For most of my life, I’ve been a big believer in polls.
OK, there was that time back in ’48 when George Gallup and the rest of the Western Hemisphere, including editors at the Chicago Tribune, wrongly predicted a Dewey landslide.
Gallup fixed its reporting methods – mainly, it began surveying poor people and minorities – and has been pretty much spot on ever since.
Then came 2012. Gallup (and several other polls) had Mitt Romney narrowly winning the presidential election – and could not have been more wrong. President Obama won 51.1 to 47.2. Not exactly a mandate but enough to make Gallup gulp.
Digital Access for only $0.99
For the most comprehensive local coverage, subscribe today.
Even Romney was so confident he would win he planned a big fireworks display on election night. When he lost Ohio despite GOP predictions – a disbelieving Karl Rove even questioned his own Fox News bean counters on TV – he and Gallup went quietly into that dark night.
This primary season polls have been all over the place, to the point where it's difficult to know what's going on.
This primary season, polls have been all over the place, to the point where it’s difficult to know what’s going on.
In Michigan, an average of polls taken in that state had Hillary Clinton winning by 21 points, but when the votes were counted Bernie Sanders had won by 1.5 percent.
Nate Silver, whose FiveThirtyEight.com website correctly predicted the 2012 outcome, called the Michigan experience “among the greatest polling errors in primary history.”
Gallup itself did a post-mortem after Michigan and found, according to a story in The Week magazine, that two things have been contributing to a skewering of the polls.
First was the growth of cell phones. Ten years ago, about 6 percent of Americans had cell phones; in 2014 the number had rocketed to 60 percent. It’s more today.
Worse, according to The Week, federal law forbids autodialing cellphones, so any cellphone polling has to be done manually, which is prohibitively time-consuming and expensive.
Gallup acknowledged that in Michigan it misidentified likely voters, under-counted Democratic regions and dialed only listed numbers on landlines, meaning older voters.
A second problem is that fewer people are willing to answer surveys – a problem that accurately describes my wife and I. We simply check the incoming number and pass on it if we don’t recognize it. Apparently a few million others do the same thing.
In fact, we disconnected our landline because of all the telemarketing calls, and when a Marquette University pollster complained that “telemarketers poisoned the well” for political polling, we understood.
I’ll still pay attention to the polls; it’s the only game in town.
But I believe from now on I’ll be reading them with a large grain of salt. Because if they tell me Dewey’s going to win, I’ll just go crazy.
Contact Bob Bestler at email@example.com.