Don’t trust the polls this year

Polls are the air, bread, and water of politics.  Campaigns plot their strategies around them, news outlets conduct, report and pontificate on them incessantly, and political junkies like me formulate our analyses based on them.

But what if none of the polls are reliable?

In 1948, every poll in the universe predicted that Republican Thomas Dewey would defeat incumbent Democrat President Harry S Truman by a comfortable margin.  They were all wrong (which famously left the Chicago Daily Tribune with egg on its face).  The reason all of the polls were wrong is that they were polling people by telephone, but a large percentage of low-income families didn’t have phones and therefore were never part of any of the polling samples.  Those low-income voters gave Truman his margin of victory.

Jump ahead to 2016.  Nearly everyone has a mobile phone.  They have become so integrated into our lives that the government has deemed them a “necessity” and gives them away to low-income people.  They have become so much a staple of our lives that I recently read a feature news article about a man who doesn’t have one.  That’s right:  One guy whom none of us know decides not to own a mobile phone, and in today’s world that’s international news.

Yet, for the most part, pollsters do not call mobile phones.  They still call land lines almost exclusively.  This despite the fact that as of two years ago, 40% of homes had no landline, and that number has to have gone up since then.  (Disclaimer:  I got this number from the Chicago Tribune, but this time it appears to be well sourced….)

Pollsters don’t call cell phones because of FCC restrictions and higher costs associated with the underrepresentation of households with no land lines.  But a study released last June shows that polls that were based only on calls to landlines badly underestimated support for Barack Obama and overestimated support for Republican Mitt Romney in the 2012 election.  This finding was in contrast with similar polling in 2004 and 2008 that was largely accurate.  So, it is demonstrably true that the landscape has changed dramatically in just the past the past few years with regard to how polls must be conducted in order to be credible.

Pollsters do not seem to be taking this issue to heart, though.  Just yesterday I saw several polls of people who plan to vote in today’s New Hampshire Republican primary.  All of these polls showed Donald Trump with a large lead over the rest of the pack.  But behind Trump, one poll showed Marco Rubio remaining in 2nd despite a disastrous debate performance last weekend. Another showed John Kasich in 2nd.  Another showed Jeb! Bush in 2nd, but yet another released on the same day showed him in 5th.

And we all know that the polls in Iowa predicted a yuge victory for Trump, who ended up losing to Cruz and tying with Rubio.

The take-away from this situation is to take the polls you see this year with a large grain of salt.  Polling might – might – be able to show large trends, like Trump comfortably leading the pack in NH and Bernie Sanders comfortably leading Hillary Clinton in the Democrat race, but they are all over the map where the races are closer.  And in Iowa, even the large trend consistently shown in most polls of Trump having a comfortable lead was not borne out at the caucuses.

In the absence of reliable polls, campaigns are going to be making strategic decisions that could end up fruitless or even backfiring.  But more importantly, the news media, who in modern times are more interested in making news than reporting it, are going to be crafting and incessantly hawking narratives that are based on unreliable polls.

So, don’t take this year’s polls too seriously, and don’t take the news media’s analysis of them seriously at all.