Rally Points

Nate Silver says the math doesn't add up when it comes to Republicans backing Trump, in particular conservatives who've long hated him and found him distasteful, or something.

If a realignment is underway, then it poses a big empirical challenge. Presidential elections already suffer from the problem of small sample sizes — one reason a lot of people, certainly including us, shouldn’t have been so dismissive of Trump’s chances early on. Elections held in the midst of political realignments are even rarer, however. The rules of the old regime — the American political party system circa 1980 through 2012 — might not apply in the new one. And yet, it’s those elections that inform both the conventional wisdom and statistical models of American political behavior. 

This doesn’t necessarily mean we’ll be completely in the dark. For one thing, the polls — although there’s reason to be concerned about their condition in the long-term — have been reasonably accurate so far in the primaries. And some of the old rules will still apply. It’s probably fair to guess that Pennsylvania and Ohio will vote similarly, for example. 

Still, one should be careful about one’s assumptions. For instance, the assumption that the parties will rally behind their respective nominees may or may not be reliable. True, recent elections have had very little voting across party lines: 93 percent of Republicans who voted in 2012 supported Romney, for example, despite complaints from the base that he was insufficiently conservative. And in November 2008, some 89 percent of Democrats who voted supported Barack Obama after his long battle with Hillary Clinton.

But we may be entering a new era, and through the broader sweep of American history, there’s sometimes been quite a bit of voting across party lines. The table below reflects, in each election since 1952, what share of a party’s voters voted against their party’s presidential candidate (e.g., a Democrat voting Republican or for a third-party ticket). There’s a lot of fascinating political history embedded in the table, but one theme is that divisive nominations have consequences.

ELECTION	DEMOCRATS	REPUBLICANS
1952	23%	8%
1956	15	4
1960	16	5
1964	13	20
1968	26	14
1972	33	5
1976	20	11
1980	33	15
1984	26	7
1988	17	8
1992	23	27
1996	15	19
2000	13	9
2004	11	7
2008	11	10
2012	8	7

Silver has a mild point.  Reagan Democrats in 1980 and 1984 did make a difference, as did the Dems who jumped ship on McGovern in 72. and those elections certainly broke that mold, but look at the last 4 presidential elections.  

There's very little party-flipping, and what does happen effectively cancels out.

So no, I see something very close to what we've seen before, somewhere around a meager 10% of voters switching up, and it happening on both sides, effectively neutralizing the phenomenon.  90% of voters are going to stick with their party in November.

Another massive Democratic defection like 1972 or 1980 isn't going to happen in our heavily partisan body politic.  Worst case scenario is 1992, where about 25% of voters switched up, but it happened on both sides, and that was with Perot clogging up the works.

I don't think you'll see mass defections on either side.  Too much tribalism.