We are being conned by the polls – the Tarot Cards of politics

loading animation
Thursday, 1 March 2012 16:08 by Ad astra
Imagine this – a world without opinion polls. Then ask yourself whether in such a world the leadership contest played out this week would have occurred at all. Consider on what it was based – a decline in the polls for the Government, in Julia Gillard’s popularity, and in her popularity compared with that of Kevin Rudd and Tony Abbott. Over and again we were told how important the polls are in the fate of PM Gillard and her Government. If they slip any further, she approaches political death, or is irretrievably doomed; if they recover, resurrection might be possible, although in the view of many, that is out of the question.

In our culture numbers carry great weight. Descriptors such as ‘declining’ or ‘improving’, or the more dramatic terms journalists prefer to use, have nowhere near the potency of numbers, not just because they are imprecise, but because we assign so much more power to numbers. They are subject to calculation and comparison in a way words are not. More of this later.

I can hear you saying: ‘But there was more to the Gillard – Rudd contest than just polls numbers’, and of course you are right. From all the accounts we have now had the chance to read, there had been white-anting of Julia Gillard and the Government by Kevin Rudd for over a year, but this was not unambiguously reported in the media until the crisis was upon us. What was reported in the polls, week after week, month after month, was the Government’s score in the TPP stakes, PM Gillard’s standing in the popularity stakes, and most significantly her standing compared with Rudd in the ‘preferred Labor leader’ stakes and with Abbott in the ‘preferred PM stakes’, and to make sure we got the real message, her standing in the ‘who is most likely to lead Labor to victory’ stakes.

It was the numbers with which we were assailed, and every deviation from previous numbers was tediously analysed, and learned interpretation of the deviation offered. It was almost totally a numbers game, a poll numbers game.

In New Matilda on 27 February Ben Eltham, commenting on the outcome of the leadership ballot in a piece: Gillard won, Labor lost, had this to say: ”…in modern politics, winning elections is no longer enough. Staying ahead in the polls appears to be just as important. It’s a fortnightly test of political legitimacy that creates constant pressure on under-performing leaders and parties. Few leaders can survive a sustained run of bad polling without at least some rumblings within their own party. If Labor was leading the Coalition in primary vote polling, Julia Gillard’s position would never have been in doubt. For Gillard, it’s hard to see this as much of a victory. She has certainly proved the support of her colleagues. But while her polling figures remain dire, her position will always be in doubt.” and later ”…it seems almost impossible to imagine that Labor’s poll figures could recover to an election-winning position.” For Eltham, polling numbers seem to be crucial. He is not alone. For Dennis Shanahan, who regards himself as the supremo in poll interpretation, numbers are everything. Even small deviations, within the usually accepted margin of error of three percent, are given prominence, especially when they signal a decline in the PM Gillard’s or the Government’s position. While not psephologists, the likes of Michelle Grattan, Peter Hartcher, Matthew Franklin, even Andrew Bolt will try their hand at poll interpretation, and use the numbers to make their points. It’s always the numbers.

Obsession with the numbers would be reasonable if they were reliable and if they actually meant something. It is tacitly assumed by the public, and unfortunately by many of the poll ‘experts’, that they carry important meaning, significant pointers. Before accepting this with such touching faith, we need to ask how valid and reliable are they.

Last month I wrote about polls in How opinion polls poison politics The part of that piece relevant to this one reads: “Without going into tedious detail, polls are only as reliable as the quality of the sampling and the size of the sample…Getting a sample that is truly representative of the opinions of the entire Australian electorate is the greatest challenge to pollsters.

A sample size of around 1,000 carries a margin of error of around 3%; with smaller samples (some may sample as few as 600), the margin of error rises. To reduce the margin to 1%, around 10,000 would need to be sampled, but this is too expensive for the pollsters. While pollsters acknowledge these sampling drawbacks, usually in fine print, they usually do not feature prominently in any commentary, so that readers tend to regard the figures as ‘gospel’ and attribute more significance to them than the figures warrant. Even minor deviations, within the margin of error, are given credence.

The thrust of this piece is that the recent leadership battle was predicated on the polls more than any other factor, and that the reliability of some aspects of these polls was of such questionable validity and reliability that to use them to precipitate such a major upheaval was wrong.

What the interpreters of polls seldom tell us is that polls are not predictive, but the other day on ABC TV, Peter Lewis from Essential Poll said just that, followed quickly by a quip ‘that he would probably be thrown out of the pollsters league for saying that’. Yet every pollster and psephologist knows that to be so. But of course polling organizations don’t want this inconvenient truth exposed as their business and the revenue it generates depends of the results of their polling being accepted as valid, reliable, and able to predict. I have been asking how eighteen months out from the next scheduled election contemporary polls could possibly predict what will happen then. We hear the ridiculous statement, mainly on news bulletins, ‘if an election was held today, the Government would be ‘annihilated’, ‘reduced to a rump’, ‘be out of power for a decade’, or ‘lose X number of seats’; take your pick. They say that, despite the fact that an election is not being held today, and probably not for eighteen months, and knowing full well that all polls narrow before an election and that in recent times most election results have been close, with the vote often within the 51/49 percent range. In other words, polling organizations, and all who feed off them, are conning us deliberately. Why is this so?

It’s because polling organizations are lucrative, self-perpetuating businesses that have found they can rely on the public giving them credence, and because their owners, often media outlets, depend on them for easy, cheap copy that can be sensationalized into catchy headlines and startling stories. Imagine how bereft Dennis Shanahan would be without polls. What would he write about? Polls are mentioned in almost everything he pens.

So let’s not imagine that we will ever be told the truth about polls; they are too lucrative and too central to political reporting to admit that they are in reality not much better than Tarot Cards, used from the late 18th century until the present time by mystics and occultists in efforts at divination.

Who then are we to believe? I submit that we can believe only the best of our analysts, the brilliant statistician Possum Comitatus, and Andrew Catsaras, who will now have a regular slot on the ABC’s Insiders. They focus on trends, not individual polls. Possum’s Pollytrend is statistically sound and Andrew’s analysis will also give us trends, using a different statistical method.

Let’s accept then that polls on voting intentions are not predictive, but what about popularity polls? The psephologist Mumble has written an interesting article Rudd’s first demise in The Australian which is well worth reading. Amongst many other interesting observations, he says: ”I’m not a great believer in ‘satisfaction ratings are more important than voting intentions’ stories, but they do tell us something”, and referring to Rudd’s removal in 2010, “As I’ve said before, the idea of chopping down a PM because he has dipped behind in the polls doesn’t pass the laugh test. By that criterion Howard would have been gone after 18 months and Bob Hawke after three years.

Vex News didn’t think much of the polls either: ”Rudd wasn’t removed because of bad polls or mining taxes or flip-flops on carbon. We now know – in unvarnished truth – because of the carefully-considered yet brutal truth-telling of Rudd’s most senior colleagues – why he was removed.” Yet it was the numbers in the polls that were quoted over and again, and exactly the same was done as journalists and sundry ‘experts’ insisted that Julia Gillard must have ‘a spill’ to ‘clear the air’ that they insisted was continually overshadowed by the low clouds of unpopularity in the polls manifest by an adverse approval/disapproval ratio, ‘poor TPP polls’ and ‘Rudd is favoured (usually by twice as many) as favour Gillard to lead the party’. All these numerical measures were considered incontrovertible evidence, not to be denied.

When asked for approval/disapproval, how many based their opinion on previous polls, thereby succumbing to the bandwagon effect? How many have been influenced by the incessant disingenuous bagging of the PM by Tony Abbott and much of the media? When asked about preferred Labor leader, how many based their opinion on the Kevin Rudd they believed they knew, the Rudd they saw in public places in his ‘hail fellow well met’ mode, rather than the Rudd they saw exposed this week? How valid are the propositions that pollsters put to their subjects? How reliable are the responses? How predictive are their responses? Would even the most committed pollster assign the predictive potency to their polls that commentators do? Surely not! Yet the poll numbers were used as a powerful lever by politicians and journalists to insist that a leadership ballot must be held, eventually of course precipitated by Kevin Rudd’s resignation. It is not just the pollsters and the journalists that respond to the leverage of the polls, it is the politicians too. Although I suspect many of them are as skeptical about their value as I am, they have been bludgeoned by the pollsters and the press into taking notice of them. Despite their virtual uselessness in painting an accurate picture and their acknowledged incapacity to make reliable predictions, the pollsters have conned the politicians and most of the public into believing these Tarot Cards must be obeyed.

Our own Jason queried the validity of the polls when Rudd seemed to be saying: "…the polls say I'm popular so you must have me back.” Jason added “If Rudd was to be reinstated because he's popular in the polls, would he then be bound to legislate whatever else was popular because of those same polls? 

I may very well be wrong…but to be ‘governed’ by ‘polls’ is a slippery slope to get on I think.”

In response, Lyn collected a number of quotes about polls made in the Fifth Estate. Margaret Simons said: ”Is public opinion (i.e. polls) the same as public wisdom? 
Those watching the gaping hole that has opened up between the political class and the public over who should be prime minister might wonder whether there a difference between public opinion, as measured by the giant strainer of opinion polls, and public wisdom.” Jeremy Sear said: ”The polls are a poor approximation for ‘the public’. One poll counts and it's in 2013.

Others said: “"…the measure of a government is not by opinion polls or daily headlines" and ”Polls don't predict anything - they just tell us what happened recently …” And the much respected George Megalogenis has often commented on the slavish attention afforded opinion polls by the political community in this country. In Poll-driven parties put the individual first at electors' expense in The Australian on February 25 he began: The nuttiness in the Australian political system did not begin with Kevin Rudd's sacking in 2010. The main parties have been yielding to personality-based politics since at least 2003.” Later he refers to “…the hijack of public debate by opinion polling.” Read the rest of his article to see what he thinks about personality and opinion polls.

It seems many are now seriously questioning the validity of the very polls that are assigned so much potency, so much importance, so much predictive power, so much influence on crucial political events. It’s time we in the Fifth Estate called them for what they are - a confidence trick, a political Tarot Card.

How long should they be allowed to con an unsuspecting public?

What do you think?