John Hood: Summer polls are flawed predictors
By John Hood
In this year that sometimes feels like a decade, North Carolinians have yet to cast a single general-election ballot for president or other offices.
But I’m seeing lots of politicos and pundits making confident predictions about the state’s key electoral contests based on data from recent voter surveys. You can count me out of that game. I’ve seen too many North Carolina races narrow in the homestretch. Using summertime polls to predict November outcomes is fraught with peril.
This is certainly true with regard to the presidential race. North Carolina has proved to be a key battleground in recent cycles. The Tar Heel State voted for Barack Obama in 2008 by less than a percentage point. In 2012, North Carolina flipped to the Republican candidate, Mitt Romney, who won it by two points.
Across the eight publicly released polls of North Carolina voters taken during August of 2016, Hillary Clinton led Donald Trump by an average of two points. As we now know, Trump would go on to beat her in North Carolina by nearly four points (although Trump actually got a smaller share of the vote than Romney had four years earlier, due to a higher share of 2016 votes going to Libertarian and other alternatives).
Let’s also remember what happened the first time Thom Tillis ran for U.S. Senate in 2014. Across four midsummer polls, incumbent Democrat Kay Hagan led Tillis by an average of 1.5 points. In November, Tillis defeated Hagan by 1.7 points.
Even in the North Carolina gubernatorial race, where the conventional wisdom has Democratic Gov. Roy Cooper fated to defeat Republican Lt. Gov. Dan Forest in 2020, recent history makes a case for caution. At this time in 2016, Cooper was leading then-Republican Gov. Pat McCrory by an average of six points in the August polling. As we now know, Cooper would go on to victory — but by one of the closest margins in the history of gubernatorial elections, just 0.2%.
In other words, swings of 6% or so between August and Election Day are something akin to normal behavior for our state’s electorate.
As I write, the August 2020 polls have Trump and Joe Biden tied in North Carolina, Cal Cunningham up five, and Roy Cooper up eight. Naturally, it is better to be up than down. Democrats understandably feel good about their current position. But veteran Tar Heel Democrats are likely warning their younger colleagues not to get overconfident, not to take anything for granted. That would be wise.
For example, even if Cunningham and Cooper win in November, it is quite possible their margins will have narrowed considerably by then. Again, based on recent electoral history, it is entirely conceivable that Republicans would still win some other statewide contests, for Council of State and judicial offices, and keep at least partial control of the state legislature under such a scenario.
While it is true that ticket-splitting is rarer than it was a generation or two ago, when as many as a quarter of North Carolinians might vote Republican for federal offices and Democratic for state and local ones, there are still enough true swing voters to tip the balance to one major party or the other on any given Election Day.
That’s because those major-party coalitions are so closely balanced in North Carolina. Disregard party registration, which doesn’t necessarily reflect voter behavior. When pollsters for Emerson College asked North Carolinians which party’s candidates they will support for Congress — without mentioning any candidates by name — 44.6% said Republicans and 44.2% said Democrats. The latest Civitas Institute survey had it at 43% Democrats and 42% Republicans.
Will some unforeseen event tip undecided voters decisively to one side or the other? Will some voters vote “strategically,” consciously splitting their tickets so neither party exercises unchecked power? Will one party’s coalition turn out to be more enthusiastic and energized to cast their ballots when it counts?
I don’t know. Neither does anyone else.
John Hood is chairman of the John Locke Foundation.