After 8,000 calls, a new Eyewitness News pollSeptember 28th, 2010 at 5:25 pm by Ted Nesi under General Talk, Poll Results
Tomorrow is a big day around here, as we release the findings from our third Eyewitness News poll of the year. We asked Rhode Islanders what they think about every hot topic – the governor’s race, Cicilline vs. Loughlin, President Obama’s job performance, the 38 Studios and Deepwater Wind deals, the Bush tax cuts, and more.
Tim White and I have been digging through the findings with our pollster, Eyewitness News political analyst Joe Fleming, and although I can’t put out any numbers yet – under threat of death from the powerful Jay Howell – take it from me that this is one meaty poll, with a host of interesting findings.
I’ll have a preview with the results of one question tomorrow morning, and then Tim will have the results for statewide races and issues in the evening at 6 p.m. and 11 p.m., with federal races and topics following Thursday at 6 and 11. (I have my eye on tomorrow at 11, when we’ll find out how voters feel about the 38 Studios deal.)
In the meantime, here’s some background on how a survey like this gets done.
Joe Fleming is an old hand at this – he’s been conducting polls for a quarter-century. For this poll, Fleming & Associates contacted 500 likely voters registered in Rhode Island. Their phone numbers were generated randomly by a computer.
But the group has to be representative of the broader electorate to be valid, meaning it has to have the right proportions of men to women, younger voters to older ones, Democrats to Republicans to independents, etc. That’s accomplished through what’s known as “weighting” – calling and calling until you have spoken to enough young people, enough independents, enough women.
Weighting can make or break a poll, Joe emphasized to Tim and me. If you think about it, the only reason we do surveys at all is to get a sense of what people are thinking – and if the poll isn’t an accurate snapshot of the public, it won’t be an accurate snapshot of public opinion.
In addition, for this poll we wanted to speak to “likely voters,” as opposed to registered voters, so we only asked questions of people who said there was more than 50-50 chance they will cast a vote in the November election. That poses another challenge in putting together a proper sample. Plus, sometimes a call turns up a disconnected phone line or an answering machine.
How labor intensive is it to get the weighting right? I think this statistic says it all: Fleming’s team had to call nearly 8,000 people to get the final sample of 500 voters right. That’s an average of 16 calls to come up with one poll respondent – repeated 500 times.
One question Tim had was about cell phones, considering that a growing number of households, particularly younger ones, are dropping landline phone service. Joe said statistical studies show that any impact there is small, and controlling for it is part of the reason his team makes so many calls, to ensure an accurate sample.
The poll was conducted from last Wednesday, Sept. 22, through Sunday night, Sept. 26. Here’s how our 500 poll respondents break down:
- Male: 47%
- Female: 53%
- Congressional District 1: 50%
- Congressional District 2: 50%
- Ages 18-39: 23%
- Ages 40-59: 43%
- Ages 60+: 33%
- Union household: 21%
- Non-union household: 78%
- Democrats: 42%
- Republicans: 17%
- Independents: 39%