This is getting redundant, but if the FTC regulated pollsters, most would be pulled off the market.

Much has been written about how much pollsters missed the mark in the 2016 Presidential race, predicting Clinton victories in states Trump won handily and predicting Clinton landslides in states she wound up losing.

Polls didn’t get any better in Virginia this year.

Various polls showed the battle for the Democratic nomination for Governor to be close between Perriello and Northam, and a poll released four days before the election had Perriello ahead.

Northam won by 12 points.

On the Republican side, Ed Gillespie led polls for months by double-digits (as high as 28 points in one poll).

One online poll by a new firm gave Corey Stewart a one-point advantage in the final week, but the same poll predicted Perriello winning by 8 and he lost by 12.

Gillespie edged Stewart  44-43.

Bottom line – the polling in these gubernatorial primaries was wrong, wrong, wrong, and even when one of them got one race fairly close to right, its other numbers were way, way off.

There’s something wrong with the models these firms are using.  Either simply polling “registered voters” needs to be eliminated, or whatever predictive model for who is turning out needs to be overhauled.

Throughout the primaries, polls were done for who is ahead in Gillespie vs. Northam, and for months polls showed Gillespie trailing.

Suddenly, a new poll shows the race tied already.

Voter confidence (and my own) in these polls is almost zero.  The world has changed.  People don’t answer landline phones, and they are online a heck of a lot more than they used to be.

The key to polling used to be in a random sample, in that everyone had an equal chance of participating.  Is that even possible anymore?  It’s obvious that large numbers of voters aren’t being included.

As for the models, predicting voter turnout is almost laughable now.  More Republicans voted in June’s primary than in 2012 and 2015, and by a large margin.  More Democrats voted in their primary than usual, too, and pundits said high turnout was Perriello’s key to victory.  Not so.

Polling companies will make lots of money and the press will tout them as gospel, but if they don’t get an election right, sooner or later they have to be held accountable.

Don’t they?

  • I think the biggest issue is simple – these polls only include folks who answer them. I think most Stewart supporters, as with most Trump supporters, aren’t going to tell a random stranger how they’re voting.

    • John Massoud

      I’ll agree with you. I was knocking on doors for Team Gillespie in Shenandoah County, and got an awful lot of refusals – more than usual. Kinda led me to know something wasn’t right, although everyone else said they were supporting Ed.

    • BrianKirwin

      Except it’s not just Virginia. Polls in Georgia showed a Democrat win for weeks.