STEVE INSKEEP, HOST:
This is a big day for candidates, of course, and also for pollsters, who will be watching to see if their predictions panned out. Because in this age of caller ID and do-not-disturb mode, the job of a pollster is getting harder every year. Jeff Guo and Nick Fountain from our Planet Money podcast wanted to find out, just how hard is it to poll people? They asked our polling partners at Marist College to let them give it a try.
JEFF GUO, BYLINE: Marist pollster Daniela Charter taught us, above all, you need to charm the person on the other end of the phone.
DANIELA CHARTER: Smile while you dial.
NICK FOUNTAIN, BYLINE: Smile while you dial. Because if you smile...
CHARTER: You sound upbeat, you sound good and you sound like someone that the respondent is going to want to talk to.
GUO: But when we actually started dialing...
FOUNTAIN: Hi there. My name is Nick Fountain. I'm calling from Marist College.
GUO: ...We discovered this was a really tough job.
FOUNTAIN: Oh, I got hung up on. I did not smile. I did not smile (laughter).
Non-responsiveness is a huge problem for pollsters, and it comes in a variety of flavors. There is the soft refusal.
GUO: Ooh, they told me to have a blessed day. I feel blessed.
The hard refusal - answering machines.
FOUNTAIN: Oh, thank you. Is this a real person? (Laughter) It was a voicemail.
GUO: So many voicemails.
FOUNTAIN: Have you ever seen someone strike out 70 times in a row?
GUO: That's actually pretty common these days. Just to get one person's opinion, pollsters have to dial 50, maybe 100 numbers.
FOUNTAIN: Barbara Carvalho, the director of the Marist Poll, says it's especially tough to reach certain demographic groups, like rural voters or younger voters.
BARBARA CARVALHO: So if we're just dialing phone numbers, we're really going to miss a lot of people.
GUO: And if you're missing large swaths of the population, that makes your polls a lot less accurate.
FOUNTAIN: Pollsters have been looking at different solutions to these problems. Carvalho and Marist - they are surveying by text message, and also they weight their polls to account for the people who are harder to reach.
CARVALHO: So instead of counting as one interview, they might count as 1.2 interviews so that they make up for the people that we didn't talk to.
GUO: There might be a third solution too. It uses an idea that economists love called the wisdom of crowds. Researchers told us, if you want to predict an election, don't ask people, who are you going to vote for? Ask them, who are your friends voting for? Or, who do you think is going to win the election?
FOUNTAIN: And the idea is by asking people to share info from their social networks, talking to one person becomes like talking to a dozen.
So, Barbara, we have a question for you.
CARVALHO: Uh-oh.
FOUNTAIN: Would you be willing to let us test one of these methods in one of your polls?
CARVALHO: Oh, sure.
FOUNTAIN: Really?
Marist added our wisdom of crowds questions to some recent polls. And compared to the traditional who are you going to vote for question, the new results seem to heavily favor Republican candidates.
GUO: And once all the votes are counted, we'll have a better sense of which questions can better predict elections. Jeff Guo.
FOUNTAIN: Nick Fountain, NPR News. Transcript provided by NPR, Copyright NPR.
