The Polls and The Presidential Primaries
Once again the polls published before the New Hampshire primary demonstrated the difficulty of forecasting a primary. For most of his career George Gallup shunned pre-primary polling. Wise man! The problems are almost insurmountable.
During the week leading up to the 2000 New Hampshire primary an astute poll watcher would have expected a close race between John McCain and George W. Bush in the Republican contest and an easy victory for Albert Gore on the Democratic side. But none of the polls published in the primary day issue of Hotline came close to forecasting John McCain’s 18 point win over George W. Bush. The closest poll had the size of the win at 12 points. The others all showed it closer. The polls faired slightly better with the Democrats. Three polls were close to Gore’s 4 point triumph over Bill Bradley. The other three polls forecast much larger margins of victory.
Here are some of the reasons why primary polls are more difficult to get right than general elections.
- LOW TURNOUT
Turnout is usually low in primaries, even those with extensive national media attention. The Iowa Caucus had about 6 percent of the voting age population as participants in the Democratic and Republican caucuses combined. New Hampshire, which set a state record for Republican turnout had a 27 percent vote. The Delaware Democratic primary four days later had fewer than 12,000 voters casting ballots, or 2 percent of those ages 18 or older. Delaware Republicans managed to double that. Screening for likely voters is not nearly as straightforward as it is in a general election, nor is the methodology as highly developed.
In some primaries only registered members of a party can vote. Others, like Iowa and New Hampshire, permit unaffiliated voters to declare on primary day at the polling place for one of the parties. Still others, like South Carolina and Illinois, permit anyone to vote in a party primary. In New Hampshire pollsters had to "guess" at how many independents to include in their poll for each primary along with registered party members. The consensus figure in the New Hampshire pre-primary polls was about one-third. As luck had it, 40% would have been closer to what actually happened in the election. Also, the independents voted differently than registered Democrats and Republicans. They cast more of their votes for McCain and Bradley. This error by the pollsters could have accounted for only a couple of points of their errors, not all of it.
Most primary campaigns take place during a very short period of time as compared to a general election. Also, there are almost no core constituents for candidates as there are in a general election, where party loyalists line up behind their party’s candidate. In a primary almost every voter can be swayed to almost any candidate. The short campaign and the lack of commitment make for a volatile electorate. The further away from the day of the primary a poll is conducted the more likely it is to differ from the final result, even if it was accurate at the time it was taken. Even polls done up to the last day will not pick up those voters who make up their minds in the polling booth. Of course only the final poll conducted by any organization should ever be described as a forecast or prediction, and only then if it is conducted in the last few days of the campaign.
- DELEGATE ELECTIONS
In some primaries one does not vote for the candidates. One votes for only delegates. A voter typically gets to vote for several delegates. Sometimes the delegates’ presidential candidate preferences are not marked on the ballot. The New York and Pennsylvania Republican primaries are like this. If the voter does not bring a list of which delegates support his or her preferred candidate the voting becomes guesswork. The winner in these primaries is usually the candidate with the best grass roots organization. In other states voters may split their vote by choosing delegates from different presidential candidates. Polls in any of these primaries may say more about their presidential candidate preference than the primary result, but any resemblance between the results of a primary in any of these states and a pre-primary poll are purely coincidental.
- PARTY RULES
California is this year’s poster child. It will be possible for John McCain to get the most votes in the California primary and not get a single Republican delegate. All candidates from all parties will appear on the same list. Any voter can choose any one on the list. But the parties will only count votes cast by voters registered in their party. Polls for this contest will have to specify which outcome they are forecasting, the statewide vote or the party delegate contest. The Democratic Caucus in Iowa presents another apples-and-oranges result. Caucus participants have preferences that are not accurately reflected by the delegates selected. The party’s official result is measured by the share of the delegates a candidate wins. At this year’s caucus 8 percent of the participants were uncommitted, but thanks to the party’s rules only 1 percent of the delegates are uncommitted. Again, which outcome is a poll forecasting?
Most pollsters understand these pitfalls, and they do polls anyway. Our advice: caveat emptor!
For more information about this and other polling issues, contact the NCPP Polling Review Board Members.