LePage Leads in Down East/MPRC Poll

[1]

LePage with 39% of the vote, Cutler at 29%, Mitchell at 24%, Moody at 4%, Scott with 1% and 2% still undecided, with a margin of error of 4.19% (95 times out of 100) – that’s the result of a new poll of 546 likely voters conducted by the Maine People’s Resource Center and sponsored by Down East.

These results, when compared with other recent surveys, show LePage with a solid, virtually insurmountable lead and Cutler having gained significantly against Mitchell, who now sits in third place.

MPRC is a non-partisan, non-profit organization dedicated to giving people, especially those groups that are under-represented in civic life, the skills to participate in the decision-making processes that affect them. One of the most prominent examples of this effort has been the biannual publication of the Citizen’s Guide to the Maine Legislature, a book with detailed, accessible biographies of each of Maine’s 186 legislators.

This poll is another step in that direction.

Much of the public discussion of issues and politics lately has been turned into a numbers game. There are fewer political reporters and media outlets have fewer resources to devote to covering issues in depth. At the same time, the number of polls released during and between elections has markedly increased, as new technology makes it easier than ever to gather public opinion.

This means that elections and even policy debates are now viewed more and more as horse races, while at the same time polls are presented to the public without a full understanding of what they really mean and where their numbers come from.

This first poll from MPRC isn’t meant to fix the first of these problems, the lack of a broader scope of research into Mainers' issue-based civic participation (it is, in fact, a simple horse-race poll), although I hope that later surveys might begin to do so. It is, however, designed to be an example of transparency and honesty in public opinion reporting and the democratization of political research.

So here are the full results of the poll, and below that a full accounting of the methodology used to conduct it, and some of the raw data. This level of transparency exceeds the Principles of Disclosure [2] recommended by the National Council on Public Polls.

Frequeny Results and Crosstabs [3]

Basic Method:

The poll was conducted using Interactive Voice Response or IVR calls (often affectionately or pejoratively known as robo-polling). There has been debate in the polling industry about the efficacy of IVR, but the technology has proven to be very effective when used correctly. Some of the most prolific and most accurate [4] national pollsters use IVR, including Public Policy Polling, Rasmussen and SurveyUSA.

Rasmussen and PPP have surveyed the Maine race several times and their results track closely with other pollsters. SurveyUSA usually releases a Maine poll today, the Sunday before the election, but will not do so this year.

MPRC’s IVR hardware, software and phone infrastructure was developed specifically for public opinion polling and is owned and run by the organization. No outside group was contracted to run this poll.

Timing:

The survey was conducted from Thursday, October 28 through Saturday, October 30 at a variety of times of day to maximize our chances of reaching a wide range of people.

Sample:

A poll’s sample is the list of people or phone numbers that are sought out for their opinions. Some pollsters use random phone numbers. This poll used a more targeted list.

A random sample of phone numbers was drawn from a list of registered Maine voters who had voted in at least one of the last three elections. Respondents were also asked a further three questions to verify their registration status and discern the likelihood of their voting.

Questions:

In an attempt to make our first survey as simple and straightforward as possible, we limited the number of variables by only polling one race and asking only eight questions [5].

The questions were recorded by a professional phone canvasser. One of the advantages of IVR is that every respondent hears the questions in exactly the same way.

Weighting:

The people who respond to a poll by telephone do not necessarily accurately represent the broader voting public. Pollsters therefore use a variety of methods to shape their results so they represent the population that will actually vote in the election.

A well-known issue for this kind of polling is much greater response rates from women than men. To combat this, we gathered a larger number of responses than necessary and randomly discarded a mathematically determined number of responses from female respondents.

The sample was then mathematically weighted based on the age of the respondents and their county of residence. Because this is an election sample of likely voters, the information we sought to match was from previous elections. Specifically, the US Census Department’s Current Population Survey for voting rates by gender and age and the Maine Secretary of State’s figures on voting by county.

We used average figures for the last four gubernatorial elections, going back to 1994. Interestingly, these rates have been remarkably consistent over time. For instance, according to the CPS, the Maine electorate has been 53% female and 47% male for each of the past three gubernatorial elections.

This poll was not weighted by party identification.


Possible areas of criticism of this poll:

- The use of IVR rather than live callers, discussed above.

- We identified the media sponsor of the poll at the beginning of the survey, in part to make sure people knew this was a legitimate poll. Certain people might be more likely to respond to a poll sponsored by Down East than others, creating some bias in the sample.

- The numbers of likely voters in each political party may seem a bit off – specifically our sample shows more Democrats planning to go to the polls than other recent independent surveys (although his doesn’t seem to have affected the candidate preference numbers, which are in line with other polls).

We did not weight by party, in part because party identification is a fluid thing over time and in part because it’s impossible to predict the party turnout on Election Day with complete accuracy. More information [6].

That being said, here is our raw data in CSV format [7], screened only for likely voters and gender. Feel free to use it to make a prediction based on your own model of what you think the electorate will look like.

- No question or answer rotation. To make things simpler, we only asked the questions and answers in one order – this may have introduced some bias.

- No track record. MPRC hasn’t done this before and we've likely made some mistakes. Feel free to ask questions or make suggestions in the comments below.