Mumble

Nicholson in the Oz. Click for larger.

Michelle Grattan in the Age,  tables (regular questions only) here. Table and questions below.

Tony Abbott has a similar boost Malcolm Turnbull received, but these sorts of 2pp voting intention numbers are still only about Simon Crean (2001-3) on a very bad week.

However, good news for the opposition in 45 percent preferring his climate change solution to the government’s on 39. Then again, 43 percent like Kevin Rudd’s climate approach more broadly versus 30 percent for Mr Abbott’s.

Nielsen’s John Stirton reckons, paraphrased by Michelle, that ”the apparent contradiction probably reflected voters’ low level of understanding.”  It might also reflect voters’ opinion that Tony doesn’t really believe in climate change. And one question is about the government and opposition, while  the other uses leaders’ names.

See drop in support for ETS (although still decent majority in favour; also note the changing wording) in this table:

ACNielsen support for ETS

Respondents were also asked: “Thinking about climate change policies generally, which do you prefer – Kevin Rudd’s broad approach to climate change or Tony Abbott’s broad approach to climate change?”  43% preferred Mr Rudd’s approach while 30% preferred Mr Abbott’s broad approach.

Respondents were then told: “Both the Government and the Opposition have climate change policies that aim to reduce greenhouse gas emissions by 5 per cent by 2020. The Government’s policy is to introduce an Emissions Trading Scheme. The Opposition’s policy is to set up a fund that would be used to purchase initiatives aimed at reducing greenhouse gas emissions.” Asked which policy they preferred, 39% preferred an Emissions Trading Scheme and 45% preferred an Emissions Reduction Fund.

As pollies drag themselves to federal parliament for the start of an election year, we should get a Newspoll in tomorrow’s Australian, Fairfax hopefully cranks up ACNielsen soon, Galaxy will start appearing in News Ltd tabloids, Mr Morgan never takes a break, and then there’re the online ones …

There’s a conundrum in the poll-producing process. I like opinion polls as much as (more than) the next person and wish to see lots of them. But they are inevitably over-interpreted, and it can’t be any other way.

Newspapers usually pay pollsters oodles to conduct them; if you or I commissioned a similar stand-alone survey (and not just tacked onto an omnibus) just once, the figure someone mentioned to me a few years ago was around $20k. However, my understanding is that the pollsters supply them at cut-price to papers because they love the publicity. And for good reason: Newspoll, Nielsen, Morgan and Galaxy all earn the vast bulk of their income from non-political market research, but who would have heard of them if it weren’t for their political polls?

The grapevine also reckons only Newspoll charges a close to full rate to its client, the Oz, and as they’re both Rupert-owned (or at least half in the pollster’s case) the dough largely remains in the family.

The Oz runs these federal polls every fortnight, sometimes more often than that, sometimes with extra questions. The annual bill must be huge, and so they’re obviously not going to bury them on p.8, about where they belong in any sensible consideration of their news value. Nor will the paper caution readers not to get carried away because this is only one modest dip in the public opinion ocean.

No, they’ve paid good money and are going to put it on the front page, get their top political writers onto it and squeeze what they can. It’s the same for all the papers, albeit with lesser frequency.

And the rest of the media plays along, repeating whatever theme Mr Shanahan or whoever has nominated, so that by days end every political class member, including many pollies themselves, has internalised and is repeating it.

So first item on my wishlist is an admittedly forlorn one: that individual opinion polls are treated a little lightly. Instead, the trends are important.

Newspoll and rounding

On the subject of Newspoll and the Oz, wish number 2 is that they reconsider their presentation of “swings”. This is particularly relevant as the election approaches and they measure the various geographies. What they currently do is round their findings to the nearest integer, and then calculate a swing from the actual election result (to one decimal place), so giving an unrounded number as the swing.

So, for example, if Newspoll finds 55 2pp percent for Labor, it is presented as a “2.3 percent swing” from the 2007 result of 52.7. This is mathematically silly and misleading, because their “55″ was really anything from 54.5 to (a touch under) 55.5 and so the “2.3 percent swing” is really anything from 1.8 to 2.8.

The solution: work out a swing using unrounded poll numbers and then round that. (Or even not round it.)

Continuing on rounding ….

Funny behaviour in the provinces

A couple of newspapers (in Adelaide and Perth) work out their notional 2pps like this. They take their pollster’s primary vote data, which is rounded to the nearest integer (and usually from very small samples) and calculate 2pps based on flows at the last election (basing on flows at the last election is good) and present that to one (or even two) decimal points.

 !

Recommendation: do what the big outfits do. Get the pollsters to estimate two party preferred numbers from their unrounded numbers and then round those.

Roll out the online polls

Online polls have appeared in recent years and we’ll doubtless see more this year. They are much cheaper to run and may eventually spell the end for phone surveys. I wrote about Nielsen’s for Crikey in 2007.

Essential Research, which seems to have a relationship with Crikey, (apparently not, see comments) produces weekly polls from an online panel. I have misgivings about their panel, as I wrote here. The crucial link there is now broken, but their online panel looks likely to be skewed towards in an environmental/community (ie left-leaning) persuasion.

How they go during the campaign will perhaps be the test.

Nielsen ran online polls alongside their phone ones at the last two elections and will likely do so again this year.

The debilitating effect of opinion polls

We’ll get a record number of published polls this election year (as we always do). This is a good thing for observers, but for the politicians themselves, particularly incumbents, it can mean more caution, self-consciousness and even perhaps dysfunction. 

Nothing we can do about that.

Tables at bottom of Ms Maiden piece in the Oz.