Sunday, April 24, 2016

How Many Lower House Seats Can The Xenophon Team Win?

Note: For links to updated analysis re NXT written since this article, scroll to the bottom.

I have had a fair few questions about the Nick Xenophon Team (NXT) and their seat prospects for the Lower House in South Australia, because my seat projection nationwide still shows only four seats for Others.  This rather dry and mathsy piece (about Wonk Factor 3/5) explains my initial attempts to model what sort of vote NXT needs in South Australia, all else being equal, to win one or more of the 11 Lower House seats in the state.  These are very broad-brush attempts that don't take account of candidate factors, because until there is specific seat-polling we don't know anything objectively about how well (or badly) specific NXT candidates are going to campaign or be regarded.  Comments based on detailed local knowledge are welcome.

In summary, I estimate that NXT probably need a statewide vote in the very high teens to win seats.  Once they get well over 20 they start to win multiple seats, and something in the mid-20s could result in spectacular seat gains that would make a national hung parliament quite likely.  However there's no reason yet to believe those higher votes will actually happen, based on what little current public polling of the NXT vote exists.

The NXT Lower House run was initially seen as an insurance strategy.  In 2013 NXT nearly polled enough vote for two Senate seats but was ganged up on by other parties in preference-dealing, resulting in Labor preferences putting Family First's Bob Day into the Senate instead of NXT's second candidate Stirling Griff.  Although how-to-vote cards will be less important in the Senate this election, by running Lower House candidates Xenophon can cause seat losses for any major party that tries to damage his preference flow in the Senate, giving him an insurance strategy he probably wishes he had last time.

It's worth briefly noting a history of near-wins of third-party seats in SA.  The Australian Democrats polled 15.3% statewide in 1990 but did not come all that close to winning any seat.  Janine Haines (leader) missed second place in the cutup in Kingston by 7.4%.  In Mayo the Democrats candidate missed second place by just 2% but it was irrelevant as Alexander Downer had an outright majority.  A famous closer miss came in 1998 when the Democrats' Jon Schumann came second in Mayo (although the statewide Democrats vote was only 10%).  Schumann needed 86% of preferences to beat Downer and got 81% of them, losing 51.7:48.3.  The 1998 result provides a template for what NXT need to do to win seats: outpoll or overtake one major party then win on their preferences over the other.  It is assumed that both majors will preference NXT ahead of each other, or that if they do not do so then their voters will do it for them.

What is NXT polling?

The only national polls that currently include NXT in the "readout" are Morgan and, recently, ReachTEL.  The national vote for NXT isn't relevant because the party is only going to run in selected non-SA seats and isn't remotely likely to win any of them (an amusing shot at a certain northern beaches relic in NSW notwithstanding).  What is of interest is the SA vote.

Morgan have released seven fortnightly readings of the NXT support level starting in late January.  These have been (in order) 22.5, 15, 16.5, 18, 18, 22.5 and 22.5, an average of 19.3%; the sample size would typically be a few hundred making these results very wobbly.  The single SA ReachTEL reading for Morgan so far (from their April 14 poll) was 17%.  I consider ReachTEL more reliable than Morgan, especially as Morgan's combination of methods (SMS and face-to-face) are both prone to overcapture enthusiastic voters, and ReachTEL has the better track record than either of Morgan's current methods.  However we will need to see more ReachTEL polling in view of sample size issues to see whether this particular sample was representative.

Other polls that lump NXT with "others" in their SA samples could well be under-recording the NXT vote, but it's worth noting that they generally do point to even lower NXT figures than the above before that is taken into account.

There has been one seatpoll of Mayo in January in which NXT were on 16.3% after redistributing the undecided.  That ReachTEL poll had them competitive for second place, but requiring 91% of preferences assuming they could come second, which is too much.  Still, a winnable position would be within that poll's margins of error, even then.  That is a very old poll now in the context of a new party campaign.

Other polling reported has often showed NXT in an extremely competitive position in Mayo and Sturt especially, but this has always been union polling or internal party polling and none has been released in an adequately documented form.  Also, this result hasn't been unanimous, with an AMWU-commissioned ReachTEL showing NXT on only 14.5% in Sturt and 12% in Hindmarsh (add a percent or two for undecided voters).  Seat polling is never very reliable at the best of times. A recent tweet from commercial radio station Fiveaa (5AA) is of interest here:

(By comparison, NXT polled 28% in both seats at the last Senate election.)

The "David" in the tweet above is David Penberthy, who has many more such reports (including a suggestion that major party polling has NXT set for four seats with the Greens winning zero - which I find hard to believe given that the Greens shouldn't need a full quota).

Poll Bludger reports seeing an Australia Institute commissioned ReachTEL that has the Team at 24.8% in the Senate but only 16.1% in the Reps and notes that such a disparity seems unlikely.

Where do NXT votes come from?

Modelling whether NXT can win seats is not just about predicting their vote share and using past evidence of seat-based performance to project the size of the NXT vote.  It is complex because it is also necessary to model how much damage NXT might do to Lower House major party votes and hence whether NXT can get into second.

So, how do we get a handle on which parties NXT might take votes from?

Across the nation but excluding South Australia, the major parties typically polled a few points worse in the Senate than the Reps (as they usually do), while the Greens polled about the same in both (except in Tasmania where their Reps vote was smashed by Andrew Wilkie).  Obvious non-SA exceptions were NSW where the Coalition did 13 points worse in the Senate, and WA where the Coalition did 8 points worse.  Both were caused largely by confusion with the Liberal Democrats who had good ballot draws in both those states.

In SA Labor polled 13.1 points worse in the Senate while the Liberals polled 17 points worse.  The Greens polled 1.2 points worse and Family First polled 1.7 points worse.  SA also had a Liberal Democrat issue, so it may seem that Xenophon was taking votes more or less equally from the major parties and very few from anyone else.

What I found, however, is that there is a reasonably strong correlation between how badly the Labor Party does in the Senate compared to the Reps in SA electorates, and how much the total vote for Others increased in the Senate compared to the Reps (where Others in this case excludes both majors, the Greens, FF, NXT and LDP).  Thus, Labor was not only bleeding Senate votes to Xenophon, but also to a massive list of micros.  This was less so for the Liberals with the exception of the LDP.

On this basis (and independently via regression) I determined that Xenophon probably had more Lower House Liberal voters on board than Lower House ALP voters.  But this isn't because Liberal voters are more attracted to his ideas, it is just because there were more of them.  Supporters of each party were about equally likely to vote NXT in the Senate.

I also found that a strong predictor of the extent to which the major parties would underperform in the Senate compared to the Reps was simply the size of their Reps vote in a given electorate.

Modelling NXT damage to major party vote shares

So in trying to model the Xenophon primary vote in electorates, at a first pass, I assumed it would be proportional to his Senate vote in 2013 (Mayo, Boothby and Sturt highest, Grey lowest).  To model the damage he would do to the other parties, I allowed 24% of his vote to come from non-major sources and allowed the rest to come more or less proportionally from both majors (actually it seems Labor voters were trivially more likely to have backed him in the Senate), with the amount lost by the majors also proportional to the Xenophon state vote share.  On this basis I can then input an estimated NXT state primary and get a crude projection for each seat, that ignores things like sophomore effect, candidate quality, the possibility of Labor running dead in some seats, regional effects of submarine construction debates and so on.  Most importantly, the crude version ignores any possible state swings between Liberal and Labor.  (The current evidence from Bludgertrack is that they're minor, but we'll see if that remains so closer to the election).

To get Xenophon's team in a winning position in most seats, it's necessary to get them above Labor and with the Coalition far enough short of 50% that the mountain can be climbed on preferences.   In some seats, it's the other way around. Provided that the NXT candidate stays in second after Green preferences (which might cause Labor to overtake them in some cases - a big unknown!) the NXT candidate should get almost every Green or Labor preference ahead of the Coalition and might win.  But preference flows of well above 80% might be hard for NXT since many SA seats have quite high Family First votes and these may favour the Coalition, or at least split sufficiently between NXT and the Coalition to dampen the overall preference flow.

Here are some example outputs of the basic projection.  The "Share Req" column gives the percentage of preferences NXT would need ahead of the leading major party in that seat, on the assumption that they reach the final two.  In some cases this share might be low but they might not have a chance of making the top two, in which cases the share required becomes irrelevant.  I can't stress strongly enough that these projections don't really aim to predict results for specific seats (there are all kinds of factors not included in this model, and reasons why it might be inaccurate in given cases) but they aim to give a reasonable overall picture of the conversion of votes to seats.

Because of variations in seat patterns, if I project NXT as ahead of a given party by a small margin off a certain statewide vote, it could be that they really end up slightly behind, and vice versa.

1. NXT statewide vote is 17%:

In this scenario, NXT places second ahead of Labor on primaries in Mayo and (by 0.1%) in Barker.  However the mountain is too steep in Mayo (86.5% of preferences needed) and especially in Barker, where the "rest" includes the Nationals.  On this basis it seems difficult for NXT to win any seat off 17% of the statewide vote, though a very concentrated attack on Mayo or perhaps Sturt, with good candidates, might pull one off.  Boothby is also of some interest because the preference share required there is lower and NXT aren't too far off second.

17% is by no means the lowest possible State vote for NXT; I just use it as a starting point for where they could start being competitive.

2. NXT statewide vote is 20%:

In this scenario, NXT should win Mayo, since the preference share they need is more or less exactly what the Democrats got.  They probably don't win Barker given that the "rest" column includes the Nationals.  They're also competitive for second in Boothby and Sturt and should at least win Boothby on preferences if second there.  So, probably one or two seats.

3. NXT statewide vote is 22%:

This starts to show how dangerous NXT become if they make it into the low 20s in the Reps (bear in mind that in 2013 they polled 24.9% in the Senate).  They are second in four seats in this projection, and while some of those are close and winning Barker is still dubious, there's also the strong possibility of passing the Coalition on preferences in Kingston and winning there.  So, probably three seats, maybe four.

4. NXT statewide vote is 25%:

This is the mayhem scenario.  Off a state Reps vote that is basically what they got in the Senate last time, NXT would be second in six seats and fighting for second in two others (in one of which they could well move into second), with only Adelaide and Wakefield (where the major party votes are too high) and Grey (where NXT has little traction) not in play.  They could win 6-8 seats off a quarter of the primary votes.

Note: The "mayhem scenario" is very sensitive to the balance of major party votes.  Reducing the Labor vote and increasing the Coalition's makes NXT's chances much lower.

Caution Required!

As usual we need to be careful about the more exotic results being suggested in these cases.  At this stage the public polling generally has NXT well below its Senate result from last time in the Reps.  There is also concern about whether the Xenophon Team candidates will poll well when voters realise they are just voting for the Xenophon ticket in the Lower House and not for the man himself.  Xenophon tickets which don't have Xenophon as a candidate don't do nearly as well (as seen in the last SA election where the "Independent Nick Xenophon Team" polled 12.9%.  Maybe this will be different when he is running for the Senate at the same time.

Xenophon notoriously inflicted eight years of Ann Bressington on the SA upper house, and there is good reason to be cautious about his team being a magnet for anti-vaxxers or suffering from other quality control issues, as fourth parties often do.  I suspect his candidate-screening will prove better than PUP's, but that is not saying much.

The issue of Green preferences in cases where NXT fights Labor for second is a tricky one.  My suspicion is that even if Green voters do prefer Labor to NXT, they won't do so strongly enough to allow Labor to catch NXT from more than, say, 2-3% behind, and that this could also be softened by micro-parties preferencing NXT ahead of Labor.

The other possibility is that one major party gets sick of NXT and decides to scuttle them by either putting them behind the Liberals, or issuing an open ticket in seats that major party cannot win.  If Labor for some reason does this then most of the possible NXT wins go completely off the table, requiring NXT to poll very well indeed to maybe take a seat or two on Liberal preferences.  However it makes more strategic sense for majors to preference NXT and force each other to waste resources fighting for otherwise safe seats.  This only changes if NXT looks to be winning a given seat easily.

This model is all highly experimental and feedback on its doubtless many flaws is welcome.  But until something better comes along, it shows the sort of approach I'm going to be using to converting polled NXT vote shares roughly into expected seats.  At the moment the available neutral polling is not showing NXT in a position where it would be all that likely to win Lower House seats.  However, that could quickly change.

I'll add links to other attempts as I come across them.  For starters, Henry Schlechta presents the case for zero Xenophon MHRs, including that dastardly Liberals might even vote 1 Labor to try to eliminate NXT!

(This piece was updated on 28 April to add a few more commissioned poll results.)

Note added 1 June: See more updated coverage of the NXT's prospects here.

Note added 10 June: A rather remarkable seat poll of Grey has NXT winning that seat, which is amusing given that Grey is the worst seat for NXT in the above projection!  Of course, the projection does not consider local issues that may have increased NXT's appeal to voters since the 2013 election, so we should keep in mind that these projections are only all-else-being-equal and that in reality NXT might win a seat anywhere in SA.  Seat polls are quite unreliable so this extraordinary result should be treated with caution.

Note added 16 June: Some more Xenophonia over here, based on a detailed Morgan release.

Note added 29 June: Yet more added here.  NXT's own internal polling is interesting when applied to the model because the model is not so keen on its chances even from a primary vote of 24%, if the balance between the major parties is altered slightly in favour of the Liberals.


  1. Have we ever had an independent Senator get someone into the Reps before? It's been hard enough for established minor parties like the Greens and the Democrats to get a single seat in the Reps let alone an independent.
    The other interesting thought is what happens if the NXT puppy catches the balance of power truck?

  2. Prior to Xenophon there had only really been five independent Senators elected as such (not counting Senators who left their parties mid-term and were defeated at the following election). None of these attempted this kind of franchise expansion so there is really no federal precedent for what NXT are trying here.

    It isn't clear yet on what basis NXT would make a balance of power decision in the event of holding seats in a Lower House hung parliament, since the party has no natural leaning to either side. If the election looks close and the NXT are threatening to win seats I would not be surprised to see pressure placed on them to declare their hand, or else attacks being made on them claiming that a vote for them is a vote for chaos.

  3. Interesting work as per usual, Kevin. Our household will vote for the Xenophon candidate in Sturt, simply because it is the best option to get rid of Pyne, but we will all vote for Sarah Hanson Young in the Senate. The Greens should get a quota, at least. A lot of voters do not like Pyne even though he is from the moderate faction of the Liberal Party.

  4. In modelling the Xenophon vote shares, I'm not sure I follow the basis for the assumption that his ticket would draw a quarter of its votes from what would otherwise be minor party votes. Unless that's a very rough guess in lieu of being unknowable, is there a basis for choosing that figure? It looks like you're drawing on some historic basis to then go on to assume that the major parties then contribute proportionately, but I'm not sure if there's a basis for assuming first that between them they supply three quarters of NXT votes (regardless whether NXT polls 17%, 22% etc)?

  5. The basis for the estimate of 24% from minor party/independent voters is that that was what was left over after taking out the amount that appeared (based on regressions) to be taken from the two major parties combined. I tried to find a way to determine which minor parties and independents specifically it might be coming from, but because their vote shares are so small there was much too much noise.

    Of course it could be that if NXT's vote varies from the Senate figure last time then the mix of parties he draws votes from changes. That applies especially if the difference between his Senate vote and, say, a House vote of 17%, came very heavily from major party voters who were happy to back him in the Senate but not in the Lower House.

  6. Interesting work as always, particularly to an economic refugee from South Australia. Nick has blinked twice in the past two weeks: the Libs got him to effectively announced he'd run open tickets and Labor got him spooked on penalty rates. He can be damaged. I find it odd that the only lower house specific stunt he's pulled so far that got any attention was in Boothby when there's a bigger anti-Liberal vibe to be tapped into in Sturt and Mayo. I suspect at the end of the day he'll do well in the Senate because of his name on the paper and fall short in the Reps because he simply won't have people on the booths with his HTVs. He has also devalued his currency terribly by running candidates in other states. How could candidates from them put SA first if elected? The majors have barely begun to open up that front.

  7. Don't believe everything you read in the Murdoch press - Damian Carey, the NXT candidate for Kingston, is not an anti-vaxxer but his position (he is certainly a practitioner of alternative medicine) gives the Oz enough wriggle room to get away with saying he is. He has written a detailed reply to this accusation on his own website.

  8. I regularly refute things appearing in the Murdoch press, as with the Fairfax press, the Guardian, Crikey, the ABC, commercial TV (etc) so it is hardly the case that I would believe anything on account of it appearing in a particular outlet. That is not, however, to say that I would never be too credulous of a media report from some outlet or other.

    Anyway, Mr Carey's reply is here:

    While the reply states Carey's own views and on that basis disputes that he is himself an anti-vaxxer, and also claims that the article was "completely untrue", there are some conspicuous omissions. Especially, he does not challenge the article's claim that he "posted an article on his Facebook page that contains a series of extreme views about vaccination." There was a particularly extreme article attacking Bandt on vaccination doing the rounds on social media at the time, from a blog called Septic Skeptics. Nor does he dispute any of the quotes attributed to him - he just says his actual views are not anti-vaccination.

    If Mr Carey did indeed act as an enabler of anti-vax sentiment by sharing or linking that anti-Bandt piece and also by giving the quotes in the article, then he gave the Oz all the wriggle room it could have possibly needed by himself.

    Also while his reply mostly seems to get him off the hook as concerns his own views, it does throw the anti-vaxxers quite a lot of bones along the way (eg by defending the expression of their concerns as natural mothering instincts and even potentially valuable to the scientific process).

    Anyway Carey's views on this and other issues are largely beside the point that candidates with eccentric views (whatever they are) could be a liability for NXT. That said I am quite happy to have Mr Carey's views and those of other candidates discussed here and for any apparent errors, however incidental, to be challenged at any time.

    1. I'm pro choice. Vaccination is a choice. Therefore I'm voting for NXT regardless what the media says about NXT.

  9. Kevin,
    Many thanks for this analysis, as I was about to ask for your thoughts on the topic. The comments also have added some useful context. I think he is likely to get a significant proportion of votes which have previously gone to minors as an anti-two big parties (and even do some damage to the Greens on that account).
    Following Christian Kerr's point, does he have resources to staff polling booths, and of increasing importance pre-polls? I was in Adelaide a week prior to the 2013 election, and was astonished to see the number of street signs for Xenophon. It suggested that he has some people to do the ground work, and I was at a loss to think where such motivated people might come from.

  10. I had a quick look at 2013 results and it seems that he did OK at within-electorate prepolls (these appear as ordinary votes but the booths are labelled as PPVCs) - more or less comparably to the on-day vote at equivalent booths. There is no way of knowing from these results whether that means his booth coverage is equally good or equally bad at both, but I suspect it means both are well covered; SA residents may be able to add more info on this.

    However, he did remarkably poorly on out of electorate prepolls (16.6% cf total 24.9%), which I suspect reflects that his voter base is not especially mobile/transient compared to Green and Liberal voters.


The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.