The Eagleton Center for Public Interest Polling at Rutgers University has been in backpedaling mode recently, trying to explain why its polls on the two biggest political races in New Jersey last year were so far off.
In the October special U.S. Senate election between Democrat Cory Booker and Republican Steve Lonegan, a Rutgers-Eagleton poll showed Booker winning by 22 points. He actually won by less than 11 points, half that margin.
In November's gubernatorial race, the poll showed Republican Gov. Chris Christie beating Democrat Barbara Buono by 36 points. Christie's actual victory margin was just over 22 points.
Last month, Rutgers-Eagleton released the results of a study it commissioned to try to figure out the problem with its numbers. The report concluded that the problem was that the polls asked respondents about their attitudes toward candidates before it asked them whom they intended to vote for. The combination primed voters to respond differently than they otherwise might have.
The report points to the problem of putting too much stock in election polls, and it raises questions about how these polls may have influenced last year's races.
The demand for political polls has increased, in part, because they play into the media's preference for covering elections as if they are sporting events - who's ahead, who's behind and which demographic will make all the difference.
There are plenty of reasons why polls may be inaccurate. Most polling is done by phone, and pollsters usually only reach people with land lines, not exactly a representative sample in the cellphone age. Pollsters also only talk to people who agree to take part in a poll.
But the more intriguing question is how the Rutgers-Eagleton poll may have affected the elections. Poll results can alter turnout. Maybe some Christie and Booker supporters stayed home, reasoning that their candidates were cruising to easy victories and didn't need their votes.
Did the exaggerated margins of victory in these polls make it more difficult for Buono and Lonegan to raise funds - or to be taken seriously by undecided voters?
We'll never know. But we can take this as a reminder not to put more importance on opinion polls than they deserve. In politics, the only poll that counts is the one on Election Day.