Polls badly done, yield inaccurate election forecast

Polls are rolling out almost daily. They persistently show Republicans Donald Trump at or near the top of his party’s contenders and Democrat Hillary Clinton running away with the Democratic race.

But the polls are probably wrong. In fact, they may be so far off the mark that we will have to wait for the primaries to know the real frontrunner. Unfortunately, beliefs about candidate strength, based on dubious polling, could influence voter turnout and preferences.

One of the major factors strongly influencing the value of a poll is the ability of the polling organization to develop a sample of people that validly represents the “universe” of people whose opinions are being measured.

The best way to achieve this goal is the purely random selection of participants from a large number of people. For example, picking every tenth name from a list of 1,000 people would produce a “random sample” of 100 people.

But let’s say that of the 100 people called, 90 refused to be questioned. The polling organization would then have to pick 90 more and so on until it had 100. But its sample would not be nearly as truly random as the first 100 selected.

While refusing to participate in a survey used to be rare, the example of only 10 percent of the original selection agreeing to be questioned is reportedly not unusual these days.

Most surveys are conducted by phone. Some people don’t have phones. Some people won’t answer the phone except for calls from known callers. Many people primarily use cell phones, which by law cannot be automatically called randomly. All this makes it more difficult to get a fair sample.

And pollsters typically give some people interviewed more weight than others. For example, if the 100 people happen to include 60 women and 40 men, the value of any person may be changed to produce a result better reflecting the proportion of women and men in the survey universe.

Increasingly, polling depends on people, like those with cell phones, volunteering to participate and the pollster later weeding out answers until it gets to the correct number and type of people for the sample. That would hardly pass the traditional “scientific” survey standard.

And, in the case of the 2016 presidential contest, the matchup between candidates assumes the election takes place today. What’s missing is the key final and usually heated phase of the campaign when many people decide whether they will vote and, if so, for whom they will vote.

It’s likely that many people have not yet paid much attention to the campaign. Those that have followed it have heard more about campaign tactics than the issues. That could cause them to decline to respond to polls or give answers that will change as they approach the election.

In addition to all of the weaknesses of polling, which are increasingly undermining their accuracy, this year in the Republican race, there is a major complicating factor: too many candidates.

Trump may be leading in the polls, but about 70 percent of those surveyed favor somebody else. The rest of the field is split among more than a dozen others. The race might look different, if it Trump faced only one or two others.

Will the field thin out? Often by this point in previous races that had happened, as poor polling results discouraged contributors from supporting candidates who appeared to have little or no chance.

Now, thanks to the Supreme Court’s Citizens United decision, immensely wealthy people can contribute unlimited sums to support their candidates. So long as one billionaire sticks with one hopeful, that candidate can stay in the race. Of course, Trump bankrolls himself.

Because several candidates have political sugar daddies, they stay in the race despite low poll numbers. That reduces the likelihood of any of them emerging as the alternative to Trump. Facing a divided field, Trump has remained the frontrunner.

In other words, big money, most of it going to other candidates, serves to keep Trump at the top of the race.

The voters themselves may have to weed out the field. Recently, in the Canadian elections, when the anti-Conservative vote appeared to be split, which would have allowed the Conservatives to stay in power, voters defied the polls and just before the election settled on one of the two alternatives, giving it a big victory.

The weakness of polling and the misleading influence of money suggest that just who is the frontrunner may be a lot less clear than it seems.

Gordon L. Weil

About Gordon L. Weil

Gordon L. Weil formerly wrote for the Washington Post and other newspapers, served on the U.S. Senate and EU staffs, headed Maine state agencies and was a Harpswell selectman.