Unless you were completely oblivious, you saw what happened with political polling in the 2016 election. In a word, it was an unmitigated disaster.
In evaluating the polls in the run up to the election, various organizations analyzed the results, and in so doing made bullish predictions about the likelihood that Hillary Clinton would be the next president.
The Princeton Election Consortium was so certain of the polling that they gave Clinton a 99 percent chance of winning.
Huffington Post, rather famously, gave her 98 percent chance of winning.
The New York Times, somewhat less optimistic, gave her an 85 percent chance.
One of the most favorable evaluations of Donald Trump’s chances at becoming president, Nate Silver’s FiveThirtyEight, still said that Clinton had a 71 percent chance of winning.
Obviously we know what happened. The polls were wrong. Trump won.
This is hardly the first time this has happened.
When I still lived in Virginia, I was involved in the 2013 gubernatorial election there. Throughout that race, Democrat Terry McAuliffe led Republican Ken Cuccinelli by a wide margin, and no one in the country thought it was a real race. Two weeks before the election, McAuliffeby double digits. One poll had him up 11 points. Another had him up 15 points. Another, 17 points.
On election day, McAuliffe did win, but he only won by 2.5 percentage points. Had national conservative groups understood that the race was actually close, they likely could have helped push Cuccinelli over the top.
Here in Maine, I’ve seen election after election be polled wrong.
In 2010, for instance, the final four polls of the election — all of which concluded a week or less prior to Election Day — showed then-Waterville Mayor Paul LePage with a lead of 14 percentage points, 6 points, 12 points and 19 points. LePage ended up winning by just 1.8 points.
In the 2014 gubernatorial race, the final four polls of the election showed the race tied twice, and Democrat Mike Michaud with a small lead twice. The final results gave LePage a 5-percentage point win.
Over in Maine’s 2nd Congressional District that year, the results were no better. In late October, Pan Atlantic released a poll showing Democrat Emily Cain with a lead of 1 point over Republican Bruce Poliquin. Poliquin ended up winning by 6 points, even with 11 percent of votes going to a conservative independent candidate that year.
And, 2016 was no better. In late October of that year, the Portland Press Herald commissioned a poll of Maine’s 2nd Congressional district that showed Cain beating Poliquin by 1 point. In the actual election, Poliquin won by nearly 10 points.
That same poll showed Clinton “surging” in the 2nd District in her race against Trump, showing her with a lead of 3 points. Trump ended up winning the 2nd District by 10 points.
Understand why this is happening.
Right now, as widely respected polling expert Nate Silver himself has said, there is ain political polling. It is getting less and less accurate, and is frequently predicting the wrong outcomes and margins on a number of races across the country.
For a poll to be accurate, it first must sample the correct universe of people to give a true snapshot. Pollsters build models based on assumptions — educated assumptions, I grant you, but assumptions none the less — about who they believe will vote in the coming election.
This is even more important today than 10 or 20 years ago, as response rates to polls have fallen off a cliff, leading pollsters to rely increasingly on.
In short, pollsters often times aren’t polling the right people. This is why we see them wrong so often.
Which begs the question, why do we still care? Why do we still take them seriously?
More importantly, why does the media insist on obsessing over them? No matter how many times we see the same story play out, we start it all over again the next election cycle. Inaccurate and misleading polls are conducted. The media salivates over them. The public gobbles them up, and narratives about the state of campaigns are created out of thin air, often in direct contradiction to the reality on the ground.
Rinse. Wash. Repeat.