I watched the debate last night. It looked like a Presidential debate – but the volume of the President’s falsehoods left me frustrated and angered. From a political standpoint Trump’s “new” behavior was better. But Joe Biden did better as well, able to actually lay out his ideas for America. Scorecard: Biden held his own, and Trump just spoke to the base – Biden wins.
Voodoo
As you might imagine, I become an “election geek” during Presidential contests. That’s been true since Mom pinned a Kennedy button on my sweater in 1960. I followed campaigns from “All the Way with LBJ” to my boss Jimmy Carter’s “Why Not the Best”, to Reagan’s “Let’s Make America Great Again”, to John Kerry’s “Let America be America” (or Bartlet be Bartlet) to Obama’s “Yes We Can”.
So I spend a lot of time with polls. Getting a read on what American voters intend to do is important. But there’s always a caveat, a statement of caution: what about 2016? The pollsters seemed to be so wrong then? How can we trust their numbers now?
Well there are a couple of answers to that concern. First, polling is science not voodoo. Pollsters take the opinion of a limited number of people, and project that into a conclusion on what everyone is thinking. So how that limited number, the sample, is selected, will control the accuracy of the data that ultimately comes out.
Historic Mistake
There’s a famous picture of a smiling Harry Truman holding up the front page of the Chicago Tribune the day after the 1948 election. The headline reads “Dewey Defeats Truman”. Tom Dewey, the Governor of New York, was the Republican candidate for President, and as is apparent from Truman’s beaming smile, Dewey did not win.
So what happened? The Tribune did a last minute telephone poll that showed an overwhelming move by voters to the Dewey camp. They were so confident in the result that they set the headline for Wednesday morning, and went to bed.
The key to the poll’s failure was in the sample. It was 1948, and not every American household had their own phone. In fact, even when I moved to Pataskala thirty years later, there were still “party lines” where several households shared a phone line (I know, it sounds like Mayberry in the Andy Griffith Show). So the sample that the Tribune spoke to was slanted towards people who could afford to own their personal phone line. And that’s exactly the kind of folks that voted Republican. It shouldn’t be a surprise that they were voting for Dewey. But they weren’t a representative sample.
Sampling is critical. And some of the sampling in 2016 was based on the Obama 2012 and 2008 election turnout. Not surprisingly, Hillary Clinton didn’t generate the same kind of excitement as the first Black President did, and didn’t get the same kind of turnout, particularly in minority communities. It wasn’t that she didn’t win those communities; she did, overwhelmingly. But the numbers of people didn’t come to the polls. Think about Detroit, Milwaukee, and Philadelphia: the lower turnout there changed the outcome of those state elections, and who ultimately became President.
Snapshot
Polls in 2016, like the current polling for Biden today, showed a strong Clinton lead through early October. But that month in 2016 was tumultuous to say the least. There was the Access Hollywood video, followed immediately by the Podesta email dump by Wikileaks, then the Comey letter reopening the Clinton email investigation, followed by Comey closing the investigation again.
While we were distracted, the polls significantly narrowed. And that brings us to the second point. Polls are a snapshot of how people feel at that moment. They don’t signify what they believed yesterday, or what they are likely to do tomorrow. In the unstable political environment of October 2016, things changed.
Democrats remember a summer and fall when Hillary had a seemingly insurmountable lead. Even the Trump campaign seemed resigned to defeat. But all of the October craziness tightened the race. What would have been an electoral landslide for Clinton in the first week of October (that snapshot) dwindled to a marginal lead in November. Democrats then, and now, remembered the polling of a month or even week before. But the snapshot changed.
Margins
Marginal is a significant word in polling. Pollsters take maybe 1000 responses, and extrapolate the results over millions of voters. There is always some room for mistakes, called the “margin of error”. When we saw Clinton with 46% to 43% for Trump with a margin of error of 3.5%, it meant that Clinton wasn’t really ahead of Trump. Within that margin Trump could be ahead of Clinton by ½ a percent or more.
And that’s where we were when the actual snapshot was taken and we held the election. Almost all of the polls except for the “national polls” were inside the margin. And Trump won by a razor thin margin in the states he had to take: Wisconsin, Michigan, and Pennsylvania. By the way, the “national polls” were right – Hillary won the national popular vote.
Sample
So who gets polled is critical to the accuracy of the data. The Rasmussen poll, for example, has been slanted towards Trump for the past five years. It’s not because Rasmussen is a “Trumpster” (though he is), it’s that the way he selects the sample favors Trump. So when you see Rasmussen polls favoring Biden (Arizona – Biden 48%, Trump 46% 3.5% margin of error) then you know that Biden is doing well under the “worst” polling conditions.
Generally polls look at three major variables: likely voters, registered voters, and adults of voting age. “Adults of voting age” is the least important. Only twenty-one states have “same day” voter registration, where you can register and vote on the same day. The rest close registration early, so in those states, measuring unregistered adults is meaningless. And even in the “same day” states, folks who haven’t already registered are less likely to actually vote.
Registered voters seems like a “good” sample. 79% of adults of voting age are registered to vote. But in the 2016 election only 58% of eligible adults actually voted. That means that 20% of voters who were registered still didn’t vote. Even though a person is registered to vote, there still a one in five chance that they won’t. And that’s the “turnout” question that Democrats and Republicans debate. “IF” Hillary had a better turnout, she would have won. “IF” Biden can motivate his “base” he will win.
And finally there’s the “likely voter”, the registered voter who consistently votes in election. And that’s the sample that most pollsters like to survey.
Political Geeks
Want to join in the “geekdom” of poll watching? Nate Silver was one of the few who predicted a Trump victory in the last few weeks of 2016. His website, Five-Thirty Eight, not only lists the latest polls, but rates their quality. “A” and “B” polls are regarded as high quality, “C” and “D” polls not so much. If you want to quote polling, look for those “high quality” polls (sounds like Bobby Boucher from The Water Boy).
And if you want more of a cross section of polling, check out Real Clear Politics. While the editorial content slants conservative, their polling averages are right on. Like most general elections, it’s likely the polls will narrow as Election Day approaches. The “undecided” will make their choice, and this year, fully a third of the US vote will be cast before we even get to November 3rd. Their votes, like mine, are in.
As we used to say after the last hard track workout, “the hay is in the barn” for early voters. And for the rest of you: get “your hay in the barn” (go and vote) then watch the polls. Then, in the days and probably weeks after November 3rd, we can sweat out the actual count.
