About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« Obama and the Celebrity Culture | HorseRaceBlog Home Page | Here We Go Steelers, Here We Go! »

Gallup on Nationwide Partisanship

Over the course of 2008, Gallup conducted an enormous number of interviews with voters nationwide. Such a large dataset could be used for more than just tracking the horse race, and Gallup has begun to deploy it for a broader purpose. Today, they have published an article on partisanship in the 50 states based on all of their public polling conducted through 2008.

First off, thanks to Gallup for using this data for something more than tracking changes in Obama-McCain in June. Hopefully, we'll see more publications like this.

So, what's the upshot of Gallup's findings? Unsurprisingly, they find that the country has moved left. Below is a reproduction of their partisanship results from 2002, 2006, and 2008.

Gallup Partisanship 2002-2008.jpg

Clearly, the shift is uniform and not insignificant. Gallup spends a good deal of time discussing this, and I won't repeat their key findings. I encourage you to read their write-up carefully.

Instead, I want to focus on two points.

First, this should serve as a cautionary note for those with a habit of finding new, permanent majorities in recent election results. As these maps make clear, it does not take long for partisan identification to shift one way then back again.

That's not to say that any given election isn't the starting point of a permanent shift - it's just that it takes time to differentiate it from movement generated by the mood of the country. For instance, back in 1953, one could have examined the elections of 1948 and 1952 and determined that the GOP was reestablishing permanent majority status in the Northeast, and breaking up the Democrats' permanent majority in the South. Only one of those statements would have turned out to be correct - and you wouldn't know one way or the other until you had several more elections go by.

The bottom line is that there is an underlying stability to partisanship that can shift over time - as we have seen, for instance, in Connecticut and Mississippi in the last 80 years. However, partisanship is a more complicated concept. People can shift their partisan orientation because of the national mood - so that when that mood changes again, so also does the partisanship. These pictures make that clear.

Second point. These pictures offer a warning about interpreting public opinion polling. Relative to election results, there appears to be a bias in their partisanship data. I don't think it's sufficient to say that it's a pro-Democratic bias, as one might infer from just examining the 2008 results. Instead, it might be better to say that it's a pro-majority party bias. In other words, this data overstates the electoral power of the majority party.

For its part, Gallup sees this problem, too. They have a sensible explanation that is worth breaking down into smaller pieces:

There are several reasons for possible disparities between the party affiliation data and the voting outcomes in a given state. First, turnout has typically been an equalizer in U.S. electoral politics because Democrats almost always have an advantage in identification, but Republicans have been competitive in national and state elections over the last three decades because Republicans are usually more likely than Democrats to vote.

This is a good point, but it wouldn't account for some of their 2002 findings. After all, Gallup has persistently Democratic states like Michigan and Vermont (!) leaning to the GOP that year. They have the Democratic advantage at less than 5% in states like Pennsylvania and New Jersey. We can't explain these by virtue of a Democratic bias. That's why I would suggest a majority party bias in the results.

Second, one's partisan leaning is not a perfect predictor of voting in a presidential election, in which candidate-specific characteristics can influence a voter's choice.

This is true, but we need to expand on it. Partisan defections are not uniform, and this points to a fundamental point about our two-party system.

Look at the states in what the Census Bureau calls the South Central divisions (Oklahoma, Texas, Arkansas, Louisiana, Kentucky, Tennessee, Mississippi, and Alabama). They are not the most Republican-leaning states on the map when it comes to partisanship, but in the last few cycles they have been when it comes to votes. Why? White Democrats in these states have been highly prone to defection in recent cycles, as Sean Trende and I argue here. In the South Central divisions, McCain won 33% of white Democrats - including a whopping 60% in Louisiana.

This indicates an underlying reality about the Democratic Party. It is a very broad political party. I personally doubt that it has been smaller than the GOP at any point since 1932 (regardless of what polling data might say for 2002-2004). That enables it to compete in congressional elections in most districts. In most places, there is a solid core of people who are amenable to the Party of Jackson. This is why Maxine Waters and Travis Childers are in the same caucus.

However, breadth carries with it political problems in a diverse Republic such as ours. Namely, it is difficult for the national party to craft issue positions and emphases that appeal to all Democrats. The same goes for national candidates. Take Barack Obama for instance. He won a smashing nationwide victory, the largest we have seen in 20 years. Yet he could not hold all quadrants of his party's voters, losing large portions of self-identified Democrats in the South.

It's a big party that is difficult to unite - which in turn enables the GOP to win handily states that Gallup identifies as solidly Democratic. This is why Arkansas, Kentucky, and West Virginia have, in recent years, voted strongly Republican, but still exhibit strong partisan ties to the Democratic Party.

Third, the party affiliation data reported here cover all of 2008, while presidential election voting was limited to Nov. 4 or the weeks leading up to it.

The idea behind this argument is that partisanship shifts through the course of the campaign, as many political opinions might. Respondents at the beginning of the cycle are more partial to one party, but become less so as the campaign goes on - so that ultimately there is a difference between June polls and November votes.

I wholeheartedly agree, and I have consistently argued against using polling data from the spring or summer. Polls done significantly before an election have precious little value, and I think polling analysts have a bad habit of overemphasizing them. Gallup shoulders a good share of this blame, for they have contributed to this overuse. What is the need for a daily tracking poll in June, for goodness sake? What possible value can it have? I can think of no reason except that political junkies demand it, and they drive web traffic and thus advertising dollars.

I can appreciate taking periodic tests of public opinion so we can get a general sense of the national mood, but that daily tracking poll gave the wrong impression about the value of those data points. The implication is that there is a reason to track day-to-day changes when there simply isn't. This had the effect of spoiling political analysis in the summer, I think, as analysts were too dependent on the numbers.

So, when the next cycle rolls around, and you find yourself obsessively checking Gallup's new daily tracking poll at 1 PM in the middle of the summer - just remember what Gallup told you today: polls taken in June don't necessarily mean much for elections held in November!

-Jay Cost