Commentary: There are limits to polling in primaries

By Berwood Yost

There are lies, damned lies, and statistics, as the saying goes. Then there are polls, the most damnable source of statistics, if you believe their critics.

Election results in Pennsylvania's recent primary seem to support some skepticism. Polling averages predicted easy victories for both Donald Trump and Hillary Clinton, which did happen.

But the individual polls used to create those expectations were scattered. Polls in the last month showed a Trump lead ranging from eight to 26 points (he won by 35) and a Clinton lead ranging from six to 28 points (she won by 12). In the end, the polls overestimated Clinton's margin of victory and underestimated Trump's.

This reveals a fundamental truth about primary polling: It is much less reliable than polling conducted during a general election. And it becomes more accurate the closer to primary Election Day it is conducted.

In the 2012 primaries, for instance, polls conducted the day prior to the election were somewhat accurate, missing by about four points, according to the website FiveThirtyEight. Polls conducted three days prior to the election missed by about seven points, and those that were conducted 10 days before the election missed by even more.

This variability may support polling skeptics, but only if one ignores three realities that pollsters confront during primary elections.

First, voters pay less attention to primary elections, so they are more likely to be influenced by election reporting, new information, and notable campaign stories. Even perceptions of momentum can swing voters from one candidate to another. Swings in polling are one result of this inattention and uncertainty.

Second, voter uncertainty in primary elections is increased because people are voting for candidates who represent a single party. The competing candidates' issue positions are much more similar in a primary than they are in a general election, when candidates from opposing parties face each other. In short, guidance provided by issues and parties is mostly absent in closed primaries. This increases voters' uncertainty about which candidate they will vote for and also makes their preferences less stable and less certain.

Third, voters in primaries are less certain about whether they will vote. The most important factor in getting accurate poll results is determining who will vote, but primary elections make it harder to figure this out because far fewer people vote in primaries than in general elections.

For example, primary turnout in even years since 1996 has exceeded 30 percent only twice among Republicans and only three times among Democrats in Pennsylvania. Compare this with the general elections during this same period, in which turnout has never been less than 40 percent of registered voters.

These three features of primary elections, each a product of voter uncertainty, help explain why primary polls vary and, on occasion, pick the wrong winner.

Poll consumers should remember that even the best-designed and -executed polls are really just pictures of the moment when they are conducted. In primary elections, these snapshots can become quickly outdated because voters are uncertain about so much.

So it seems reasonable to use primary polling to understand voters' characteristics, interests, motivations, issue priorities, preferences, and how these things change. Just don't expect them to determine the final margin of victory.

The bottom line is that primary polling is fluid, has always been fluid, and will always be fluid. We need to accept that.

Berwood A. Yost is the chief methodologist for the Franklin and Marshall College Poll, and director of the Floyd Institute for Public Policy Analysis and the institute's Center for Opinion Research. byost@fandm.edu