I'm A Phone Pollster This Election: 5 Realities

Greg remembers one Democratic employee who, during a heated local race, "would word some of the questions in a way to make them seem better than the Republican options. It wasn't blatant, either. He would do it in a tone of voice -- a little lighter and warmer for Democrat candidates and darker and more monotone for Republicans. We asked about how positively or negatively they were viewed too, and he added details not in the script. One question was on the candidate's name, and he worded it like, 'Which name sounds more like a governor's? Is it the esteemed or the other ?'"

That kind of chicanery, whether Trump-blatant or Democrat-manipulative, has no place in legit polling practices. And yet it is, in fact, the entire point of many polls. Called "push polling," these operations maintain only the thinnest veneer of objectivity, while their real purpose is to influence opinion, not record it. It has a long history in politics, right up through this year's primaries, when South Carolina Republicans received an automated polling call that berated them if they chose anyone besides Ted Cruz. Who could possibly be behind such a thing? It seems pretty obvious ... Someone trying to make Cruz look bad! At least, that's what Cruz himself said, and he has such a trustworthy face.

According to Ron, push pollers are about as seedy as they sound. And not only has he never done what they do, but his company also wants nothing to do with anyone who does. "They would rent a room in a hotel and hire a bunch of temp employees to make calls with cheap cell phones," he says. "They don't need to match quotas, or store or analyze data, so they don't need the expensive hardware and software or staff training a real polling company needs."

A push poll may be mostly indistinguishable from a legitimate poll, but there are a few tells. "A push poll will typically be done in the days immediately prior to an election and have no demographic questions," Ron explains. "The poll will then go on to ask you questions that make one side look good and the other side look bad. The may not even be recording your answers, because they are useless."

Basically, listen for someone who sounds like they're making jerk-off motions while you talk instead of diligently typing your answers.

If the only person who's ever called to yell at you about politics is your grandpa, it may be because you're part of nearly half of the population that no longer has a landline. Obviously, ignoring the iPhone-carrying masses makes for much less accurate forecasts. Remember Bernie Sanders' "shocking" win in Michigan? That wouldn't have been a surprise had pollsters used a cellular/landline mix. Why not simply use the internet? Then you're missing a different side: the 13 percent of the population that doesn't cotton to these newfangled innertubes.

There are two ways pollsters attempt to solve this problem. The first is building statistical models. For example, let's say they talk to 1,000 old white folks, a handful of people from other races, and if the probability gods really smile upon them, one or two college students visiting their parents ("the Holy Grail for landline-only surveys," Greg calls it). In that case, the data is "weighted" -- some people's answers count for more than others. This is why we wind up with polls showing wildly different results during the same time period. It all depends on the company's specific methodology, which largely depends on the idea that everyone in a given demographic group is a scary Village Of The Damned hive mind.

They might also use a quota system, which requires pollsters to keep calling until they get every demographic. That's a special circle of Hell -- it can take Greg months to meet his quota on the 18-30 demographic. He'll talk to a whole lot of middle-aged white mothers on the way, but there's an easy fix to prevent them from skewing polls: Once they've gotten enough of the "still not over Kevin Sorbo" group, pollsters stop entering in their responses.

We conducted our own little poll to find out how middle-aged white mothers feel about that, and they overwhelmingly stated that they "just like having someone to talk to." Eighty-two percent further amended, "You don't call enough. What are you, a rock star or something? You can't call?"

"Statistically, we should be asking for more people in their 20s, 50s, and 60s, since they are more of the population," Greg says. "But we don't, because those are nice round numbers, and it's easy to do the math." This flawed logic is known as "age distribution inflation," and it's downright deceptive.

Then there's an approach we call "any port in a storm": "The easiest surveys were those that asked to get 1,000 people, regardless of age or gender," Greg says. "We still recorded who they were, but those would have the final results of mostly older people. The easiest one ever only had a few questions, and they were all about voting. They were 'Are you planning to vote?' 'Did you vote by mail?' and 'Why did (or didn't) you vote?' Surveys usually take several days to finish, but we got that one done in eight hours. I'll let you decide if that's a good or bad thing."

We'll go with "good," because it was the first option and we're eager to call it a day.

Greg gets a randomized list of landline numbers that his company draws from an online phone book, which is fairly standard practice. To hit on the occasional unlisted line, he may be given numbers with the last two digits randomly generated. This tends to irritate people on the Do Not Call list. Pollsters are exempt, baby.

Ron's organization does things a little differently. Many of their clients are running for office, so they can rent access to political party and affiliated group databases -- information that has given Democrats a considerable edge in the past couple of elections.

"Since campaigns are already paying for access to these databases with phone numbers of all their voters, they usually just send us the numbers they have already paid for, rather than pay for a company to provide us with random numbers, as was common in the past," Ron says.

It's a damn disgrace, those cheap politicians running mom-and-pop random number generators out of business. Be sure to tell them what you think about it when they don't call you.

Of course, that's assuming you can get people to respond to your questions at all. Greg says it's easiest around the primaries, when his research group enjoys a response rate of around 13 percent. In general, it hovers between 7 and 9 percent. And you have to be very careful to even get that low level of interaction. Greg has been instructed to never use the words "poll" or "Obama." Instead, he says "survey" or "president."

If it helps, we have a few more suggestions. Instead of "politics," say "interstate death battle." Instead of "voters," say "sexy beasts." And instead of "hello," say "ARE YOU READY TO RUMBLE?!"

Evan V. Symon is a Personal Experience team interviewer, writer, and interview finder guy at Cracked. Have an awesome job/experience you would like to share? Hit us up at tips@cracked.com today!

For more insider perspectives, check out Why Tech Support Sucks: A Look Behind the Scenes and 5 Idiotic Misconceptions About Calling Customer Service.

Subscribe to our YouTube channel, and check out What If Donald Trump Is Just An Elaborate Prank?, and other videos you won't see on the site!

Also, follow us on Facebook, and let's be best friends forever.