Friday, March 23, 2007

The usual attacks...

Since we only started the blog this week I am going to have to catch up on a lot of topics. Here is something from February.

Growthmatters.org, the organ for the Triangle Community Coalition (aka the real estate/development lobby), attacked a poll done by our sister company, OnPoint Polling and Research. Well, they really didn't attack a poll, so much as, attack the whole method of Interactive Voice Response (IVR) or automated polling. We're pretty used to these attacks, and I will post more on the reliability and accuracy we have achieved in coming weeks.

Let me go through some of what they have to say:

How do we know they don’t get a bunch of teenagers or 12 year olds punching numbers? We don’t. And that’s the problem.

Is that really the biggest problem you have with our polls? We employ screening questions such as "are you a registered voter" to try and weed out any youths. We also toss out nonsense responses or people who simply press #3 or #5 for every answer, for example.

The fact is we see no evidence that 12 year olds have been corrupting our polls. And this complaint ignores problems with traditional polls that we don't have, for example, interviewer bias-- live interviewers may ask questions differently, like tone or voice inflections, that could impact responses. Also, the anonymity provided by a recorded voice also increases the likelihood that the respondent with answer truthfully.

Most polling firms we talk to say the results of robo-polls are not reliable. The IVR method is cheap. Traditional polling is more expensive— but is also more reliable.

Of course they do! IVR polling is a new competitor to traditional telephone pollsters and often they think of us as a threat. They don't want to relinquish their place at the top of the field and so they denigrate our form of polling. Back in the 60s-70s, "traditional pollsters" criticized the new technology of "telephone polling." They thought you could only be accurate with tedious door-to-door interviewing techniques. IVR is simply the next stage in polling technology. We won't replace live interviewer polling, but we will stand beside them.

In October we went into the field with a poll on the upcoming Wake County School Bond, while the News & Observer hired a traditional pollster to do their research. The results were striking. We got it right, the N&O's pollsters did terribly.

PPP said:
49% for
41% against

Research 2000 said:
35% for
54% against

Election Day
53% for`
47% against

Who is more reliable?

In the end, the real story here is not our polling methodology but the poll results. The real reason growthmatters.org attacked OnPoint is that polling revealed that 69% of likely North Carolina voters support giving county governments the authority to charge impact fees (PDF). They must be operating under the rule: if you don't like the message, attack the messenger.

4 comments:

Will Cubbison said...

An interesting analysis from 2004, indicating automated polling did at least as well, maybe better.
http://www.slate.com/id/2110860/

Pollster.com had some interesting discussion on this. It sounds like the conclusion is that automated polling could end up being more accurate all the time, will just take people like you continuing to fine tune your methods. Not that surprising of a conclusion.

Justin Guillory said...

I have seen 3 or 4 analyses of election year polling that concluded IVR polling is at least as accurate, if not more accurate than traditional polling.

There has not been, however, a definitive mathmatical study of the issue.

I'm waiting.

Anonymous said...

Justin,
You never addressed the NY Times policy on IVR.

Justin Guillory said...

Anonymous,

Sorry I missed your comment. I disagree with the NY Times policy.

Other national news sources regularly cite polling from Survey USA and North Carolina newspapers including the Charlotte Observer, the N&O and the News & Record report our survey results.

The NY Times needs to join the 21st century.

 
Web Statistics