MOUNT VERNON — After learning about what actually goes on in a Gallup poll, Mount Vernon High School students in Jeff Jarosz’s class put their knowledge of statistics to work and conducted their own poll to try and predict the outcome of the Democratic primaries in the state of Ohio.
When the News went to the classroom on Friday to get the results, the class data indicated that Barack Obama was likely to win the primaries with 52.8 percent of the vote. The students were “very confident” in their results.
Class members called a total of 1,617 people, said Emily Cassell, and got 180 responses, which is only an 11 percent response rate. Although not very high, that was enough responses to complete the weeklong polling project. Of the 180 respondents, Cassell said, 85 were likely to vote for Hillary Clinton, and 95 reported they would probably vote for Obama.
Because the class did not have access to a list of registered voters, they “googled” the list of the most popular surnames in America. The pollsters then considered 219 municipalities listed as cities in Ohio; no hamlets or villages were included. Twenty-five of those cities were chosen as survey “targets.” After the cities were selected, said Drew Cheek, the assigned pollster picked a common last name, like Smith.
“Then,” he said, “ we looked on line for four people with the last name of Smith from that city and wrote down their phone numbers. And just called those four people.”
The class goal was to get 400 total answers for prediction purposes. That meant everyone in the class had 96 people to call, hoping to get at least 20 responses each.
“We called people and asked them if they would be willing to participate in a two question survey,” Joe Stuart said. “We then asked them if they would be voting Republican or Democrat in the primary. If they did say ‘Democrat,’ we would ask them Obama or Clinton. It was relatively quick when we did get responses.”
The students agreed they would not like to be a professional poll taker. Some of the selected phone numbers didn’t work; some were disconnected or were busy. Some respondents were just plain rude, said Erin McIlvain.
“They would just hang up before you got to talk, or they would just say, ‘I’m not going to answer that.’ Some seemed to be afraid the information would be used against them, and they didn’t realize it was just an experiment for our class.”
Matt Chacey explained the margin of error percentage related to confidence intervals, and showed the class how those ratios affect the poll’s predictions. Statistically, a margin of error decreases as the sample size increases, and the confidence level refers to how many respondents give a particular response out of 100 individuals questioned.
Chacey concluded that with using a 98 percent confidence interval, Obama would receive between 44.2 percent and 61.4 percent of the vote, and Clinton would garner between 38.6 and 55.8 percent of the vote.
“We are very confident in this prediction,” Chris Jaymes said, “because we are allowing our data to have a margin error of almost 9 percent.”
“True polls,” said Jarosz, “have a plus or minus margin of 4 percent. With our 9 percent, that gives you a large window to miss by. That’s why we wanted more people so that window would be smaller. I would think anyone off the street could give you a good answer plus or minus 20 percent. We got it narrowed down to 9 and 7, which isn’t bad, but we wanted to look professional and get it to 4 (percent). But with the number of people who satisfactorily answered our survey, we just couldn’t do it.”
If more people had responded to the phone survey, the MVHS data would have been more accurate.
“The professional polls,” Austin Kasserman said, “said that Clinton is leading in Ohio. Our findings contradict that. If our data is correct, it would be surprising that a group of high school students would be more accurate than people who do this for a living.”
While the students’ were partly correct — they predicted Obama would receive between 44.2 percent and 61.4 percent of the vote (he got 44 percent) and Clinton would garner between 38.6 and 55.8 percent of the vote (she got 54 percent) — their conclusion that Obama would win was erroneous.
“Any prediction coming from a random sample is never going to be ‘spot on.’ That’s why we use the margin of error,” Jarosz said. “Unfortunately for us, we were at the very, very end of our margin of error. Our data didn’t show very good accuracy of the true vote in Ohio.
“I think if we would have had more respondents, we might have been more accurate. If we could have gotten another 300 people to respond, it would have been more likely that we would have gotten more Clinton people. With such a limited sample, 180 people, it is statistically extremely rare that we got such a high amount for Mr. Obama.”