Pollsters were convinced Clinton would win in 2016. Can they be trusted in 2020?


According to the polls, Democratic candidate Joe Biden is the frontrunner to win the 2020 election against his Republican opponent, President Donald Trump. As of this writing, FiveThirtyEight.com, which aggregates and analyzes polls, gives it a 76 percent chance of winning; All eleven of the most recent polls listed on RealClearPolitics predict a victory for Biden with an average gap of nearly six points.And yet – as anyone who watched the 2016 election remembers – pollsters heavily favored Clinton to beat Trump while she was the Democratic candidate. Clinton supporters who were confident of his victory on election night in 2016 are allowed to feel once bitten, twice shy about trusting the polls again. Should we all feel equally suspicious of polling companies as the 2020 election looms?

To answer this question, one must first understand the science of polls – and how it works (or fails, in the case of the 2016 presidential election).

“It’s sampling theory, which is simply the idea that if you interview a subset of the population [and] if the sample is drawn at random from that population, then the distribution you get from the responses should reflect the characteristics of the entire population within a known margin of error, ”said Alan Abramowitz, professor of political science and polling expert at Emory University “It’s science and it can be used not just for human populations but in many other areas like quality control where you could have an assembly line and draw yourself a sample of the products to test for defects. It is used in many fields. ”

As Abramowitz explained, there are good reasons to question the accuracy of polls.

“The problem with public opinion polls these days is that since the response rates are so low in telephone polls, you start using internet polls. [and] you have to use a lot of special techniques to try to find a sample on the Internet that matches the characteristics of the population, ”said Abramowitz. It is very, very difficult to say how accurate the polls are. ”

Abramowitz pointed out that the post-2016 polls have been quite good. “They’re not always right, but in the 2018 midterm election if you came down and looked at the polls average for all races – like the governor, the senator – and where they were at the polls for the House districts, on the whole they were pretty specific. ”

Dr. Christopher Wlezien, professor of government at the University of Texas – Austin, held similar views on contemporary polling practices. Wlezien said the story of the U.S. poll began in 1916 when Literary Digest asked readers to let their editors know by mail how they were going to vote – an unscientific approach, as Literary Digest readers did not represent a random sample of the population. More scientific polls began to be used in 1936, after Gallup correctly predicted that President Franklin Roosevelt, a Democrat, would defeat Republican challenger Alf Landon using scientific methods. (Literary Digest sadly predicted a Landon victory.)

At the same time, Wlezien explained, the challenges of conducting accurate polls, as the main methods of conducting them have changed. Correspondence surveys have been replaced by door-to-door surveys, telephone surveys, and online surveys over time, with each practice having certain sets of advantages and disadvantages. The overall goal, again, is to ensure that the sample size is large and diverse enough to accurately represent the larger population, so that the data pollsters provide to the public are statistically likely to match the real public sentiment.

So, did a systematic error arise when Clinton was seen as the likely winner against Trump in 2016?

“The American Association for Public Opinion Research did an analysis of the 2016 polls,” Wlezien said. “And a part was published in a large report and there is a smaller part which was published as a newspaper article. I happened to be on that committee of that working group. We know a little bit about what we did and what the analysis basically shows is that the national polls were pretty good. In fact, they were probably better than they were in the past. ”

He added that while many state polls were good in 2016, “a lot of state polls weren’t, especially those of key states. And we learned a lot in this report about what polling organizations did wrong with their polls and it has a lot to do with the underweighting of non-university types. And those who got that right were a lot more specific than those who didn’t, which was a pretty wrong number. And even to the extent that some were doing it right, a lot of the withdrawal organizations withdrew from certain states early, so we couldn’t really pull together the late movement, which was towards Trump. ”

He noted, with a bit of luck, that “my understanding is that the traction organizations are aware of what went wrong in 2016 and that states, and at least some states, have corrected that. ”

Ultimately, the key to fully understanding polls is realizing that they are not meant to be prophetic.

“The problem is, of course, that polls are not predictors,” Dr. Allan Lichtman, an American University historian who studies the science behind elections, told Salon. “I repeat: polls are not predictors. They are abused as predictors. The easiest thing in the world is to write an article about polls. You don’t even have to get up in the morning, just read the polls and write the story. And of course we also have survey compilers like Nate Silver [the statistician who founded FiveThirtyEight.com], whose forecasts are no better than the polls themselves. So that was the real problem with 2016, it was using polls as predictors when they are not predictors. ”

Lichtman pointed out that state polls can be unreliable and “tend to have large margins of error.” He also noted that “the national poll was only a point or two out of the plurality Hillary Clinton compiled” in 2016, meaning they weren’t far off the mark.

Lichtman also revised his own presidential election prediction system, a system that has successfully anticipated the results of every presidential election since 1984. It uses a series of true or false statements that anticipate whether the outgoing party’s presidential candidate will be elected in the future. a given period. year. If six or more statements are false, the incumbent will lose; if less than six are wrong, he or she will win. His system led him to make predictions that were rejected by expert consensus at the time: he predicted that Republican candidate George HW Bush would beat Democratic candidate Michael Dukakis in 1988, even when Dukakis was ahead in the polls by 17 points, and was one of the few experts to publicly predict that Trump would beat Clinton in 2016.

“I’m sitting here in my office and I have a note over my shoulder written about the Washington Post interview where I predicted Trump’s victories. And the note says, ‘Professor, congratulations! Good call! ‘ in a large Sharpie letter signed “Donald J. Trump”. ”

Yet as Trump appreciated Lichtman’s prediction in 2016, the professor noted that “he did not understand the deep meaning of the keys, which is that governing, not campaigning, matters. And when you’re the outgoing president, you’re going to be judged on your case. And rather than meeting these challenges in a substantial way, [Trump] returned to his 2016 playbook when he was a challenger and the result was a disaster for Trump’s re-election prospects. ”

In 2020, Lichtman predicts that Biden will beat Trump, citing a poor economy, widespread social unrest, his scandals, his lack of major foreign policy successes, his losses in the 2018 midterm election, and the fact that he failed is not popular (or “charismatic”) beyond its own base. Lichtman’s only caveat was that, because Trump slowed down the post office and embarked on voter suppression, the election could be “stolen.”


Please enter your comment!
Please enter your name here