Long term Presidential election trends (and other stuff)

[Nichlemn]

Another interesting exercise (although probably somewhat more time consuming) would be to combine this with your work on US regions, by doing trend graphs for regions. (Actually, even raw percentages would be interesting for regions).

[Nichlemn]

Also, I like the moving average of four better than a moving average of two (that is, CPVI). I think it would do a better job of predicting Congressional voting, though as a practical matter it would be difficult to calculate for most CDs (e.g. needing to work out 2012 numbers on 1990s Census lines).

[Antonio the Sixth]

I intend to start working soon on compiling results for US regions, but that will certainly take some time. In the meantime, I have a few stuff that migh still interest you.

[All Along The Watchtower]

Good work! Virginia is interesting, there seems to be myth that it has gone from "Solid R" to "swing" recently, but looking at the graph there's not much change, with VA being about the same as it was in the 80s.

Let's be fair here. If we exclude Carter's years, Virginia has been a fairly republican State since the 1960s, and the dem trend started after 1996 is pretty outstanding. Even if we count Carter, Obama actually beat him in 2008 ! Quite a performance if you compare with the trends in places like Arkansas or West Virginia...

Virginia has NoVA though, which has grown dramatically in the past few decades. The Republicans get most of "Real Virginia", the Democrats get Arlington, Alexandria,and Fairfax County,  and Norfolk and Newport News in the South, and Loudoun and Prince William are swing counties (especially Loudoun, which voted for Obama 53-45). I imagine in those three wealthy counties, you have a lot of people who are socially liberal, so they lean Democratic on the one hand, but on the other hand they don't want to pay higher taxes, which makes them tilt Republican.

[Nichlemn]

I've decided I prefer exponential smoothing to a moving average, partially because I think it gives a more accurate picture and partially because it's more difficult to work out intuitively, so it'd be more interesting.

I think α (the weighting given to the most recent result) should be about 1/3. That means ~55% of the weighting is given to the past two elections, ~80% to the past four and ~90% to the past six.

[Nichlemn]

And another thing - I like the idea of halving the margin of victory, to make it more easily comparable with CPVI.

[Nichlemn]

Here's my attempt at graphing New Jersey by the above rules:



"Half of Democratic margin" seems weird and arbitrary, but "difference in Democratic percentages" isn't accurate for these numbers.

I picked New Jersey because I noticed your graph had a moving average Dem trend from 2004 to 2008, despite the Republican trend from 2000-2008. With the exponential smoothing, there is still a Democratic trend but it's only 0.25 percentage points. Also, your graph has a more or less constant Democratic trend since 1992, while this one shows it tapering off after 2000, which makes sense.

We can also see how the state gets credit for its Republican past, as the 2008 PVI was still more Democratic than the trend.

[Antonio the Sixth]

That's certainly interesting. Indeed the problem of moving average is that there are always arbitrary : their swing is determined by the difference between the present election and the 5th last election. For example, in 2008, the moving average suddenly ceased to take into account 1992 (when Clinton performed quite poorly in NJ), giving the impression that the situation is improving for democrats there.

I encourage you to keep making graphs like this one. As for me, I'll keep with the classical way because I don't want this to get too complicated (I have an excel feature that authomatically generates trendlines, but it doesn't have the function you want to use).

[Nichlemn]

One issue is that I don't have the relative margins for states you haven't made graphs of. So, to save me from having to calculate them myself, could I please borrow your numbers?

I'd like do:

Arkansas
Georgia
Utah

The first because I'm curious to see if Clinton's numbers give it a reasonable-looking PVI, the latter two because they're states that I'm suspicious about the alleged existence of Democratic trends in.

[Antonio the Sixth]

Of course. You can ask for whatever graph you want, I'll be happy to show you it. Here are those you wanted :



Good but not great performance for Clinton, especially if you compare to Carter just 16 years before. Also interesting to see how Obama got utterly crushed there, whereas Kerry and Gore got pretty decent results there. It seems like Clinton's legacy survived for almost a decade before suddenly disappearing.




Definitely no long-term dem trend by any way. Georgia is more or less stable since 1988, with only a dem peak in 1992. If there will be a democratic trend, it hasn't begun yet.




Democrats clearly improved their showing in 2008, but like in the rest of the inner West (see the 1st page of the thread), this is more of a recovery from the dem's demise in 2000. That's why Utah is still more republican than it used to be in the 1990s.
Interestingly, Utah has been the most republican State (in terms of relative margin) in every election from 1976 to 2004, and in 2008 it has been outpaced by Wyoming and Oklahoma.


Here are two others that interested me.



Pretty incredible stability : PA has almost always been a D-leaning State since 1952 (with a few peaks like 1964 and 1984, and a few flirts with the GOP like 1976 and 1996)




I'll always be stunned by Dukakis' performance there : IA was its second-best State, after RI. Since 1992, it is mostly a swing State with a slight dem tilt. Note that Obama was expected to perform far better in Iowa than he eventually did (Nate Silver had it 5.7 points over the national margin).

[Nichlemn]

That's nice and all, but what I really wanted was the raw numbers so I could copy and paste them into Excel .

[Penelope]

Can I get trend maps from 1932-1968, and 1968-2008?

[Antonio the Sixth]

Can I get trend maps from 1932-1968, and 1968-2008?

1932-1968 :

[Antonio the Sixth]

1968-2008 :

[🐒Gods of Prosperity🔱🐲💸]

Impressive stuff here, Antonio. I'm enjoying it. I'd be interested in seeing a graph for NH.

[Mr. Illini]

Who are the Northeasterners that voted for Gore '00 but not Obama '08?

[I Will Not Be Wrong]

Who are the Northeasterners that voted for Gore '00 but not Obama '08?

Those in western PA, and in Litchfield County, CT, and some in Massachusetts.

[Antonio the Sixth]

Bump! Here are a couple maps based on the most recent election.

2004-2012

2000-2012

[Antonio the Sixth]

1996-2012

[Antonio the Sixth]

1992-2012

1988-2012