Here's an idea for post-Election 2014 polling reform: take all the Virginia pollsters and launch them into outer space, because they keep getting everything wrong.
In last year's Virginia gubernatorial race, pollsters showed Democratic candidate Terry McAuliffe leading Republican Ken Cuccinelli by about 6 percentage points on average. He ended up winning -- but only by about 2 percentage points.
Even more infamously, earlier this year, the political world missed then-House majority leader Eric Cantor's shocking primary defeat -- blowout, really -- at the hands of economics professor Dave Brat in Virginia's 7th congressional district. (Brat won his general election tonight, by the way. Congrats!) There wasn't much polling of the race, but the polling that there was was off. Cantor's internal pollster had him up 34 percentage points. Even the Daily Caller's poll from the weekend before the election showed Brat losing handily. Brat won by 11 percentage points.
And then there's tonight's Virginia Senate race. Mark Warner has hardly lost yet, but it's surely sending him into cold sweats. The polling average had Warner up by about 10 percentage points heading into the night. It's not that confusing. He is popular there. Ed Gillespie is a former lobbyist for unsavory clients and a George W. Bush-era Republican operative. Virginia is a tilt-blue state. C'mon.
Nevertheless the race is still too close to call and could go on all night. Gillespie has the paper lead with about 50+% of precincts counted -- although there are still a lot of outstanding votes in the heavily Democratic suburbs of D.C. for Warner to pick up. He can't have a letdown there. Gillespie is doing far too well in rural areas.
Nate Silver still doesn't think the upset's likely. Why, it would break our system of mathematics! Robot heads would explode, computers would think it's 1900, etc:
Warner is still the favorite to pull out the race: Exit polls aren’t necessarily all that informative as compared with pre-election surveys. We’ll be getting more real votes in soon. But for what it’s worth, if Republican Ed Gillespie wins, it would be one of the bigger polling errors in recent memory in a Senate election — a once-or-twice-a-decade type miss. The FiveThirtyEight model gave Gillespie just a 0.5 percent shot (1 chance in 200) of winning based on its pre-election forecast.
Even if Warner does eke it out, it's not going to be by 10. It's going to be much narrower than that.
Can pollsters please figure out why they keep getting Virginia stuff wrong? It's just... Virginia. It's a pretty simple state. You're not polling Alaska or some distant moon colony.