The Long Term

My new National Trendline code.  Details below!

Our democracy just (barely) survived a major assault, and as many have pointed out, it's unlikely to be the last. Even beyond the issues of the 60+ legal challenges presented in court, we've had serious talk among state legislators of trying to overturn the will of the voters, members of county and state voting boards refusing to accept the outcome, 139 US representatives – the overwhelming majority of Republicans in the chamber – and 7 senators, formally challening the vote.  And lest we forget, an actual armed insurrection on the capitol, aimed at disrupting the electoral vote certification.  And having failed at all of those, legislatures in Arizona, Pennsylvania, Georgia, and many other close states are now hard at work either making it harder to vote or to make sure they have the option of overturning the election next time.

That is, to put it bluntly, very bad.

But there's an even deeper, anti-democratic (but wholly constitutional) flaw in our system as well. Biden, as you know, won by over 7 million votes and 4.4% nationally, but yet if it weren't for a combined total of about 43,000 votes in WI, AZ and GA – roughly 0.03% of the 150 million votes cast nationwide – he would have lost. The electoral college would have been deadlocked 269-269 and by the insane rule set up by the founding fathers, the majority of states would have given the election to Trump.  Or put another way, Biden won the tipping point state, GA, by only 0.6%, meaning that there was a 3.8% inefficiency, or mismatch, between the popular vote and Electoral College.  By contrast, the inefficiency in 2016 was only 2.9%.  

The Electoral College, and the Senate for the matter, are an affront to basic conception of democracy and in a broad sense, they're getting worse, (though articles on this that note that 70% of the population are slated to live in only 15 states by 2040 fail to note that 66% lived in 15 states in 2010).  And the majority of the smaller states are more rural, meaning that they lean red.

But in some ways, the story isn't quite so simple. We've seen in Arizona and Georgia that consistent red states can flip blue.  Or consider the popular vote in general.  Democrats winning nationwide isn't a new thing. There's been a 50 year trend (following the political realignment following the aftermath of the 1968 Voting Rights Act which threw huge numbers of Southern white voters to the Republican party) of Dems improving, year by year.  Republicans have only won the popular vote once in the 21st century, in 2004, a reelection in the wake of 9/11, where W still only won by a modest 2.4%.  

More generally, plotting the results from the last 14 elections, we see a pretty clear trend:

D-R margins in the Presidential Popular Vote (1968-2020). Third party candidates in 1968 (George Wallace) and 1992 (Ross Perot) took more than 10% of the total vote, making popular vote shares in those years less meaningful (hence grey and not included in the model).  An argument could also be made for 1976 where Carter's strong performance was likely a response to Ford's pardon of Nixon.

Excluding a few special cases, the trend is for Dems to improve their margin by about 0.4%/year, albeit with a large scatter (about 7%).  The expectation from this simple model is that a "generic" Dem should have won the popular vote against a "generic" republican by 7.5%*.  Even though the actual election was anything but generic, it still produced a result within a few points of that. 

I've written "Trendline," a new numerical widget to explore how the long term trends in national vote – along with the relative movement of individual states.  You can see a snapshot up at the top of this page, but the basic idea is that for most states, the relative movement compared to national has been steady in one direction or another for decades.  Consider my home state of Pennsylvania:

Pennsylvania presidential election votes since 2000 compared to the national popular vote.

The trend is quite clear.  Until maybe a decade ago, PA was to the left of the country as a whole.  Over time, it's trended about 0.5%/year redward, to the point where it soon won't be a tipping point.  On the other hand, consider TX:


Texas presidential vote since 2000.

which is trending decidedly blueward. Again, as a reminder, this is the vote relative to national, so the state can be redder than typical, but still vote Dem, as TX seems poised to do in 2024 or 2028.

There are, of course, complications. Arizona is trending blue as well, but the pattern doesn't look like a simple linear model.  The biggest outlier is 2008, when popular senator John McCain ran for president pushing the state redward.

Arizona relative presidential votes.

But for most states – and especially most swing states – this model works quite well.  

Combining those two results, relative state trends and national, we can can get an estimate for how different states will trend overall over the next decade or so:

The modeled net trend of the US presidential voting.

As you can see, the South and Southwest are trending either blue, or mostly neutral, while West Virginia, Iowa, and the Dakotas continue to trend red.  The bluening of Utah might be a surprise to some, but by my estimate, it's voting about 2% (!) more democrat each year.  What will the electoral map look like in 8 years?  Well, there's uncertainty on this, of course, but we can put a stake in a best fit model:

Model for 2028.  Surprise! The rust belt and Florida are still going to be tossups, but on the other hand, TX, GA, AZ and NC are likely to be firmly blue.

In terms of the electoral college, this gives a trendline:**

Modeled electoral college breakdowns for the next two presidential cycles, with errorbars.  Despite increased electoral college inefficiencies, Dems dominance on the national scene will likely cause them to increase electoral college margins moving forward.

This is pretty great news! 

We can also get a sense of how the Senate is headed over the next decade or so.  Notwithstanding Susan Collins (R) in Maine or Joe Manchin (D) in West Virginia, or a few others, for the most part, there's a pretty strong correlation between how a state tilts in the presidential race and senate representation (in 2020, Maine was the only state to vote differently in the two races).  We can use the Trendline widget to see how many states tilt in either direction. For the default national trend values:

Using our simple model, Dems are expected to slowly gain in the senate over the course of the next decade.

But we can also test our assumptions.  Suppose, for instance, that we've been overly optimistic, and that all elections moving forward have Dems winning by even 4 1/2 points nationally, but assume that stays constant:

Dems would maintain a decent chance to hold the presidency throughout the decade – and again, this ignores incumbency advantage and the strengths and weaknesses of individual candidates.  But those complications only seem to matter to the tune of a few points.  The worse news, though, is that Dems could expect a modest decline (while still winning the popular vote) in the senate, with perhaps only 45 or so seats being typical, if we don't continuously increase our popular vote edge. 

Please take a look at the widget, play around, and by all means, send me (or comment below) any and all recommendations for improving it.  Also be sure to follow on Facebook and twitter.

* A note to modelers: yes. 2020 data is included the model, so you can't use  it to model the 2020 result. If you take out 2020 from the data set, the model predicts a slightly higher D-R of 8.7 for 2020.  Higher, but still within the 7 point scatter range of the true value. 

** A Technical note: For 2024 and 2028, I'm using the 2020 electoral college allocation.  Likely, a few seats will change, but I won't update the modeler until the map is finalized.