Wednesday 9 November 2011

Vicious Spiral

Spiral Of Silence: The ability of opinion polls to materially shape the outcome of elections is strongly indicated in the academic research. If the Election Night result suggests that our major polling agencies somehow got it wrong, a great many New Zealanders will have every right to feel extremely angry.

IT’S GOING TO BE one of the most interesting “results” on Election Night, and quite possibly one of the most devastating. How closely do the “snap-shots” taken by the major opinion polling agencies resemble the actual election outcome?

It will be interesting because for the past three years the conclusions of the major polling agencies (Colmar Brunton, Reid Research, Research International, DigiPol) have sharply contradicted the intuitive assessments of many ordinary voters. To these folk the idea that John Key’s government is supported by more than 50 percent of decided voters is ludicrous. Again and again, you will hear people say: “I simply don’t believe National is that high, or that Labour is that low.”

Not surprisingly, the academic experts in the field of statistical research have reacted rather condescendingly when confronted with this response. Their stock reply has been to invite the doubters to consider the historical evidence. And why not? Averaged out in the so-called poll-of-polls, the agencies’ final “snap-shots” of the electorate have consistently, and accurately, mirrored the actual election outcome.

There are a whole host of personal and ideological reason why people may not want to believe the poll results, say the experts, but that does not mean that they’re wrong.

But what if they are wrong?

Over the past week there have been at least three statistics-based depictions of the electorate’s political preferences.

The latest, from Research International, confirms the findings of the other major polling agencies. National rides high with 52 percent; Labour flounders at 26 percent; and the Greens, on 12 percent, continue their impressive upward climb. No other party comes even close to crossing the 5 percent MMP threshold.

A very different picture emerged from the second survey. Conducted by a trio of Weekend Herald journalists led by Simon Collins, it showed National on 43 percent, Labour on 31 percent, and the Greens on 14 percent. The sample amounted to just over 500 persons, and the interviews were conducted face-to-face on the streets of the nation. Collins and his team readily acknowledged that the sample was skewed in favour of younger voters and that this was likely to advantage Labour and the Greens.

But what about the major polling agencies’ reliance on telephone interviewing over land-lines? Doesn’t this skew their results in favour of mature, middle-class, settled suburban voters willing to talk to pollsters? And doesn’t that person sound a lot like your typical National Party voter?

Isn’t it possible that the methodologies of the major polling agencies are consistently over-stating the level of public support for the right-wing parties? Let us suppose for a moment that they are, or that it is, at the very least, a possibility. Is this possibility of disortion acknowledged by the newspapers and networks who commission these polls? Or does each media outlet report “their” poll’s findings without the slightest reference to potential methodological shortcomings or contradictory data?

Which brings us to the third survey, released on 2 November by Horizon Research Ltd. This poll, conducted within methodological parameters radically different from those employed by the majors, presents a dramatically different picture of the 2011 election campaign. In Horizon’s poll, National stood at 36 percent, Labour at 30 percent, the Greens at 14 percent, NZ First at 6.5 percent, the newly-formed Conservative Party on 4 percent, Act on 3 percent, Mana on 1.8 percent, the Maori Party on 1.2 percent and United Future on 0.8 percent.

If those numbers are accurate, then the NZ First Party and its leader, Winston Peters, will once again play the role of Kingmaker in post-election negotiations.

It’s right about here that things turn ugly.

Using the Colmar Brunton poll as their guide, TVNZ decided to exclude NZ First from the minor party debates. Other media outlets (including Radio New Zealand and Sky Television) have done the same. Indeed, the reporting of the entire election campaign is being shaped, to an unhealthy degree, upon data supplied almost exclusively by land-line telephone interviews. Not surprisingly, a National victory is taken for granted. Labour is painted as a party of losers. The Greens are said to be making impressive gains. And Winston Peters is dismissed as irrelevant.

Overwhelmingly, this is the picture of the campaign being presented to the electorate. It is, therefore, highly likely that a large number of voters have already succumbed to the psephological effect known as “TheSpiral of Silence”.

Now consider this counterfactual.

Imagine that the newspapers and broadcasting networks have based their coverage of the election on the Horizon poll. That all the talk has been about the “knife-edge” election. That Winston Peters has found himself and his party at the centre of media attention. Public engagement in the campaign has been considerably higher – and so has the turn-out on Election Day. In short, the Horizon-based picture of the contest has been dramatically more inclusive than the major agency-based picture.

Which brings us back to Election Night.

How devastated centre-left voters will feel, and how justifiably furious, if the actual voting statistics indicate that National’s support was grossly exaggerated, and that Labour has fared considerably better than all of the major polling agencies were suggesting. Imagine their anguish if NZ First, in spite of being almost entirely shut-out of the mainstream news media, wins 4.9 percent of the Party Vote. And if NZ First’s desperately narrow failure to crest the 5 percent threshold turns out to be the difference between a National-led and a Labour-led government.

If that is the way the numbers fall on Election Night, hundreds-of-thousands of New Zealanders will not only be perfectly entitled to say “bugger the pollsters”, but they will also be entirely justified in asserting that the election has been stolen from them by a news media which placed far too much faith in what has proved to be the major polling agencies flawed methodology.

This posting is exclusive to the Bowalley Road blogsite.


Tiger Mountain said...

Agree Chris, though some say youngish professionals ditch landlines too these days in favour of ‘naked’ broadband and smart phones, not just the young and poor. But the real poll will tell the story.

The polling accuracy or possible discrepancy will be right up there alright for political followers with the general election and the referendum result.

Just because your breakfast has not poisoned you to date does not mean it never will. Conditions change.

Gem said...

Chris – this post is sure to spark some interesting discussion. I’ve already posted this in a comment on your blog a couple of weeks ago, but I think it’s worth restating. I contacted TVNZ to express my concern about the continued sidelining of Winston Peters and New Zealand First. A TVNZ Corporate Affairs Spokesperson, Georgie Hills, justified Peters’ absence in terms of the results of September and October Colmar Brunton polls.

In response to my expressing similar concerns in comments responding to a post, someone under the username Jellygirl (who presumably has a professional association with TVNZ) advised “It's [debate participation] based on Colmar Brunton polls because they are the ones designed and paid for by TVNZ and therefore the one TVNZ knows is robust. The Horizon poll, for example is self-selecting and offers prizes to those who sign up - it's not taken seriously.”

The spiral of silence is fascinating. I do believe that National, particularly John Key, has genuine popular support. However, as per the spiral of silence theory, the mainstream media has inflated this popularity by constant, sycophantic cheerleading. If gleeful media representatives tell you repeatedly and emphatically that Phil Goff is a weak loser and Mr. Key is popular, it becomes the truth. Presumably, the desired effect is for people’s “I don’t want to back a loser” response to kick in, thereby further widening the gap between Mr. Goff and Mr. Key.

Take, for example, Paul Holmes, Duncan Garner and Guyon Espiner. These men fawn over Mr. Key in particular, and are reduced to salivating stupidity in his presence. Perhaps Mr. Key has a mesmerizing effect on them. Or, do man crushes reduce these journalists to bootlickers? Whatever – in any case, it is disheartening to see such blatant partiality in the media, particularly, as far as Mr. Holmes and Mr. Espiner are concerned, from the state broadcaster. Naturally, people have political and other biases. However, it would be in the interest of journalistic integrity for political reporters to put their preferences to one side and provide balanced coverage of political issues.

Mark Wilson said...

Given the relative accuracy of the polls at previous elections the left is clearly clutching at straws. What fascinates me is Labour's personal attacks on Key at the last election and again this time and they are getting worse yet it rebounds on them and thye can't see it. Dumb and dumber.

Dave Kennedy said...

Excellent post, Chris, let's hope the election result will be a catalyst for a media shake up!

Ed said...

Thanks for this article. Please publish this in one of your newspaper articles and speak on this issue when discussing the election on the mainstream media.
Questioning the use of polls to influence the electorate is an important step towards an improved democracy.

Ed said...

Thanks for this article. Questioning the methodology of polling and their use to sway voters is an important step to an improved democracy

Anonymous said...

Nota comment but 2 questions:
1) What is different about Horizon polls?
2) Was Horizon polling in the previous election and if so, how did their polls stack up?

Loz said...

This will be an interesting election, if only to assess Horizon’s polling formulas. Horizon adds weight to responses (presumably) by gender, age, physical location, ethnicity etc. to get a better view of “types” of voters. This relies on previous trends being “universal constants” for the maths to work. This weighting system guarantees that Horizon results will be different from polling results of other firms.

Interestingly, Horizon's formulas not only suggest the return of New Zealand First (in a big way) but also an 11% swing against the government - a backlash of the order experienced by Rogernomics Labour in 1990 (a 12.8% swing)... which was the worst election defeat of any government in New Zealand's history. It's a little difficult to accept.

If the other polling firms are right and support for NZ First and Act has evaporated since the last election you could expect that vote to be absorbed by National. Based purely on the votes cast from 2008, National could expect around 52.65% of the vote. This is exactly where the latest Colmar Brunton, Fairfax and Herald Digipoll surveys believe National's support to be.

Alternatively, in the last election, Labour and the Greens shared a combined vote of 40.71%. Again, the latest Colmar Brunton, Fairfax and Herald Digipoll surveys have (within the margin of error) Labour and the Greens sharing that same level of support. Labour's low level of support is proportional to the increased support of the Greens.

All that the polls currently suggest is that those who supported the National bloc in the last election still do while the Greens support base is growing at Labour’s expense.

Anonymous said...

"If that is the way the numbers fall on Election Night, hundreds-of-thousands of New Zealanders will not only be perfectly entitled to say “bugger the pollsters”, but they will also be entirely justified in asserting that the election has been stolen from them by a news media which placed far too much faith in what has proved to be the major polling agencies flawed methodology."

Labour cannot win this election. They could be correct on every issue and have Bill Clinton as leader and they still would not win. Voters are more conservative in times of uncertainty. This is dumb, but so are they.

I honestly do not know why people are making a fuss because the result has not been in doubt since 2008. Labour might as well not have bothered. Only an idiot would want to win this election.

Boss Hogg said...

How can anybody be concerned about Winston not getting some lime light - Please. And Banks. Surely we can move on from these relics. If these are the only friends that Labour and National have then they both live in a very sad and lonely world.

The polls will not be so wrong and you are worried, but please do not encourage deals with the Devil.

Carol said...

Loz, the Fairfax poll:

Headlines National at 52.5%. But this doesn't include the "undecided" figure of 14.3% down the bottom (do the maths).

So, unless all the undecideds either go National or don't vote, then really National could only count on 44.9% of the votes going by that poll.

Stuff/Fairfax are guilty of poor analysis of, and misleading headlines of this poll.

MPledger said...

The move to MMP has really changed our need for high quality data because people can use their vote strategically.

For example: If National are over 50% in party vote then some people will vote ACT to keep them alive but if National are under 50% then they might vote National to hope to break the 50% threshold.

The landline polls have done a really good job in the past to match election day outcomes. Part of that may be because they shape how (or if) people vote rather than being very accurate in of themselves.

Landline based polls should tell us their response rate (and how they calculated it) because that would be a much better indication of how much we should trust the polls.

Anonymous said...

Thanks for this post Chris. I would like this issue raised in the media but the chances seem to be about zero. How many media outlets are going to allow criticism of polls that they pay for? I must admit i'm getting serously P***** off at being told what I think by the media.

Sanctuary said...

I don't know anyone who regularly uses their landline anymore except my parents, who are their 70s.

Even amongst those who still own phones, how many still answer them? The absolute plague of marketing calls means I don't answer my landline anymore (I only have until we get an unbundled line) unless I am told to expect a call (a message via Facebook or something) from someone at a specific time.

I know the polls were broadly accurate three years ago, but three years is a very long time in telecommunications technology. One would think that the polling companies would not want to have their polls and polling methods thoroughly discredited - it is how they make a living, after all.

Time will tell, I suppose.

Loz said...

Carol: In the weeks just prior to an election, undecided voters generally distribute their votes the same way as committed voters (or decide not to vote at all). It's unlikely that all of the undecided voters will head toward a single party and fundamentally alter polling results. Point taken though.

There used to be an observation that "National voters go out in the rain". Low turnouts (for whatever reason) tend to be at the expense of the left wing and Labour in particular. With this election, more than a quarter of 18-24 year olds aren't enrolled. Nearly one in ten of us aren’t registered to vote either. The past three elections also show that 20%-25% or registered electors won’t actually vote on the day.

The fear of the left shouldn’t be that National's current poll standing is accurate. The real nightmare is that the polls are overstating the number of Labour supporters who, regardless of intentions expressed to pollsters, will end up voting at all.

paul scott said...

John Key is losing political capital to sell assets, this is a disaster for him,
while labour and green will get some little claw back over no assets sales.No sane New Zealand person would sell assets.
Naturally as ever the very astute Winston is selling this fact, and an equality fairness politic.

Carol said...

Yes, Lolz, but there's the influence of the poll itself and the way the MSM present it, that can influence what the undecideds do on polling day, with the kind of results that other commenters here have referred to.

By not factoring in the undecideds, and by not pointing out that their eventual voting behaviour is not a certainty, the media/Stuff are presenting a false sense of finality.

Also, some poll results count in the indecideds and some don't, creating a false equivalence.

I incorrectly said in my first comment that National would stay at 52.5% if all the undecideds decided to vote National. In fact, for the National vote to slip below 52%, it would require less than half of the 14.3% undecideds to vote National - very possible in an MMP environment.

Not to mention, just going on the party vote, doesn't take into account some of the complexities of MMP.

In the end, I would much prefer, as Chris has said, that the media focused on discussing and fairly analysing policy, and got away from this rather superficial preoccupation with polls - treating it more like a game, where the punters aim to choose winners, rather than the democratic intent of voting for the preferred people, parties and policies

Mark Wilson said...

There is a reason that the left have difficulty getting their supporters out on polling day - the same reason that keep them in the losing camp - they can't be bothered.

Than said...

Chris, if using landlines is biased towards National surely this could be shown by comparing previous years polls versus actual election results?

You said it yourself - "Averaged out in the so-called poll-of-polls, the agencies’ final snap-shots of the electorate have consistently, and accurately, mirrored the actual election outcome." The reliance on landlines is a weakness, and one that will only get worse as more and more people give up landlines in favour of cellphones/Skype. But so far nobody has come up with a better way. Horizon's methodology is deeply flawed, and doesn't produce reliable results.

KiwiGee said...

I wonder if the media polls published so far have been misleading because they report only commited responses. Those staunch amongst us have already decided and those supporting the encumbant will have no concerns. Those with concerns will be undecided and as they do decide then the polls will move accordingly. If the undecided are those that swung last election then methinks the next 2 weeks will see that swing reverse to some degree.