The changes in voting behavior in recent years have a distinct geographic flavor. In 1988, George H. W. Bush won with 53% of the vote; 16 years later, his son got 51% nationwide, pretty close to the same. We should be able to figure out something from the comparison, as both Democrats in the race were ‘Massachusetts liberals’ and both Republicans were Bushes, although the son has a very different persona than the father. If George H. W. Bush had run against Dukakis again in 2004, that would give us the most information about changes in voting behavior, but let’s take a look at what we do have:
Click to enlarge. Below, more on trends in voting and population growth in recent years.
Cross-posted at Open Left and DailyKos.
Despite similar nationwide vote percentages in 1988 and 2004, the maps above show striking and not-too-surprising changes. Just like in the distribution of the House seats, there is a widening gulf of increasing polarization, although this is likely in part an artifact of the definitions we are using (percent vote for President). In general, the Northeast and Midatlantic, Southwest, and the industrial Midwest trended towards Democrats. The Plains states and the Mountain West trended towards Republicans. Intriguingly, the South generally stayed about the same.
Let's look at change by county (voting data available here): the maps below show the intensity of a shift towards Republicans or Democrats (from 0 to 30 points between 1988 and 2004). This doesn't mean Kerry won the blue counties, it means he did better than Dukakis; the darker the blue, the greater the shift. The lightest color for each represents an increase of 0-5%.
Click to enlarge.
Kerry increased in performance over Dukakis solidly in the Northeast and down the East Coast to Virginia. Ohio, Michigan, Florida and South Carolina also show some consolidated increases. There's a checkerboard through the South and Southwest. The Plains show almost uniform increases for Bush II with a very solid pattern in the Mountain West and much of the Midwest and Ohio River valley. Texas, which had a horse of a different party in each race (Lloyd Bentsen in 1988), has the most dramatic changes, although note it is not uniformly red.
What portion of the changes we see can be attributed to the persona of Bush the Texas Rancher versus Republicans in general can't be sorted out. For now, it appears that Bush's fake cowboy swagger has infected the entire Republican brand, so the information may be relevant for the short term at least.
But this is not much of a story: if we are interested in voting behavior changes, we have to know how many votes we're talking about. So let's look at cartograms of the counties, using the method of Gastner and Newman, where the size of the counties are proportional to their populations instead of their areas (thanks to BentLiberal for the suggestion). The US map ends up looking strangely like Asia:
Click to enlarge.
Again, these maps show change in support, not who won. The pattern familiar from conventional wisdom now becomes clear: urban versus rural. Not only are urban areas generally Democratic strongholds, they have become even more so in recent years. There are some exceptions: rural New England, and some urban areas in the Plains and Appalachia, for example.
This still isn't the whole picture: we're looking at change, so we want to know how fast these counties are growing.
Click to enlarge.
If growth patterns were exactly the same between 2004-2014 as between 1990-2000 (an assumption which probably isn't reasonable), and voting patterns remained the same in every county as in 2004, the political split would still be about the same – the Republican vote would increase by 0.6 points for a hypothetical 2014 election in this hypothetically growing country. We do know, however, that since 2004 the number of people that identify as Republicans has decreased, indicating they're probably less likely to vote for a Republican for president; that alone throws off the entire hypothetical scenario. We might be able to think of it as a worst-case scenario, however.
So what can we take away from these maps? Well, one thing is to look at extremes in growth. First, those counties that lost population; the map on the left shows Bush's performance in 2004; the map on the right shows the change from 1988-2004, where blue indicates an increase in Democratic performance, and red an increase in Republican performance.
Click to enlarge.
The vast center swath of the country that looks so large on an area-based map has aligned with Republicans on the presidential level even more closely since 1988. But it doesn't have many people and is in the process of losing them. There's also a fair number of strongly Democratic and Democratic-trending urban centers that have lost population. Another interesting note, the Appalachian region increased its Republican support to a draw, more or less, as it lost population.
Next, only those counties that grew by 25% or more between 1990 and 2000.
Click to enlarge.
The high growth counties show us a bunch of circles in the South, Midwest, and West. These are the exurbs: they vary mostly from strong to very strong Republican support, but they are not all places of unbridled Republican growth. In a fair number of exurban counties – outside Chicago and Northern Virginia, for instance – Republican support actually decreased from 1988 to 2004.
Where do we go from here?
At this point let me reiterate the cautions necessary in interpreting these changes: they are based on presidential vote, which can be quite different from local politics, and may have much to do with the individual candidates involved. And past behavior is no guarantee of future behavior.
Because populations don't just grow, they also change. We all die, and young people start voting, and people who move somewhere don't necessarily act like those who are already there, and communities change. And, both individuals and populations change their attitudes. All these factors favor Democrats on the national level over the long term; in the short term, the Republican brand is in the doghouse.
First, according to Pew's surveys, not only is the country as a whole – even Republicans – becoming less socially conservative, but each successive generation is less socially conservative than the last, and each generation itself is becoming less socially conservative over time. Pew measured this as the average number of socially conservative responses on six questions. Pew also found a recent increase in support for various government programs.
Communities change as well. Over the long term, exurbs and suburbs are becoming more like urban central cities in their problems and concerns, and, one might think, more like urban centers in politics as well. The housing crisis may very well accelerate this transformation.
Young people are far more progressive than the older voters they are replacing at the polls, as mentioned above. Republicans cannot use fear of non-Christian gay brown people to get them to vote Republican, because they and their peers are increasingly non-Christian, non-white, and openly gay themselves.
Finally, the population increase in this country is 5-15 times higher for Democratic-trending minorities than non-Hispanic whites, translating to a stunning 4 out of every 5 new residents. By focusing on appealing mainly to rural and exurban white Christian voters, Republicans may well write themselves into irrelevance over the long term.
Selected relatively close elections of the past ~100 years. Note that who was allowed to vote also varied over time….
Click to enlarge. Data source.
which was useful for showing which parts of the country were shifting so far from the Dems that Kerry couldn’t top Mondale.
In DeKalb county, GA, Reagan won 58/42, and Bush lost 73/27, which should be a sign of obvious demographic changes. (DeKalb went 50/49 for Dukakis)
In Forsyth County, GA, Reagan won 75/25, and Bush won 83/16. But Forsyth cast 9K votes in 1984 and 57K in 2004.