Expert Commentary

Voting in presidential primaries: What we can learn from three decades of exit polling

2008 paper from Northeastern on the U.S. presidential nominating process and the role played by voter judgment on incumbents' performance and party affiliation.

A 2008 paper from Northeastern University, “Voting in Presidential Primaries: What We Can Learn from Three Decades of Exit Polling,” published in The Making of the Presidential  Candidates 2008, analyzes interview surveys taken directly after a sampling of voters exited the primary voting booth between 1976 and 2000 and evaluates which underlying factors had the strongest correlation with voter choices.

Using this historical data, the paper explores the idea that presidential primaries “function as a referendum on the incumbent’s performance in office” — what political scientists call “retrospective voting.” In addition, the author looks at the under-appreciated role of party affiliation and partisanship in predicting primary voting outcomes.

The paper’s findings include:

  • Analysis of the 1976 Republican primary race between President Gerald Ford and Ronald Reagan shows that “the effect of presidential approval on vote choice is at least four times greater than that of any other variable,” such as ideology, religion, income level or policy preferences.
  • In the 1980 Democratic primary contest between President Jimmy Carter and Edward Kennedy, a voter’s judgment on whether Carter had handled such issues as the economy, foreign policy and the Iran hostage crisis appeared decisive: 94% of voters who approved of Carter’s handling in all three categories in exit polls voted for him in the primary, but only 9% of voters who disapproved of Carter’s job on those matters voted for him.
  • In the 1988 Republican race, the judgment of voters on sitting Vice President George H.W. Bush was strongly correlated to their judgment on President Ronald Reagan. Those who approved of Reagan voted for Bush 59% of the time, with Bob Dole getting just 22% under that scenario. The same is true in the 1992 Republican contest between Pat Buchanan and then-incumbent President Bush: “Those who approved of Bush’s performance voted for him by a 93 percent to 5 percent margin.”
  • In the 2000 Democratic contest between sitting Vice President Al Gore and Bill Bradley, how voters viewed President Bill Clinton “had a major impact” on their primary voting patterns. Those who approved of Clinton voted for Gore 81% of the time, to just 17% for Bradley. Those who disapproved of Clinton gave 58% of their ballots to Bradley, with just 35% for Gore.
  • Political scientists often focus on ideology — whether voters are liberal, moderate or conservative — as an important factor to consider in evaluating how primary voters behave. This variable can indeed show some correlation with how voters choose candidates, the author notes, but the pattern from 1976 to 2000 varies widely. For example, in the 2000 primary, Al Gore won between 68% and 76% among voters from all three types of ideology; the Republican race in 1992 saw a similarly balanced pattern for Bush. But in other races, as with Ford in 1976 and Carter, Kennedy, and Reagan in 1980, there were sharp splits among liberal and conservative wings of the parties.
  • There is a more consistent relationship between a party’s identifiers — those who are full-fledged registered Democrats and Republicans — and how they vote in the primary. Independents, who are allowed to vote in many state primaries, have often made up a fifth to a third of voters. It appears that “voters appear to recognize gradations of partisanship among the candidates. Some candidates embrace partisan symbols and themes quite readily; others have a more independent or even antipartisan appeal.” Thus, in 2000 both Bill Bradley and John McCain were attacked for being insufficiently loyal to their parties and subsequently lost; both did better with independents, but party identifiers went in mass to Gore and Bush.

Tags: presidency, Iowa, New Hampshire, polling

About The Author