The polling industry is suffering from a lack of credibility for very good reason. For the second time in a row its predictions for a U.S. Presidential election were off the mark.
America’s pollsters predicted that Hillary Clinton would win the Presidential election with around 3.2% of the popular vote, Real Clear Politics average of polls indicates. Instead, she finished the contest ahead by .2% of the popular vote a 3% margin of error. That enabled Trump to get enough votes in the Electoral College; which ratifies American presidential elections, to win.
Despite the media’s claims this is actually a replay of what happened in 2012. The Real Clear Politics average of polls predicted that Barrack Obama would win that contest by .7%. He actually won by 3.9%; so the margin of error four years ago was actually slightly higher than it was this year.
Nor is this just an American phenomenon something very similar occurred in the last British general election in May 2015. Polls had predicted that no party would win enough votes to form a government. Instead David Cameron’s Conservatives won a comfortable majority.
It looks as if the polling industry was getting it wrong and nobody; expect possibly Donald J. Trump and Reince Preibus seemed to notice. What went wrong? How did the pollsters fail so badly?
Why the Polls were wrong
The pollsters’ failure probably rests in the methods they use to gather their data. Pollsters use a method called sampling in which they find a small group of people who are supposed to represent the entire population and question them on issues or candidates.
There are two big things that can go wrong with this strategy; the sample might not reflect the population, and those questioned might not give honest answers. A strong possibility is that both of these problems occurred in 2012 and 2016.
Why Pollsters Missed a Lot of Americans
Most of the polls in the United States use landline telephones to contact participants. That’s a huge problem because 47% of the U.S. households surveyed by the Centers for Disease Control and Prevention (CDC) in 2015 had no landline. A related problem is that a percentage of those using landlines rely on Voice over Internet Protocol or VOIP phones.
This means that pollsters might be missing a lot of Americans which can badly skew their numbers. It is problematic in Presidential elections because those most likely not to own a landline are working class whites (likely Trump voters), or working class African Americans and Hispanics (likely Obama voters).
A related problem is that working people are more likely to be at work when the pollster calls. Persons with working class jobs; like waitress or truck driver,are more likely to be working the swing shift and not at home if pollsters call in the evening.
That means the only persons they hear from are the retired, stay at home moms, the unemployed and middle class office workers with nine to five jobs. That sample is not representative of the population of the United States.
The situation is made worse by the widespread availability of caller ID. On such devices pollsters’ numbers will look like those of telemarketers or debt collectors; two calls nobody wants to take. That means they simply ignore the call or hang up when they see such a number.
How Dishonesty Hurts Polls
The other problem the pollsters face is that of honesty. Their whole methodology is based on the presumption that the answers will be honest.
That was problematic in an election like 2016 when two unpopular and controversial candidates were on the ballot. An African-American, a Hispanic, a Mormon or an educated white who was planning to vote for Trump might have lied about it; because Donald is widely viewed as a racist. Likewise a working class white who was planning to vote for Clinton might have lied about it; because the guys on the factory floor like Trump.
All it would take is a small percentage of those sampled to lie to skew a poll and there would be no way to know about it. Likewise, pollsters would only have to miss a small percentage of the population to get it wrong.
Given these variables we must ask ourselves why anybody still believes the polls. It looks as they are completely unreliable and possibly little better than tea leaves in a cup.