So now the Zogby poll, which had seen more Kerry strength than most, is picking up a Bush surge -- while the Rasmussen poll, which trended for Bush, is seeing a Kerry bump.
I'm glad that I am out of the business of polling U.S. elections.
The difficulties of accurately surveying presidential prospects have become ever more daunting with each new election cycle. The growing difficulties account for the wide fluctuations in survey data.
The main variant in the data relates to the likelihood of a person actually voting. With half of the potential electorate normally forsaking the franchise, pollsters have great difficulty in identifying who will actually participate.
But every indication is that this year's turnout will dwarf our recent experiences. With record numbers of new voters coming on the rolls and get-out-the-vote operations generously funded by political committees released from any effective controls by the so-called campaign-finance reform, participation is likely to soar.
The closeness of the 2000 election is generating new voters. With party feelings approaching fever pitch as the election nears and emotional issues like terrorism, gay marriage, recession and the like dominating the campaign, turnout seems likely to be very high.
The truth is, no modern pollster really has any idea what to expect if turnout reaches the high 50s or closes in on 60 percent.
There's no good way to estimate turnout. Most people won't admit to not planning to vote.
Who will the new voters be? The conventional wisdom says downscale Democrats. When one compares the vote in presidential election years with off-years, the difference between the two usually consists of low-income, poorly educated and minority voters who tend to back Democrats. The increased interest of younger voters in this contest -- likely due to the war -- is also probably a Democratic advantage.
Techniques for getting out the vote have also changed as many states, Texas for example, encourage early absentee voting. Workers of both parties throughout Bush's home state are regularly visiting nursing homes or college campuses collecting absentee votes during the three-week window before Election Day during which balloting is permissible. Mail-in voting, pioneered in Oregon, is also dominating the counts in many states. In assessing trends, pollsters must find out who has actually already voted, making changes in their opinions irrelevant.
Cell phones are part of the problem. Political survey research is based on a proper geographic distribution of the sample. But cells are hard to pin down.
Finally, people are fed up with direct phone marketing. The era when voters welcome participation in a survey is as distant as the age when people looked forward to getting mail.
The Internet, which gets around the concerns of phone polling, still under represents the poor and the elderly. And even online polling can only reach those willing to take the time to participate, so the sample is still suspect.
As Election Day approaches, two trends will dominate, neither good for Bush. Downscale voters will begin to pay attention and be included in surveys of likely voters. Since they largely vote Democratic, this is good for Kerry. Also damaging to Bush's prospects is the fact that undecided voters tend to break in favor of the challenger at the very end.
But, again, all that holds in a normal year. We won't know if it's still true until next Tuesday night (if then). It's a tough year to poll.
Dick Morris is a syndicated columnist.