“Imagine polls don’t exist. Show me evidence Hillary is winning?” That was the challenge that Charlotte radio host Bill Mitchell posed for his Twitter followers just one week ago. It was an interesting request, given that polls have been used to predict elections with relatively strong reliability for almost a century now.
To cast them aside would require a seriously better and more reliable option for election tracking, preferably with a representative sample, no bias, and a good track record. Although we surely don’t advocate for doing away with the polls, let’s look at some of the alternatives in a world with no polls:
{mosads}Social Media: During this election cycle especially, there has been a large push to track social media followings of the candidates. This includes Twitter followers, Facebook likes, subscriptions to supportive subreddits on Reddit, etc.
When not given a ton of scrutiny, on the surface, this might seem like a good idea: the number of people who like or support a candidate online SHOULD correlate with overall support. For several reasons though, it doesn’t.
First, not every person who likes or follows a candidate’s social media account actually supports them. There are people who just want to hear what that candidate has to say, even if just for laughs.
Second, users of social media are a select group: people who have to be able to afford an internet using device and have enough spare time to maintain a social media presence.
Third, remember not all likes and follows are genuine potential voters. Candidates can be liked be citizens of other countries, kids under 18, and even bots and click farms.
Crowd Size: Reference to enthusiastic crowds comes up in every Presidential election. In 2012, Mitt Romney was stunned to see that he had lost Florida despite the size of his rally in Jacksonville. In 2016 Bernie Sanders supporters prepared for a New York upset because of his record setting rally turnouts in the state, and yet Hillary Clinton took 58% of the vote. So why did the method of gauging with crowds fail?
Remember, candidates get to choose where they hold rallies, which usually means in places that already strongly support the candidate. A large crowd in a place that already leans towards the candidate speaking probably doesn’t say much about the overall or even state electorate.
Also, consider that a crowd is only representative of a candidate’s most fervent supporters, and that plenty of voters don’t feel as strongly about the candidate that they choose to support.
Betting Markets: While not suggested by Mitchell, betting markets were often referred to during the lead-up to the Brexit vote because people felt that it represented the sentiments of the electorate, in part due to the fact that people could lie during polls, but only bet if they truly felt a certain way.
Again, the same theme emerges in that bettors are not a representative sample of the electorate. Only a small wealthy portion of voters can afford to gamble.
To make matters worse, bettors usually only have the same information that everybody else does, so they are either making the mistake of betting based on anecdotal evidence, or based on the same polls (that in our imaginary world we have done away with) that everybody else is viewing.
Finally, bettors often bet just to hedge against an unpleasant result. If a bettor is concerned that a candidate would hurt his/her business, they could bet on that specific candidate as a form of insurance.
Fundraising: When a person is willing to not just express support for a candidate but put their money behind a candidate, one might think this is a sure sign of deep support. So, are total fundraising dollars raised by a candidate a good measure of support?
For much the same reason as betting markets, we don’t think people who are giving money to a candidate are representative of the broader electorate. It is not hard to imagine certain very wealthy donors supporting one candidate in a manner that skews the numbers. And, with more money comes more campaign rallies, advertising, and other campaign activities — which could subsequently effect the race.
All in all, polling isn’t perfect. Each poll has its biases and flaws, as well as a level of statistical uncertainty that is often greater than what we might prefer in a perfect world. But this doesn’t mean that we should cast polls aside without a better option in place, and currently there is no better, more reliable option available.
And especially when every poll, regardless of pollster bias, trends in the same direction like they all do right now for Hillary Clinton, we can be reminded that there is still a science behind these polls.
John H. Johnson is President and CEO at Edgeworth Economics and co-author of Everydata: the Hidden Misinformation in the Little Data You Consume Every Day. Adam Dettelbach is a Polling Researcher at Edgeworth Economics.
The views expressed by Contributors are their own and are not the views of The Hill.