An emerging genre in popular commentary on politics is the use of statistical models to predict election results. Once the domain of academics writing for the scholarly journal P.S., it has become very widespread in recent years. And now, the Washington Post’s Ezra Klein offers up his own model:
Political scientists have long known that you can predict most of what will happen in a presidential election with just a few key pieces of information: how the economy does, for instance, and the incumbent’s approval ratings in the summer. If you have those two numbers — even before you know the opponent, the campaign strategies or the issues — you can usually call the winner…
The final model uses just three pieces of information that have been found to be particularly predictive: economic growth in the year of the election, as measured by the change in gross domestic product during the first three quarters; the president’s approval rating in June; and whether one of the candidates is the incumbent.
That may seem a bit thin. But it calls 12 of the past 16 elections right. The average error in its prediction of the two-party vote share is less than three percentage points...
I find Klein’s model to be particularly unpersuasive, but all these models seem to share a similar problem: they take the blowout elections of 1952, 1956, 1964, 1972, 1980, and 1984 not as historical peculiarities with little relevance to today, but as central tendencies. To put this in plain English, the three variables Klein elaborates (or, for that matter the variables in any model I've ever seen) cannot account for the wide gulf between Eisenhower v. Stevenson and Clinton v. Dole. Those were campaigns waged in different ages, yet the the models never acknowledge that and end up basically forcing square pegs into round holes.
The reason the contests of then versus now are so different has to do in part with the changing nature of the party coalitions. The huge victories won by Ike, LBJ, Nixon, and Reagan all depended in part on their peeling away significant chunks of the opposition’s vote – due not only to the economy, presidential job approval (or whatever), but also the instability of the political alignment for decades after the Great Depression.
What instability am I talking about? Well, consider the difference between the electoral map of 1904 and that of 2004. Where we once saw stark regional lines, we now see much more of a geographical hodge-podge. What happened is that the nation began transitioning from regional (i.e. North v. South) parties to ideological (i.e. liberal v. conservative) parties in the mid-1930s, but the process was slow and tortuous, with both sides being exposed to significant structural weaknesses at varying points. This process really only worked itself out recently, around the midterm of 1994, in fact. This meant that the party coalitions were in flux for more than half a century, and during that period the vote ceiling for a party could be greater than 60 percent while the floor could be under 40 percent, which made for a relatively large playing field.
Today the party coalitions are much more stable, and the battle is fought almost entirely between the 45-yard lines of the field. We have not seen anybody win less than 45 percent in terms of two-party presidential vote in twenty years, and it has only happened once in the House vote (to the GOP in 2008). This means that both sides have secured a solid base of 45 percent, and the range from cycle-to-cycle in terms of two-party vote share is now half of what it once was: the average difference in two-party vote share from 1948 through 1984 was 10.9 percent; since 1988 it has only been 5.8 percent. What’s more, between 2000 and 2008 a total of 10 states voted Republican and Democratic for president at least once, but between 1964 and 1972 forty-three states voted for both sides at least once.
(In Congress, this transition from regional to ideological parties has created the polarization that Beltway pundits regularly bemoan. Really, it is just a consequence of “Democratic” now meaning “liberal” and “Republican” meaning “conservative.” Fifty years ago, that was not necessarily true.)
So I would say that 90 percent of the vote is pretty well set. And this is the biggest reason that I am skeptical of these predictive models -- they usually fail to account for the fact that there were simply more gettable voters for Ike in 1956, LBJ in 1964, Nixon in 1972, or even Reagan in 1984. They assume that a president today can still win 60 percent of the two-party vote -- even though this was a regular occurrence before 1988 but has never happened since. And it has not happened since because the two parties have finally, after years of struggle and back-and-forth, locked down roughly 45 percent apiece.
This is a lesson not just for the wonky backwaters of predictive modeling, but a good lesson for moving forward through this presidential cycle. If the only real swath of persuable voters amounts to maybe 10 percent of the electorate, then we need to be careful in how we look at the horse race. After all, we are talking about a group of people that have virtually no partisan or ideological attachments, pay very little attention to politics, and often create the crazy swings we see in the horse race polls during the course of the cycle. They are at the least fickle and at the worst maddening, as they regularly tell pollsters they have settled opinions when in fact they do not!
That’s why I’m keeping an eye on the fundamentals – rather than the horse race polls – until relatively late in the cycle. My instinct is that this swath of 10 percent or thereabouts is going to “break” late, but they already have pretty well-formed opinions about Barack Obama, especially regarding how he's handling the biggest issues of the day.