Skip to content
Link copied to clipboard

Commentary: Polling industry faces challenges, but still worth the investment

Polls prior to the March 8 Michigan primary were showing Hillary Clinton with a double-digit lead over Bernie Sanders. However, as we now know, Sanders eked out a win over Clinton in the state.

Polls prior to the March 8 Michigan primary were showing Hillary Clinton with a double-digit lead over Bernie Sanders. However, as we now know, Sanders eked out a win over Clinton in the state.

Professionals in the survey research community are fully aware of the challenges facing public opinion polling in general, and preelection polling in particular. It is no longer simply a matter of selecting a sample of respondents that is representative of the population and contacting them.

A half-century since polling became standard operating procedure for governments, companies, and other organizations, there are three huge hurdles posed by this social science in 2016.

First, with the proliferation of cellphones, it is estimated that roughly half of the American voting population consists of "cellphone only" users. This is problematic because cellphone numbers are portable. This means that a respondent may have a Michigan telephone exchange but currently be living in California, for example.

Second, polling experts cite low response rates as a possible cause of unreliable polls. Polling response rates, the percentage of those contacted who actually respond to a poll or survey, were 70 to 80 percent 50 years ago, but today the percentages are in the single digits. Moreover, those who do respond may be quite different in their attitudes and behaviors from others in the population from which the sample is drawn.

And third, in preelection polls, the ability to identify those in the sample who are likely to vote on election day is, quite simply, impossible. This problem is particularly acute in statewide caucuses and primaries, in which voter turnout is typically much lower than in nationwide general elections.

Such problems no doubt influenced the Gallup organization, for decades the preeminent polling firm in the United States. In October, Gallup announced that it was suspending its 2016 preelection polling.

Yet in spite of all this, public opinion polling in general, and preelection polling in particular, continue to be extremely reliable. This is no less true of statewide caucuses and primary polling.

Aside from Michigan, polling this primary season has had a solid record. To illustrate, let's look at the primary and caucus results in Iowa, New Hampshire, and South Carolina. And in the interest of simplicity, let's look only at the first- and second-place finishers:

In Iowa, on the Democratic side, Clinton and Sanders virtually tied. The polls were showing 48 percent for Clinton and 44 percent for Sanders. On the Republican side, Ted Cruz got 27 percent of the delegates and Trump got 24 percent. The polls were showing 23 percent for Cruz and 28 percent for Trump.

In New Hampshire, on the Democratic side, Sanders got 60 percent of the vote and Clinton got 38 percent. The polls were showing 55 percent for Sanders and 41 percent for Clinton. On the Republican side, Trump got 35 percent of the vote. The polls were showing 30 percent for him.

In South Carolina on the Democratic side, Clinton got 74 percent of the vote and Sanders 26 percent. The polls were showing Clinton at 60 percent and Sanders at 28 percent. For the Republicans, Trump got 33 percent of the vote and Marco Rubio got 23 percent. The polls were showing 32 percent for Trump and 18 percent for Rubio.

Inherent flaws notwithstanding, polls continue to provide highly accurate estimates. When polls miss the mark by a wide margin, as in Michigan, it makes the news. But it's an exception to the rule. Criticizing polls may be one of America's favorite hobbies, but they are here to stay.

From Wall Street to Main Street, organizations of every kind use survey research for a variety of reasons. And virtually every serious candidate for a House or Senate seat or the presidency uses polls extensively - at not insignificant costs. If polls were unreliable, they would not be wasting the money.

Len Champney is a professor of political science at the University of Scranton. champney@scranton.edu