Q&A: Orr explains how Brown University’s poll was conducted

October 9th, 2013 at 12:38 pm by under Nesi's Notes, On the Main Site

Marion_OrrBrown University released a new opinion survey on Wednesday morning showing Treasurer Gina Raimondo leading Providence Mayor Angel Taveras in the Democratic primary for governor. The findings are strikingly different from those of the poll Taveras commissioned and released last month that put the mayor 19 points ahead.

Marion Orr is director of Brown’s Taubman Center for Public Policy and Frederick Lippitt Professor of Public Policy & Political Science. He talked with WPRI.com on Wednesday about how the new poll was conducted. The interview has been lightly edited and annotated for length and clarity.

What are the headlines to you out of today’s new Brown poll?

The headline out of this poll is that Gina Raimondo is leading pretty good among likely voters in the Democratic primary. This is a fairly good lead, I think, within the margin of error.

Now, the lead narrows a bit – that is, her lead narrows a bit but she still leads – when you focus only on those people who tell us that they typically are Democrats. But she still leads. But the lead narrows.

What I’m suggesting here is that when you add in, say, independents who could perhaps vote in a Democratic primary, her lead increases. So I think that’s one thing.

The other thing is, I was somewhat surprised by Mayor Allan Fung. I know he’s a mayor of one of our cities but I don’t see him as one of these statewide officials who everybody knows. So I was really surprised by how well he did in this poll. Moreover, I was surprised how well he did when we put him up against a potential Democratic nominee of Gina Raimondo – there it’s, like, a tie. So I suspect that the Fung people will find this poll to be quite interesting.

Now the other part of this of course is that Angel Taveras, the mayor of Providence, on a head-to-head with Mayor Fung, Angel Taveras would win. So for me I think this is clearly good news for Raimondo. She’s doing good among independents, which is really, really important because, as you know, the breakdown among registered voters is more unaffiliated voters than Democrats or Republicans.

I had some methodology questions for you. First, how do you draw your sample?

Registered voters. We get the most recent registered-voter list and we do a random sample of registered voters.

And do you get the phone numbers you call using the numbers people write on their voter registration cards?

Yes, exactly. Only way we can call them is – we identify the registered voters, they have a number, and we know we’re talking to someone who is a registered voter.

So the overall survey sample of 638 voters – those 638 are all registered voters in Rhode Island?

No question about it.

Now, with the sample, did you do anything to it in terms of looking at only calling people who voted in the ’06 or ’10 primaries?

No. No.

OK – so what is the voter screen to get the 433 likely Democratic primary voters?

The voter screen is just what we released. The first question was: “How likely are you to vote in the 2014 Democratic primary for governor?” And those people who said “very likely” or “somewhat likely” are the people who come up in the [primary] responses.

Do you have a breakdown, for all 638 registered voters in the poll, of how many were women and how many belonged to each party?

Yes. It was male 320, female 318. [That's 50.2% male, 49.8% female.]

Great. And for party affiliation?

In our full sample of 638: 11% Republican, 31% Democrat, 48.3% independent. And then we have a category called “other” and “no response,” which makes up the difference of 9% or so.

[Official voter registration in Rhode Island was 10% Republicans, 40% Democrats, 50% independents and 0.2% Moderate Party members as of Oct. 31.]

And do you have the same breakdowns for the 433 Democratic primary voters in the poll?

This is what’s interesting. You’ve got to be mindful, you have people who tell you they’re going to vote – I’m very serious, people can tell you they’re likely but do something else, and it’s very, very important to understand that.

Among the 433 people who said they’re [very likely or somewhat likely] to vote in the Democratic primary for governor, we have 21 Republicans, 173 Democrats, 199 independents, and the difference is “don’t know or other.”

[In percentage terms, the survey's likely Democratic primary electorate is 46% independents, 40% Democrats, 5% Republicans, and 9% "don't know or other."]

Thanks. And what’s the gender breakdown for the primary voters?

I have 201 males, 225 females – that comes to 426; I can’t tell you where the other seven are, but it’s not a significant difference I don’t think. It must be some missing data – there must be a missing variable there, because you should have 433. So there’s seven missing here for the female/male, which shouldn’t change the analysis.

[In percentage terms, the survey's likely Democratic primary electorate is 46% male, 52% female, 1.6% unknown.]

Did you do anything to look at where in Rhode Island the voters live?

No, no.

Some of the pollsters I talked to today said they weren’t sure if it was correct to take a sample of all registered voters, and then pull out two-thirds as the Democratic primary electorate – to do one poll to test both universes of voters. What do you say to that? Why did you do one poll?

This is why we asked the filter question about one’s likelihood of voting in a Democratic primary. Plus, we have people’s registration, plus what they say they normally vote as. So I really don’t understand the logic behind that question.

We wanted to get a survey of all registered voters – so we did the 636, which comes out pretty close to the regular numbers – and then we wanted to know, among those registered voters, how many are likely to vote in this important Democratic primary. And so we tried to isolate those people who said they were very likely or somewhat likely to vote in the primary. And then you could sort of take those people, the 433, which is a very sizable number to work with.

Who does the interviews? Is it students?

We train and pay – and let me underscore, pay; we don’t have volunteers – we train and pay largely Brown University students. I think in this instance all the interviews would have been students. Sometimes, we add interviewers and sometimes we hire a private company to get some people. But most of the time, almost all the time, we’re using Brown University students.

Finally, when is the next time you think Brown will do a poll?

Very soon. We’ve got to do something on the city [of Providence]. We always do an annual poll of just the city, and the next one will be on the city solely. And then we’re going to soon come after with a statewide election poll. •

• Related: Brown U. Poll: Raimondo leads Taveras in Dem gov primary (Oct. 9)

This post has been updated.

(photo: Brown University)

Tags: , , , , , , , , , , , , ,

7 Responses to “Q&A: Orr explains how Brown University’s poll was conducted”

  1. George says:

    Sam Bell, Do you know the percentage of voters are Spanish language only speakers that are citizens? I may not understand how it all works, but I thought to be a citizen you needed to speak English.

    1. Steve says:

      Not necessarily, people born in Puerto Rico are American citizens by birth.

  2. [...] Orr, the poll’s methodology was called out almost immediately. WPRI’s Ted Nesi has an interview with Orr explaining the methodology; here on RI Future our editor Bob Plain has a quick list comparing the [...]

  3. Bill says:

    Asking registered Republicans and Independents if they are likely to vote in the Democratic primary is not a verified prediction tool. It is easy to say you will vote in another party’s primary, but harder to go to town hall and disaffiliate 90 days before the primary. But it could happen in this case. Gina had, in her push to cut public pensions, the support of the “astro turf” organization Engage RI which allowed Republicans and the 1% to create press and personal pressure from the powerful to push through their agenda. The same folks could mobilize and spend to get Republicans to hijack to Democratic primary were only about 10% of eligible voters show up. Of course in the general election these crossovers will abandon her and vote for a real Republican and many real Democrats will not vote for her as a “Democrat in name only,” as this poll predicts with Fung beating her, but not Tavares.

  4. [...] primary electorates in 2010 and 2006, respectively – much, much higher. Brown’s Marion Orr is standing by his projections, telling me: “This is why we asked the filter question about one’s likelihood of voting in [...]

  5. [...]  The dubious primary results, which show Raimondo beating Taveras in a Democratic primary where most voters aren’t Democrats, have been widely mocked.  (For comparison, a Taveras internal by a reputable pollster put Taveras [...]