Last week, three national polls came to three different conclusions about the viability of the two Democrats running for their party's presidential nomination. (In each poll, the respondents were asked to assume that their opponent would be GOP front-runner George W. Bush and that the election was being held today.)
The CNN/USA Today/Gallup Poll showed Bush leading Vice President Al Gore by nine points (a smaller lead than he had in several previous measurements). However, former New Jersey Sen. Bill Bradley trailed Bush by only two points. The natural conclusion is that Bradley could be the stronger Democratic nominee, which has been the conventional wisdom for months.
But the CNN/Time magazine poll found Bush leading Gore by only five points, while the Texas governor led Bradley by nine points. The conclusion here would be that Gore could be more electable.
Finally, a Newsweek Poll had both Democrats lagging even further behind Bush. As in the CNN/Time Poll, Gore trailed Bush by fewer points than Bradley did.
So who is more electable?
All of the polls were conducted at approximately the same time, with a reasonable number of respondents, using generally accepted telephone polling techniques. Despite the pollsters' best efforts, what we learn from these polls is not necessarily who's ahead, but the difficulties inherent in asking hypothetical questions.
In the summer of 1992, for example, Ross Perot always did better in hypothetical match-ups against Bill Clinton and George Bush (the elder) when respondents were first asked to choose between only Mr. Clinton and Bush, than when they were first asked for their three-way preference.
Polls conducted during primary campaigns give distorted results when questions go beyond the primary into the general election. Many of those who are supporting one of the primary contenders say they could not vote for his opponent in a general election. But typically, by the time the fall election comes around, they do support them - sometimes with enthusiasm.
In one case, the order of the questions affects the hypothetical outcome; while in the other case, the timing has a clear impact. Respondents know that despite how the question is worded, the election is not being held today.
National polls conducted in January before the official start of the primary season may be extra-susceptible to these order and timing issues. George W. Bush has many advantages in this campaign, and one is his enormous lead in national polls of likely Republican primary voters. That means - at least for now - fewer potentially disaffected Republican voters (that is, Republicans who aren't partial to Bush) answering questions about what they might do in a general election. Gore's lead among Democrats nationally doesn't reach Bush's national Republican support level.
Other reasons for the differences this year probably include several "ordr" effects. It would be helpful to know, for example:
- Which pairing was asked first? (Bush vs. Gore or Bush vs. Bradley?)
- Did the general election pairings follow or precede the questions about choice for the nomination?
- What other potentially related questions were asked and when? (e.g., party registration, overall approval ratings, opinions about the individual candidates).
The difficulty in interpreting the hypothetical horserace questions now is not that respondents misrepresent their preferences. Rather, it is that pollsters sometimes expect too much of the poll and the voters. They assume that answers given to the hypothetical question are more than simple, immediate, short-term or reflex reactions. Indeed, sometimes the first time the voters are confronted with such choices is in that poll!
© 2000 CBS Worldwide Inc. All Rights Reserved. This material may not be published, broadcast, rewritten, or redistributed