The defeat of House Majority Leader Eric Cantor in the Republican primary on Tuesday night came as a surprise to probably everyone outside of the Cantor campaign — and perhaps them, too.

There are three immediate reasons that the result was unexpected. First, the defeat was unprecedented; no House majority leader had ever before lost a primary challenge. Second, there was a tacit assumption that, since Cantor was a veteran, his campaign was doing what it needed in order to win, just as so many other incumbents manage to do each year. And, third, the available polling was severely limited, including a poll released by the Cantor campaign that showed him winning by a massive margin.

That internal poll, released last week and showing Cantor with a 34-point lead, was thoroughly mocked in the wake of the loss. The National Journal talked to the pollster, John McLaughlin, to try and figure out how he got it so wrong. McLaughlin made three arguments in his defense:

  1. Turnout was higher than expected.
  2. Attacks on immigration hurt Cantor.
  3. Democrats ran a concerted effort to vote against Cantor on primary day.

Only the second argument is very good. As a pollster, it's a critical part of McLaughlin's job to anticipate the likely electorate; if he can't do so, he can't judge whether or not the sample of people he polled is any good. And as for the "Democrats voting" thing, well, read this. McLaughlin's polling has come into question before, after a lengthy series of missed calls in 2012.

But if we take only one polling-related lesson away from this surprise, it is this: Internal polls don't tell you anything about the state of an election.

Depending on how much money they have, campaigns (both run by candidates and by outside groups) regularly run polls to gauge how they're doing. There are big polls, testing messaging and the effectiveness of attacks. In many high-profile races, there are regular "tracking" polls which are used to get a picture of the race on a daily basis. All of these polls are generally closely guarded secrets. And when they aren't, when they're released, it's to make a messaging point, not because the campaign has suddenly decided to be generous with its information. They release the results to raise money or to discourage opponents or any one of the many reasons that campaigns do anything. It's not for your benefit.

And, perhaps as a result, they're unreliable.

  • An internal poll released by the campaign of Republican Steve Lonegan last October had him within three points of Cory Booker. Booker won by 11 points.
  • An internal poll released last June showed Gabriel Gomez trailing Ed Markey by only three points. He lost by ten.
  • Indiana's Richard Mourdock released an internal poll a few days before the 2012 election showing him up by two. He lost by five. (Pollster? McLaughlin.)
  • That same year, the Desert Sun ran a headline for a House race in California: "Mary Bono Mack, Raul Ruiz both claim to lead polls." Ruiz's poll was right.

They're not always wrong, of course. In 2012, Heidi Heitkamp's campaign released an internal poll showing her up three points in her North Dakota senate race. Somewhat unexpectedly, she won. The point is that they are largely unreliable and highly selective.

The Cantor poll that came out last week was conducted in late May. His campaign probably released it to counter another poll conducted for The Daily Caller that came out on the same day showing a closer race. The interesting question is why Cantor's campaign decided the best message to send was one of dominance, even as they spent heavily on advertising against his challenger. Sometimes, of course, the internal polls fool the campaigns, too.