Some correspondence available on my recent series (see Part 1 and Part 2) on ACT's poll ratings:
Hi Geoffrey
I know you prefaced you comments by saying that you are not a statistician, but I think you should leave the commentary on polling to people who are a little more informed.
Your statement:
"ACT supporters are perhaps actually overrepresented in opinion polls, due to the fact that they are likely to be older and thus more likely to be at home when rung by polling companies"
shows your complete ignorance of polling techniques. Polling firms use quotas to ensure that they obtain a representative sample - once we have enough old people in the sample we stop ringing them, while young people are more difficult to contact due to the prevalence of cell phone only households, we continue to ring until we reach quota, or weight responses.
There is an argument that the green vote is actually over stated, due to the low turn out of younger voters which is not factored into the some of the demographic models that some NZ pollsters use.
As you are student of political science, I would suggest that you take a few undergrad papers in opinion polling before you express an opinion on something you know little about.
Regards
Stuart Wilson
Director
Lysander Research
---
Hi Stuart,
It's tempting to fight fire with fire. I welcome genuine feedback, but I don't think it's fair to make patronising comments to me, especially when your own background is hardly an independent one. I am upfront about my research interest in ACT, you should be upfront about your partisan one.
To clarify, Lysander Research is a front for ACT. I found the following information revealing (source: http://groups.google.com/group/nz.general/browse_thread/thread/fc7dff3374919d2f?pli=1)
"There is a website lysanderresearch.com that never seems to work. They seem to have done "polling" for Dick Quax who stood for Parliament as an ACT Party candidate. Now they appear to be doing "polling" for Rodney Hide. The pollster was evasive when I asked if he was connected with a polical party. So if Lysander Research phones anyone, ask if they are polling for ACT. And where's their office and why don't they have a website that works."
"Companies Office records show the involvement of a Stuart Wilson in
Lysander Research Ltd. A whois lookup for lysanderresearch.com shows the registrant as Stuart Wilson, ACT NZ, Suite 5, Level 2, 309 Broadway, Newmarket."
"Thanks, I thought as much. I did ask the interviewer if he was
connected with a political party and he said no. A typical ACT liar."
It sounds like your own polling methods are somewhat questionable. And your criticism of my commentary does not even focus on the key point: ACT has polled better in opinion polls immediately before elections since 1999 than it actually does on election day. I do not need to have taken "a few undergrad papers in opinion polling" before making this observation.
The statement you chose to critcise was merely an attempt at a possible, plausible explanation for this phenomenon. Notice that I caged this with "perhaps", precisely because this was merely a supposition.
As for your explanation of polling company methods - my "completely ignoran[t]" view is this. Ringing landlines to "reach quota" doesn't replace actually ringing mobiles - you are just getting youth who have access to landlines, which is not the same thing. Given overseas pollsters ring mobiles using random dialling techniques, I'm surprised your company doesn't do this.
And seeing you are a professional, please offer some supporting evidence about your "argument" about the Green vote being overstated by "some" polling companies.
Regards,
Geoffrey
Subscribe to:
Post Comments (Atom)
2 comments:
Geoffrey
Lysander Research is an independent market research firm. We specialise in quantitative and qualitative political research for centre right candidates and parties. A recent NBR article ( http://www.nbr.co.nz/article/hide-opens-big-lead-epsom-36601 ) noted that we have “an impressive track record”.
While ACT has contracted our services for this election campaign, we have a number of satisfied clients who are not currently involved with the ACT party, including two successful mayoral candidates, a regional councillor and a number of other local government candidates. The market for political research in New Zealand is pretty small, so it is unsurprising that some of our clients may have been involved with the ACT party.
Our website has not been a priority for us this year (I am far more concerned with ensuring that my client has the very best advice to ensure they maximise their vote) and was setup by a former business partner. He has moved overseas and I did not realise that the whois information was pointing at ACT’s Head Office address, and I will have this amended to our business address.
To your substantive points – while you are correct that the results of polls completed in the last 4-5 days before an election are very close to the result that ACT has achieved on the election (One News poll had us at 7.0% on 24 Nov 1999, we got 7.05% 27 Nov 1999; they had us at 7.0 on 24 July 2002, we got 7.1 on 27 July 2002 and in 2005 they had us at 1.9 on 14 Sept and we got 1.51 on 17 Sept – all within the margin of error on a sample of that size), ACT has historically increased its polling during the election campaign and I invite you to look at polls taken 3-4 weeks before the election and contrast that with the result on election day.
There is a general feeling that many of our potential voters in 2005 decided not to give us their party vote in 2005 because they didn’t think Rodney would win the seat of Epsom and their vote would be wasted. We now have polls that tell us that Rodney has a strong lead in the Epsom seat and we hope that as this message is communicated to voters they will be more confident that they vote will count towards a change of government and a change of direction. This was the point that was being made in the comment.
On your surprise that we don’t poll cell phones as “like overseas pollsters”, I recommend you have a look at the latest research on the issue – particularly the 2006 Pew Research study ( http://people-press.org/report/276/ ) and the comments by John Zogby in the latest Campaigns and Elections Journal ( http://www.zogby.com/cemag/view.cfm?id=13 ). Both state that the attitudes of cell phone only households “are not different from land-line voters in the same demographics, so missing them would not adversely affect the survey results”
On your challenge regarding the green vote being over estimated, I suggest you look at the green result in the TNS –TV3 poll – 1999 last poll: 6% actual 5,2%; 2002 last poll: 9%, actual 7%; 2005 last poll 6.9%, actual: 5.3% . The Colmar Brunton shows a similar pattern. I haven’t got the results for the Morgan Poll in front of me, but I suspect you will see a similar pattern for their 2005 result.
Finally, I think academic research into political parties is very important for the democratic process and I personally value the insights that can be gained from those who dedicate their lives to taking an impartial and long term view of the political process. ACT goes out of its way to provide access and data to researchers who are interested in studying our party because we are genuinely interested in the conclusions you will draw. However I find your work, and particularly this blog, to be of a low quality and little more that an ill-informed rant. You will often take issue with a single line in a communication, without considering the context or the long-term strategy that the party is perusing. Your amateur insights into political strategy have very little merit.
Regards
Stuart Wilson
Please see http://douglastodancing.blogspot.com/2008/10/question-time.html for a follow up to this comment
Post a Comment