Skip to main content
ABC News
A Brief History of Primary Polling, Part II

This is the second part of a three-part series exploring the reliability of early primary polls in forecasting the winner of a presidential nomination.

The first article, posted last Thursday, focused on Republican primary races going back to 1976, and found that polls conducted as far in advance as a year before the primaries did a remarkably accurate job of forecasting the nominee. In fact, the Republican who was leading in national polls of primary voters in the first six months of the year before the primaries won the nomination on 6 out of 7 occasions; the only exception was John McCain, who was placing a reasonably close second to Rudy Giuliani at this point in 2007 but overtook him to become the standard-bearer.

Focusing only on the Republicans, however, would lead one to be overconfident about the accuracy of these surveys. The Democrats have tended to have far more exciting primary races, producing several surprising winners. One such election was the race in 1972, which is where we’ll begin our journey.

George McGovern began the race with lower name recognition than some of his competitors, and he was pulling only about 5 percent support in the earliest polls. His nomination was an upset in a field that included both halves of the Democrats’ 1968 ticket in Hubert Humphrey and Ed Muskie. In the end, Mr. McGovern attracted slightly fewer total primary votes than Mr. Humphrey did, but he won 21 states to Mr. Humphrey’s 5 and captured the nomination, in part by having a superior understanding of the new primary rules that he had helped to design.

Perhaps equally surprising is that the Democrats did not revert to a more top-down process for picking a candidate after Mr. McGovern went on to lose to Richard Nixon in a 23-point landslide.

The 1976 Democratic field is the one that reminds me the most of this year’s Republican one.

The circumstances ought not to have been bad for Democrats. Richard Nixon had resigned in the face of impeachment, prompting big gains for the Democrats in the 1974 midterm elections. And there were plenty of brand names bandied about: of the 20 potential candidates whom pollsters inquired about in early 1975, a half-dozen had name recognition of roughly 80 percent or higher.

But all of them seemed flawed in some way, including the nominal frontrunner, George Wallace, who would have been an extremely problematic general-election nominee. Meanwhile, quite a few of the well-known Democrats declined to run. The candidate who eventually emerged was Jimmy Carter, who had only 1 percent of the support in early polls.

If you buy into the analogy with 2012, here are the lessons: First, the absence of a true frontrunner makes it easier for a dark-horse candidate to emerge — and that means not just a lesser-known name like Mitch Daniels, but also somebody whose chances are not being taken seriously at all so far. Second, though, the Republicans are not necessarily doomed in the general election just because their field looks weak right now: Mr. Carter did, after all, win the general election.

Then again, Mr. Carter won by only 2 points against Gerald Ford, an unelected vice president whose approval ratings spent most of their time in the low to mid-40s (and who barely survived a primary challenge). Mr. Carter got the job done, but there is a good case to be made that he underachieved and should have been a clearer winner.

In the modern primary era, the only candidate to consistently achieve at least 50 percent support in polls of primary voters and then fail to win his party’s nomination was Ted Kennedy, who had roughly a 24-point lead over Mr. Carter, who was then the sitting president, in early 1979. Mr. Kennedy’s standing began to erode, however, after an infamous interview with the CBS correspondent Roger Mudd, and in the end he did not come all that close, winning primaries or caucuses in just 12 states to Mr. Carter’s 37.

The first time in our study period that the early polls actually predicted the Democratic winner correctly was 1984. Walter Mondale, who had a modest early lead over John Glenn, won reasonably easily, after Mr. Glenn never clicked as a presidential candidate. Instead, Mr. Mondale’s most vigorous competition eventually came from Gary Hart and Jesse Jackson.

It was a bit of an underrated field in 1988: it contained two future Vice Presidents in Al Gore and Joe Biden and a future House Majority Leader in Richard Gephardt. And a couple of pollsters were ambitious enough to ask about an unheralded governor from Arkansas named Bill Clinton (who would wait another four years to run).

The early leader was Gary Hart, who was winning around 35 percent support in polls before abruptly dropping out of the race after news of an extramarital affair surfaced. Mr. Hart eventually re-entered the race, but not before he had been lapped by the field.

This is one case where adjusting the results for name recognition really makes a difference. Michael Dukakis was getting around 7 or 8 percent in polls — not great — but he was doing it despite being known to only about 30 percent of primary voters. In other words, about 1 in 4 primary voters who were familiar with Mr. Dukakis were already prepared to vote for him, a considerably better figure than that, for instance, for Jesse Jackson. Once Mr. Hart dropped out, Mr. Dukakis was at least as likely as any other candidate to win the nomination.

It might seem as though Republicans are off to a slow start this year, but it’s nothing compared to what Democrats endured in 1991, when the eventual nominee, Bill Clinton, did not officially declare for the presidency until October. More remarkably, the top 9 candidates in early polls that year all declined to run, out of worry about the extremely high approval ratings for George H.W. Bush in the wake of the Gulf War. The frontrunner among them, Mario Cuomo, may be the politician in modern times who would have been most likely to win the presidency if only he had bothered to run for it.

To put this in perspective, the situation for the Democrats in the 1992 race would equate to a 2012 Republican race in which Mike Huckabee, Mitt Romney, Sarah Palin, Newt Gingrich, Ron Paul, Tim Pawlenty, Mitch Daniels, Donald Trump and Chris Christie all declined to seek the presidency, leaving the nomination to be contested by the likes of Haley Barbour, Rick Santorum and Michele Bachmann.

Mr. Clinton had very little support in the earliest polls in 1991, but quickly began to gain it once the field consolidated. Because this primary race was so unusual, we’ve also examined the poll numbers for the second half of 1991, not just the first half of the year. By late 1991, Mr. Clinton was polling at about 8 percent, despite being known to only about 30 percent of voters — he was doing as well as any other candidate, after adjusting for name recognition.

Mr. Clinton faced no primary challenge in 1996, but his vice president, Al Gore, did in 2000, in the form of Bill Bradley:

Mr. Bradley never polled all that close to Mr. Gore, but his numbers were more impressive in consideration of his average name recognition, which we estimate to have been about 50 percent in early 1999. Still, the polls were right, and Mr. Gore swept all 50 states on his way to the nomination.

Another year where accounting for name recognition would have been helpful was 2004:

Joe Lieberman was the nominal leader in surveys in early 2003, but after adjusting for name recognition, he was trailing both John Kerry and John Edwards — the eventual Democratic ticket. The other noteworthy thing about this field is that almost from start to finish, it left Democratic voters unsatisfied; about 30 percent of them were undecided in early polls.

In contrast, Democrats had a star-studded field in 2008:

Somewhat contrary to the conventional wisdom, which holds that Barack Obama suddenly burst onto the political scene, the polling shows that he was already reasonably well-known to voters in advance of the 2008 primaries, largely as a result of his speech at the 2004 Democratic National Convention. His name was recognized by around 60 percent of primary voters by late 2006, and that figure quickly ramped up to 80 or 90 percent after he declared for the presidency in February, 2007. (Mr. Obama was assisted, perhaps, by receiving somewhat more media attention than Hillary Clinton did, even in the early going.) Although name recognition did account for some of the roughly 15-percentage-point lead that Ms. Clinton held on Mr. Obama early on, she still led by about 10 points even after adjusting for it, and Mr. Obama had a real deficit to overcome.


In contrast to the Republican primaries, where the early frontrunner won in all but one instance, Democrats have had a considerably more varied set of results. Of the 9 primary cycles in our study, the Democrat who was placing first in the early polls won just twice (1984 and 2000). Although the second-place candidate won on three other occasions (1980, 2004, 2008), there were years (1972, 1976, 1992) in which the eventual nominee emerged from deep in the early field.

Results are considerably improved, however, by accounting for name recognition. By doing that, you would have had Mr. Dukakis and Mr. Kerry as favorites or co-favorites by early 1987 and early 2003, respectively, and Mr. Clinton as a co-favorite by the time he entered the race in October, 1991.

Is there something systematically different about the ways Democrats and Republicans tend to pick their nominees? We’ll explore that question — and take a detailed look at this year’s Republican field — in the next installment.

Nate Silver is the founder and editor in chief of FiveThirtyEight.