Chris Trotter: If polls matter, they must be robust

Twenty-seven per cent: the results from the One News-Colmar Brunton Poll fell upon Labour's hopes like a sledge-hammer.

Party members and supporters who, just days before, had been feeling proud and confident stared at the numbers in blank amazement. Everything they'd experienced since their tax-policy launch had reassured them that the party's plans were being well-received by the public. And yet, there they were, those wretched numbers: the worst in a decade.

It was all so demoralising. Why go on fighting? What was the point? When, in spite of most experts hailing Labour's capital gains tax (CGT) as not only a courageous and much-needed fillip to economic development, but also as a statement of basic fairness, the voting public simply turned their backs.

Such a savage rejection must surely mean that Labour and its supporters constitute an eccentric, almost cultish, subset of the New Zealand electorate? A sort of political Amish, whose beliefs and policies appear as outlandish to "Mainstream New Zealand" as the Amish's horse-drawn carts and buggies?

That is certainly what Labour's opponents would like its supporters to think. Which gives rise to the very troubling question: "Are opinion polls being used, quite deliberately, to demoralise the Government's opponents?"

Prior to the events of the past fortnight in Britain, such a question would have reeked of the political paranoia usually associated with conspiracy theorists. But, after the revelations exposing the corruption and collusion that has for decades defined the relationship between Britain's news media and politicians, it is a question that merits some sort of answer.

Let's begin with Colmar Brunton. Since at least 2004 there has been persuasive academic evidence that a statistical bias in Colmar Brunton's sampling methods causes it to consistently overstate the support of the political Right.
The company's telephone survey is conducted Saturday to Thursday evenings (inclusive). But, by conducting its polling mostly on weeknights, Colmar Brunton's academic critics (Rob Salmond, Keith Rankin) argue that the company is much more likely to make contact with higher income-earners  men and women whose electoral choices traditionally favour the more conservative political parties.

This statistical bias could be eliminated if Colmar Brunton's data was weighted to overcome the preponderance of higher income-earners among its respondents. But, although Colmar Brunton weights its results to match Statistics New Zealand's population data on age, gender, household size and ethnic origin, it does not weight them to match the income-spread of New Zealand's voters.

And neither, it appears, do the other major polling agencies: Reid Research, DigiPoll and Roy Morgan.
There is, however, one polling agency that does weight its data for income  with startling results. The latest HorizonPoll, taken shortly after the May 24 Budget shows a New Zealand political scene that is radically different from the picture presented by its competitors.

In HorizonPoll's survey, support for the "Left Bloc" (Labour, Greens, NZ First), stands at 43 per cent; with the "Right Bloc" (National, ACT, United Future, Maori Party) only just behind them on 42.7 per cent support.

Admittedly, there are more differences in the HorizonPoll's methodology than simply allowing for voter income. Even so, a news media motivated by a genuine desire to obtain the most accurate description of the electorate's mood would surely be concerned by the startling discrepancies between the HorizonPoll's results and the findings obtained using "standard" methodologies.

Indeed, the only plausible explanation for not being concerned is that, from the perspective of this country's largest media corporations, the "standard methodologies" are exerting a consistent and positive influence over the public's political perceptions and voting intentions.

By portraying the contest between Right and Left blocs as a hopelessly one-sided horse-race, the "standard methodologies" obviate the need for a more detailed examination of, and a more even-handed debate about, the political alternatives currently on offer.

But surely, if opinion polls possess the power to alter our perceptions (and they do), then the methodologies sanctioned by their sponsors must be above suspicion?

The Dominion Post