Pollsters Generally Had It
WEINSTEIN and CARL BIALIK
National pollsters accurately captured the broad outlines of the presidential election, but surveys in some closely contested states missed the target and early exit-poll numbers led observers astray.
Though some missed calls were couched by large margins of error, and voter sentiment may have swung after the final polls were tallied, there were nonetheless some early winners and losers among the big pollsters. SurveyUSA and Rasmussen Reports did well, while Harris Interactive's online unit was wrong in key battleground states and nationwide. Pollster John Zogby's late-Tuesday forecast of a Kerry victory also proved wrong.
In Ohio, where Mr. Bush appears to have prevailed by two percentage points, several polls gave him a pre-election lead, but both Harris's online poll and Gallup saw Mr. Kerry with a four point lead in the campaign's final days. In Florida, where Mr. Bush won by five percentage points, Harris Online and Fox News found Mr. Kerry had pre-election leads of four and five percentage points, respectively. Harris correctly placed Pennsylvania in Mr. Kerry's corner, but a CNN/USA Today/Gallup showed Mr. Bush ahead in that state by four percentage points.
Most major national pre-election polls correctly predicted Mr. Bush would win the popular vote, though the race wasn't as close as many pollsters predicted. Mr. Bush was leading Mr. Kerry by three percentage points, 51%-48%, on Wednesday morning, with 99% of national precincts reporting. That squared with polls taken in the campaign's final days by CBS News, Rasmussen, Gallup, Reuters/Zogby, Wall Street Journal/NBC News, ABC News/Washington Post and a Harris telephone poll. They all showed the Bush-Cheney ticket ahead by one to two percentage points, within the polls' margins of error.
"The consensus of the national polls was reasonably close to the final result," wrote pollster Mark Blumenthal on his Web log MysteryPollster.com. Mr. Blumenthal posted a rundown of the final pre-election polls compared with actual results.
There were two notable exceptions: a Fox News/Opinion Dynamics survey showed Mr. Kerry holding a 48%-46% lead, and a Harris poll conducted over the Internet gave Mr. Kerry a 50%-47% edge.
Results Mr. Zogby provided for a poll of battleground states that ran regularly in the Online Journal since late May came fairly close to the mark. The final poll, which included both Internet and phone polling, correctly called the winners in 13 of 15 states. (Mr. Zogby incorrectly said Iowa and Florida would go to Mr. Kerry.) One state, New Mexico, was considered a tie in the poll -- votes were still being tallied in that state Wednesday afternoon, but Mr. Bush appeared to have a narrow lead.
Two of the big winners -- SurveyUSA and Rasmussen -- use recorded voices to deliver polling questions over the phone, something they say gives them an edge because it eliminates variations in inflection and pronunciation. Rasmussen publisher Scott Rasmussen said the long-term consistency of its computer-generated polling, plus adjustments it made in methodology -- including weighting by party identification and expanding its definition of "likely voters" -- resulted in one of its best years to date.
"The reason we feel so good about last night is that our state-by-state numbers were great," Mr. Rasmussen says. "We had the president up in Ohio and Florida and we had Kerry up by a little in Minnesota, Pennsylvania and Michigan."
SurveyUSA's methodology adds a twist: The closely held company, based in Verona, N.J., has local television news personalities record the questions it asks. Jay H. Leve, editor of SurveyUSA, calls the method the company's "magic sauce," adding, "when you are called by SurveyUSA, you are talking to a man or woman you have grown up with." He added that the firm continued to crunch the numbers from races around the country, but so far it seems to have made at most one wrong call -- in still-disputed Iowa, which SurveyUSA called for Mr. Kerry -- out of about 70 races.
Harris's online poll results marked a turnaround from the 2000 election, when the polling firm's results stood out from the pack after it spent a year and $1 million conducting monthly polls of 1,000 voters over the phone and 10,000 people over the Internet. Both polls led Harris to the same conclusion in 2000: The firm called the 2000 race dead even, giving each candidate 47% of the vote. As it turned out, Mr. Gore won 48% of the popular vote; Mr. Bush, 47%. (Harris does polling for both the print and online editions of The Wall Street Journal.)
For this year's election, "the phone poll was great, but the online poll was a little off the mark," said Humphrey Taylor, chairman of the Harris Poll. He noted that staffers would be looking at the online poll's weighting system and tweaking it if need be, since the errors in Florida and Ohio reflected errors in the national poll.
Mr. Zogby, meanwhile, issued a prediction on his Web site late Tuesday that Mr. Kerry would win the election with at least 311 electoral votes. On Wednesday he had removed the forecast and posted the following statement: "Our predictions on many of the key battleground states like Ohio and Florida were within the margin of error. I thought we captured a trend, but apparently that result didn't materialize." Mr. Zogby added, "I've called this the Armageddon Election for some time -- a closely-divided electorate with high partisan intensity on each side."
Shawnta Walcott, a spokeswoman for Mr. Zogby's firm, Zogby International, echoed those sentiments, and added, "Using the standard techniques we've used in the past that have proven to be reliable, we crafted our prediction. And apparently the old standard simply didn't work yesterday."
As for the exit polls, conducted as voters left the polls, the results were mixed. In the closest states, the final exit-poll numbers -- supplied to media organizations by a consortium they formed called the National Election Pool -- were mostly on target. In Florida, Ohio, Pennsylvania, Iowa and Mew Mexico, the polls -- posted on news sites like CNN.com and MSNBC.com -- generally were within one or two percentage points of the results.
However, the early exit polls showed Mr. Kerry holding a nationwide lead, shaping the early tenor of TV broadcasts and, as Mr. Bush showed greater strength in early returns than was indicated, creating a backlash from TV commentators against the exit polls.
Proving particularly unreliable were early estimates of the exit-poll numbers leaked to bloggers, who posted the numbers early Tuesday afternoon. The data generally showed Mr. Kerry ahead in nearly every swing state, including Florida and Ohio.
Contact Us Help E-Mail Setup Customer Service: Online | Print