American Elections, Part 2: Counter-Cyclical Movements

(I said they would be irregularly spaced!)

Imagine you're a state. Imagine, moreover, that you are a "safe state" for one of the two parties. And in Election Year X, you dutifully cast upwards of 60% of your popular votes for the candidate of that party, only to watch helplessly as your party's lousy candidate gets just 45% of the national vote. But then, four years later, rejoice! Your party has found a much stronger candidate, the incumbent is disliked, and your party wins the national popular vote 53-47! That's a net 16-point swing nationally. And like the rest of the nation, you're excited by this new stronger candidate, and you give even more of your vote to that candidate than you did last time.

Except, you don't give a net 16 points more of your vote to that candidate.
Over the past eighty years, there have been 12 elections in which the national popular vote swung more than 9.7% (since that's what Obama did) toward one of the two parties. I examined all but two of them (1940 and 1952) for the correlation between the relative margin in each state in the previous election and the Trend in that state in the big-swing election. And here's what I found. On average, in these big-change elections, one extra point of margin the last time around nets your party -0.4% of Trend. (That's slightly rough-estimate, since I include a couple of elections both with DC included and without; it tends to skew the line, being so effing far toward the limits of possibility.)

That's what I mean by counter-cyclical: when there's a big political cycle happening, it erodes the relative standing of the winning party's strongholds. That is to say, in big Democratic years the pre-existingly blue states contribute less than their fair share to the Democratic shift, and vice versa. To consider why this is so, take Washington D.C.. The absolute margin in Washington, D.C. has, since it started voting in Presidential elections, stayed within the rather narrow bounds of +56.5% in 1972 and +86.0% in 2008. Excluding these two, which are outliers, the range is from 61.5% to 79.9%. The percent Democratic vote in D.C. has ranged from 74.9% to 92.5%. That's a very narrow window. During that same time, the national popular vote has swung from 61.1% Democratic, 1964, to 37.5% Democratic, 1972.

Barack Obama won Washington, D.C. with 92.5% of the vote in 2008. Now consider the following: if the economy is getting better, the war in Afghanistan is going relatively well, and the Republicans nominate Sarah Palin for President, Obama could easily post rather large gains on his 2008 performance. Indeed, for a President to be re-elected by a narrower margin than his first election is quite rare, having happened only in 1812 and 1916. But there is a hard limit to how much Obama can improve in Washington, D.C. by! He's up against the 100% boundary! You just can't get 101% of the vote!

More generally, I think it's hard to win a state by a lot. A "landslide" is, by and large, something in the 60%-70% range. It's damn hard to move beyond there. Hell, in Hawaii, where the voting patterns suggest somewhat of a personality cult, Obama only got 71%. It's just plain hard to win a state by truly massive margins. So when a seriously blue state gets caught up in a serious blue wave, it's up against some serious drag. It's a lot harder, I think (and the evidence seems to back me up on this), to go from 60% to 70% of the vote than to go from 45% to 55%.

Which relates to what I said about FDR's 1932 performance in the Southern states. He made back a lot of the ground that Hoover had taken in 1928, but not all of it. Well, yeah. As it was he won 92.7% of the vote over the four Gulf Coast states. Think about that for a while: 92.7%. If Roosevelt had made up all the ground Al Smith lost the Democrats in 1928 in those states, he'd've had to win 103.9% of the vote. It was literally impossible for him to make up all that ground. In point of fact he won a considerably higher percentage of the vote in the Southern states than Davis had done in 1924. But he was running 43 net points ahead of Davis' 1924 performance nationally, and there just wasn't that much room left in the South.

Now, all of this raises the question of how party strongholds ever develop. After all, whenever there's a big swing each party's strongholds seem to be regressing to the mean. Well, we'd like to find that in non-swing elections the opposite happens, and in fact this is the case. In fact, in the 1996 election, in which the national vote swung 2.9% toward Clinton compared to 1992, each extra point of Clinton vote in '92 was worth an extra 0.29% of Trend in '96, excluding D.C. (in 2004, a -3% swing from 2000, the correlation was slightly negative and essentially nill). Results from similarly static elections prior to this do not seem to confirm the hypothesis, though: in 1956 and 1948 the negative correlation still applied, though (if one removes the Strom Effect states from 1948) weaker than in most of the big swing years I've seen. I think there's an explanation for this, though.

Remember, the entire period from 1928 through about 1992 was one in which the two parties' electoral strongholds switched. We should expect that there was a lot of parties losing relative votes in their respective strongholds over such a period, even when the nation as a whole looked fairly static. That 1996 followed the pattern I would expect makes me think I might be on to something; if 2012 is a fairly similar overall result to 2008 I'll be mighty curious to see if the same holds then.

Oh, and one final side note as to the reason why I'm using relative margins instead of absolute margins. Basically, it's difficult to see the real trends in the underlying electoral landscape if you just look at overall results and don't filter out the national margins. The effect of alternating landslides and massive shifts is that the rather steady progression of the last eighty years, the South's switching from alarmingly Democratic to solidly Republican while the Northeast and Pacific Coast have gone from swing regions to solidly Democratic, keeps getting obscured. And we see, for instance, Kennedy winning many Southern states, when in reality the 1960 election saw the Democrats losing a whole lot of strength in the South.

There's a further justification, too: there's a lot of political science evidence that national overall results are about 75% a function of "fundamentals," like the economy and military conflicts. If so much of the election is almost mechanistically predetermined by the underlying situation of an election, then most of the real action in a given election is going to be in how the states move relative to one another. That is, if what changes between a 1964 and a 1972 isn't really ideology or anything like that but rather just how the world is doing, it makes a lot more sense to look at what the political landscape looks like if you ignore the effect of those fundamentals and just focus on relative strengths.