Monday, January 16, 2017

GetUp! National And Centrelink Poll Reporting Is A Trainwreck

GetUp! ReachTEL (undecided redistributed) Coalition 37.1 ALP 35 Green 9.8 PHON 10.6 
Published 2PP 54-46 to Labor
2PP by 2016 preferences 52.1 to Labor
Verdict: Go back to sleep

My little eyes lit up when I saw in my Twitter stream that I had somehow missed the release of a national ReachTEL at about midnight last night.  There has not been a national ReachTEL since before the July election. Given the relative paucity of polling data since then, and the Centrelink and ministerial "entitlements" issues currently affecting the Turnbull Government, new data concerning where the government was standing could be quite interesting.

Unfortunately it turned out that this was not a new Seven or Fairfax ReachTEL, but rather one commissioned by the lobby group GetUp!  Moreover, the level of immediate publication of the poll's details has been abysmal.  Rather than it being promptly released with full details either on GetUp!'s website or the ReachTEL site, what seems to have happened is that it has been sent (in part or full, who knows) to a range of media agencies who have then presented us with a partially digested dog's breakfast of the findings.

Thus for instance from the SMH we learn that:

"While primary support for the Liberal party and Labor was deadlocked at 32 per cent, the opposition captured the majority of the "undecided" support 33.3 per cent to 19.4, with 58.5 per cent of those still making up their mind answering they would preference Labor higher."

For the Turnbull Government to be behind the Labor Opposition on the primary vote (after redistributing the undecided) would be startling news indeed, given that the Government has held a primary vote lead - albeit generally a feeble one - in all polls since the election but for one Newspoll and five Essentials that showed ties.  However, from the figures published by Poll Bludger it becomes clear that the Government is ahead on primaries in this poll too, because the figures published by the SMH are, as stated, a reference to the Liberal Party only, and ignore their Coalition partner the Nationals.

Poll Bludger publishes the primaries as 37.1 Coalition, 35.0 Labor, 9.8 Green, 10.6 One Nation.  GhostWhoVotes on Twitter additionally gives Other as 7.5, which doesn't quite sum to 100 but the loss of a few tenths here and there in rounding will make little difference to the picture.  The published 2PP is 54-46 to Labor.  However, by last-election preferences, this poll comes to only 52.1% to Labor.  So, unless One Nation voters have suddenly developed a massive preference for the Labor Opposition over the Coalition (which is possible, but would fly in the face of the entire history of One Nation preferencing to date) then we are probably looking at the normal problems with respondent preferencing: it skews to Labor overall, but it also makes the estimate of the 2PP vote in specific polls more volatile.

If the primary votes are really as stated in this poll then the conclusion to be drawn would be that the Government has suffered no damage on a 2PP front from the Centrelink debt recovery atrocity or the Sussan Ley scandal.  Compared to the BludgerTrack aggregated primaries as at the end of last year, which produce exactly the same aggregated 2PP as my own estimate, the only real action to see is a drift of about a point each from the Government and Opposition to the seemingly rampant One Nation.  The One Nation surge in this ReachTEL could be a product of sample noise or sampling method rather than a genuine shift, but you shouldn't bet your bottom dollar on either of those at the moment.

Even if the poll did show a primary vote shift that matched the reported 54:46, this would not amount to evidence that the government had taken any hit at all (it could just be sample noise), and even if there was such a hit it would not show it would be caused by anything in particular.  So GetUp! has got the story it wanted about the Centrelink debt recovery atrocity damaging the Government's standings, but the evidence behind the claim is flimsy.

This is no way to conduct national polling releases.  If GetUp! want to be taken remotely seriously they should release their polling properly rather than only feeding it (in an unknown form) to the media.

Centrelink: A Personal View (Rant Warning!)

Readers may have noticed that I describe the Centrelink debt recovery system as an "atrocity" above. From what I have been able to glean from the large amount of often confused coverage of this matter, Centrelink recipients present and past are sent letters advising that data matching has suggested possible discrepancies between what they were paid and what they should have been paid, and asking them to verify their details online.  I would have no problem with this whatsoever if it provided an easy road to resolution for those who were in fact paid the correct amount.

However, recipients who earned income for a part of the year while they were not on benefits are in some cases confirming that their annual income is correct, and then for having the temerity to naively make a true statement, are being plunged into a nightmare in which a bogus "debt" is raised against them and they have to challenge it and prove they don't owe money.

Challenging is onerous because of the difficulties of collecting information from several years ago and severe problems with being able to contact Centrelink in a timely fashion without getting wrapped up in red tape and incompetence.  Also, fortnightly pay slips often mismatch fortnightly reporting to Centrelink, in part because of different fortnightly schedules.  Centrelink's bizarre practice of asking recipients to report earnings in advance (what work you did for which you will be paid) rather than as they are received is a further complication here. It is a classic case of trying to use data because you have it, without considering whether the data are fit for purpose.

For those having to challenge a false debt, this can become what in a busy world is a nightmare: an open-ended time-sink in which one must waste an unknown amount of time for which one is not compensated, in order to (maybe) prove that one is not guilty of a debt for which one has been presumed guilty on insufficient evidence.  It doesn't help that in some cases if the original payments were incorrect, this was in fact as a result of Centrelink incompetence with their own rules rather than any wrongdoing by the recipient, and yet the recipient will be made to do the donkey work.

This should be a national disgrace as it is totally contrary to supposed Australian values of a fair go. But it is not an isolated case; it is just the latest ramping-up. The administration of welfare payments, in which jobseekers who are financially and psychologically vulnerable are humiliated and dehumanised by being compelled to apply for jobs they have no chance of getting, has been a disaster zone for at least 20 years, with bipartisan support from the major parties and little meaningful opposition from anywhere else.  Even lobbying from the Greens, for instance, has far too often focused on the financial levels of payment (which is an issue) and not on how recipients are treated.

Centrelink: The Polling

Much as I would like to report that GetUp!'s polling shows that Australians share my views of this appallingly clumsy and nasty debt recovery scheme and Centrelink in general, the GetUp! polling on Centrelink isn't very useful.  As related by Poll Bludger (the exact wording of the questions being elusive in all the mainstream media coverage) recipients were asked:

“The Turnbull Government recently started using an automated system issuing tens of thousands of letters to Australians about alleged debts from Centrelink overpayments. The Government admits that at least 20% of these letters are incorrect, but the burden is on Centrelink clients to correct the information or pay the debt. Do you support or oppose the Government stopping the automated debt collection system?”

In undeserved fairness to the Government on this point, the Government in fact disputes that 20% of the initial letters are "incorrect" (let alone that it has said they are) on the grounds that the initial letters merely raise a potential issue and ask for confirmation of income.  (The sleight of hand in this claim concerns what happens when a person agrees that their annual income was as stated without realising this will then result in an incorrect debt.)

In any case even if the claims in the preamble are true, the placement of claims favouring one side of the debate and not the other (for instance, that the scheme is successfully recovering some amount of money) could potentially bias the poll in favour of a negative result for the scheme.  But the double negative in the final question (asking recipients whether they support or oppose stopping the scheme) might confuse some respondents.  Moreover, some respondents might oppose stopping the scheme but might support serious changes.  So the net finding of 46.2% supporting stopping and 31.8% opposing stopping is simply number salad - you can make of it anything you want.

A further question reported is:

"The Turnbull Government has acknowledged significant errors in the Centrelink automated debt collection system. Where there are potential errors, do you think the burden should be on Centrelink to verify their claims against information they already have on file or on the individual to defend themselves, which may include accessing pay slips and employment records from up to five years ago?”

Here 78.6% agreed with the former option while 21.4% disagreed, and it seems that we have another problem here, as it seems forced choice was used for an issues question, which is inappropriate as it forces respondents who have no opinion to concoct one or abandon the survey.

It is also questionable that the Turnbull Government has acknowledged significant errors - Alan Tudge still doesn't. But that aside, the problem with this question is that it is likely to be primed by the previous questions and therefore the result is unreliable.  This highlights the need for full detail to be released in such polls immediately including the order of all questions asked and their full wording.

There is also the usual question about whether the scheme has made people more or less likely to vote for the government (49.8% less likely, 14.4% more likely.)  The effect size is a little larger here than the 30% that a question of this kind will produce for pretty much anything, but the level of priming from the other questions in the poll contaminates it.

The Centrelink debt recovery scheme is a very serious matter, and the letters sent out so far are only the tip of an intended iceberg.  The impact of the issue on the ability of the agency to even function could be very severe, with consequences beyond those who are even affected (the numbers of which alone stand to become very large.)  The issue deserves much better than to be used as a tool in GetUp!'s usual game in which scientific polling is something you muck about with in order to convince the media that the electorate feels a certain way.  I can only hope that the Centrelink debt recovery matter will feature in properly designed national polling commissioned by neutral sources, so we can get a better idea of what voters actually think.

A nearly final word on this belongs to the GetUp! campaigns director Mark Connolly, as reported by Sky News:

"Treating Australians like crap is going to get you crap poll results."

No, in this case it has only got us a crap poll.

Thursday, January 12, 2017

How many federal electorates have you visited?

When I travel outside my home state I like to pay attention to which federal electorates I'm in at the time.  A few months back I thought it might be fun to try to work out how many federal divisions I had actually visited.  Because I live in Tasmania and do not drive, my score is not as high as it might otherwise be - indeed for a nine-year period covering most of the 1990s I didn't leave Tasmania at all.

For this task I set a couple of ground rules.

Firstly I imposed an age limit - anywhere I went before I turned fifteen doesn't count.  15 is the age from which my travel decisions were generally independent.  Before age ten I travelled around Brisbane and up and down the eastern seaboard with my family a lot, which would add many electorates to this list, but there's no hope of me remembering everywhere I went.  

Secondly, I exclude any electorate I was just passing through (or over) for travel purposes between points outside that electorate and without staying overnight in the process.  Last year I spent three hours in Moscow airport because flying from Baku to Dubai via Moscow was much cheaper than flying there direct and didn't get me home any later, but that doesn't really count as having "been to Russia".  

Thursday, January 5, 2017

2016 Ehrlich Awards For Wrong Predictions

It's time for the fifth annual giving of the Ehrlich Awards, which round the start of each year go to the most amusingly or staggeringly wrong predictions I observe in any field of interest relating to the previous twelve months.  The Ehrlichs are named for Paul Ehrlich, the evangelist of ecological end-times who put me on a path to a lifetime of health scepticism of dark green gloomery when he not just lost a famous bet with economist Julian Simon but also gave poor excuses for the defeat. For the past editions click the Ehrlich Awards tab at the bottom, and for the ground rules see the 2012 edition.

As usual I should note my own predictive efforts are hardly perfect, but I had a pretty good year in 2016, missing by only two on the Coalition's national Reps seat tally (for example).  I did wrongly predict two Reps seats in my own home state, which was embarrassing, though I did indicate those were quite uncertain.

Two areas of widespread predictive failure that will dominate these Awards were the US election and Senate reform.  They weren't the only ones I noticed.  For instance the late Bob Ellis made a bold bid for posthumous glory with "It is likely, though not certain, that Malcolm Turnbull will lose his seat" - Turnbull won his seat 68:32 with a trivial 1.2% swing against him.  But I think the two I've mentioned are the most interesting ones and I've decided to make the US election case the dishonourable mentions, saving Senate reform for the medals.

Saturday, December 31, 2016

2016 Site Review

Another year done for a site that is now just over four years old.  When I started this site I intended it just as an interim site while I explored other options, but in spite of some very minor frustrations (like difficulty in stopping spammers from even submitting comments) I've seen no compelling reason to move.

This year was a federal election year and the activity pattern for the year looks like this (the units are sessions per week):


The federal election and its post-count events basically swamp everything else.  Senate reform performance reviews maintained a high level of reader interest through to mid-August and after that the rest of the year was quieter, with little spikes for the ACT and NT elections. The massive federal election spike meant that this year had about 128% more traffic than last year, and about 60% more than the previous busiest year (2014).

Sunday, December 18, 2016

What Chance One Nation Seats In The Tasmanian Parliament?

Recently there have been some noises about Pauline Hanson's One Nation trying to break into Tasmanian state politics at the next election.  The next state election still could be 15 months away, if the parliament goes full term, and before then we'll have the WA and probably the Queensland elections, which both look like fertile ground for the resurgent party.   But anywhere might be fertile ground if a recent ReachTEL in Victoria that had the party on 9.4% there is to be taken even half-seriously. One Nation's current national surge might fall in a big heap by the time Tasmanians next go to the polls, but let's suppose it doesn't.  I thought it was worth a detailed look at the sort of chances the party might have, supposing that it makes a serious effort.

The case for One Nation as a threat is pretty easily stated.  The party very nearly won a seat in the state at the Senate election, albeit when competing for one of twelve seats rather than one in five per electorate.  Its primary vote was low (2.57%) but it received about another 2% in preferences from micro-parties that might reasonably be expected not to contest the state election.  Throw in regional variation and it's easy to project One Nation above 6% in both Lyons and Braddon.  Throw in that the party's national polled support is running close to double what it polled in the Senate and something like 10% in these seats starts to look pretty viable.

One Nation might appeal to some voters who are displeased with the current state government but would hate to go back to another minority government where the Greens hold the balance of power.

Tuesday, December 13, 2016

Poll Roundup: 2016 Year In Review

Aggregate 52.0 to ALP (+0.1)
Labor would win an election "held now"

It's about the time of year when polling shuts down for a few weeks, so as usual I will post an annual recap.  If there are any late polls then I will edit the text to add them in.

In the three weeks since the last roundup, we've seen some evidence that the anti-Coalition trend in polling has softened, but not all that much.  Newspoll has come in from 53:47 for Labor to 52:48, while Essential's last three readings for Labor have been 51, 52, 53.  I aggregated the Newspoll at 52.1 and the Essentials at 51.3, 52 and 53.1.  There was also the Ipsos discussed in the update to the previous piece.  All up I have things at 52.0 to ALP, down from 52.4 three weeks ago:


Whatever is going on in February and March is likely to shape polling much more than whatever (if anything) is causing the very small comeback that we see here.

Wednesday, November 30, 2016

SA's Voter Choice Bill Could Be Better

Earlier this year, the Australian federal parliament successfully passed Senate voting reform, which abolished the Group Voting Tickets that had trashed the 2013 election, and returned control over preferences to the voters.  I've had plenty to say already about how successful that was (performance review part 1, part 2, JSCEM sub (PDF) etc).  However, while Group Ticket Voting has been put in the bin at federal level, hopefully to remain there for good, it remains in place in the upper houses of Victoria, South Australia and Western Australia. (For completeness, NSW fixed its system after the 1999 debacle, Tasmania has a single-member-per-seat Upper House and Queensland has no Upper House at all).

I'm not going to expend a huge amount of energy on trying to get state upper house systems reformed in states other than my own.  But with South Australia the first state where a move away from GVTs is the subject of legislation since the 2016 Senate outcome, I think it's interesting to have a look at what is being proposed.  In attempting to get rid of Group Ticket Voting, the SA Labor government has come up with a near-polar opposite.  The proposed alternative, while still much better than keeping GVTs (as is just about anything really) is so inferior to the Senate and NSW systems that I wondered for a moment if it was built to be voted down.  It appears this isn't the case, so hopefully parties in the SA Upper House (where the Government holds just eight out of 22 seats) will be able to amend the legislation to improve it.  An interview with the Attorney-General in July did suggest he was open to different models provided that preference harvesting got the chop, which is commendable.