Labor’s position continued to weaken in the opinion polls over the second week of the election campaign.
For Labor the best news came early. Essential Research’s poll released on Tuesday 13 August showed no change in the two-party-preferred vote, with the Liberal-National Coalition on 51 and Labor on 49.
The same poll showed a slip of two points in Tony Abbott’s approval rating, but a slip of five points in Kevin Rudd’s, compared with those of mid-July.
Looking at the Essential data from July and August, it seems that more people are making up their minds about Mr Rudd (there are fewer undecideds) and they do not seem to like what they see.
It seems that more people are making up their minds about Mr Rudd (there are fewer undecideds) and they do not seem to like what they see
He is also down three points as preferred Prime Minister – although at 47% still a healthy 12 points ahead of Mr Abbott, who is unchanged on 35%.
The news for Labor got worse at the end of the week, with Newspoll reporting a shift of two percentage points in the Coalition’s favour on the two-party-preferred vote, opening up a gap of eight points (54-46) compared with four points (52-48) a week earlier.
While technically the shift is within sampling error, taken with the other available data from Newspoll, Essential, ReachTEL and JWS Research, the shift to the Coalition looks real, even if the exact size can't be nailed down.
The rest of the week’s published polling was largely concentrated on marginal seats.
In Victoria, according to ReachTEL, Labor might recover the seat of Melbourne from the Greens, but lose Deakin and Corangamite to the Liberals. The Liberals also appear to have strengthened their hold on Indi, and according to JWS Research seem likely to win Aston.
For Labor, that adds up to a net loss of two seats in Victoria. For the Coalition is adds up to a gain of three.
In New South Wales, JWS Research shows Labor losing Lindsay and Banks, the Liberals strengthening their hold on Macquarie, and Greenway being too close to call. Also in New South Wales, Newspoll shows the Coalition winning the seats of New England and Lyne, previously held by the retiring Independents Tony Windsor and Rob Oakeshott.
For Labor, that adds up at best to a net loss of one seat in New South Wales. For the Coalition it adds up to a gain of at least four.
In Queensland, JWS Research is showing the Liberals likely to retain the seats of Forde and Brisbane. In Forde, despite the parachuting into it of the former Labor Premier of Queensland, Peter Beattie, the sitting Liberal, Bert van Manen, seems to have turned a marginal seat into quite a safe one.
So the arithmetic from this week’s polls in marginal seats is:
Labor: +1; -5 = -4
Coalition: +7; -0 = +7
Too close to call: 1
When we are asking about issues that have high salience in an election campaign we need equivalently high standards of questionnaire design
So much for the arithmetic. Now for the matter of questionnaire design.
In addition to the standard questions about voting intention, approval of leaders and preferred Prime Minister, Essential Research’s poll released this week contained a battery of questions about national debt and government spending. It provides a case study in the difficulties of polling sensibly on complex issues, especially when there is so much election rhetoric in the air.
The first question was: Compared to other developed countries, do you think Australia’s national debt is higher, lower or about the same?
It turns out that 25% thought our national debt was higher than that of other developed countries, 48% thought it was lower and 10% didn’t know. And voters’ responses differed according to their voting intentions: Labor and Greens voters thought it was lower, and Liberal-National voters were split on the question.
This suggests that voters’ answers were not grounded in knowledge – they didn’t think anything about this -- but were impressionistic and coloured by political preference.
It also became clear from the overall line of questioning that the survey was really asking about government debt, not combined government and non-government debt. A little more precision would have improved the question.
The question would also have been more precise if it was made clear that it was people’s impressions that were being sought: did they have the impression that Australia’s government debt was higher, lower or about the same as that of other countries like ours?
If we know – and competent pollsters like Essential do know -- we are asking about impressions, then for the sake of validity and reliability, we should ask about impressions.
Then having asked about levels of debt, the survey went on to ask whether it was “more important to reduce debt or maintain government spending on services and projects”. Unsurprisingly, 48% said it was more important to reduce debt, while 35% said it was more important to maintain government spending, and 17% didn’t know.
It is unsurprising because “debt” is a very specific term, and in the current political debate it has high salience and negative connotations. By comparison, “services and projects” is a vague term, and even though the poll went on later to ask people what should be cut, this setting up of debt versus spending was actually telling us more about the cut-through quality of Mr Abbott’s campaign message than about people’s policy priorities.
This was illustrated by the fact that there had been a shift in attitudes away from spending and towards reducing debt since Essential last asked that question in May. (In that time, as a matter of fact, Australian government external debt fell, though it remains high by historical standards.)
If a question were asked about reducing debt even if it meant less spending on something specific like schools or hospitals, then it would be a real test of voters’ policy priorities.
Another issue presented by this battery of questions was that respondents were then asked: Do you think the Government should raise taxes or cut spending to reduce the national debt or should it do neither?
This begs the question that the national debt should be reduced, a question that was not asked. Offering the option to “do neither” doesn’t solve this problem. It still presupposes that the national debt should be reduced.
This critique is not saying the questions were no good, only that they could have been done better and that the information we take away from it needs to be qualified.
When we are asking about issues that have high salience in an election campaign we need equivalently high standards of questionnaire design.