Sunday, January 22, 2017

Nate Silver on how and why the pollsters blew it

Nate Silver has some interesting observations and questions this week.

First, what happened with the polls?   Having been impressed the accuracy of their predictions last time out I followed Silver and his FiveThirtyEight.com's coverage of this race rather closely, and strange as it may seem Silver's analysis now is consistent with what he was saying throughout the campaign. Over and over, FiveThirtyEight emphasized that even though the polls made Hillary Clinton's victory almost certain, there was a high volatility in the race and their prediction should be taken with a large grain of salt. Sure enough, large numbers of people who ordinarily don't vote did this time. These people by definition weren't among the people the polls included. And lots of people made up or changed their minds at the last minute. Bizarrely in my view, people who agreed with me that neither Clinton nor Trump were qualified to be president broke for by far the less qualified candidate- Trump- by two to one!

Although Silver's post-election analysis strikes me as being for the most part right on the money, I might register two notes of dissent. First, while this election did oddly seem to vindicate Trump's strange conviction that charisma rather than an effective "ground game" determines the outcome of elections, campaigns do matter. It's easy to over-interpret unexpected election results like this one. 2016 (please, God!) was a fluke. The degree of anger and frustration among the American people was unusual. It's a tautology to say that it's unusual for people who usually don't vote to vote in large numbers, but that fact should be a warning not to generalize too much from this result. While Silver is probably right in his observation that campaigns matter less than events and the mood of the electorate in determining the outcome of elections, it's easy to overdo that point. Gerald Ford might well have beaten Jimmy Carter if he hadn't tripped over his tongue during the debate while trying to say that Poland didn't accept Soviet domination, Michael Dukakis appeared to be headed toward a landslide victory over George H.W. Bush in 1988, but a savvy Bush campaign and an inept Dukakis one turned the tables, and who knows how many missteps by Al Gore in 2000 cost him his home state of Tennessee and the election?

The second point is that George W. Bush had a far better claim to a mandate in 2000 than Trump has today. I continue to be amazed at how few analysts and historians remember one of the biggest stories of the 2000 election: how a new and (as it turned out) wholly unreliable model for using exit polls to generate projections caused the networks to continue to project Al Gore as winning Florida almost throughout the entire evening on Election Night. There can be no reasonable doubt that the consequences were profound.The projection came just as peak polling hours were beginning in the Western part of Florida, where the greatest Republican strength lay. We will never know by how much the Bush vote in Florida was suppressed by the news that it was now over and the bulk of the state's Republicans needn't bother wasting their time by voting, but it seems unlikely that the result in Florida would even have been close if the networks hadn't made their blunder. Further, we will never know by how much the Bush vote nationally was suppressed by the network's stubborn refusal to admit that they had called the state universally expected to decide the result of the election hours early. But however unwilling the media even now may be to admit it, t seems unlikely that Bush would have lost the popular vote in 2000 if voters in most of the country hadn't been told until relatively late in the evening that in effect Gore had already won.

And then, there are the polls which even now consistently show Trump as the most unpopular incoming president in recent history and having made a bad impression with his handling of the transition. NEOTUS* responded characteristically to those polls via Twitter:


Of course, they were not "rigged" before, and the paranoia inherent in the tweet illustrates anew just what a wild ride we have ahead of us with our first conspiracy theorist president. But self-serving as NEOTUS's* analysis might be, there is a legitimate question as to how seriously we can now trust the polls which underestimated the erratic new president's support in November.

Silver, again, does a good job of breaking down what the polls actually said, and the story of the pollsters in 2016 does look less damning the more closely it's examined. But Donald Trump thrives on people not reading past the headlines, so his barb is apt to be effective especially with his tinfoil-hat wearing followers.

Still, there's a lesson to be learned from 2016 that the pollsters are unlikely to willingly admit. Polls are only as good as the sample they get, and will always be vulnerable to people showing up at the polls who are not accustomed to doing so. Further, polls are weighted in order to give what pollsters believe will give a more reliable picture of the likely outcome. Demographic groups who are seen as more likely to vote are generally polled in greater numbers than those who are seen as more likely to stay home.

There is no doubt that the reputation of political polling took a hit- and a well-deserved one- from the outcome of the 2016 election. And we will undoubtedly be more aware than we have been of the possibility that people who ordinarily don't vote might be motivated in any given election to do so. But as Silver demonstrates, there is a great deal that can be said in defense of the pollsters and pundits if one reads past the headlines to what their analysis actually said.

But it strikes me that the most important thing to remember about the 2016 election is the degree to which it was driven by irrational anger. I suspect those numbers showing Trump as so decidedly unpopular are still influenced to some extent by the bias in favor of habitual voters which helped skew their prediction of the election results. But how long can that anger be sustained? What will happen when Trump has a record to defend? Will Trump's boast that he wouldn't lose any voters if he committed murder in broad daylight on Fifth Avenue in New York continue to hold? Or will his failure to achieve results cause his supporters to turn on him, or at least to descend back into apathy?

Who knows? The Trump administration is likely to be largely a low comedy. But the real story will probably be how long Trump, in Lincoln's words, can fool some of the people all of the time.

Preliminary indications are that he might be able to do it for quite a while, since his followers remain so very eager to be fooled.


*NEOTUS- National Embarrassment Of The United States

No comments: