31 Jul 11
Ex-patriate political scientist Rob Salmond has told Radio New Zealand's Media Watch programme Horizon Research is making good moves in polling undecided voters.
On the July 31 programme Mr Salmond was discussing the large differences between DigiPoll, Colmar Brunton and Reid Research poll results - and ones being produced by Horizon.
Mr Salmond said the other polls have proved reliable in the past. Online polls by U Gov in the UK and USA had produced reliable results, but "we don't know yet about Horizon".
Asked if Horizon's probing of don't know voters could help explain differences, Mr Salmond said asking the undecided about their intentions was "very important" and New Zealand had been "somewhat of a backwater" on this.
"Horizon is making good moves in this direction," he said. "It's one good aspect of Horizon."
In June political economist Keith Rankin of Massey University told TVNZ's Media7 programme that other polls are misleading and Horizon would tell a better story between now and the November 26 general election.
Links in this report also include ones in which Horizon has recast a DigiPoll result, by including up to 30% of respondents who either say they don't know, won't say how they will vote or are choosing not to vote. Horizon says this gives a complete view of the political marketplace: http://www.horizonpoll.co.nz/page/135/political-sc
The other published polls are reporting results for less than 100% of respondents. They report the decided voters only as a percentage of 100. This pushes National over 51%. However, if undecided and other voters were to be added back in, no party would be likely to get over 50% or can govern alone.
Horizon does not criticise the others' polls or methods.
Horizon asks different and additional questions - and then filters results differently.
In this way it qualifies respondents as registered to vote, likely or definitely going to vote - with a decided party vote or preferred party (if undecided).
Horizon understands it is also the only pollster weighting results on 2008 party vote. In the DigiPoll analysis, removing party vote weighting resulted in up to a 5% support shift for National.
All Horizon political polls are also weighted by six factors to ensure the respondent sample represents the population aged 18 or older. More than 95% of Horizon's nationwide panel members are recruited by invitation, to represent the population at the 2006 census.
Questions for the other pollsters and those who publish them include:
The close nature of left and right wing party bloc voting support apparent in most of Horizon's polls indicates why the main parties have been making significant moves to ensure MPs are elected for minor support parties.
A good question for DigiPoll, Colmar Brunton and Read Research: will you publish results please for decided voters, undecided voters with a preference, the won't says and those choosing not to vote?
What will the support numbers be then for the main parties? And will you gather samples of 2000 or more, so results have lower margins of error and the position of minor parties is shown more clearly?
If they do this, will National be able to govern alone, as most recent headlines have said (even though with the don't knows at about 12% or more put back into the pool, no party will have more than 50%)?
Would this type of complete reporting of the political marketplace show a closer race - and therefore a more interesting one for media to cover? But more importantly better reflect the current state of the political marketplace?
Why are some publishers and editors, investing heavily in political polls, not insisting on presentation of full results?