| || || ||
| ||ONLINE EXTRAS/What Americans Think|
Into Home Stretch, Gore, Bush Neck-and-Neck
By Richard Morin
Washington Post Polling Director
Monday, Sept. 11, 2000
Finally, the home stretch. And finally a race: The first post-Labor Day poll by the Washington Post and ABC News finds the presidential race a dead heat. Gone are George W. Bush's double-digit leads that carried through his convention in July. Gone, too, is Al Gore's bountiful post-convention bounce, which propelled him into a similar double-digit lead in some late-August polls conducted after the Democratic National Convention in Los Angeles.
While political scientists and their forecasting models were confidently predicting a big Gore win, the first round of post-Labor Day national surveys were saying something altogether different: Don't bet too big, too soon. Unlike previous years, and despite a whopper economy and popular incumbent usually signs that the incumbent party's in for a big win this race is up for grabs.
Historically, the first polls of September have been remarkably good at picking the eventual winner. In trial heats conducted by the Gallup Organization, there have been only two instances in which the eventual loser blew a significant lead after Labor Day. Jimmy Carter was eight points up in October 1980, and Ronald Reagan won. Thomas Dewey was likewise ahead by eight points in 1948, but Harry S. Truman won.
Gallup reports there have been two other instances in which the eventual loser had a slight numerical lead after Labor Day, albeit an insignificant one: Gerald Ford was up one point in November 1976 and Carter won. Nixon likewise was up by one in September 1960 and John F. Kennedy won. (A full and fascinating accounting of post-Labor Day poll trends may be found at the Gallup Web site, www.gallup.com).
There are 11 instances where the leader after Labor Day never trailed. But as ABC News polling director Gary Langer notes, "three factors temper even that notion."
Seven of those races never were close, including the 1936, 1952, 1964, 1972, 1984 and 1996 campaigns. It's easy to pick a winner when a popular incumbent is running against an unpopular challenger (or a war hero is in the race, as was the case with Dwight Eisenhower in 1952).
"In the rest, the leader's 'lead' was, at times, small-to-negligible: Roosevelt +4 in October '40 and +2 in September '44; Nixon +1 in November '68, Bush +3 in October '88, Clinton +1 (Gallup) or +3 (ABC) in October '92," Langer reported in a summary to ABC political reporters.
Finally, few post-Labor Day polls were done before 1988-anywhere from two to seven, Langer reported. "In close races, there may have been moments when the loser had a lead, but the sparse polls missed it." Now, with everybody polling 24/7, it's hard to miss every jot and squiggle in the horse race.
So what's the forecast for 2000, based on the post-Labor Day polls? Delightfully uncertain and impossible to predict. "A review of Gallup polling conducted prior to every presidential election since 1936 shows only two situations in which early September polling was dead even, and each had a different outcome," Gallup analysts wrote.
In 1960, the first Gallup poll conducted in September had Richard Nixon at 47 percent of the vote and John F. Kennedy at 46 percent a dead heat. "The closeness of that race persisted throughout the fall, and in the end the two candidates were virtually tied in the popular vote, with Kennedy winning by less than 1 percent of the popular vote," Gallup reported.
Two decades later, Carter and Reagan were tied in the first September Gallup poll, 39 percent each. That race generally remained close in polling conducted in September and October, but as the election drew near, Reagan pulled away from Carter and went on to win by a 10-point margin, 51 percent to 41 percent, Gallup analysts wrote.
So keep watching this space. While conventional wisdom suggests that the front-runner in the first September polls comes out first in November, not always particularly when the race is close at summer's end. "[W]hoever leads usually wins," Langer wrote. "But in close elections, all bets are off."
Those famously bold pundits who want to tempt fate and declare a winner based on early and inconclusive polls also might want to consider the great public shame that sometimes accrues to those who get it wrong like Gallup did in 1948 "the most famous example of changes between early polls and Election Day," according to Gallup analysts.
Dewey led incumbent Truman by eight points in Gallup's first poll after Labor Day in 1948, and Dewey continued to lead by healthy margins in every poll that followed. Truman went on to win an upset victory over Dewey, by a 50 percent to 45 percent margin. (In those days, people thought presidential preferences were fixed early in the campaign. As a consequence, Gallup's last poll was taken a full three weeks before election day, missing the surge to Truman. Live and learn.)
It is ironic that Gallup's overall error in 1948 was even larger than Literary Digest's famous blunder in 1936, in which it predicted Alf Landon a big winner over Franklin D. Roosevelt. And just as 1936 made Gallup literally a household name, 1948 practically wrecked Gallup and momentarily killed off public confidence in the infant science of public opinion polling.
After the debacle, Gallup was sent a book titled "What Dr. Gallup Knows About Polls." It contained nothing but blank pages.
Gallup's own family tells the story of how he was driving the wrong way down a one-way street in Los Angeles soon after the 1948 election and was stopped by a police officer.
The policeman took one look at his driver's license and said. "Wrong again, Dr. Gallup."
Archive: What Americans Think