March 4, 1999

Does an early lead in the polls usually hold up?

A stream of candidate announcements and polls showing Texas Governor George W. Bush leading Vice President Al Gore have spurred interest in the 2000 presidential contest. But with the nominating conventions still more than 16 months away, what do these early polls mean for would-be nominees?

A look back at nearly 40 years of early primary polls suggests that the Republican front-runner is a good bet to capture the nomination. The same cannot be said of the first horse in the Democratic pack.

In six open Republican contests since 1960, the early front-runner has held on to win the party nod five times.(1) But in six open Democratic contests since 1960, the early leader has won the nomination just once. That was Vice President Walter F. Mondale in 1984.

While this may be good news for Bush, it is not necessarily bad news for Al Gore. The Democratic front-runners who did not win the nomination had not been vice presidents. Further, they were not usually blocked by lack of voter support. One decided against a run, one withdrew in the face of scandal, and one dropped out in the wake of a now legendary dirty trick.

Why do Elizabeth Dole and George W. Bush look strong, when the GOP looks weak?

While Gore may be a good bet to win the nomination, recent nationwide polls find him trailing Bush and former Cabinet Secretary Elizabeth Dole among registered voters. And a recent Pew Research Center survey found more respondents saying they would consider voting for the two GOP front-runners than for the Democratic Vice President (72% for Bush and 64% for Dole, compared to 52% for Gore). These are surprising findings given Clinton’s 64% approval rating and the fact that the Democratic party has a better national image (58% favorable) than the Republicans (44% favorable).

Gore’s standing in the polls is reminiscent of former Vice President George Bush 12 years ago when he ran behind several Democratic contenders at this point.(2) But Vice President Bush’s weakness in the polls then may have reflected mixed views of the administration, while Vice President Gore’s position in the polls today may have more to do with his own image problems.

Gore’s favorability ratings today are significantly below those of Vice President Bush at this point in the Reagan administration. Gore is regarded favorably by 58% of the public today, compared to the 67% who had a favorable opinion of Bush in April 1987. Moreover, still stinging from the Iran-Contra scandal, President Reagan’s job approval rating was a paltry 47% in 1987, compared to President Clinton’s lofty approval score today.

Tepid response to Gore among many Independents and even among Democrats is leading a significant number of these voters to consider voting for George W. Bush or Elizabeth Dole in 2000. Nearly half of Independents (49%) and 16% of Democrats say they have ruled out voting for Gore. In contrast, 76% of Independents and 53% of Democrats say they would consider voting for Bush. The numbers for Dole are nearly as high: 69% of Independents and 49% of Democrats would consider voting for her.

How well did the independent statewide polls do in November?

The 1998 state election polls should put to rest the criticism that the polls consistently underestimate Republican strength. Last year, more than two-thirds of the independent polls reviewed by the Pew Research Center overestimated the Republican vote.

This failure led to some criticism of the state polls, since several national polls caught the Democratic surge the weekend before the election. But despite a consistent pattern of underestimating the Democratic vote, the state polls were for the most part accurate: well over half of those reviewed correctly forecast the elections.

To get a picture of what the polls showed and why they may have gone wrong, the Pew Research Center gathered information about 34 independent polls conducted for the news media in key Senate and gubernatorial races. All of the polls were published within 10 days of the election, and only races that were close or predicted to be close were considered.

Looking at the spread — that is the difference between the Democratic and the Republican candidates — 10 of the 34 polls missed the mark. They missed the spread by over 8 percentage points, more than the margin of sampling error for these polls. Eight of these polls also missed the victor. Another 14 polls missed the spread by between 7 and 4 percentage points, a fair showing, and 10 were on the money, calling the spread within 2 percentage points.

No identifiable methodological patterns emerge to separate the better polls from the rest. Nearly all had adequate sample sizes for an individual state (600 or more) and almost all looked at likely voters. Almost none report weighting their sample by party identification (which fluctuates) rather than demographics. And while over half did not ask follow-up questions to undecided voters, pushing them to make a choice between the candidates, this variable turns up just as often on the accurate and inaccurate polls.

The only pattern that does emerge is partisan. The spread on only five state polls underestimated Republican strength, while the spread in 26 of the 34 underestimated Democratic strength. Eight of the 10 polls that fell outside of the margin of error fell into the latter category, and the other two involved the three-way Minnesota gubernatorial contest won by Independent Jesse Ventura.

But the reasons the polls missed the Democratic surge are not clear. In some instances, estimates of Democratic voter turnout may have been off. Of the eight that missed the race outcome by a large margin, four involved elections in which black voter turnout was slightly higher than in the last midterm election: the Georgia, South Carolina, and Maryland gubernatorial contests. Without an African-American candidate or ballot issue that can boost minority voter turnout, the pre-election polls may have estimated that black voter turnout would be like the 1994 elections. This proved wrong in some states, as Democratic Party groups made a major get-out-the-vote effort among blacks in 1998.

In other places, voters who made up their minds late appear to have gone Democratic. Two of the polls that were wrong involved the Iowa gubernatorial contest, won by Democrat Tom Vilsack. Both Mason-Dixon and the Des Moines Register showed the Republican leading the week before the election, but polls throughout the summer and fall showed a trend that hinted at a Democratic upset. The Republican was below 50% in both polls and had not gained any ground during the fall, while the Democrat steadily climbed 20 percentage points in this time.

The remaining two polls involve the only two Senate races that state polls missed: New York and California. Zogby International predicted a dead heat in New York; the Democrat won handily. In California, Mason-Dixon forecast a razor-thin Democratic margin; in fact, the Democrat won comfortably. The starkly partisan tone of the two contests and the relatively large minority populations in each state might account for part of the mistake. But other polls conducted around the same time did call these two races accurately, and the two pollsters who missed the races accurately forecast the winner in other close contests.

Endnotes

1. The early front-runners are candidates who drew the strongest support among all potential nominees in national polls taken more than a year before the presidential election. Questions about Republican nominees were typically asked of Republicans and Independents who lean Republican; questions about Democratic nominees were typically asked of Democrats and Independents who lean Democratic. Based on past surveys by the Pew Research Center, the Gallup Poll, and CBS News/New York Times.

2. For example, Gary Hart led Bush, 47%-38%, in an April 1987 Times-Mirror survey.

Cite this publication: “Does an early lead in the polls usually hold up?.” Pew Research Center, Washington, D.C. (March 4, 1999) http://www.people-press.org/1999/03/04/does-an-early-lead-in-the-polls-usually-hold-up/, accessed on July 23, 2014.