Monday, March 23, 2015

New South Wales: The Final Week

NSW Primary Aggregate (updated 27 Mar) Coalition 45.4 Labor 34.0 Green 10.7 Others 9.9
2PP By 2011 Preferences: 55.8% to Coalition
Estimated 2015 Preferences: 54.0% to Coalition
Seat Projection (estimate): Coalition 53 Labor 36 Green 1 Ind 3

It's time to kick off my rolling post for the last week of the NSW election, which will be updated an unknown number of times through the week.  Before posting too much detail and getting into some of the finer modelling detail - I've posted provisional figures above, but these will be revised through the week - I want to get something up about the big picture of where the polling is at.

From their position last week I thought there were three things Labor needed to do all of to still win the NSW election:

1. Reduce the primary vote gap to the Coalition compared to what polling was showing at that time.
2. Increase their preference flow from the Greens and minor parties compared to what polling was showing at that time.
3. Get lucky on the distribution of seat swings.

There is some evidence that 2 may be occurring.  There is still some evidence that 3 is unusually likely.  But as concerns 1, the four polls out in the last week don't provide cause for hope; indeed they suggest the gap is widening.  No amount of preferencing or uneven seat swing magic will win it if the primary vote gap is too big.  And so, based on current polls, Labor starts the final week in a losing position and needs a late swing back to have a realistic chance.  The extent to which dramatic swings occur in most election campaigns is overstated and Labor's task if they want to win is very challenging now.

Why is it so?  I reckon it is mainly Baird.  The voters, mostly, see him as a human being, and nothing said against him seems to stick.

It's interesting to observe Labor trying to counteract the Baird factor by telling voters that if they ditch Baird on Saturday, Tony Abbott will be gone on Monday.  This line might be good for motivating the faithful, but I don't think it plays well in the wider electorate.  A voter who likes Baird but dislikes Abbott (a very common species) might receive this sort of line from Labor as exactly the sort of whatever-it-takes mentality that got the party in such a mess in the first place.

Galaxy and Ipsos

The fifth NSW Galaxy of the year came out with primaries of 45-36 with 10 for the Greens and 9 for Others.  Galaxy has again been a remarkably steady poll throughout the campaign.  Successive Galaxys have had the primary vote gap between the majors at 8, 9, 7, 8 and now again 9 points.  The released 2PP was 54:46 by 2011 preferences.

The Ipsos, the second of the year, had a massive 15-point primary gap, 47-32, with 13 for the Greens and 8 for Others.  These represented the highest primary gap, highest Coalition primary, lowest Labor primary and highest Green primary of any poll all year.  The previous highest primary gap was 12 points in the February Ipsos, which also had the previous highest Green vote. The last-election preferences were 58:42 to the Coalition, and respondent-allocated, 54-46.

To some extent these results reflect apparent leanings of this pollster, since their polls in Victoria and federally have also tended to show stronger results for the Coalition and in Victoria's case Greens, and weaker results for Labor, than other pollsters.  But even if the poll is adjusted by moving, say, 1.5 points from one side to the other, the result is still bad for the ALP.

Someone might say that if we assume this Ipsos is a bit wobbly on the primaries, and remember that Ipsos skews a bit, and expect its respondent preferences to flow at full value, we could drag the 54 down to something sub-52 and get Labor over the line.  However, that doesn't work.  One reason the preferences in this Ipsos do bring a last-election 58:42 down to a respondent-allocated 54:46 is that this Ipsos has such a high Green vote as a share of all third-party votes, and this is something no other poll has had, and that I'm very confident won't happen on the day.  If you want to adjust Ipsos for perceived pro-Coalition house effects you have to adjust it for apparent pro-Green house effects too, and at that point the preferences aren't so Labor-friendly anymore.

That's one of the causes of the massive gap between Ipsos's 58:42 by last-election preferences and their respondent-allocated 54:46, and another one is the way Ipsos do their last-election preferences.  In common with several other pollsters, they use batched preferencing such that the composition of the third-party votes is treated as being the same as in 2011, even when it isn't.  So although the poll has primaries of 13 for the Greens and 8 for Others, the 2011 preferences treat this as if it was about 9 to the Greens and 12 to Others.  That difference is usually no big deal, but in this case it's worth about 0.8 of a point. The last-election flow without batching off these primaries is more like 57:43.

All up my aggregate for NSW starts the week with a 10.2 point lead to the Coalition (45.3-35.1).  The preference distribution implied by last week's respondent-allocated preferences in ReachTEL points to a 2PP of about 53.7 off those primaries (55.1 by 2011 preferences).  There's a case that I should allow for a difference between respondent preferences and (my) 2011 preference estimates that is more than the 1.4 point gap suggested by ReachTEL, especially since Ipsos implies a gap of about 2.5 once its methods and the Green vote issues are accounted for.  There's also the usual case that respondent preferences overstate things.  The 2PP off the primaries I have could reasonably be argued down into the low 53s but probably not that much lower.  Anyway, it makes little difference: on current primaries the Coalition wins, and changing the 2PP by half a point or so while it is in that sort of range only makes about one seat difference to the expected outcome.

The projection at the top of the post arbitrarily allocates one Coalition seat to an Independent.  I believe this is a reasonable median reading of the strength of the anti-CSG rural backlash but I would like to see voting-intention polling of Tamworth especially to be more confident of that.

I'm intending to repost my seat model later this week, if time permits and after giving it a bit more panel-beating.  The main thing I want to do with it is vary the seat swing slightly based on a seat's Green vote, to try to get it to smell some of the coffee in seats like Lismore.  But in any case whatever changes are made they tend to only affect the overall projected outcome, for a given 2PP, by about a seat or so, and there are so many seats we have no polling data for.

Ipsos predictably has Baird with a still massive +38 netsat (60-22)  and a 56:27 preferred Premier lead.  Luke Foley scores a reasonable +5 (37-32). Partial privatisation of the electricity network continues to get the thumbs down (31-62) but if voters are primed with the information that the money will be put to its intended use that turns around to an even 48-47.

Seat Polls

ReachTEL has issued three seat polls which were published in the Sun-Herald (the Sunday edition of the SMH).

Newtown is a new notionally Green seat.  However, it is notionally Green in part because Labor's primary vote performance in 2011 was so bad, and it does not have a Green incumbent.  ReachTEL shows Labor leading the Greens 56.5:43.5 (respondent-allocated) off primaries of 37.7:33.3 with 19.3% for the Liberals.  Apparently the vast majority of Liberal respondents say they will exhaust their preferences.  As the case that the Greens would win this seat based on its notional status was dubious anyway I am now treating it as a Labor seat in my projections pending further evidence.  (I do think the margin flatters Labor a little bit though; it is unlikely only 7.6% of preferences will actually go to the Greens.)

In Ballina the ReachTEL has the big swing on with Labor winning 52.2:47.8 off primaries of 29.7% (ALP), 39.2% (Nationals) and 19.8% (Greens).  Depending on the rate of exhaust, this would probably require Labor to get 80% of non-exhausting preferences, but since most of those preferences are Green that seems quite plausible.  However the margin is small for a seat poll and this is not conclusive evidence that Labor will win the seat; it just shows that the seat is in play (which itself is amazing enough).

In Strathfield, Labor is only just over the line 50.8:49.2 off primaries of 45 Liberal, 41.1 Labor and 9.8 Green.  For a seat on a feeble 6.4% that's not too flash, but on the other hand the seat is harder than it looks because the incumbent has a double sophomore effect on his side.

There are also three Galaxy seat polls showing the Coalition up 52:48 in Coogee, 51:49 in Campbelltown and 51:49 in The Entrance.  I haven't seen the primaries yet but if these are using 2011 preferences then these slender leads are meaningless and lineball in all three is a better reading.  It should be noted here that the Coalition has a double sophomore advantage in Coogee (8.3%) and effectively single sophomore in Campbelltown (6.8%) and The Entrance (11.5%).  I've also pointed out that the loss of a troubled first-term incumbent, as in The Entrance, doesn't necessarily spell disaster in the seat.  Labor should be mildly disappointed not to have Campbelltown on toast; that aside, nothing surprising here.

Betting Markets

Once again the markets are looking so much like the modelling that I believe they mainly reflect the views of the modellers working for the bookies rather than the views of the punters.  Labor has blown out to 9.50 at Sportsbet and 15.00 at William Hill, which is a pretty fair reflection of the difficulty of their task.  William Hill has a median of 52 Coalition seats with a slight skew towards higher figures in that distribution.

William Hill has the Coalition favoured in Gosford, which is still a tossup on Sportsbet; aside from that I don't think there are any differences in which party is favoured in each seat covered by both.  I also can't see any changes in which party is favourite in Sportsbet seat betting from last week although the Nationals' edge in Ballina is now seen as trivial (1.91 v 2.00).  The Sportsbet vote total odds point to about 44.9 for the Coalition (including 9.5 for the Nationals), 35.4 for Labor, 10.5 for the Greens and a 2PP around 53.4.

Comments will be posted through the week as new polls emerge or when I have more detailed modelling to add.  There may or may not be an election-day post; there will be coverage on election night.

Morgan (24 March)

Another Morgan SMS poll delivers another nasty result for Labor with the Coalition on a 56% 2PP (last-election presumably) and the Coalition with a 13-point lead, 45.5 to 32.5. This is Labor's worst result of the campaign from anyone other than Ipsos.  Morgan SMS is not very highly weighted in my aggregate but even so I've flicked the Coalition another 0.2 points two-party-preferred and another projected seat.  Perhaps 53 seats to 36 is a little on the generous side given the far north coast uncertainty but I haven't had time to adjust my model to take Green votes by seat into account yet.

On the matter of internal polls, Labor today claimed it had left-wing union independent Arthur Rorris under control in Wollongong.

25 March: My time is limited today for work reasons so I refer readers to Poll Bludger.

Lonergan, Essential: Today has seen a couple more polls out with a second Lonergan recording a thumping 57:43 off primaries of 47 to the Coalition, 31 to Labor and 11 each to Greens and Others.  The previous Lonergan was slightly bullish for the Coalition compared to other polls out at the time and this one with a sample size of 3,251 is even more so (indeed on primary votes it has the largest gap of any poll so far).  The poll was conducted from March 20-22. Not much is known about the properties of Lonergan robopolls when applied across a full state like this so it remains to be seen whether this result really represents a blowout in the Coalition's favour and/or house effects similar to those apparent in Ipsos.  Lonergan again found Mike Baird with a hefty lead as preferred premier, 52-25.  Results on various issue questions can be found at the Guardian.

Essential's result of 53-47 off primaries of 44-36 with 9 for the Greens and 11 for Others was quite benign by comparison.  The Essential poll was taken from March 13 to 23 with a smallish but usable sample size of 659.  It does therefore miss much of the recent apparent upsurge in Coalition support.

Essential's issue questions appear more damning for Labor, with the ALP leading the Coalition in only its strongholds of health and education, and these only by 4 and 3 points respectively.  However on close inspection this is partly caused by the inclusion of the Greens in the issues poll. It is very likely many of those preferring the Greens' approach to both majors in pretty much any policy area would also prefer Labor's to the Coalition, which would give Labor leads of several points on health and education, a small lead on environment and a tossup on unemployment and jobs.  Voters are overall split (33-33) on which party they would be better off under.

Both polls have adopted their own solutions to the issue of how to predict preference flows at this election.  Lonergan are using 2007 preferences while Essential are using flows from "previous elections" (according to Bernard Keane in Crikey, "several" of them).

Adding them to my aggregate does nothing, but only because I am tentatively assuming a one-point house effect in the Lonergans.  Should established polls like Newspoll, Galaxy and ReachTEL come out with double-digit primary gaps I will remove it (in which case add another 0.1 to Coalition).

Heathcote CSG poll (26 March): Most of a slew of commissioned polls on coal-seam-gas and mining issues (that invariably show almost everyone in an electorate claiming concern about said issues) have been bereft of voting intention results but I've finally found one that has something: in Heathcote, a commissioned CSG poll found results of 41.4% Liberal, 31.7% Labor, 9.3% Green.  It's possible those primaries include undecided; if they don't the 2PP comes out in the 54-55 area for the incumbent depending on preferencing assumptions.

Galaxy Says Game Over (27 March)

A Galaxy of 1300 voters statewide taken over the last few nights has snuffed out whatever remotely realistic hopes Labor might have been clinging to, with its highest Coalition primary vote lead of the year, 11 points (45-34).  While hardly debunking the idea that Lonergan's 16-point lead had a bit of house effect on board, this result strongly supports the appearance that the Labor campaign is flopping as the finish line approaches.  The poll had the usual preferred-premier lead for Mike Baird over Luke Foley (53-25, noting again that incumbents tend to have large advantages in such polling), but the party breakdown was especially illuminating: Labor supporters preferred their man only 53:26 while Coalition supporters preferred Baird almost unanimously (88:3).   (I infer that Green and Others voters combined prefer Foley about 27 to 22 with the other half of them indifferent).

The Daily Telegraph (not averse to a scare or three itself) probably correctly blames "Labor's scare campaign, which tried to frighten people with outrageous claims about Chinese interests leasing part of the state's electricity network" for the result, and reports that the buoyant Liberals are trolling Foley in his own seat (I suspect to little effect other than random jollies, though Foley's seat has a substantial Chinese population).  They also have Labor's candidate for the theoretically marginal Seven Hills in yet more campaign trouble.

There has been some suggestion that federal Labor's compliant form on data retention might be causing some inner-city vote loss to the Greens making the latter's position in Newtown more competitive than this week's polling showed.  We'll see.

The Australian (paywalled, title "Private polls indicate rural swing to ALP") reports that Labor based on internal polling is claiming it will win Ballina, is competitive in the other northern rivers seats, and that there will be large but inadequate swings in Bathurst and Barwon, while Monaro remains tight despite its small margin.  The track record of these kinds of internal-poll claims before elections is far from great but I note them to see how they go.

Morgan and ReachTEL (27 March)

It's getting like an auction to see who can get the biggest primary gap among the less established pollsters!  Morgan has released an SMS poll taken last night with a 20-point gap between the majors, with the Coalition on 49, Labor on 29, the Greens 12.5, the Christian Democrats 3 and others 6.5.  Morgan gives the 2PP for this as 57.5; I get 60-40 by last election preferences.  This result is possibly rogue; while there is clear evidence of a late move to the Coalition it challenges belief that it could be anything like this much.  Morgan SMS polls were very volatile in the Victorian election and this may be the case again here.

ReachTEL meanwhile has put out a result that lands right in the middle of the fairway so far as my aggregate is concerned: 54:46 to Coalition (respondent-allocated) off primaries of 45.8 Coalition, 33.8 Labor, 10.5 Green, 10.2 Other.  ReachTEL's detailed surveying of preference intentions suggests that preferences have moved Labor's way to a degree worth about half a point compared to their previous survey.  Mike Baird gets a Good or Very Good rating from 48.8% of voters (up 7) and the thumbs down from 23.3% (up 4.3); by ReachTEL standards his net rating of +25.5 is exceptional.  Luke Foley has lost 0.4 points from the good side (23.2%) but the real concern is that the seven point decrease in voters who don't know him has gone directly to the negative side, as has some of his previous Satisfactory rating.  His Poor/Very Poor rating is up from 23.8 to 35.2, for a net rating of -12 (still not that bad for a new leader on this scale.)

Voters do say the scare campaign about Chinese leasing has made them less likely to support asset leasing, but they don't say whether this impacts on their vote.  The proportion of voters most concerned about asset leasing is up from 8.8 to 13.4% while the proportion most concerned about health is down five points.

Most likely we await just Newspoll of which first figures won't be far away.  I'm still working on some modelling to put up late tonight but this is looking like an underwhelming recovery for the ALP.  It may be that not a lot of seats above an 8% margin fall, that the swing falls mainly in "safe" rural seats jagging the odd one but being mostly wasted, and that even some seats below 8% stick with the Coalition.  Thus, my current projection of 36 Labor seats may even be generous.

And Finally ... Newspoll (27 March)

Newspoll has come out with a 55:45 by last-election preferences (52:48 by respondent preferences) off primaries of 44 Coalition, 34 Labor, 11 Green and 11 Other.  The poll was taken over four days (Mar 23-26) and hence does not attract the same weighting in my aggregate as the Galaxy and ReachTEL which are more recent.

Based on the Newspoll primaries and 2011 preferences the swing in Sydney is 5% and that outside Sydney is 16%.  This is much more polarised than Ipsos which had the swings at 4% and 10% respectively.  Both, however, are consistent with the swing to Labor being potentially partly wasted in unwinnable seats.  I tried a variant of my seat model in which I altered the swing according to these results in all non-Sydney seats, but what is likely actually going on is a very strong swing in some non-Sydney areas.  (The adjustments to my model on a uniform basis had slightly radical results, giving the Coalition Londonderry, Strathfield and Campbelltown and giving Labor Kiama, The Entrance, Port Stephens and Gosford.)  Anyway the most important point is that the predicted seat totals didn't move.

Also, even if I used the Newspoll's respondent preferences fully and its Sydney-vs-non-Sydney divide, I still get 51 Coalition seats.

Newspoll has Baird with a still-huge +28 netsat and Foley with +1, and Foley with one of his better Better Premier readings (he trails 27-54).

Anyway I have tried all kinds of fiddles to my seat model and keep ending up in more or less the same place: 53-54 Coalition seats, 35-36 Labor, and 4 Others - though there are always some odds for individual seats that I don't trust.

While the Greens' hold on their inner-city seats is quite precarious and the case for them winning Lismore still quite speculative, this may for left-wing voters be becoming the kind of "meh!" election in which the party tends to do reasonably well.

I think it is still very unclear where the 2PP will land for a given set of primaries this election, and even if my aggregate has the primaries about right I wouldn't be greatly surprised to see the 2PP up to at least a point out.  However, in seat total terms, it doesn't make a lot of difference.

Live comments here from when I catch the first sniff of an exit poll, or 6 pm, whichever is earlier.

Final bids

Lastly, here is a table of what I think are the final polls for each pollster (unless Morgan spits out yet another one tomorrow morning!).

If you want to compare to Victoria and Queensland check those tables here:



No comments:

Post a Comment