Where does the electoral pendulum swing? Ask the voters who stayed home last time

The greatest obstacle that Hillary Clinton faces is winning a third consecutive term for the same party. In the last post we saw that everyone who pulled that off was an incumbent in the outgoing administration, and a high-ranking member like President, VP, etc. Hillary is a non-incumbent, which hinders her serving as a bridge of continuity, and she was Secretary of State, a role that has not been a launching pad toward the Presidency since the early 1800s, when it was of greater prestige than the VP.

Now we're going to look more at the idea of momentum that carries the same party from winning one election to the next. The basic idea is that the party has become more popular from the first to the second, an upward trend that could deliver victory in the third election.

The common way to measure this is by looking at the share of the popular vote -- if the party got 51% the first time and rose to 57% the second time, they seem to be on a roll. If they won the first time with 57% and fell to 51% in their second win, that suggests their popularity is evaporating and will not last to secure a third win.

But it turns out this is not entirely accurate. I'll be looking only at the elections back to 1972, because the data-set that the new model is based on only goes back that far. We'll just look at who won the popular vote, since that usually means they won the election. When we say "won," we mean the popular vote.

There have been 6 elections back through 1972 that were the second term of the same party, which allows us to see if they went on to win a third time. How good of a prediction is "gaining share of popular vote"?

In 1972, the Nixon landslide was a major improvement over his narrow win in '68. That suggests soaring popularity for Republicans, which ought to have led them to victory again in '76 -- but the Democrats won instead.

In 1984, the Reagan landslide was up even higher than his earlier landslide in '80. This predicts a win for Republicans in '88, and that did in fact happen.

In 1988, Bush's popular vote was lower than Reagan's in '84, and about what it was in '80 (adjusting for the third party vote). His Electoral College win was smaller than either of Reagan's. That predicts a loss for the fourth election in '92 -- and they did indeed lose it to the Democrats.

In 1996, Clinton's re-election improved from his first win, suggesting a third win in '00 -- and Gore did indeed win the popular vote (though not the EC).

In 2000, Gore's popular vote was down from Clinton's in '96, and far lower in the EC, suggesting that the Democrats would lose the next time -- and in '04 it was Bush who won the popular vote.

In 2004, Bush improved on his showing in '00, predicting a third term in '08 -- yet the Republicans got blown out.

So, the measure works in 4 out of 6 cases, which isn't bad but isn't great either.

For what it's worth, Obama declined in the popular vote (and EC) from '08 to '12, predicting a loss for Democrats in '16.

I stumbled upon an even better measure by investigating the behavior and the effects of voters who are "new" to a party and therefore represent changing directions driven by those who are not partisan loyalists -- having crossed over from the other party, having sat the last election out, or having been too young to vote last time. I'll be writing much more on these topics, but for now let's look at how they predict what happens in the attempt at a third consecutive win for the party.

It turns out that we can predict the fate of that third attempt by looking at the second win, and ask how the "infrequent" voters leaned -- meaning those who could have voted in the last election, but did not, and are now getting off the sidelines to participate. Something motivated them beyond typical partisan loyalty, and that something could be a sign of which way the wind is blowing.

We're using the General Social Survey because it allows us to get the voting behavior of a person across two separate elections.

We'll start with the level of support that, say, the Democrats enjoyed among the "frequent" voters -- those who voted in the last election, and the current one as well. That captures where the zeitgeist already is, whereas the infrequent voters will tell us where it is headed. So we take the difference in support levels for the Democrats, comparing the infrequent against the frequent voters. Compared to the usual voters, how much more Democrat-leaning are the infrequent voters?

For example, in 1972 the frequent voters went 35% for the Democrat, while the infrequent voters went 52%, for a difference of 17 points favoring the Democrat among the infrequents. The graph below shows how Democrat-biased the infrequent voters were, compared to the frequents, where negative values mean they were Republican-biased compared to the frequents.

Let's look at those 6 elections again, and see how this measure of momentum performs. It makes the same correct predictions as the "increasing popular vote" measure does for '84 (R in '88), '88 (D in '92), '96 (D in '00), and '00 (R in '04).

But in the two that the crude measure missed -- '72 and '04 -- it correctly identifies flagging momentum. In '72, the share of the popular vote may have shot up for the Republican, but those who came off of the sidelines actually favored the Democrat (and in an election when he would get demolished). That predicted the failure of the third attempt in '76.

Later in '04, when Bush had improved his popular vote, the voters who had sat it out in '00 decided that they couldn't sit idly by this time, and joined in the effort to dump Bush. As in '72, they failed at their current task, but this counter-movement showed the way forward, and the Republican would lose big-time in the attempted third term in '08.

For what it's worth, this model shows a loss of momentum between Obama's first and second wins. Those who sat it out in '08, but showed up in '12, were more Republican-leaning than the frequent voters. That suggests discontent with the Democrat administration already by their second win, making a third win all but impossible.

This model does not perfectly predict whether a first win will lead to a second win. For example, the infrequent voters in '76 were Democrat-leaning compared to frequents, yet they lost massively in Carter's re-election bid in '80. The infrequents in Reagan's first win were also Democrat-leaning, yet he won re-election handily in '84.

Still, in 8 of the 10 elections, the behavior of infrequents does predict who wins the next time. But those bungled predictions from '76 and '80 are rather damning failures -- missing the Reagan landslides in '80 and '84.

Why do infrequents perfectly predict from a second to a third, or a third to a fourth, but not necessarily from a first to a second? I think these folks who tend to sit things out will let the first term of the party have a chance to do well. After having four years of results to judge them by, they'll come out and give a referendum.

Remember, this does not block the incumbent party right away -- they were biased toward McGovern in '72, Dukakis in '88, and Kerry in '04. Instead, it reveals a stalling momentum, that the people who generally pay little mind to politics are upset enough with the status quo that they're actually participating this time.

On the other hand, if these infrequents make a point of leaving the house this time around just to support the incumbent party, that means even the election-averse citizens are eager for more of the same, and the party has enough momentum to win another term the next time. The infrequents were biased toward Reagan in '84 and Clinton in '96, which heralded popular vote wins for the incumbent party in '88 and '00.

For predicting the 2016 outcome, both measures -- trend in popular vote and bias of infrequent voters -- agree that there is no momentum going from '08 to '12, putting a third win out of reach for the Democrats.

GSS variables: presYY (year), voteYY (year)