Do the national polls matter? I heard that question a lot over the past year from readers and reporters who wondered whether we should pay more attention to national presidential primary polls or those from early primary states. For most of the past year, the answer has been easy. If we wanted to follow the status of the campaign, the polls in Iowa and New Hampshire were more important. The voters there were more likely to be engaged in the campaign, and national polls were likely to give a deceptive read until the early states had rendered their verdicts.
New conventional wisdom has grown up over the past two weeks: "Bounce" is a thing of the past and, therefore, Iowa and New Hampshire have lost their disproportionate influence. One can certainly argue that 2008 has not seen a repeat of the pattern established by Al Gore in 2000 and John Kerry in 2004, in which each candidate swept Iowa and New Hampshire and proceeded to run the table in the contests that followed.
But note that as of this writing, the two New Hampshire winners are also the front-runners in national polls. surged to that position as a direct result of his success in New Hampshire, and 's surprise victory in that state helped her maintain her national lead. Although he's still trailing Clinton, has gained roughly 10 percentage points in national surveys in the wake of his victory in the Iowa caucuses. So the notion of primary campaign momentum that boosts the winners of early primaries is still very much with us.
Now, finally, we have reached the stage where the national trial-heat polls matter, with 25 states holding primaries or caucuses on Feb. 5. Survey organizations have been actively polling California, New York and New Jersey, but polls in the other Super Tuesday states have so far been few and far between. As such, the national surveys may be our best guide to the evolving preferences of Feb. 5 voters. But how reliable a guide will they be?
The answer may depend on how national polls select the "primary voters" that answer their presidential trial-heat questions. While media pollsters expend considerable effort on the "likely voter" models they use for the general elections and some early statewide surveys, they tend to take a simpler approach to the primary-vote preference questions asked on national polls. They typically ask their primary trial-heat questions of registered voters that identify or "lean" to a particular party (I catalogued the way national pollsters select primary voters on Pollster.com last summer).
That approach creates a potential problem. Consider the Democratic primaries and caucuses of 2004, in which 16.8 million voters participated -- a number that amounts to 8.3 percent of adults eligible to vote in the fall of 2004. The true primary electorate of 2004 was a small fraction of the 35 percent to 50 percent of adults that typically answer trial-heat questions on national polls, those classified as "Democrats" or "Democrats registered to vote." True, Democratic turnout has been much higher this year than in 2004, but it is not likely to be that much higher.
Would the results of national primary trial-heat results look different if they were based on a narrower definition of "likely primary voters?" When pollsters examined that issue last summer, they found little evidence that a "tighter screen" would make much difference. In August, for example, Gallup's Frank Newport saw "very little difference at the national level in candidate preferences even when we analyze smaller groups of more hard-core voters."
Of course, narrower screens might make more of a difference now. Both McCain and Obama do better in their respective party primaries among voters who think of themselves as independents, so looser screens may work to their benefit, at least in theory. On the other hand, many true primary voters think of themselves as independents (whatever their registration status).
To be fair, the national pollsters face an impossible task in trying to find a one-size-fits-all definition of "likely primary voters" for national surveys. Registration procedures, primary eligibility requirements and historical turnout patterns vary widely from state to state. National surveys may also provide a misleading read of Feb. 5 states that end up on the receiving end of a TV advertising blitz from some candidates, which may produce different vote patterns.
Still, the difference between true primary voters and those who make it through the screen on national surveys may be critical over the next few weeks. Hopefully, national pollsters will give this issue more attention in their next round.