Gillard hardly needs enemies

NEXT time I'm invited to lecture on the subject of sound polling practice, I'll take along this week's claims by Essential Media as an example of how to stuff things up.

"There are three kinds of lies: Lies, damned lies and statistics," former British prime minister Benjamin Disraeli has been quoted as saying. It has a simple meaning: statistics can be contrived to suit outcomes. That is what Essential's boss Peter Lewis did this week, spruiking the notion that his research shows Greens and Labor voters agree on most things, therefore Julia Gillard should move Labor closer to the Greens as the solution to Labor's quagmire.

The thesis falls down on many levels. Methodologically, the data being used has been compiled and interpreted in a way that skews the results. (For a start, it needed to be disaggregated.) It also draws conclusions about so-called Labor voters that are overly narrow given Labor's vote is depressed. Essential needed to look at those polled who have switched main parties since Kevin Rudd was elected in 2007. And Lewis's opinions don't necessarily follow from what the data indicates.

That the bulk of Labor and Greens voters may agree on certain policy positions says nothing specific about swinging voters, the people who decide elections. It also doesn't enlighten us about which issues register as vote changers or whether a Labor move closer to the Greens would win back previous Labor support (from the Left or Centre).

A debate has been raging about whether the government's alliance with the Greens is a perception problem for Labor. Gillard has sought to put distance between her mainstream party and the conglomerate of protest elements that makes up the Greens. Doing so is a response to the growing perception (if not reality) of the Greens as a collection of radicals.

The strategy by Gillard is a reasonable one. Labor's base is deserting it and she must find a way to re-engage with working-class Australians who in the modern age are as likely to be drawn to the aspirational elements of the Liberal Party as Labor. Elections are won and lost in the mortgage belt outskirts of cities. Academic research shows a de-alignment in voting intentions is occurring across the country. This explains the willingness of one-time Labor voters in NSW to punish the party as strongly as they did at the recent state election.

In contrast, the Greens are largely a party of inner-city Sydney and Melbourne, where they dominate local councils and regularly threaten to gain state parliamentary representation. Adam Bandt won the seat of Melbourne from Labor for the first time at the federal election last year. Lewis has long advocated the line in a polemic sense that he is now running with data to back it up. But his views have directed the research, not the other way around.

For a start, Essential polls online, which immediately marginalises some citizens from its findings. Older voters, time-poor voters and lower socioeconomic status voters are less likely to participate in an online poll. Conversely, people hauled up at internet cafes in Newtown or on Chapel Street (or playing with their iPads and iPhones) are disproportionately likelier to engage in such a poll, just as their views will disproportionately mirror Greens policies. I am sure Essential adjusts its data for demographic equivalence, but that can also increase the margin of error as an unintended consequence, especially when sample sizes are low.

Perhaps the biggest methodological flaw in the Essential research is the assumption that if its polled Labor voters mirror the Greens on issues of interest, that is a benchmark for the views of traditional Labor supporters. But Labor's primary vote is at its lowest ebb in its history. Across polling agencies it is in the very low 30s. For every two voters preferencing Labor first, three are voting for the Coalition. What Labor needs to grapple with isn't what voters sticking with it in the worst of times think. It needs to re-engage with those who have deserted the party, namely those who have bolstered the primary vote for the Coalition. Essential ignores this.

Deserters have gone in two directions: to the Greens and to the opposition. Greens voters will come back courtesy of our compulsory preferential voting system (other than in the odd inner-city seat Labor may lose to the Greens).

Leading political scientist Ian McAllister from the Australian National University has long spoken of the importance of understanding the influence of compulsory preferencing on our system.

In recent years the percentage of Greens voters who come back to Labor has jumped from about 70 per cent to 90 per cent. In other words, the argument that Labor needs to worry about "Greens leakage" has diminished.

But working-class voters who leave Labor to support the conservatives because they don't like the radicalised association of Labor and the Greens aren't coming back, certainly not if Gillard follows Lewis's advice. They need to be the target of Labor positioning if it wants to win elections.

My old PhD supervisor, Campbell Sharman, used to say it didn't matter what issues people identified as of interest to them, it mattered whether they were vote changers. The Essential research claims that, from the carbon tax and nuclear power to same-sex marriage, Labor voters and Greens supporters largely see eye to eye. When asked a simple question perhaps they do, but if those questions are refined to burrow deeper into which issues will change people's votes, often very different results are found.

For a start, the correlation between the collapse in Labor's primary vote and Gillard's personal ratings on the back of announcing a carbon tax tells you that issue is a vote changer for many (or at least her broken promise is).

Lewis, however, is only including those polled who have stuck with Labor despite this shift when he claims Labor and Greens voters see eye to eye. While Essential found that on the issues mentioned Greens and (what remains of) Labor voters agree, I am willing to bet that the passion of those on the Greens side of the ledger is stronger than that of Labor supporters. That passion will prevent them passing on preferences to Tony Abbott just as the passion among disaffected one-time Labor backers will lead them to refuse to support a Labor Party seen as being too close to the Greens.

Essential Media ran the Your Rights at Work campaign. One of its former employees is now Bob Brown's chief of staff. It has a commitment to the left of politics. But there is a difference between research informing strategic moves and ideological views informing what to do. This week Lewis let the latter inform his opinions.

Lewis reflects a misreading of the electorate common in Labor circles: that the electorate is becoming more progressive, not less, and that sooner or later everyone will come around to its agenda. His research also reflects a growing misuse of data as the quantitative limitations of polling are set aside in the name of opinion.

Any politician who underestimates the innate conservatism of the electorate is headed for a fall. With friends like Lewis, Gillard has no need of enemies.

Peter van Onselen is a Winthrop professor at the University of Western Australia. He has written about opinion polling for The Australian Journal of Political Science.