ABOVE: Media Take presenters Toi Iti and Russell Brown
I've finally got around to watching Maori Television's valuable new series Media Take. Given that I helped kick-off the recent social media debate on the Undecideds (albeit with a relatively brief analysis), I was particularly interested to see Russell Brown's interview with Colmar-Brunton's Andrew Robertson and UMR's Gavin White. I'll be dealing with the issue of Undecideds in a later post, but at this stage I just want to correct one or two minor points on other matters.
First
In the opening sequence of the final segment on opinion polls, Brown suggests that "Until 1971 the only way to tell how New Zealanders might vote was to go out and ask them personally. Or to wait until we'd actually voted. We were late to political polling."
Well, yeah, compared to the US, the UK and even Australia we certainly were late.....but not quite that late. In fact, the first bona fide, nationwide pre-Election poll was carried out as early as 1960. In that year, a polling company calling itself Gallup New Zealand (not, apparently, a subsidiary of the famous American Gallup Poll organisation) carried out its first pre-election poll. Gallup New Zealand employed the (now) old-fashioned quota method, with interviewers fanning out into city streets and elsewhere with the aim of finding a representative sample by filling quotas for various demographics in the same proportion found in the census.
Click on Read more for rest of analysis
They apparently conducted their polls on spec and then attempted to sell the results to New Zealand's leading Newspapers, sometimes successfully (as in 1960), sometimes not. Although I'm relying here on a memory of something I read a few years ago, I'm pretty sure the company carried out polls for all three elections between 1960 and 1966. And they appear to have been reasonably accurate. Ultimately, however, Gallup New Zealand went under, not least it seems because much of the 1960s New Zealand Print Media remained unconvinced about both the science and the quality of opinion polls. The National Party and Social Credit League, incidently, each began to carry out private polling in the early 60s, but - due to both a lack of financial resources and (like the print media) a highly sceptical attitude among its leadership - Labour didn't move in the same direction until the 70s (I'm still not sure of precisely who conducted these polls for National and SC).
And then, in 1969, the National Research Bureau (still an active research company, though these days no longer bothering with political opinion polls) carried out its first two pre-election polls (September and November 1969). They covered everything from Party Support, to Party Leader Attributes, to which Party would be best or most effective on a range of different issues. Heylen began at roughly the same time, but it was a few years before they managed to carry out their first pre-Election poll. And it wasn't until 1975 that all of Heylen's polls were conducted on a nationwide basis. Before then, their sample was often (though not always) confined to the greater Auckland area.
Even before 1960, however, some enterprising university-based political scientists managed to carry out a small number of pre-Election polls in individual electorates. The first was in Wellington's Mt Victoria seat in 1949. There were at least two more, one in 1957 (by R.S Milne in Wellington Central) and one in 1960. They tended to choose highly marginal seats on the basis that, given their limited resources, these could be considered a rough approximation of the Country as a whole. And their detailed data yielded some quite valuable insights. It had long been believed, for instance, that women were more socially and politically conservative than men and thus more likely to vote National. (Hence, for instance, James K Baxter's "National Mum and Labour Dad" in his Ballard of Calvary Street). Milne found that, although this common assumption was indeed true in Wellington Central, things were a little more complex than that. His poll suggested single people were at the extremes - single men far more likely than anyone else to vote Labour, single women far more likely than anyone else to vote National, with married couples much closer to each other in their political sympathies (albeit with married men still a little more Labour-leaning than married women). (I'll have more to say about the party support gender-divide in an up-coming post, because obviously things have changed dramatically since then). Austin Mitchell and one or two other political scientists continued to carry out pre-election polls based on 1, 2 sometimes 3 or 4 electorates throughout the 60s.
Second
Later in the interview, Russell Brown points out "Winston Peters complained that the polls always underrated him and Winston was right, wasn't he ?" UMR's Gavin White replies: "Yeah, Winston is right, it's consistent that New Zealand First always does better at the election than they do in the polls before that." White was talking, here, about New Zealand First's Election-day Party Vote, relative to their support in the final round of opinion polls conducted a week or so earlier. He makes the same suggestion in a post on his SAY it blog (here... http://sayit.co.nz/blog/when-does-49-become-5 ) and argues this consistent pattern goes right back to at least the 1999 General Election if not before ( here... http://sayit.co.nz/blog/what-political-polls-tell-us ).
Well, Gavin White is certainly voicing a common assumption here but, in fact, it's not entirely true.
While there's no doubt NZ First's support was indeed under-stated in the run-up to the last two General Elections, that certainly wasn't the pattern in 2005.
In 2011, every one of the final round of polls put NZ First below the 6.6% they actually received on Election day. Roy Morgan came closest with 6.5, while 3 News Reid Research was the most errant with 3.1. What's more, none of the 58 polls of 2011 recorded NZ First as high as 6.6. Indeed, well over half of them had the party down below 4%, often as low 1.5 or 2%.
Three years earlier, at the 2008 General Election, just one of the Five major public polling companies (again Roy Morgan - 4.5%) put NZ First above its Election day result of 4.1%. The other polls had them anywhere between 2.4 and 3.9. And only 5 of the 63 polls carried out in 2008 put Winston Peters' party above their Election day result.
At the 17 September 2005 General Election, New Zealand First received 5.7% of the Party Vote. And in stark contrast to 2008 and 2011, all but one of the final round of polls actually placed NZ First higher than this:
Roy Morgan Early September 6.5
3 News TNS Early September 6.8
Herald-DigiPoll Early September 4.5
One News Colmar-Brunton Mid September 6.0
Fairfax Media-Nielsen Mid September 7.0
I might add that roughly half the polls conducted throughout 2005 put NZ First up in the 7 - 11% range. While it's certainly true, then, that opinion polls tend to understate (for whatever reason) the support NZ First ultimately receives at the subsequent Election, there's clearly no hard and fast rule here.
