What More Than 40 Years Of Early Primary Polls Tell Us About 2020: Part 1

How much can we learn from early primary polls? Back in 2011, FiveThirtyEight Editor-in-chief Nate Silver set out to answer this question and found that early national polling is at least somewhat predictive of who will win the nomination, especially when the results are adjusted for each candidate’s name recognition.

Now, eight years later, FiveThirtyEight has collected more polling data, plus there are two more presidential election cycles — 2012 and 2016 — to look at, so we felt it was time to update the series. In the first two installments, we’re just going to look at what the polls say for competitive presidential primaries for both parties, starting with early primary polls from 1972, which is widely thought of as the start of the modern primary era. This first installment runs through 1996 — analyzing primary polls for seven presidential elections is enough ground to cover in one article — and the next installment will start with the 2000 presidential primaries and run up through the 2016 presidential primaries.

As for how this series will work, we took all surveys from the calendar year before each election — so for the 2016 presidential primaries, that means all polls from 2015 — and then split them by whether they occurred in the first half of the year (January through June) or the second half (July through December). In the 2011 version of this series, Nate only looked at polls from the first half of the year before the election, but we decided to include the second half of the year as well because this helped us capture how a candidate’s standing changed during the course of the year and let us include candidates who jumped into the race on the later side. We also included anyone we had polling data for, no matter how likely or unlikely they were to run or win.

Our tables in this series include two key metrics that help us better understand the primary field and how things changed throughout the year we’re examining. The first is an average of each candidate’s (or potential candidate’s) numbers in all the polls for that half of the year — candidates were counted as having zero percent support in any poll they did not appear in.¹

The second is an average of each candidate’s standing in the polls that is adjusted for how well-known the person was at the time. To do this, we divided a candidate’s polling average by their level of name recognition, which helped us identify folks who might have had a small national profile but were doing relatively well among voters who knew about them.² But because pollsters aren’t consistent, our methods of estimating name recognition had to be treated as rough estimates. To reflect that inherent imprecision, we sorted candidates onto a somewhat subjective five-tier scale to sum up their level of fame.³ We combined polling averages with a few educated guesses to produce the name-recognition scores. For example, a candidate like Hubert Humphrey in 1972, who had been the Democratic Party’s presidential nominee in 1968 while serving as vice president, would almost certainly be extremely well-known, while someone who had lost the nomination to Humphrey in ’68 might not be as well-known but would still be more widely recognized than someone bursting onto the national scene for the first time. These name-recognition scores are represented as five square boxes in the table below (more black boxes means higher levels of name recognition).

And it just so happens that in the first primary we’re looking at, the 1972 Democratic primary, we have an example of how our adjusted polling average can reveal a potential winner. Take Sen. George McGovern, who was polling at around 4 percent in the first half of 1971. He wasn’t very well-known, but he was polling relatively well among those who had heard of him, so his adjusted polling average for that same six-month period was 6 points higher, at 10 percent.

However, he trailed Sens. Humphrey, Ed Muskie and Ted Kennedy — the former two had made up the 1968 Democratic ticket, and the latter was … well, a Kennedy. But McGovern still managed to win the nomination — that’s why he’s bolded in the table above — in part because he had a good understanding of how the reformed nominating process worked. After all, he led the committee that wrote the new primary rules after the calamitous 1968 Democratic convention, where party bosses helped Humphrey get the nomination even though he had not contested any of the primaries. And even though McGovern was crushed by President Richard Nixon in the 1972 general election, Democrats kept their new primary rules. McGovern would not be the last Democrat to become the nominee despite low early polling numbers.

We’re skipping the 1972 GOP primaries because Nixon was running for re-election and faced no serious challenger from within his party — remember, we’re only interested in competitive nomination processes. So moving on to 1976, the Democratic field was comparable to 2020 in that it was crowded and support was fragmented. In the first and second halves of 1975, four or five potential Democratic candidates averaged 10 percent or more in our adjusted polling average. But three of them — Humphrey, Kennedy and Muskie — didn’t wind up running. Segregationist Alabama Gov. George Wallace was the nominal frontrunner in those early polls, but he was a divisive choice for obvious reasons, which left things wide open for a dark horse candidate. And in the end, one of the least-known candidates — former Georgia Gov. Jimmy Carter — claimed the nomination. The early primary polls don’t really capture Carter’s success, but he did set up shop in Iowa, which voted first, and his strong showing there gave his campaign a boost that helped him gain enough momentum to win the nomination. It also cemented Iowa as a premiere early state alongside New Hampshire.

The 2020 Democratic race could be primed for a similar upset. The polls are currently split between two early frontrunners — still undeclared former Vice President Joe Biden and Sen. Bernie Sanders — and it’s not yet clear if they will continue to sit atop the leaderboard, effectively blocking lesser-known candidates, or if their lead over the rest of the field will shrink as other candidates garner support, which could make the Democratic field look as wide-open as it did in 1976.

As for the GOP in 1976, the party had two clear frontrunners from the start. President Gerald Ford and challenger Ronald Reagan battled all the way to the national GOP convention, where Ford narrowly edged out the former California governor. The 1976 Republican primary marked the start of a streak of early polling frontrunners winning the nomination.

Four years later in 1980, the early national surveys had the incumbent president, Democrat Jimmy Carter, trailing badly against Sen. Kennedy, who had finally decided to run. Based on our data, Kennedy holds the dubious distinction of being the candidate with the highest adjusted polling average to not win a presidential nomination, at least for the 1972 to 1996 period. In the end, Carter was renominated, but he lost the general to Reagan.

Reagan’s nomination in 1980 marked another GOP contest won by the early frontrunner, but it also started a trend of the Republican runner-up winning the GOP nomination in the party’s next competitive primary cycle. His principal foe was former CIA Director George H.W. Bush, who was relatively unknown in the year prior to the primaries. Bush advanced from about 4 percent to 11 percent in our adjusted polling average from the first half to the second half of 1979, and even though he beat out Reagan to win the Iowa caucuses, Reagan went on to win the nomination (although it wasn’t entirely a bust for Bush, as Reagan did make him vice president).

Reagan sought re-election in 1984, and to face him Democrats nominated their early polls leader — former Vice President Walter Mondale. This marked the only time between 1972 and 1996 that the Democrats nominated someone who led in the raw polling average a year before the primary, although he still had to sort things out at the convention against his main opponents, Sen. Gary Hart and civil rights leader Jesse Jackson.

In 1988, Democrats returned to form by not nominating the early frontrunner, and instead picked Massachusetts Gov. Michael Dukakis. Yet our adjusted polling average indicated Dukakis was a lot stronger than early surveys suggested, as he enjoyed a fair bit of support while being relatively unknown. It didn’t hurt him that Hart, who was campaigning again and one of the early frontrunners, was undone by an extramarital affair. While Dukakis went on to clinch the Democratic nomination, he lost in the general election.

Opposing Dukakis in 1988 was George H.W. Bush, who had served eight years as Reagan’s vice president — yet another GOP runner-up who went on to win the next open primary. Bush also led most early polling, with Sen. Bob Dole in second. And while Dole did win the Iowa caucuses, Bush recovered in New Hampshire and won most contests after that. But this wouldn’t be the last Republicans saw of Dole; he would get his shot at the presidency later on.

In 1992, Bush became the third president to fend off a notable primary challenger since 1972. Initially, Bush’s approval rating had been high — close to 90 percent after the Gulf War. But a faltering economy and broken promise not to raise taxes weakened Bush and helped spur a challenge from the right in the form of pundit Pat Buchanan. We only have data for the second half of 1991 — at first it looked unlikely that Bush would face a serious challenge, so pollsters weren’t asking about the Republican primaries much — but we can see that despite Bush’s massive lead in the polls, Buchanan’s position was more robust than it looked. Even though his name recognition was low, those who did have an opinion of him seemed to support him. Bush won the nomination, but Buchanan embarrassed him by grabbing 37 percent of the vote in New Hampshire, and ultimately, Bush lost re-election.

Facing Bush in the general election was Arkansas Gov. Bill Clinton, who was still not that well-known when he announced his run for president in October 1991. Nonetheless, Clinton’s adjusted polling average improved substantially in the year leading up to the primaries. He started at around 3 percent in the first six months of 1991 and jumped to about 19 percent in the second half of the year. New York Governor Mario Cuomo led the polls, but at the last minute decided against running, and Clinton went on to defeat former Sen. Paul Tsongas and former Gov. Jerry Brown for the nomination.

In 1996, the GOP chose Dole to face off against Clinton during his re-election bid, making Dole the third runner-up since 1976 to win the GOP nomination at the next opportunity. Dole seemed to have an insurmountable edge in the Republican primary polls, but he still had a rough start — he barely defeated Buchanan in Iowa and then lost to him in New Hampshire before recovering to win most other contests. Dole went on to lose in the general against Clinton.

And that’s a wrap for now. In Part II, we’ll look at early primary polls in contests from 2000 to 2016. And as we’ll see, Republicans broke with their habit of nominating the early polling leader, while Democrats nominated an early frontrunner twice, after doing so only once between 1972 and 1996.

Additional contributions by Laura Bronner.