Our Q&A, conducted by email, is below.
WHAT MATTERS: There’s so much polling out there. What are three (or four?) rules of thumb people should consider when they read about a new poll?
ENTEN: Three rules.
View it in the context of other polls in that same race. If the poll is an outlier, you should ask yourself whether anything in that race has changed for an outlier to anything more than a statistical oddity.
As with anything else in life, you should know who you’re doing business with and if you don’t (i.e. you’ve never heard of the pollster), the result isn’t worth anything.
Polls are tools and only tell you what’s currently cooking. They aren’t predictive. The margin of error is there for a reason and recognize that it applies to each individual candidate. You have to double the margin of error to know the full scope of uncertainty when looking at the difference between the candidates.
Polling averages are valuable
ENTEN: Averages are simply more accurate over the long haul. In any race, some polls will beat the average. Few pollsters beat the averages consistently.
What are signs of a poll to avoid?
WHAT MATTERS: Not all polls are created equal and we don’t include all polls in our average. What should set off alarm bells about a poll?
ENTEN: I hinted at this earlier, but if you haven’t heard of the pollster or they aren’t transparent about their methodology it should be sirens blaring.
Polls sponsored by a group affiliated with a campaign or a cause are, when released publicly, often biased to favor that side. And of course, outliers happen. So always look at other polls in a race, when available.
What are signs of a good poll?
WHAT MATTERS: What are some things to look for that suggest a very good poll?
ENTEN: Transparency is a big one. Good pollsters have little to hide, and those who are transparent are on average more accurate.
There are plenty of good pollsters who don’t follow this one, but live interview polls that call cell phones and weight by age, education, gender, race and region tend to be the best. One pretty safe rule is that if a major news organization is sponsoring the poll chances are the poll is pretty good.
How pollsters are reaching people
WHAT MATTERS: I don’t have a landline telephone and I don’t answer calls from numbers I don’t recognize. How are pollsters reaching people now?
ENTEN: I don’t either. Pollsters use an array of methods from live interview polls to internet polls to text polls. Response rates are about as low as ever, but the good pollsters make multiple attempts to reach those who don’t answer the poll on the first go around. The good news is that the people who do answer the polls are as likely to be Democrats as Republicans, once weighting occurs.
What good are national polls?
WHAT MATTERS: The current polling suggests a decent lead for Joe Biden at the national level over President Donald Trump. But it’s not a popular vote that wins the election. Why pay attention to national polls?
ENTEN: A few reasons and stealing a bit from an article I’m writing.
First, we get a very good understanding of the issues on the voters’ minds (e.g. the coronavirus) from national polls.
Second, we get a good understanding of how groups of voters (e.g. those under the age of 30) are deciding on who to vote for.
Third, national polls are more accurate on the whole than state polls. This is important because we know, for example, that if Biden is up by say 8 points nationally, he probably is up something close to that. The chance anyone loses the electoral college when up 8 points nationally is basically nil.
Look past the horse race numbers
WHAT MATTERS: Take us past the horse race. What are some key things you look for that signify strength or weakness in a candidate?
Who is going to decide this thing?
WHAT MATTERS: Who will decide this election? Undecided voters? Committed partisans?
ENTEN: The simple answer is both. The longer answer is undecideds. For instance in 2016, Hillary Clinton probably would have barely squeaked by with similar Black turnout as that occurred in 2012. She would have crushed Trump, however, if she would have just held onto Barack Obama’s numbers with whites without a college degree. I think the same is true this time around. Persuasion is more important than turnout, in my mind.
This election is not over
WHAT MATTERS: Is it possible for a politician with a 43% approval rating (Trump) to win reelection. If so, how?
WHAT MATTERS: Trump’s approval rating has been remarkably consistent. What’s something that’s changed in recent polls that’s surprised you?
ENTEN: Underneath the hood, there’s been some movement. Trump’s doing poorer with White and older voters than he did in 2016. At the same time, he’s doing better with Black and Hispanic voters. I don’t know if you read a lot about that or would expect that.
It’s a big year in polling
WHAT MATTERS: Name one key misconception about polling you’d like to fix right now?
ENTEN: There is ZERO proof that Trump voters are answering the polls at a lower rate than Biden voters. There is zero proof that Trump voters are lying about their answers to pollsters.
What if the pollsters are wrong?
WHAT MATTERS: What’s one question I should have asked but didn’t?
ENTEN: What happens if the pollsters are right in 2020? What happens if they’re wrong? The answer to the first is nothing, probably. The answer to the second would be a lot more interesting. The race is close enough where I could foresee a polling miscue that allows Trump to win. The aftermath of that would be interesting.