Saturday, April 23, 2016

Some Recent Senate Polling And Related Claims

This article assumes, for the sake of analysis, that the Senate election will be a double-dissolution under the new Senate system.  Neither of these things are yet confirmed, but both appear highly likely.  

I thought that there had been no polling at all of the Senate races yet, and innocently told a journalist so this week, but to my surprise reports of not one but two Senate polls have surfaced (one since I made that comment).  There are also many reports of an (apparently unpublished) Australia Institute analysis that claims that from five to nine non-Green crossbenchers could get up at a double dissolution, apparently based on commissioned Senate polling from ReachTEL and Research Now.  The Research Now (an online panel poll a la Essential) polling has been published but the main thing I can find on the ReachTEL is an AFR report from a month ago (!) that had somehow escaped my notice.

Figures at the lower end of this five to nine seat range should not be surprising, and even the higher end, while a stretch in my view, isn't completely wacky.  It's generally agreed that the Nick Xenophon Team has enough gas in the tank to claim at least two SA Senate seats, quite probably three (about 19% of vote required, depending on votes from other parties - guaranteed at 23%) and perhaps even four (27% could be enough, a little larger than it polled last time).  NXT also persistently polls about 4% in states outside SA and on that basis seems to have some chance of snagging another Senate seat or two somewhere.  However it remains to be seen if this vote holds up when voters are confronted with a full list of options and when individual NXT candidates are scrutinised.

Of the remaining crossbenchers Jacqui Lambie is generally considered highly likely to be re-elected,  as Tasmanian independent Senators quite often have been in the past, though I've seen no public polling of her support level.  I couldn't rule out her winning more than one seat, though at this stage I doubt it.  Bob Day must be a reasonable chance given that he would have got up had the 2013 votes been cast at a double-dissolution under the new system.   Opinions vary widely on both Ricky Muir and Glenn Lazarus.  The remaining three (David Leyonhjelm, Dio Wang and John Madigan) are likely to struggle, but none can be written off absolutely.  Other high-profile names being speculated about by sources other than Glenn Druery (who shouldn't be taken seriously after his nonsense projections for the final Tasmanian seat) include Pauline Hanson (Qld) and Derryn Hinch (Vic).

It's interesting to see a rising awareness on the non-green left that quite a few crossbenchers could be returned, given that the non-green anti-Senate-reform sections of the left foolishly hitched their cart to Bob Day's nonsense about how Senate reform would exclude everyone except Coalition, Labor, Green and NXT.  Now that reform is a done deal (subject to High Court confirmation) that line of pseudo-psephology has suddenly gone in the trash-can, where it in fact belonged from the start.

Senate polling is a rare beast and The Australia Institute doesn't have a really obvious axe to grind in this case (beyond that their implied argument will generally be that the Coalition is stupid or out of touch in some way) so let's see how what has been released goes when examined in detail.

Issues with this polling

Starting with the Research Now polling, a few obvious issues stand out.  Firstly the Queensland sample has been offered "Liberal" and "National" as separate choices, but in Queensland neither of these parties exist and what needs to be asked for is a thing called "Liberal National Party".  Secondly the don't-know rate for the Senate sample is extremely high (18%), probably reflecting a common drawback of online polling in which "don't know" must be offered explicitly as an option on the same level as others.  It's hard to say who these don't-knowers are, except that they appear to be about two-thirds female.  I suspect that they are disproportionately soft major party voters, especially soft Labor voters.  As a general rule voters who are undecided almost never end up voting Green.

Thirdly the Reps and Senate votes in this poll are suspiciously similar! In the Reps they have Coalition 42, Labor 33, Greens 12 and Others 12, and in the Senate Coalition 42, Labor 31, Greens 12 Others 14.  In fact, even excluding the various strange distortions in the 2013 election, it has been normal for many elections for both major parties to underperform in the Senate compared to their Reps vote by 2-3 points nationwide.

Sometimes the Greens are partial beneficiaries, but mostly the benefit goes to the gaggle of micro-parties, care of some voters who don't want their preferred major party to have a Senate majority, or whose preferred micro-party isn't running for the Reps in their electorate.  This sample even has the Coalition at 38% in the Reps but 42% in the Senate in Victoria.  Polling that shows a major party overperforming its Reps vote in the Senate doesn't make the slightest sense except for the don't-know factor - perhaps left-wing voters from the Reps are more likely to say they don't know who they'll vote for in the Senate. It is the sort of thing where the full cross-tabs frequently seen in British polling would easily show what is going on beneath these numbers.

So there are a lot of reasons to treat these numbers carefully, but this is normal for Senate polling, so let's see where they actually lead if accurate (bearing in mind that the sample sizes for each state are very small and margins of error hence high). To read polling under the new Senate system, firstly divide each party's total by the quota (which for a double dissolution is 7.7%).  Many seats will fall to Labor, the Coalition or the Greens (and in South Australia, NXT) on raw quota alone.  Then of the rest it is a case of largest remainders win, except where there is good reason to expect that someone will beat someone else on preferences.  Here we should bear in mind that major party voters will be often distributing preferences following their party's card, while Greens voters will probably often preference likeminded parties and Labor, whatever the card tells them.  Micro-parties will find it harder to corral their supporters' preferences, since they will lack the resources to do so.


Queensland is going to be a very interesting Senate race because of the large number of vaguely competitive micro-parties.

Taking the Queensland votes, if exactly the Research Now votes were received then that would be Coalition 5.72 quotas, Labor 3.64, Green 1.30, NXT .65, Lazarus .52, PUP .52 (etc).  Five to Coalition, three to Labor, one to Green straightaway.  The Labor remainder would be good for four especially on Green preferences and especially given the points above about the don't know rate.  Then it's a free-for-all between micros and the Liberals for the remaining two.  However the online Research Now method is probably overstating the vote for micros as well as Greens, so let's say six Liberals and one micro.

(There would be no way to usefully say which micro, since the margin of error for each of these micros in the Queensland sample would be a few percent even if the sampling method was otherwise accurate. For the major parties the margins of error would be about 6% in Queensland - nearly a whole quota making this sort of sampling very unreliable!)

The ReachTEL figures reported by AFR however show uniformly lower support levels for the significant Queensland micros - NXT 3.4%, GLT 3%, PUP 2.9%.  Not enough details of the poll are published to say if it used a forced-choice method but if it did that could well explain the difference - respondents who will say "don't know" if offered the option will often pick a party if they have to do so to continue the survey.  Even on these numbers and with no other details there would seem to be a good chance that one micro or another beats one of the "big three" for the final Queensland spot.

Declaring the final seat in Queensland to be a race in two between NXT and GLT, as the AFR piece does, off a small sample size is extremely premature, as is making any deal of NXT's 0.4% lead in the sample. It's not even clear if all competitive parties were listed.

Survey underpolls unlisted parties

Another point to make on the Research Now polling concerns the vote for remaining micros who they describe simply as Independent/Other.  In NSW this is given as 6%, but this would have to include both the Liberal Democrat vote (likely to crash at this election) and voters for the dribble of micro-parties who stood in NSW.  In 2013, 13.5% of the NSW vote was scored by a gaggle of about forty micro-parties that each scored under 2%.  Some of this vote will go to Lambie Network or NXT, but given the implosion of the PUP vote there's just no way this gaggle will only be good for 6% between them.

In Victoria it's a similar story: if the poll is right then 12.8% of micro party voters, plus some refugees from PUP, plus voters attracted by Muir from the majors (if any) minus whichever of the 3% of NXT voters were already micro-party voters (probably very few) equals ... just 5%.  No way!

And in Queensland parties not listed in the poll scored 13% last time but are supposedly only good for 5%, which seems too low even if the poll's combined PUP+NXT+JLN+GLT vote is four points higher than PUP polled in 2013 (with the other three not on the ballot).  Katters Australian Party polled 3% in 2013 and isn't even on the list!

I think a real problem with this sort of polling is it doesn't accurately simulate a Senate ballot, where voters have dozens of parties to choose from and where votes do bleed from major parties especially to the micro-parties.  Under the new Senate system micros that poll below 2% won't have a ghost of a chance, but their votes will come from somewhere, and to what extent they flow back where they came from (rather than eventually exhaust) is an unclear question.  There is no sign so far of micro-parties being chastised into retirement by the new Senate system, but after they all realise that polling even 3% in a DD when you are up against dozens of other micros is very hard, a lot of them should get the message and merge or quit by next time.  So are the micro-party voters saying they don't know who they will vote for, or if they are naming a party, which parties' votes are being overstated from the readout as a result?

Overall there are so many issues and unknowns in this sort of polling that drawing meaningful conclusions from it is really difficult.  Both polls seem (on the detail provided) to support the idea that non-"big three" parties (including NXT) are in the hunt for one or perhaps two seats in Queensland, but they don't support it all that reliably, and it's not news anyway.  If there is news here it's the evidence from both polls that Glenn Lazarus might win but isn't currently likely to do so easily.   On the other hand, writing off Muir or Leyonhjelm (or for that matter Hanson, Hinch or KAP) on the basis of the Research Now poll alone would be foolish, since they apparently were not named in the "readout", and since forces not named in the readout are clearly undercounted in these results.  The PUP figures from Queensland are also interesting, but the sampling predates the recent Queensland Nickel news which should really have been the final nail in chances of PUP polling a non-risible Senate vote.

If anything, online polling has the most potential to poll Senate voting without problems, because an online survey can include dozens of party names.  It will be interesting to see if we see any more Senate polling that gets around the issues in the poll published by TAI.

We also need to bear in mind that Senate polling is very sensitive to late surges, as with Palmer United in 2013.  Senate speculation is lots of fun, but it is so sensitive to changes of even one or two percent that it's never a very reliable game.  It is a bit easier now that we no longer have group tickets, but that doesn't make it simple or reliable.

Some other comments

Finally, I make the following non-poll-based comments concerning Hanson, Hinch and Muir.  In 2013 a One Nation ticket with Hanson at its head received little over 1% of the vote in NSW, while the party in Queensland (albeit without Hanson as a candidate) polled barely half a percent.  Hanson is also a notoriously poor performer on preferences because of her image on race issues and because she is still - even now - poison to almost any party that openly preferences her.  Given the extremely crowded field of right-wing and populist micros in Queensland it is really hard to see Hanson polling the kind of vote needed to avoid being run down on preferences or just plain beaten on primaries by one or more of them.  She might well need 4 or even 5%.  An article by the Saturday Paper discussing the Hanson issue at great length just doesn't reference how much Hanson's vote has fallen over time.  I suspect some left-wing sources are talking up Hanson's chances in yet another of her endless comebacks because they want to mobilise concern about and opposition to her. [NB See note on Hanson at end of article.]

An SMH piece concerning the Australia Institute polling report has declared Muir to be "toast" but this seems to have no basis other than assuming that because he polled next to nothing last time he will poll next to nothing this time.  A counterpoint to this comes from the career of Xenophon, who was first elected in SA on preference harvesting off less than 3% of the primary vote, but look where he is now.  Another write-off for Muir comes from Glenn Druery, who is hardly neutral on this given that Muir dismissed him as a staffer.  Druery claims that "[..] Ricky will be hidden under the name of the Motoring Enthusiasts party and people won’t find him"  Druery seems to have forgotten that a picture of Muir will appear as his party's logo!

On the other hand, Nick Casmirri on Twitter has suggested that awareness of Muir is largely confined to the political classes, most of whom won't vote for him anyway, and that there is little awareness of him outside people who pay close attention to politics.  This is the most plausible argument as to why Muir might bomb completely.

As for Hinch, there is no doubting his profile, cult following, and understanding of and connections within the media, and he is clearly making an effort.  Unless there is polling clearly showing he lacks sufficient support he shouldn't be written off, but I doubt that he is fighting Ricky Muir for a single position.  I think they'll probably each win or lose pretty much independently.  Hinch and Muir won't compete that much with each other for the same votes, and if Hinch does very well it will probably be at the expense of the majors and especially elderly Coalition voters.

In Victoria the combined fourth-party vote is rather low (16.5% at the last election) meaning that the "big three" should have eleven quotas either locked in or else accessible on left-micro preferences.  If Hinch and Muir both poll competitively (and this is a very big "if") then that will probably mean the overall fourth-party vote has gone up a lot, and in that case they could help each other win.  Really, any speculations about micro-party wins in Victoria (by Hinch, Muir or anyone else) need to not get too excited at this stage.  If the total fourth-party vote there doesn't rise, then getting enough of it to win will be a tough task for anyone.

Anyway, I've started another Not-A-Poll in the sidebar so you can have your say on how many of the eight current crossbenchers will be returned!

Update (April 24): Metapoll Analysis:

Metapoll has a projection out with the Coalition winning 31 Senate seats, Labor 28, Greens 11, NXT 4 all in South Australia, Lazarus and Lambie, assuming a Lower House 2PP of 51:49 to the Coalition.

The usual difficulty with Metapoll arises in that, without any published details of what they used to get this projection, it's impossible to know whether this includes some of their own polling or whether it is based entirely on other published polls plus the odd assumption.  But 11 Greens strikes me as generous given that the Greens barely polled a DD quota in the NSW Senate race last election and polled below a DD quota in both SA and Queensland.  Even assuming a swing to the Greens of half a DD quota, which would see them on 12.5% nationwide in both houses, they would still have to be reasonably lucky to snag two of those three.  And the Greens are just not getting 12.5% nationally at the moment (the polls that have them that high were untested at the last election at which even the established polls all had them too high - and none of the latter currently have them above 11).

I'm also struggling greatly to find 28 ALP seats in a Senate that would have 11 Greens and 4 NXT in SA.  Labor is probably not going to increase its Senate primary all that much.  There are realistic prospects of five in Victoria and Tasmania, but I don't see how five Labor plus two Green flies in NSW on current numbers and I don't think four NXT, four ALP and one Green will all get in in SA either.  So I think this projection is overstating things for the Left overall by at least one seat.

The scenario of Labor plus the Greens winning 38 seats (a blocking majority) in the Senate despite the Coalition narrowly winning the Lower House is possible if the Left gets really lucky, mainly because Tasmania gets 12 Senate seats despite its small population, and from time to time the Left does very well there.  Labor and the Greens combined would have won seven seats in Tasmania at a double dissolution in 2001, seven again in 2007 and eight in 2010.  The local Liberal preselectors are currently hellbent on giving the Left as many Tasmanian Senate seats as possible again.  So while I think Metapoll have overcooked it slightly, I don't think their broad conclusions should be ignored.  A narrowly returned Turnbull government might indeed face a Senate at least as hostile as the current one, and numbers to pass bills through even a joint sitting could indeed be touch and go.


[Note added 22 May re Hanson: One thing my comments re Hanson did not take into account was the extent to which One Nation support in 2013 may have been gouged by the now-more-or-less defunct Palmer United.  A pre-PUP One Nation polled 0.91% in Queensland and 0.57% in New South Wales, but that was without Hanson as a candidate.  It might be assumed that Hanson's presence on the One Nation candidate list adds votes simply because Hanson is famous, and that on this basis for One Nation to revert to, say, 3% without PUP is quite plausible - which would place her in the mix for a seat, though I doubt that it would be enough.

But lest anyone blame the new Senate system for that, I point out that Truth Seeker simulations soon after the 2013 election showed that Hanson would probably have been one of about 15 crossbenchers elected at that election under the old rules had it been a DD - and this off 1.22% of the vote!]


  1. Kevin, I have noticed that Senator Abetz as lead Senator for the Liberals manages to lose about 15,000 votes when he leads the Liberal team. He has an ability to destroy the Liberal left with his RWNJ attitudes and they generally head towards an independent at his rotation of half senate elections. I imagine this will have a big effect at a DD election and may reduce the Libs down to barely making four senators. In 2013 a Colbeck led team attracted 127,000 (Abbott) in 2010 Abetz led the team 109,000 (Gillard/Abbott tie)and again in 2007 Colbeck 123,000 (Rudd) and in 2004 Abetz 146,000 (Howard). Abetz was the new Senator at this election. Given that Turnbull is no Howard and the 2004 result is probably not currently reliable. My hypothesis is that Abetz is actually an electoral liability and that the Liberals will only achieve a Senate vote of 98,000. The necessary votes needed in the preferential system will this time around be approximately 24,500 (DD). This gives the Liberals 4/12 Senators and no possibility of a fifth.
    The greens may score two Senators, neither Whish Wilson nor McKim are vote catches and results without Brown and Milne will be down. The ALP may have 4 or possibly 5, more likely 4. In any event there is the possibility that Lambie can acquire two Senate seats or another independent catching the imagination.

  2. I think it is not so clear Abetz was the main liability in 2010, and not even conclusive that he damaged their vote much at all. The Coalition is considered to have run a very bad campaign in the state generally in 2010, with a swing of 4.62% against it in the House compared to 4.42% in the Senate. Of course it's possible that Abetz caused the swing in both houses, but I think it's unlikely (bad broadband policy was a more obvious culprit). If Abetz had been the major culprit then the most likely outcome would have been a much bigger swing in the Senate than in the House. Yes there were factors increasing the swing in the House (such as competition from Wilkie in Denison and loss of sitting member votes caused by defeats in 2007) but a counterbalancing factor in the Senate was the 2% vote for the Shooters and Fishers, which may have come mostly from the Liberals.