Politicians like to say that the only poll that matters is on Election Day. That's starting to be more true.
Polls in the Kentucky governor's race consistently showed Democrat Jack Conway with a slight lead over Republican Matt Bevin. Not only did Bevin win last week, but it wasn't even close. Bevin took 53 percent of the vote to Conway's 44 percent.
The day after the election, the Lexington Herald-Leader announced it would dump Survey USA as its pollster.
"We might as well buy monkeys and dartboards vs. what we had here with Survey USA," tweeted Scott Jennings, a Republican consultant in Kentucky.
The problems aren't limited to the Bluegrass State. Last year, polls around the country underestimated the Republican strength in several Senate races, as well as the governor's race in Wisconsin. Conversely, in 2012, the Gallup Poll showed Mitt Romney beating Barack Obama in the presidential election.
In October, Gallup announced it won't conduct horse-race polling in 2016. Instead of tracking candidate strength, it will attempt to measure voter concerns and values.
"I've been saying that this will be the last cycle for polling as we've known it," said Karlyn Bowman, an expert on polling at the American Enterprise Institute. "With responses for the best designed surveys below 10 percent, the challenges are really great."
It's a problem all over. England, Greece and Israel all held elections this year with outcomes that turned out to be a surprise -- if you had believed the polls. Pollsters in Canada, however, did better last month.
The list of problems is long. Fewer and fewer people have landlines. Buying lists of cellphone numbers, which are harder to target by geographic area, is more expensive for pollsters. Getting people to answer the phone and stay on the line to answer questions is increasingly difficult. Meanwhile, pollsters haven't yet figured out how best to reliably gauge opinion through online surveys.
"People don't want to talk because they don't know who's going to use that information," said Susan MacManus, a political scientist at the University of South Florida. "You almost have to be partisan to get business, and to keep business, you have to favor one party."
There's added scrutiny for pollsters in general, with more criticism coming from social media and election forecasters whenever they get things wrong. Increasingly, pollsters are nervous when their own numbers appear to be at variance with what their peers are showing.
"Pollsters seem to be increasingly engaging in something called poll herding: a tendency to either re-weight an outlying poll to fall in line witih other pollsters, or to fail to publish outlying polls altogether," elections analyst Sean Trende wrote in Politico. "In 2014 alone we saw evidence that PPP, Rasmussen Reports, Gravis Marketing and Hampton University all refused to release polls."
As much as people like to belittle them, polls play a huge role in politics.
For one thing, they offer donors a sense of whether it's worth sending more money to a candidate. For another, they can give candidates an idea of whether their messages are breaking through to the right audiences. Media outlets also rely heavily on polls to offer a nominally objective sense of who's winning. Without reliable polls, "we'd be back in the pre-TV days, when reporters and analysts had to rely on instinct and feel for their work," said Henry Olsen, a senior fellow at the Ethics & Public Policy Center.
Bowman predicts that exit polls might not be conducted in 2016 as they have done in the past. With more states allowing voting by mail, follow-up surveys to get a more representative sample might cost more than the sponsoring TV networks are willing to pay.
Exit polls -- although they have been problematic in predicting electoral outcomes -- remain widely used for their demographic data and the snapshot they offer about public opinion on issues.
The media, corporations, academics and everyone else concerned with public attitudes use polls to get a sense of how mass opinion is breaking down. If the presidential race is too expensive or problematic to poll, what hope is there for people wanting to gauge opinion on a local ballot measure?
Still, not everyone is convinced polling will go the way of the teletype. Some polling outfits will spend the money necessary to broaden their samples and get better results, said Larry Sabato, who directs the Center for Politics at the University of Virginia. Pollsters will also figure out what works online.
Where there's a need, there'll be a way, he suggests.
"We need solid, reliable measures of public opinion during campaign seasons, and the industry will experiment, adapt and provide them," Sabato said. "Woe is not us."