Despite his repeated assertions that he isn't a candidate his supporters keep hoping he'll enter the race. In the national polls he has consistently been in 3rd place, ahead of Edwards, and winning about 15% of the vote as of recently. But he says he isn't a candidate. So what should pollsters do about including him or not in the polls?
Whenever a survey question offers an added choice, that option must win at least some votes. In the primary contest, a well known and popular (among Democrats) figure like Gore is likely to attract support from voters not happy with the top candidates and not familiar with the bottom tier. The result is that Gore runs well even as he is not running.
On the Republican side, this issue has been somewhat less of a problem. Newt Gingrich and Fred Thompson have both made clear they are "considering" runs making the rationale for including much more compelling.
Some polls have asked for second choices and reallocated the Gore vote to those second choices as a way of measuring the race both with and without Gore. But the practice is quite inconsistent and many polls report only results with Gore. A smaller number report polls only without Gore. The result is a mix of questions asked across different survey organizations and across different states. For example, a majority of Florida polls have included Gore, while only one California poll included him. How should we handle this variation across polls?
We could try to standardize only on polls without Gore, and use 2nd choices when polls include him. But that would eliminate all the many polls that don't report 2nd choices without Gore. More national polls have included him than not, and so we've used that as the standard. If a pollster chooses to leave Gore out of the candidate list, we go ahead and use that poll, despite the difference in candidates. It is a considered judgment of the pollster to measure candidate preference either way and we yield to the pollster's judgment.
There are two other reasons to keep Gore in, so long as many pollsters (at least) keep his (non-)candidacy alive. First, his name provides some measure of how many Democratic voters are still looking for Prince Charming. Given the low probability of a Gore run, it is hard not to interpret his support as essentially a statement of dissatisfaction with at least the top three other Democrats. So for that reason the Gore line is revealing even if he isn't a candidate. Once he does clearly decline to run and is dropped from the questions, we'll see the other lines respond as voters face the choices they actually have. But we should just wait and see that happen in the real world rather than force it through excluding questions that include Gore.
Second, we prefer to report results from questions that present voters with a single set of choices rather than reallocate their second choices. It isn't that voters can't give a second choice, but that second choice is more hypothetical than their first. Also, we don't know if the second choice is a close second or a distant second.
So long as some pollsters include Gore and others exclude him, we can't avoid mixing some apples and oranges here. Our decision rule has been to take the first question that asks presidential preference and use that, no matter who is included or not, and avoid subsequent second choices, or in some cases a subsequent addition of candidates. This best reflects what the pollster thought their best question was, and allows the majority decision of pollsters to dominate our data.
(Mark Blumenthal analyzes the effects in the variation of how pollsters test a potential Al Gore candidacy in recent New Hampshire primary surveys.)