The Realignment Myth
With a candidate as strange as Donald Trump, it is tempting to speculate that the usual red-state and blue-state assignments may not hold. Trump ...
The Second Phase of Realignment: 1976-2012
Shifts in American political geography (“realignments”), which I wrote about on Thursday, can be viewed at a glance using the following diagram. It allows us to see just how little change there has been in recent years – including 2016 so far.In this diagram, the closer two points are, the more correlated the national pattern of state-to-state voting. Distance between points approximately represents “unshared variance” between two elections, which is defined as 1 minus the correlation coefficient squared. The reason that unshared variance can be plotted visually is that mathematically, it acts like distance: changes in successive elections can gradually add up, eventually altering the political landscape. Which appears to be what is meant by a “realignment.” Roughly speaking, left-to-right measures the Civil-Rights-Act-triggered realignment, so that’s the South going Republican plus maybe Northern states trending Democratic. Up-to-down reflects the second realignment.
(Math note: To make this plot I used multidimensional scaling, in which I plotted three of the elections to get the graph started, then added more elections one at a time, re-optimizing the graph each time. My collaborators and I have used this approach before to analyze how brain architecture has changed over the course of evolution. The scripts and data are here. Okay, back to the politics.)
Changes in partisan alignment do not coincide with transitions in power. 1976 to 1980 (Carter to Reagan), 1988 to 1992 (Bush I to Clinton I), and 2004 to 2008 (Bush II to Obama) were all momentous transitions, but in no case did the political map rearrange itself. On the other hand, 1960 to 1964 (Johnson’s re-election) showed a massive upheaval in which Democrats lost their previously strong grip on the South. As I wrote the other day, this change took three more elections to play out, culminating in 1972, the year of Nixon’s re-election and his famous use of the Southern strategy. I would therefore call 1960 to 1972 the culmination of a first great modern realignment of the parties.
This map shows a perhaps equally great second shift, starting in 1976 and indicated by the red path. Each step was small, but over time it added up. Comparison of the 1976 and 2012 electoral maps (see below) shows that during this time, two things happened: (a) the South became securely Republican (think of Gore losing Tennessee in 2000), and (b) Democrats captured Western states starting with Washington, Oregon, and California, and eventually including Nevada, New Mexico, and Colorado. These changes may be partly racial, partly economic/cultural.
Note that the distinction between “Phase 1” and “Phase 2” is somewhat artificial. Really, the major event separating the two phases was the Watergate scandal and Nixon’s consequent resignation, which caused massive damage to Republicans and led to the election of Jimmy Carter, a previously obscure Southern governor. If we leave out 1976 and 1980 from the graph above, the remaining elections look like this:
In this representation, the shift of 1984-2012 corresponds to (b) above, the loss of Western states by Republicans. Washington/Oregon/California/Nevada currently add up to 80 electoral votes, marking a substantial shift toward Democrats. Racial diversity is an obvious driving factor here, with some additional role for new-economy jobs growth.
Perhaps a broader lesson from this diagram is that much of the last 50 years has involved continual change in the configuration of the electoral map. The exception is 2000 to 2012, a static period during which polarization has been massive (multiple Republican takeovers of the House of Representatives, impeachment of President Clinton, multiple government shutdowns, partisan passage of Obamacare, decreased productivity of Congress). Against this backdrop, any change may seem like a lot.
As I wrote last week, despite the upheaval on the Republican side there might not be much change to the map. The change from 2012 to 2016 currently looks, at most, like one-tenth the size of a real realignment.
The graph makes it look like 2016’s partisan alignment might be moving back a little bit toward 1964. I would not read too much into that. I think it is probably noise. Most states are unpolled, and this could just as well be caused by not having enough data. From a numbers standpoint, so far it looks like Donald Trump is basically Bush/McCain/Romney minus Utah, which is amazing considering that all three of those people have rejected Trump in various ways.
There might be some issue with the maps. At least my device here doesn’t show the images, and opening them leads to an error page.
And on the substantive side, the movement of 2016 toward 1964 (and maybe 1972, horizontally) might be unimportant, as you say, but perhaps there’s a factor in play here after all: the particularly salient role of racism in the campaign. You had the Civil Rights Act in 64, the Southern strategy in 72 (68 had Vietnam as an even stronger dividing issue), and now the Trump circus show. 08 and 12, despite Obama on the ticket, were not centered on racism more than usual, and the same goes for all campaigns in between. And, of course, racism can be seen on the map in the States, more than in most other countries.
Yeah, will fix the maps. If you click on them they should eventually give a viewable image. Will fix later.
Agree, sorting by racial appeals is a hypothesis. Need more data!
What with Jeremiah Wright, birtherism and the “secret Muslim” hypothesis, I think racism was unusually salient in 2008 and 2012. But the Republican candidates in both years refused to exploit it very directly, which may be more to the point.
The other presidential election in which I would say it was unusually salient was 1988 (think Willie Horton). And we see a small zig to the right in that year.
If race is the cause behind the shift, then shouldn’t that already be reflected in prior shifts? Trump is running an overly racist campaign and getting basically the same support as prior GOP candidates.
Corrections/Clarifications to my prior post:
1) If the reallignment was driven in some part by race, and we continue to vote with the same patterns, then race is already taken into account and we should not expect additional shifts.
2) Overly = Overtly
Looks like McCain is falling in line and endorsing Trump! What a shame and lack of self dignity on his part!
Today’s LA Times has a story that in California senate’s top 2 to the general election rule, no Republican candidate will make it – it will be most likely between Harris and Sanchez. It will expose the extent to which California is lost to Republicans.
Similarly, I wonder if some states are now locked so solidly red, that no matter what they will vote red. This was triggered by Wall Street Journal’s solidly conservative columnist Bret Stephens saying on Fareed Zaqaria’s GPS that he will NOT support Trump, and wishes that this will be the biggest loss for the Republican party ever, to realize what it has become.
Is a 50 state win still possible in today’s electoral map?
McCain really impressed me when he went face to face with Bush and asked him to denounce the ‘swift boaters’. However when Bush wouldn’t do it McCain dropped it. To me, if you take a moral stand, you don’t abandon it for political gain. So as quickly as he rose in my opinion, he fell even faster and much further down from where he started. Seeing him fall in behind lock(goose)step with Trump isn’t surprising. I’ll give him due for his military service but he isn’t in the military now and doesn’t have to take his marching orders if he doesn’t find them morally acceptable.
I think that a 40+ state win is still possible for a Republican, as a response to some exogenous shock. Imagine, for instance, if George W. Bush were up for reelection in November 2001 rather than November 2004. He might have gotten 500+ electoral votes.
It is not possible for a Democrat. The reddest states are just too red. I think even a popular-vote win on par with Johnson in 1964 wouldn’t win as many states as Johnson did.
Does ’68 fall more in line if you count George Wallace as an honorary Republican?
I did the basic correlation calculation in the Excel spreadsheet I posted. It should have the information needed. Also, the source: https://en.m.wikipedia.org/wiki/United_States_presidential_election,_1968
The tell-tale sign is Deep South states with a Goldwater-Wallace-Nixon voting pattern There were also several Southern states that Wallace did not win in 1968, but came in second, with Humphrey third.
Dr. Wang,
i understand that the distance between two points is 1-correlation. Can you explain what exactly you are plotting on the two axes? I read that the left to right represents civil rights and the up do down represents west going democratic and south going republican (which it seems to me is also a civil rights effect), but its unclear how you are measuring these and plotting, i did try reading the scripts but couldn;t open the .xxx file for some reason
Distance is (1-R^2), not (1-r); i.e., “unshared variance”. The two axes are the first two dimensions of a multi-dimensional scaling solution, so they don’t have an obvious meaning you can specify. Sam thinks the x-axis is basically the southern white vote, so movement from left to right represents the shift of this group away from Democrats and toward Republicans. The y-axis is less obvious.
This is correct, I believe.
I played with 1-r but realized that it doesn’t have units that work naturally for plotting the data points in a plane, i.e. adding like a distance. Variance is better.
Thank you for the explanation. Are the dimensions analogous to deep features in a neural network? Are they somehow orthogonal?
The dimensions are orthogonal by definition. MDS takes pairwise distances and finds a set of orthogonal dimensions in which those distances make sense. I don’t know the relationship to deep learning algorithms.
I’d suggest the following possible two-second summary: Southern Strategy, followed by a slow but steady backlash to the Southern Strategy.
To help explore that notion–I think people hear California, Nevada, New Mexico, and Colorado, and think it is all about growing Hispanic shares. Which is probably part of it, but you have Washington and Oregon too. And another way of looking at that list of Western states in motion is we may also be looking at states which are relatively urbanized.
So–could you separate out the non-Hispanic population only, and see if the second realignment still appears?
Playing with the data some more, and not trying to tease out the national trend from the state-wise realignment as Sam is doing, it seems like the 2016 swing in Utah should not be extremely surprising. Utah seems to be more alterable than the 16 other states I arbitrarily chose as ‘swing’ states. The absolute percentage change for the D-R margin from each previous election, averaged over the last four elections (2000-2012) for the chosen ‘swing’ states and Utah, is given below:
AZ 3.0
PA 3.9
OH 3.9
FL 4.1
NJ 4.5
VA 4.6
NC 4.7
MN 4.8
IA 5.0
NH 5.2
NV 5.3
CO 5.6
NM 5.8
MI 6.0
ME 6.0
WI 6.2
UT 12.3
So Utah’s D-R margin routinely fluctuates from presidential election to presidential election more than the chosen swing states, but it is so far tilted Republican that this large fluctuation has not mattered in terms of the eventual result (except way back in 1964).
I’m not sure swings around the middle are comparable to swings in polarized states. One imagines that voters in non-competitive elections may have different motivations than voters in competitive ones. (And may start behaving differently if the vote ever looks more competitive.)
A plausible thing to do might be to run the margin through a sigmoid function before calculating differences. E.g.
score = 1/(1+exp(-(vote % difference) / 2))
and do calculations with score instead of actual difference.
(Picked a divisor of 2 because it seemed reasonable.)
I don’t think there’s a principled reason to assume that changes in swing states are qualitatively different that changes in states with a strong partisan lean.
However, mathematically, it is true that a change from 85% to 90% is not the same as a change from 45% to 50%; you do need some transformation, such as the sigmoid.
As I understand it, a sigmoid function would transform the D-R margin into a mostly binary output, with significantly lower weights assigned to the values further away from the middle. This might be useful to convert the percentages to a binary prediction, but I am not sure why it should be applied to understand the fluctuation in the vote percentage. The percentage of the state population that can be persuaded to change their vote to either Democrat or Republican just seems to be higher in Utah than the chosen swing states.
I agree that the problem with transforming to probability is that it adds a big nonlinearity. % voter support is a natural unit that is more likely to fluctuate in a uniform manner.
Motivated by the case of Utah, maybe one could think of commited D’s, independents, and committed R’s as separate groups. Utah is an interesting case because it seems linked to this year’s divisions within the GOP. If you look at that graph closely, underperformance in GOP-favoring states is larger than in D-favoring states.
I am sure, you are right about using the variance. As you well know, R A Fisher pointed it out that it is the additive property of variance that makes it fundamentally useful. I was familiar with its role in Markowitz’s theory, now I am curious to learn about your work in evolutionary neurobiology. This blog is getting more and more fascinating 🙂
This is very interesting and all, but you’re essentially just regurgitating decades-old findings from the political science literature on electoral realignments. It’s been known for a long time that 1964 was the last major realignment at the presidential level. Far more interesting is the much slower secular realignment we’ve seen occurring in congressional and state elections since then.
With the exception of a few states (Arkansas for instance), this was largely complete by the 1990s (1994 saw a major realignment in the House, as Republicans largely consolidated their electoral strength in the South). The really interesting question is why it happened so quickly in presidential elections but took so long to trickle down to Congress and the states.
Despite the truculent nature of your first sentence, I will let this through.
I was asked by a news editor whether there was a “realignment” brewing this year. The data say no, but it is necessary to put that finding in quantitative perspective. These last two posts are the result of those calculations. I am sure I could go back and read the past research. However, sometimes I find that political science literature often lacks crisp visualizations of complex data. It was less work to re-do the work in a quantitative form that allows a direct comparison with new 2016 data. For instance, we can say that this year’s change looks like it’s no more than 10% the size of the 1960-to-1964 realignment.
I invite you to give specific citations that will help people understand this topic better.
“I invite you to give specific citations that will help people understand this topic better.”
I would recommend Carmines and Stimson’s 1989 book “Issue Evolution” on the 1964 realignment, and Black and Black’s 2002 book “The Rise of Southern Republicans” on the transformation of the South in congressional and state elections.
Please excuse my truculence. While I’m a big fan of your polling aggregation and election forecasting (an area of political science that has some major deficiencies IMO), I’ve found that when you step outside this area and write about topics like gerrymandering and here, realignments, you either reinvent the wheel (see your previous post on this topic), or get it wrong (gerrymandering). I also think your criticism of data visualization in political science is well founded, although things are undoubtedly getting better on this front.
I don’t want to seem like I’m trolling. In fact, my criticism applies much more to your first post than to this one (I came back to your blog having read that post previously). I really like what you did with the multidimensional scaling in this follow up post.
Thank you for this comment. The first post was basically trying to use the calculation I did for the editor. I hate to waste an effort like that.
Beg to differ on the topic of gerrymandering. Sure, my original NYT piece was written without knowing the existing literature. But some of that literature is almost certainly wrong: for example, it was unable to account for post-2010 events in a satisfactory manner. I like the older work in this domain (including Gudgin and Taylor, who seem underappreciated). Anyway, I am pretty sure my SLR piece (see left sidebar) is a useful and original contribution.
We certainly disagree on gerrymandering, and I actually have a book about it coming out early next year. Perhaps we can have a more substantive debate on the topic at a later date. Short and dirty version: I agree with the Sides/McGhee take that your approach fails to sufficiently take into account the effects of incumbency. Reasonable people can certainly disagree on this – there are some in the political science community whose views are much more in line with yours than mine. Ultimately, I think everyone agrees that partisan gerrymandering has some effects, but the real debate is about the nature and size of those effects.
I haven’t read your SLR piece yet but I certainly will. It’s about time someone came up with some new ideas for judging asymmetry – we are approaching the 30th anniversary of Bandemer after all.
I agree, I haven’t gotten into incumbency and I think it’s an interesting topic. Probably our differences arise in part from the fact that I am not a researcher – I have normative views on this subject. I care very much about what would be a just outcome, within what is possible by means of leveling the playing field. It seems fairly obvious that partisan gerrymandering has a large effect that can be addressed by establishing a legal precedent. Incumbency is an effect that one observes in order to understand as an intellectual question – cool research but not what I am trying to accomplish.
Regression-based approaches that I’ve seen don’t do a great job of maximally capturing the rapid jumps associated with redistricting. I think my statistical measures offer a crisper quantification that could help. See the single-year jumps that I have quantified – I think my SLR piece has that for Maryland. I am confident that redistricting has effects that, when quantified properly, are large and last through multiple elections.
Voting demographics.
Older people vote more frequently – specifically in the off year elections which dictate half of the Senate (or really two thirds or one third depending on the year), the house every other session and most state governments.
They were more liberal (think New Deal) in the 1970s and 1980s – but the nation as a whole, espeically YOUNG people, were getting more conservative (a reaction I suspect to Civil Rights and related issues like crime and social spending.)
Now those young kids who gave Nixon a majority of the under 30 vote in 1972 are in their 60s and are the most reliable voters. And young voters are as liberal now as their parents/older cousins were conservative – and so you really need a unique set of circumstances to get a Republican president or a Democratic Congress.
I may have missed something, but one option that I haven’t seen mentioned is the age demographic shift in voters since the 60’s. As the WWII voters dwindle and the baby boomer generation becomes quite large, can it have something to do with shaping the shifts? Is that something that could be tested?
Thanks for this discussion, very interesting.
Marty
AP reports that Clinton has clinched.
And primary season ends with a striking failure of polling in California! After polls narrowing to the point where Clinton was only 1-2 points ahead of Bernie Sanders, she wins by 13 points. It may get relatively little attention, given the circumstances, but I’d like to understand more about that one.
Weren’t close to a third of the votes in California already cast before those recent polls were conducted? Has the effect of early voting on late polling been studied in detail?
Early voting/mail-in/absentee ballots?
Number of polls that showed Sanders winning California: Zero.
My first guess would be undersampling of Hispanics, especially males, who are hard to reach. Note that only some polls had this problem – a few were closer to the mark. One could drill into the crosstabs to learn more.
The pattern I noticed a week or two before the election was that poll method seemed to matter: traditional live-landline polls had a really close race, but Internet and even automated-phone polls showed a much wider margin, close to the actual result in fact.
But right at the end, the two camps seemed to converge on a closer margin. That was probably after a large fraction of the votes had already been cast, though.
@bks: I remember thinking that was actually odd–there was a spate of polls that showed a margin of 1 or 2 points, and the variation in their numbers seemed far narrower than the sampling MoE. Probably just a coincidence, though.
I’d guess undersampling of Hispanic and Asian voters combined with some sort of early voting effect.
Really, though, it is not a huge miss as primary polling goes.
National polls: If the national polls in the next 2 weeks show Trump gaining after his blatant racist statements about Judge Curiel dominating the news, then Trump has given voice to and gained support from the dark side of numerous American voters. Bluntly, he will have made racism a legitimate, and perhaps winning, campaign tool nationally.
Racism has been a winning campaign tool nationally for a very long time. Trump is just a bit less subtle about it than Nixon was.
I’d be cautious about conclusions like that. There is no doubt there is some sort of audience for right wing ethno-nationalism in the United States–in fact there is such an audience in every other developed country, and there was no reason to believe the United States would be immune.
However, there is also no doubt that Trump’s support is inflated in part by partisanship, which is a very powerful force in the United States (for good reason, really–it is strongly favored as a strategy by our system of government).
So you’ve got events going on like Trump securing the Republican nomination and now Clinton securing the Democratic nomination, which could be triggering partisan-based effects. You’ve also got Trump doubling down on his appeals to right wing ethno-nationalism, but disentangling those factors is not going to be easy to do on the fly.
Please read this article: https://www.washingtonpost.com/news/wonk/wp/2016/06/06/racial-anxiety-is-a-huge-driver-of-support-for-donald-trump-two-new-studies-find/
I wonder how the news of Senator David Johnson’s decision to leave the Republican party changes the calculus of this analysis. If there are major defections of this type prior to the election, what sort of re-alignment would in fact be measured?
http://www.theguardian.com/us-news/2016/jun/07/david-johnson-iowa-state-senator-leaves-republican-party-trump?CMP=Share_AndroidApp_Google
Surprising Zogby poll has Clinton > Trump in KS. Take with a kilo of salt:
http://www2.ljworld.com/weblogs/capitol-report/2016/jun/10/poll-shows-clinton-leading-in-kansas-bro/
I agree with your basic premise that voting patterns are relatively immobile. However, it’s hard to ignore the potential short term and long effects of Trump on consolidating the minority vote behind the Democrats. And once it happens, as your thesis suggests, it’s very difficult to shift. Right now Democrats consistently get 90% of the black vote and 75% of the Jewish vote. It will be cold day in hell before changes. This leaves the Hispanics and Asians who are already substantially in the Democratic tent. Romney got about 18% of the minority vote in 2012 (he got about 25% of the Hispanic vote and that bumped up because of FL). Trump will be lucky if he gets 10% of the total minority vote (even Cuba is a diminishing asset for the Republicans). This probably means in the short term that a couple of purple states go Clinton but the long term implications are very serious for the Republican party.
Hi Prof. Wang:
I recently checked the ISideWith.com map again using the “Recent” setting.
If you look at the handful of state polls available versus the map (NC, PA, GA, VA, UT), the Trump-Clinton state poll percentages seem to line up with this map provided you take ~13 points off the ISideWith numbers for Trump and give ~13 points to the numbers for Clinton.
The ISideWith map still projects what I see as a big change from 1988-2012 map. CO and VA trend and lean significantly Democratic, and PA, MI, NJ, DE trend significantly Republican and are a toss-up/lean Republican. UT trends significantly Democrat but leans Republican. WI and MN trend significantly Democratic and are as solid as IL on the map.
Still worth keeping an eye on.