About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« Another Average (At Best) Month for the RNC | HorseRaceBlog Home Page | Quote of the Year »

On the Bouncing Party ID Numbers

Gallup put out its latest numbers on party identification, and the results might have surprised some.

Gallup Partisanship.gif

Check out how bouncy partisan identification is in the Gallup numbers - not just in the picture, but in the entire history going back through 2004. Next, compare these numbers to the exit polls on Republican identifiers since 1972.

Republican Voters By Year.jpg

The GOP suffered a decline in self-identified Republicans in 1976, as Watergate and the recession pushed people to Jimmy Carter and the Democrats. They came back some in 1980, but it was not until 1984 that the GOP rebounded to the level it enjoyed in 1972, when Nixon was re-elected. From 1984 to 2004, it was pretty stable, during good years and bad for the party - but in 2008 there was a five point drop-off. This is less than what the party suffered in 1972-1976, but it is still the biggest decline in party identification for the GOP in more than thirty years.

So, the movement you'll see in 25 years of exit polling can be found in just 25 days of Gallup polling. What are the implications of this contrast?

First, I think it's a sign that the conventional wisdom that the GOP has shrunk dramatically since Election Day has been oversold. Of course, I thought that before I saw these numbers. Pundits pushing that story line were inclined to cite the AP, Pew, or CBS/NYT poll - even though Rasmussen, Survey USA and Fox News had shown insubstantial drop-offs in party identification relative to the 2008 exit poll.

Second, I do not think the party's recent improvement in Gallup is a sign that the GOP is "on the mend." Return to the exit polling, and note how little it bounces around. This implies that the 5-point drop the party suffered between 2004 and 2008 - while small compared to the media polling - cannot easily be overcome. It took Reagan's 1984 landslide to undo the damage done to Republican self-identification by Watergate. Five points in a May Gallup survey of adults might be easy to win back, but who cares? Elections are held in November and are decided by voters. They swing much less, implying that any drops a party suffers with them could be lasting.

Third, I think this highlights the difference between voters and non-voters. It wasn't too long ago that the Democrats had an eye-poppingly large lead in party identification. Now, Gallup has another eye-popping result, this time showing the two at parity. To connect the two divergent results, there's been a radical "correction" in the party ID numbers even though very little has happened in the political world.

A lot of this could be chalked up to the presence of non-voters in the Gallup surveys. The latter are excluded from the exit polls, but they should make up about 40% of the Gallup samples. Non-voters have a lot of important differences with voters. They are less likely to express much interest in politics, watch political programs, read newspapers frequently, pay much attention in general, and so on. Additionally, non-voters are more likely to place themselves in the middle of an ideological scale whereas you're likely to find voters clustering around the conservative and liberal poles.

Consider the following graphs. They depict the "feeling thermometers" for George W. Bush among voters and non-voters, according to the 2004 National Election Study. This was conducted in 2004, so obviously feelings for President Bush have changed quite a bit since then. The intent here is simply to illustrate a point about the differences between voters and non-voters. The feeling thermometer is basically a question that asks respondents to rate a person, organization, or idea on a 0-100 scale, with 0 being negative and 100 being positive.

Feelings for Bush.jpg

As you can see, voters in 2004 had highly polarized views of George W. Bush. However, non-voters inhabited the middle range, with a slight tilt toward the positive. You'll note also that fewer non-voters were willing to give an opinion on the matter. This graph does not show it - but individual voters who placed Bush high on the thermometer were likely to place Kerry low (and vice-versa) while many non-voters were inclined to put both in the middle.

All in all, non-voters often lack strong feelings on political subjects (be they ideological or personality-based), and they are also less likely to pay close attention to politics. This might make a difference in the polls on party identification. Recall the generally positive coverage of the President's first hundred days, according to the Center for Media and Public Affairs:

Fifty-eight percent of the Obama evaluations were positive on the ABC, CBS and NBC broadcasts, compared with 33 percent positive in the comparable period of Bush's tenure and 44 percent positive for Clinton. (Evaluations by officials from the administration or political parties were not counted.)

On Fox News, by contrast, only 13 percent of the assessments of Obama were positive on the first half of Bret Baier's "Special Report," which most resembles a newscast. The president got far better treatment in the New York Times, where 73 percent of the assessments in front-page pieces were positive.

A striking contrast: Obama's personal qualities drew more favorable coverage than his policies, with 32 percent of the sound bites positive on CBS, 31 percent positive on NBC and 8 percent positive on Fox.

So, the little information non-voters acquire about politics has thus been quite partial to the Democratic party in recent months. Combine this with their their lack of strong feelings about many political subjects - then toss in the GOP's inability to control the agenda in Washington - and this could take us a long way to explaining the drop-off for the GOP among non-voters in some previous surveys. Meanwhile, as the new car smell of Obama's presidency has faded, perhaps those otherwise inclined to the GOP have returned to calling themselves Republican (although I'd note that the recent Fox poll, of registered voters, shows no change from the previous one).

Before I get angry emails accusing me of irrational GOP bullishness, let me repeat: I think the electoral implications of these trends are basically nil. I have virtually no interest in the week-to-week movement of party identification - be they good for the Democrats, the Republicans, the Communists, or the Martians. I'm only discussing it today because other people have been talking about it.

There is academic or scholarly benefit in understanding how non-voters view themselves on the partisanship scale - but the electoral consequences are, by definition, zero. Ditto for actual voters, so long as we're some 19 months from the next election. I don't think it matters much what they call themselves now. It matters in November of an election year - after the two parties have spent hundreds of millions of dollars on lavish conventions and television ads to dominate the political discourse and bring their wayward partisans back to the fold. Until then, you have to expect partisanship to bounce up, then bounce down, then up, then down. Is it really a shock that during a Democratic President's honeymoon period, the numbers for the GOP would bounce especially low? Before any dramatic conclusions are drawn from one particular bounce, it is vital to remember that, in the last 25 years, the exit poll numbers on Election Day have always landed within a very narrow band of results.

The real problem for the GOP now is not that the previous AP poll shows it at less than 20% of respondents - although that extreme result makes for good copy for members of the D.C. punditocracy. The real problem is that the last exit poll showed it at 32%, down 5 points from 2004. That puts its party identification at the low end of the historical range, and previous years indicate that it might take several cycles to overcome such a drop-off.

-Jay Cost