In the Virginia governor’s race, Democrat Ralph Northam is leading Republican Ed Gillespie by 17 points, according to a newly released Quinnipiac poll. The survey shows 53 percent of likely voters backing Northam and only 36 percent supporting Gillespie. But last week, a Hampton poll showed Gillespie leading Northam by eight. Divergent results like this often confuse election watchers, so it’s worth asking—which version of the story is right?
The answer: maybe neither.
The best way to get a handle on polling is to not worry too much about individual polls and instead take the average. And according to RealClearPolitics, the average of polls shows Northam ahead of Gillespie by 3.3 points. That’s a real lead, but it’s not a completely safe one. Polls aren’t perfect, and public opinion can move late in an election— making it plausible to turn a modest Northam lead into a safe one or a Gillespie win.
At this point, I wouldn’t bet on a 17 point Northam win or an eight point Gillespie victory. While those margins could appear before the election, the sum of the current polling doesn’t yet point to that.
But that doesn’t mean outliers are all bad or useless. Surveys should be allowed to vary somewhat—good pollsters will use different methodologies, sample different people and naturally come up different results. And while a 25 point spread (the difference between an eight point Gillespie win and a 17 point Northam win) is big, it’s worth thinking before shaming either pollster if one (or both) end up far from the result on Election Day. Shaming pollsters can lead to “herding”—where pollsters adjust their results to make sure they don’t stand out too much from the current consensus. And, as the American Association of Public Opinion Research has argued, herding can reduce the accuracy of the polling averages.
