| || || ||
| ||ONLINE EXTRAS/What Americans Think|
Bush Lead Dwindles, or Does It?
By Richard Morin
Washington Post Polling Director
Monday, Aug. 14, 2000
Here's the difference a day makes: 17 points in a presidential horse race poll. Or at least that's what the Gallup Organization told us last week via USA Today and CNN.
On Aug. 7, Gallup reported that Gov. George W. Bush had a whopping 19-point lead over Vice President Gore. The next day it turned around and said Bush's lead had plummeted to 2 points, based on a one-night poll conducted the previous evening.
That's a 17-point swing -- big enough to get the poll on the front page of USA Today and featured on CNN, and certainly big enough to send the political punditocracy chattering into the night.
For Bush, the numbers suggested a collapse of epic proportions. For Gore, who on that Monday announced the selection of Connecticut Sen. Joseph Lieberman as his running mate, it was a unexpectedly bountiful pre-convention bounce.
One problem: The Gallup poll was probably wrong (note the hedge). The second problem: Its media clients were probably ill-advised (note the second hedge) to feature the results as prominently as they did. In fact, many political reporters, as opposed to the pundits, were skeptical of the second Gallup poll from the moment it was released. The Washington Post, for example, decided to make no reference to the poll in its coverage the next day.
The reason for this caution was obvious. Leads that big rarely fall that fast, at least not without good reason. And the Lieberman announcement struck many as not a good enough reason to explain why the proportion of Bush supporters dropped a full 10 percentage points, from 57 percent in Gallup's post-GOP convention poll to 47 percent in its Aug. 7 one-nighter.
The skeptics were wise to wait. By midweek, the evidence was accumulating that the Gallup's drop in support for Bush was overstated. A Post-ABC News survey found that the Republican maintained his double-digit lead over Gore at least through last Thursday, although it was slowly shrinking.
Dueling horse race numbers aside, there were other reasons to approach the Aug. 7 poll with caution. For one thing, it was done in a single night. USA Today appropriately quoted Gallup's David Moore as noting the problems inherent in doing an overnight poll -- difficulties that can but don't always skew the horse race results.
This overnight suffered from another design flaw: Gallup stopped interviewing at 8:45 p.m. EDT, or shortly before 6 p.m. on the West Coast. Who's at home at that time in Los Angeles, or anywhere else in the Pacific time zone, for that matter? Gore supporters, apparently.
Then there's the matter of party identification. The Gallup sample last Monday included a surprisingly large number of self-identified Democrats: 39 percent of their sample identified themselves as Democratic, compared with 33 percent in their weekend project. (The proportion of unleaned Democrats in the post-GOP convention Post-ABC News poll was 33 percent.)
Likewise, the Gallup poll released Aug. 7 included fewer Republicans than in Gallup's earlier project (31 percent versus 35 percent).
Those numbers raise this disquieting possibility: Perhaps the big and abrupt shift in the horse race wasn't because of Lieberman, but simply because more Democrats and fewer Republicans happened to be interviewed.
Perhaps, but probably not, says Frank Newport, editor of The Gallup Poll. Party identification is "variable, and it can shift," often in response to developments in the news, he says. "We've documented that over the years." Thus the increase in Democrats merely reflected the fact that weaker partisans or independents were feeling particularly Democratic last Monday night when Gallup polltakers called, and not because of problems with the sample.
Jim Norman, the polling editor at USA Today, also notes that the horse race number can vary significantly from day to day, which he suspects might be related to short-term reactions to news events. That might explain the big changes in the horse race. "If you dominate coverage in a positive way, it is going to give you several points on that one day. It might be ephemeral, and be gone by the next day."
Norman says they looked at the day-to-day results of their post-GOP convention poll and found Bush "had a 20-point lead among those interviewed on Friday, and a four-point lead among those questioned on Saturday." He suspects the big Bush lead was a direct result of the strongly favorable coverage Aug. 4 of Bush's acceptance speech to the convention the previous night.
Well, perhaps. One thing is clear: This has been a volatile period for the polls. On Aug. 7, The Post and ABC News reported the Republican nominee with a 14-point lead, Gallup/USA Today/CNN had him up by 19 -- and the Los Angeles Times showed the race to be statistically tied, with Bush up 2 points.
"We have certainly found a great deal of volatility," Newport says. "It's particularly unsettled during this convention period, with a lot of information going back and forth. It wouldn't surprise me in our next poll to find something else. At this point, it's a very fluid electorate out there. We'll have to wait until polling after the Democratic convention to tell us what the actual shape of the race is."
While I believe Gallup was wrong last week, the polling organization may be right this week. If Gore gets the expected bounce up in the polls from his convention, the race will inevitably tighten, at least briefly. And Gallup can say to me, "See, we were right all along."
Archive: What Americans Think