Many of us went to bed in the early hours of November 9 assuming that the United States presidential election had been another “blown call” for pollsters. When only about 60% of the votes had been tallied, CNN anchor Jake Tapper suggested that a win by Donald Trump would put the polling industry “out of business.”
To properly put in perspective, the serious flaws with Tapper’s assertion – which has been repeated ad nauseam by others – let’s use the example of the most iconic event in the 2016 Olympics: the 100 metre-dash. Imagine a sportscaster turning the television off after 6.9 seconds. Justin Gatlin is two full steps ahead of Usain Bolt. He will win and shock the world. Or will he? Not really. The race was not over.
Having successfully polled in the United States in 2010, 2012 and 2016, I know that the votes in the West Coast are not completely tallied on election night. Washington and Oregon vote by mail, and California – with 18 million registered voters – takes weeks to finalize its count. Tapper’s comments, which arrived at a moment when Trump had more votes than Hillary Clinton, fail to point out this important fact: there are millions of ballots left to be counted, and polls are meant to look at the entire nation, not just at the people whose votes have been tallied at 11:30 p.m. Eastern.
A few weeks later, as the count has progressed, Clinton will have an advantage over Trump (in the popular vote) of about two percentage points – consistent with what most national surveys showed. The poll that had Trump winning the popular vote, which was lauded by gullible journalists and comments section enthusiasts, is actually wrong. National vote polls are never supposed to be projected into the Electoral College. Claiming that they do, or should, outlines enormous ignorance.
Sadly, this is where we find ourselves after the U.S. election. As the votes were being tallied, armchair quarterbacks decided to issue statements on the state of polling. Our industry, unfortunately, has spawned cannibals, usually in the form of “table-less” pundits and pretentious wannabes who have failed in previous electoral forecasts and now moonwalk as experts who are always sure of the outcome the morning after. They are useless and only exist to offer an excuse to reporters who feel wronged when they have not understood how polling works and what it does.
We have been bombarded with media stories that claim all the polls “got it wrong”. It is evident that problems originated in the surveys conducted in three states: Michigan, Wisconsin and Pennsylvania. This might be due to factors such as turnout, stricter voter identification requirements, inadequate sampling and few or no interviews conducted in the final days of the campaign. We should let the pollsters in those states figure out what happened before jumping into conclusions and issuing blanket statements. To claim that “everybody” is wrong and that every poll is “bogus” is illogical.
We at Insights West called 14 races correctly this month: the popular vote nationally and in four states, three senate races, one gubernatorial race and five state ballot initiatives. This is hardly a “failure,” but pollsters are always a target. Our work is not meant to appease anchors when 60% of the vote has been tallied. It is designed to figure out trends and issue an assessment of the electorate in its entirety when all the votes in a specific area have been cast and counted.
The problems of the past motivate pollsters to do better. My experience has been remarkably positive. At Insights West, we have called 22 elections correctly all across North America since September 2013, including the Metro Vancouver Transportation and Transit Plebiscite, where no other polling firm dared to issue a forecast before the ballots were mailed. Yet the narrative continues to be that every poll was, is and will be wrong.
When a surgeon botches an operation, we do not see commentators declaring the demise of medicine. When Bernie Madoff was arrested, we did not witness every investment being labeled as a Ponzi scheme. Why do it in polling, even when arithmetic has proven the assertions wrong? Because it’s easy. A difference of two points in the final voting percentage for a U.S. presidential candidate is statistically valid. The difference between Trump and Clinton, however, is so shocking that a scapegoat must be found. In any case, those of us who are committed to this craft will continue to work and improve with every forecast and every poll we conduct. The interested public would expect nothing less. •
Mario Canseco is vice-president of public affairs at Insights West.
