American Elections, Part 1: Personality Effects, Debt, and Stickiness

This is the beginning of an undefinedly-long and irregularly-spaced series on my analysis of state-by-state voting patterns in American Presidential elections. As the title mentions, this installment will discuss personality effects, debt, and stickiness, which will be discussed later. First, a brief overview of what data I have and how I've been manipulating it.

This started almost entirely by accident. I wanted to figure out the relationship between how high a percentage of the electorate a Presidential candidate take and how large the variance in their performance across states. To do this, I ended up copying down the state-by-state popular vote data for every state and every candidate (with approximately 1% or more of the national vote) since 1828, the start of the popular vote in earnest, from Dave Leip's elections atlas site, which is truly wonderful. Some time later, I realized that I had just created a Microsoft Excel document with the entire history of US Presidential elections on it, to the hundredth of a percent, and that this was a gold mine.

That Excel document now has fifteen worksheets in it. "Total," the first one, is what I copied directly, the vote percentages of each candidate in each state since 1828. "Democrats," "Republicans," and "Whigs," respectively, are the data for just the Democratic, Republican, and Whig candidates; no shocker there. "Total Votes" is, again rather obviously, the number of total votes cast for President in a given state in a given year; "Vote %" is the share of the total national electorate that each state had in a given year. Two others relate to an attempt by me to figure out, for instance, how many people from each era of American politics have cast their votes for Democrats in Alabama. Then comes the interesting part. "Margins" is fairly simple: it's the Democratic margin of victory in a given state in a given year, in percentage terms. Sometimes when third parties get involved things get tricky, but it's a fairly simple concept. "Relative Margins" for a given state is the Margin in that state minus the national popular vote margin. So, for instance, if Candidate D wins State X by 10% and nationally by 3%, their Relative Margin in State X is 7%. This is about seeing where the strongholds of a party are, filtering out the actual balance of the electorate. "Swing" is fairly simply defined as the change in the Margin in a state between two elections. "Trend" (and I'm taking these terms from Dave Leip's site) is the change in the Relative Margin in a state between two elections. "Rolling Average" is the average Relative Margin in a state in the five Presidential elections prior to a given year, and "Trend Adjusted" is a state's Relative Margin minus its Rolling Average. The point of Trend and Adjusted Trend is to see a) how the party strongholds are changing over time, filtering out who's winning elections, and b) to see how well each candidate did in various states relative to the approximate recent status quo heading into that election. For various reasons, which have a lot to do with personality effects, debt, and stickiness, I found that the adjustment was useful for Point b).
So, what the devil are personality effects, debt, and stickiness? Well, there's a sort of logical and causal order, so I'll address them in that order. Personality effects are when the Trend reflects not the actual long-term trend in party regional strengths but a short-term fluctuation from that trend based on who the candidates are. Classic example: the major trend of the 20th century was flipping the two parties, moving the Democrats from South to North and the Republicans from North to South. But in the 1976 Presidential election, the Democrats posted truly monumental Trends in their favor in the Deep South, in fact the strongest and most broad-based Trends toward the Democrats in the South perhaps ever. Why did this long-term trend reverse itself so powerfully in 1976? Well, the answer is obvious, of course: Jimmy Carter. Democrats nominated a centrist Southerner who farmed peanuts and liked to talk about his religion, Republicans nominated a moderate from Michigan, and the South voted for its favorite son. That did nothing whatsoever to change the long-term trends in the South.

So what is debt? Well, it's sort of related to personality effects. Debt is the idea that when a party has posted a particularly strong Trend in a given state in a given year, it owes a sort of debt to the other party, and is going to end up having to pay that debt back. This is especially true when the strong Trend is not an actual long-term trend but rather a short-term personality effect. In the case of Jimmy Carter, this is seen in the fact that, with Carter still on the ballot in 1980, the South trended distinctly toward Reagan, especially Carter's home state of Georgia, and in 1984 when Mondale replaced Carter the South shifted violently toward Reagan again. Carter borrowed the Democrats a whole lot of points in the South in 1976. They owed them back to the Republicans. And they paid up.

This is a phenomenon that I think happens in general and not just when an acute personality effect has distorted a longer-term trend. My analysis shows that the trend in Trend (yeah, I know, it's getting too abstract; I think the results are interesting), that is to say the change in a state's Trend between two consecutive Presidential elections, has a tendency to play ping-pong. Specifically I can state that across the 1828 or so consecutive-election pairs from the 50 states plus DC since 1828 (coincidence!), 1090 (59.6%) have involved trends in the opposite direction, and just 40.3% have involved trends in the same direction. So when a party Trends in a given state in one year, it has a distinct tendency to give some of that Trend back the next year.

And as for stickiness? Well, basically that's the idea that debt is often not all payed off immediately, and sometimes all of the debt is never paid off. In 1924 the Gulf Coast states (AL, MS, LA, TX) were 90 net percentage-points more Democratic than the nation as a whole and the South Coast states (VA, NC, SC, GA, FL) 57 percentage points more favorable to Democrats, with southerner John W. Davis being clobbered by northerner Calvin Coolidge nationally but dominating the South anyway. In 1928 Herbert Hoover invented the Southern Strategy forty years ahead of his time, shifting these regions to only 36% and 14.6% more Democratic than the nation as a whole. Those were Trends of 54% and 42%, respectively, in these two regions, truly massive shifts. (Indeed, my main hypothesis about the history of American politics is that Hoover's Southern Strategy and Al Smith's relative success in the cities, and not FDR, set in motion the swapping of the two parties.)

But anyway, then 1932 hit, and Franklin Delano Roosevelt swept into office. And all of a sudden Hoover's Southern Strategy didn't seem so important to the Southerners; they were poor, they were hurting, and Hoover wasn't doing anything to help them, so they reverted to their old Democratic ways. In 1932 the Gulf Coast states were 67.6% more pro-Democratic than the nation and the South Coast states were 34.2% more pro-Democratic than the nation. Those were Trends of 32% and 20% back toward the Democrats. Those are big, impressive Trends, especially across a whole region as opposed to just one state. But notice that after the massive swing toward the Republicans in the South in 1928 and the paying of that debt in 1932, the Republicans had an overall trend of 22% in both of these regions. The shift from 1928 was sticky. In fact, of course, it was so sticky and ended up so fundamentally changing American politics that the Deep South has never gotten any closer to where it was in 1924 than it was in 1932, and now it is a Republican stronghold.

A similar thing happened with Jimmy Carter, though. In 1972 these two regions, Gulf and South Coasts, were at -17.2% and -19.8% Relative Margins. In 1976 Carter's arrival shifted them dramatically, to +3.3% and +8.5%, Trends of 20.5% and 28% toward the Democrats. Over the next few cycles, this debt was paid off, with the South Coast going back down to +2.9% in 1980, -8.1% in 1984, and -13.0% in 1988 (from where it has since slightly rebounded) and the Gulf Coast dropping to +0.9% in 1980 and -8.1% in 1984, from where it has since steadily dropped to a current 2008 level of -24.1%. So what's the point? Carter's arrival certainly did not usher in an era of dramatic change in the shape of the American political landscape; the South was solidly Republican until, ironically, the Democrats had the balls to nominate a black guy. But nevertheless, the personality effects from Carter were very sticky. The South Coast states have never gotten as pro-Republican as they were in 1972, arguably due to the emergence of Florida as a swing state, and it took the Gulf Coast states until 2000 to get back below -17.2%. A candidate's powerful personal boost to their party in a given state will not last in its entirety, and some of the debt will need to be payed back immediately, but the party will have a pretty good chance of holding on to much of those gains for many cycles thereafter.

(Arguably I'm being slightly unfair to the Southern states' debt-paying of 1932, as I will make clear in my next post on Counter-Cyclical Movements.)