The poll that lands first

Jordan Williams and David Farrar have spent twenty years learning how one poll can move another. Their Taxpayers’ Union–Curia poll just did it again.

  On Monday, the Taxpayers’ Union released its September poll. The fieldwork ran from 1 to 3 September. The headline: Labour at its lowest since July 2024, the coalition able to govern, and the Opportunity Party dumped back below the five percent threshold.

  Every other recent poll tells a different story. Roy Morgan had Opportunity on 9.5 percent. RNZ–Reid Research had it on 5.3. 1News–Verian had it on eight. Even The Spinoff noted the Curia number was bucking the trend of every other survey.

  Then look at the timing. The fieldwork dates are all public. Curia polls in the opening days of each month, sits on the numbers for a few days, then releases them precisely as the established pollsters go into the field.

  Watch how August unfolded. The Taxpayers’ Union–Curia poll, fielded from 1 to 4 August, was released on the Friday showing National ahead of Labour and the coalition clinging to power. Verian started dialling for 1News the very next day and found Labour in front. Reid Research followed a week later and put Labour ahead by three.

  Now September: numbers gathered in three days, held, then landed on Monday, just as the spring polling round begins and exactly two months from election day.

  I warned in June that releasing one poll while a rival is still dialling can nudge the next result. I documented the pattern in Unleashed. This is not a new trick.

  On 30 July 2020, Jordan Williams teased that evening’s 1News Colmar Brunton poll on Twitter, admitting he had once tweeted a result before broadcast and urging followers to “watch the 6pm bulletin tonight”. The result was talked up as ACT reaching the five percent threshold for the first time since Don Brash led National. The pollster’s own report put ACT at 4.8, rounded up on air.

  Two days earlier, National’s internal Curia numbers had been leaked to media while Colmar Brunton was still in the field. Curia polling for National kept leaking throughout 2020, always showing ACT higher than the broadcasters’ polls did. Farrar was National’s pollster.

  Both men were central figures in the Dirty Politics saga. I profiled Williams’ network of pressure groups last week.

  And in August 2024, Farrar resigned Curia from the Research Association of New Zealand while its Professional Standards Group was dealing with a complaint against his firm, a process the association took the unusual step of concluding without him. Curia remains outside the body that polices polling standards. The mainstream media report its monthly numbers as though nothing happened.

  Here is what worries me most. Over the past year, pollsters from four organisations have rung me: Reid Research, Verian, Roy Morgan, and Curia. I am a swing voter. The first three have each called between three and five times. Curia rang twice.

  When I worked for National, the party ran a “red dot, blue dot” database that categorised every voter. Farrar was key to maintaining it. A pollster with that database has a solid inkling of how you will vote before you pick up the phone. Choose who you dial, and you can produce any result you want.

  I have written about investment banks that fed one set of research to their whales and another to mum-and-dad investors. Ten Wall Street firms paid US$1.4 billion to settle those conflicts. Research that says one thing privately and another publicly is not research. It is marketing.

  So ask the obvious question. Farrar polls privately for paying clients and publicly for the Taxpayers’ Union. Can you imagine the consequences if the numbers his clients see ever differed from the numbers you see?

  A bill I drafted years ago and revised since would regulate polling and scrap the threshold that makes these games worth playing. Until it passes, treat any poll that always lands first, and always lands conveniently, with the suspicion its authors have earned.

If a pollster knows how you will vote before they ring, the poll is not measuring opinion. It is manufacturing it.