The poll was called “humiliating,” “a whopper for the ages,” a “disgrace” — I said it may have been right, at the time it was done.
Five months ago, everyone was talking about Eric Cantor’s errant poll: the one that showed him 34 points ahead, before he lost by 11. Lots of explanations were tossed around, but most of them amounted to nothing more than data-free speculation.
We now have some data, and while it doesn’t speak directly to my own conjecture, it does rule out some theories propounded by others, while reminding us of the fragility of polling results as we head into Election Day.
Disposing of some excuses was easy. One colleague, exhibiting a knowledge of statistics akin to a middle schooler, suggested that one in 20 polls is just “wrong.” In fact, the chances a poll would show Cantor garnering 62 percent of the vote while he was “really” getting 44 percent are about one in a trillion.
One of the more common explanations turned on turnout. It was about 40 percent higher than in any previous GOP congressional primary in Virginia.
But we now know that more than 70 percent of those who cast ballots in the Cantor primary had voted in some previous GOP primary. If the poll had been right for the previous voters, close to 85 percent of the new voters would have had to have voted against Cantor. That would be the biggest difference between new and previous voters ever recorded.
Which leads to the favorite conspiracy theory of GOP operatives: the Democrats did it. This claim was given some substance when a Democratic operative bragged that he skillfully plotted and planned the whole outcome. Well, if someone tried to goose Democratic turnout, they failed.
Of course, there is a definitional problem: who is a Democrat? Virginia does not have party registration, rendering the whole argument difficult to disentangle and a bit disingenuous.
Fortunately, analytics and a voter file enable us to know exactly who voted and to gain some insight into the partisanship of these voters, even without party registration.
More than 68 percent of those who turned out in Cantor’s primary were hard-core Republicans. Just 7 percent were hard-core Democrats. Even if all of the “Democrats” supported Cantor’s opponent, had not one of them voted, Cantor would still have lost.
The rest of this electorate was largely what you would expect, GOP leaners.
So if it wasn’t statistical quirks or turnout or the first successful Democratic conspiracy in history, what was it?
As I explained in June, my guess is the poll was more or less correct. Two weeks before the election, when the survey was taken, the majority leader had close to 62 percent of the vote. Six days later, another poll by Vox Populi showed his support had shrunk to 52 percent. Then, in the eight days between that poll and primary day, Cantor slipped more, to his final 44.5.
But why re-litigate this battle now? Three reasons.
First, I said I would evaluate the data when it became available, and now it is.
Second, it’s an important reminder that things can change even in a couple weeks’ time. Campaigns work assiduously and spend generously to produce changes during the final hectic days, and sometimes they’re successful (in this case without even spending money). Not every poll result will hold.
Finally, the earlier discussion around this poll was a wonderful example of jumping to conclusions without having all the necessary data. Some of the conclusions people reached have now been disproven. But while there was lots of interest back in June, now practically no one cares. The story of the Cantor election is etched in peoples’ minds and a few facts six months later will do little to change that.
Let’s be careful about leaping to data-free conclusions come November.
Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the majority leader of the Senate and the Democratic whip in the House.