/urban divide. Second, we look at “swing counties.” Finally, we look at maps.
The maps in the final section are awfully jarring, if not particularly surprising. We’d urge you to stick with us through them.
1. The Urban/Rural Divide
We first divided the counties up into six groups by population, using the convention established in the 2009 piece: Counties are considered part of a “Mega City” if they’re part of a metropolitan or micropolitan area (“CBSA,” for “Census Bureau Statistical Area”) of at least 5 million people; a “Large City” if they’re part of a CBSA of 1 million to 5 million people; a “Small City” if they’re part of a CBSA of between 500,000 and 1 million people; a “Large Town” if they are part of a CBSA of between 100,000 and 500,000 people; a “Small Town” if they are part of a CBSA of fewer than 100,000 people; and a “Rural County” otherwise. We do maintain a constant definition of CBSAs under their current 2016 definitions; we went back and forth about whether to shift, but the changes aren’t as significant as you would think, so we just kept an apples-to-apples comparison under our current terminology. It is arguably a weakness of the study. You can see how these groups shifted over time here:
The trend seems pretty clear: Stability in the cities, collapse in the towns and especially rural counties. As noted above, the East South Central Division was once the heart of the Democratic Party. Part of Bill Clinton’s wins in states like Kentucky and Tennessee came about as a result of his strong showing in rural areas. In fact, he performed better in rural areas than in urban areas. But Democrats have not performed well there since. Again, while the 2009 article notes Obama’s uptick in large cities (such as they are in this division), it notes his downward trend in the rural South and in the towns. This continued in 2012, and again in 2016.
We note too that a clear majority of votes here are cast in rural areas and towns, and that this has been reasonably stable over the past 30 years. In other words, the Republican gains in the region really hurts the Democrats.
Now consider the West South Central (Texas, Oklahoma, Arkansas and Louisiana):
In the West South Central, it is largely the same story. Hillary Clinton’s vote shares in the cities were somewhat larger than Obama’s. This is mostly a function of Texas. We don’t know entirely what to make of her showing in there – a large number of voters probably knew their votes didn’t matter, and we note that Trump ran further behind Mitt Romney than Clinton ran ahead of Obama in the urban counties. Regardless, just as urban Republicanism was the first indicator of realignment in Texas in the 1940s and 1950s, so too could this represent the seeds of a Democratic revival in the Lone Star State. We will have to wait and see.
But look at the rural areas! Hillary received less than half the vote share that Bill received in 1992 and 1996 (remember, Texas was a three-point state in 1992 and a five-point state in 1996, while Bill carried Louisiana and Arkansas twice).
In this section of the country, a little less than half of all votes are cast in rural/town counties, so this mattered for the region, albeit not nearly as much as it mattered in the East South Central. Here the rural and town populations are declining, although more gradually than many commentators suggest. Over the past 30 years or so, the “Mega City” share of the vote in the region has increased from about 26 percent to about 33 percent. This is significant, but not overwhelming.
If these trends had been isolated to the East and West South Central divisions, it wouldn’t have mattered because these states were red by the time 2008 rolled around to begin with; going from maroon to crimson didn’t matter much. But as we noted, the contagion spread. Here’s the South Atlantic region, which basically follows the Eastern Seaboard from Delaware (yes, it is part of the South, for historical reasons) to Florida.
As you can see, Clinton ran well ahead of Obama in Mega Cities (and there are quite a few in this region), though she “wasted” a lot of these votes in Atlanta. Everywhere else was a bit of a disappointment. She only tied Obama in large cities (though this is still ahead of Bill’s performances in the region), but ran behind Obama (and Bill) everywhere else. In large towns, she ran about even with Michael Dukakis, and in rural counties and small towns, she ran behind Dukakis by significant amounts. Again, rural counties and towns don’t cast a lot of votes standing alone, but they do add up.
This was enough to cost Clinton the states of Florida and North Carolina. Had she won those, she would be president.
Putting the three regions together, we can get a sense of what has happened to the Democratic coalition in the South. It has continued to grow in “Mega Cities,” which explains the party’s emerging dominance in Virginia, and its continued competitiveness in Florida. But in the towns and rural areas, it has been in a state of collapse that resists national vote trends. Obviously, there is a lower limit for Democrats here, but rural areas cast a lot of votes in the South; these voters can keep Republicans in the game for quite a while.
Finally, we look at these trends grouped by year:
This is interesting. From 1988-1996, the Democrats’ coalition was well balanced. Bill Clinton basically took the Dukakis vote, and tacked on 10 points across the board. But beginning with Al Gore, the distributions are increasingly skewed toward the mega cities. The gains there are significant, but they aren’t enough to offset the losses in rural areas.
2. Swing Counties
It is also useful to look at Democratic performance in swing counties. We will do this briefly. The 2009 piece defined swing counties as those that voted for Clinton and Bush. We noted that Democrats did well in the Clinton ’96/Kerry counties, while Republicans did well in the Dole/Bush ’04 counties. The Clinton ’96/Bush ’04 counties, however, showed a continued decline in Democratic performance. This continued in 2016 in both the East South Central and West South Central.
We thought that this definition might be a bit dated (counties that voted for Bush 12 years ago might just be “Republican” counties today), so we tried using the 2004 and 2008 elections (red are Bush/McCain, Blue are Kerry/Obama, purple are Bush/Obama or Kerry/McCain). Using this updated definition shows the same pattern of Democratic decline outside of core counties in the East South Central and South Atlantic divisions, but shows increased Democratic strength in Republican areas in the West South Central.
This probably reflects the Democrats’ increased performance in the Dallas and Houston suburbs. Again, it will be interesting to see if this is a blip, or if this is a contagion in the Republican coalition that we will be talking about in eight years.
Now we come to our favorite part of the piece. We’ve shown through bar and line charts how the Democratic and Republican coalitions shifted over time. Now we can map it out. It is a story of how, in just 20 years, the Republicans and Democrats in the South went from this:
Probably the best place to start is this map, which shows Dukakis’ vote share in Southern counties, normalized for national vote share. Whenever we say “normalized for national vote share” or just “normalized,” we mean we’ve indicated how a county leans relative to the country as a whole. It’s basically a way of subtracting out national effects, so you can view the coalition shifts:
As of 1988, the Democrats had a robust coalition in the South. They showed strength across Appalachia, in the “black belt” (named for the fertile soil), and in the Rio Grande Valley. Southern Louisiana (Catholics) and Arkansas outside the traditionally red northwest portion of the state were blue. Republicans, by contrast, were strong in the historically Republican areas of southeastern Kentucky, eastern Tennessee, and western North Carolina. They also held the cities, as metro areas like Miami, Dallas, Houston, Atlanta, and Northern Virginia were all purple-to-red.
By 1996, not much had changed. There was some polarization, but the core Democratic coalition was intact:
The most salient changes involve some reddening in West Texas, and a slight “bluing” in Miami and Northern Virginia.
In 2000, however, things really begin to change:
Texas and Kentucky were now mostly red, and Democrats lost their toeholds in Louisiana and Arkansas, though they would hold on at the state and non-presidential federal level for another decade. West Virginia began to redden, although Democrats remained strong in the unionized mountain areas in the south. Miami and D.C. are considerably bluer. This continued in 2004:
The reds here are now a deeper hue, and Democrats are almost entirely shut out of Oklahoma. A decade earlier, the Oklahoma delegation had been 4-2 Democratic; now it was almost impossible to draw a map where Republicans would lose anything. At the same time, Travis County (Austin) was now blue, Dallas County (the light square three rows down from the Oklahoma border) was a swing area, Miami was blue, and Northern Virginia was increasingly shifting from purple to blue.
Then we have 2008:
By now, the Democratic coalition is basically nonexistent within the Fall Line. The urban areas are distinct: Dallas, Houston, Austin, Nashville, Louisville, Lexington, Atlanta, Northern Virginia, Charlotte and Miami are all readily accessible if you have a basic understanding of where these areas are located in their respective states.
This can all be summarized in the following map, which shows the change over time. Democrats make some advances in the cities and black belt, but those changes are generally overwhelmed by the changes in rural and small town areas:
We can also see this at a more granular level by looking at a few state level maps. Here’s Florida, normalized, in 1996, 2004, 2012 and 2016 (read these left to right, by row):
The state was competitive in all four elections, with Democrats winning close races in 1996 and 2012, and Republicans winning in 2004 and 2016 (though with the normalization, that shouldn’t matter much). But look at the coalition shift. Miami is blue throughout, but the Tampa area (on the west coast) and Orlando (the two counties in the middle) shift toward Democrats. But this is offset by a substantial reddening in the rest of the state, particularly in the northern panhandle. This is what saved Donald Trump in the state. Again, rural votes add up.
Going further up the coast, we encounter North Carolina (as above, read these left to right, by row):
Here, we see a similar story. In 1996, Mecklenburg (the county abutting the rounded corner on the South) is actually slightly red, while Durham and Chapel Hill (the two blue side-by-side counties in the north center of the state) are light blue. Wake County (Raleigh), to the southeast of Durham, is red. There’s a blue cluster in the northeast and a blue cluster in the south central; this represents the black belt counties. While Democrats aren’t strong in rural counties, neither are they wiped out.
Over the course of the next 20 years, the urban counties become progressively bluer. By the end of the decade, Durham and Chapel Hill are dark blue, Wake County and Mecklenburg are bluer, and Burlington, Greensboro and Winston-Salem (the three counties to the west of Orange County, which includes Chapel Hill) are lighter. In the panhandle, Buncombe County (Asheville) becomes blue.
But as in other areas of the region, the rural areas become dark red. Some of the black belt counties flip; this is probably due to rural whites switching their voter allegiances. Again, this managed to overwhelm the Democratic progress in the state in 2016.
Finally, Virginia (once again, read left to right, by row):
This is more of what we’ve seen, although the changes in the southwestern portion of the state are particularly dramatic. What makes the Democratic coalition more successful in Virginia is simply that there are more urban areas for them; the blue trend in Hampton Roads, Richmond, Charlottesville and especially Northern Virginia offset the Republican gains in the rest of the state. This enabled Clinton’s five-point win in a state that had never voted for a Democrat from 1952 through 2004.
In summary: In 2009, we identified the outlines of what would become a Democratic problem: weakness among traditionally Democratic voters in rural areas and towns. In 2016, this weakness became significant enough that it overwhelmed Democratic strength in urban areas in two states that President Obama had won. As we saw above, this is significant, because while urban areas are growing, they are growing at a slower rate than many analysts seem to appreciate.
As we’ll see, this basic pattern has been repeated all over the country. In some areas, it mattered more than others.
Next up: The Mountain West and Pacific.