The Voice

A referendum to change the Constitution is always an up-hill battle, requiring the support of not only the majority of voters but also the majority of voters in the majority of states. Only 8 of our 44 Referenda have been successful—18%. It’s also an uncommon event—the last referendum was in 1999, meaning this is the first time voting in one for anyone under the age of 42.

This means that there is VERY little data to draw from, but we do know that on balance the NO campaign has the advantage.

For a referendum question to succeed, both the majority of voters and the majority of voters in the majority of states are required.

Polling

Superficially, polling indicates that the NO vote will be in the majority, which would prevent constitutional change. That is widely the consensus, and no poll has placed the YES vote ahead since July.

For the following analysis, polls are dated by the last day of polling. Polls without a published end date are not included. Polls prior to the last federal election (21 May 2022) and announcement that a Voice would be pursued by the Albanese government are also excluded.

Polling results since 21 May 2022

There has been a noticeable increase in support for the NO campaign over time, drawing from both the undecideds and YES voters. What started at less than 20% of the population now consistently commands over 50% of the vote, meaning that on these numbers even if all undecideds break for YES, the NO campaign would succeed.

We can break this down by polling house. For this analysis I have excluded sources who (1) have not released a poll in the last month, as this is difficult to extrapolate to the current results and (2) have published less than three polls on the referendum since 21 May 2022. This leaves seven firms: Essential (7 polls), Freshwater Strategy (5 polls), JWS Research (4 polls), Newspoll (10 polls), Resolve Strategic (9 polls), Roy Morgan (6 polls) and YouGov (3 polls).

 

What we see, interestingly, is that there is not a huge distinction between the polling houses: all methods seem to converge on similarly clustered results and trends. These, by their latest polls put the YES campaign at between 33% and 44% and NO between 49% and 57%. Undecideds are anywhere between 5% (Roy Morgan) and 17% (Freshwater Strategy) of the vote.

To succeed just for the vote of the total population, YES would need to take the most favourable of these offerings and win virtually 100% of the undecided vote. In reality, the existing polling shows if anything that undecideds tend to break in favour of the NO campaign (possibly as a result of the “Don’t know? Vote NO!” slogan).

If we compare the YES share of divided votes (that is, YES/(YES+NO) x 100%) to the undecideds, we find this sharply illustrated.

Trend line: Undecided = 0.3868 x Yes share - 0.0521

Extrapolating the trend line to 0% undecided, this would have the YES vote at around 13.5% of the decided votes (i.e. a NO vote of 86.5%). This seems unrealistically high, but emphasises the difficulty that YES has of securing the undecided vote.

The reason this seems unrealistic is likely because it also factors in a significant decay in the YES vote over time. It assumes that if we allowed the undecideds to resolve the YES decline would also continue. But polling day requires the undecided vote to resolve immediately which does not allow for the continued erosion of the YES vote. (Also, presumably there is a floor below which the YES vote will not drop, but that is only relevant for a gradual resolution of undecideds).

Returning to the first graph in this post, we see that the drop in YES and undecided votes is roughly the same: a rate of 0.03 percentage points per day for undecideds and 0.04 for YES. At the extremes of possibility this could reflect 0.03 pp/day from undecideds and 0.04 pp/day form YES to NO, or 0.03 pp/day from undecideds to YES and 0.07 pp/day from YES to NO. The reality is probably a break somewhere between these, with some transfer form undecided to both camps, and some YES moving to NO. Assuming NO is winning over a larger share of undecideds just as they are winning over more YES, and preferring rounder numbers, we take a wild guess that:

0.01 pp/day undecideds move to YES.

0.02 pp/day undecideds move to NO.

0.05 pp/day YES mover to NO.

This is a 2:1 split of undecideds in favour of NO based on basically nothing, but would give the NO vote something like 57% of the vote--towards the upper end of polling.

PREDICTION: NO wins the popular vote.

States

We have state-by-state polling, and we could make similar unfounded assumptions about the undecideds, but the key question is which state falls which way.

New South Wales polling has favoured NO in 7 of the 8 polls in the last month

Victoria has favoured NO in 6 of the 8 polls in the last month

Queensland has favoured NO in 9 of the 9 polls in the last month

Western Australia has favoured NO in 5 of the 5 polls in the last month

South Australia has favoured NO in 5 of the 5 polls in the last month

Tasmania has favoured NO in 2 of the 2 polls in the last month

On this rating, the states ordered from most to least likely to vote YES are:

TAS         Vic          NSW      WA/SA Qld

 

Two polls concluded on October 12. For the Newspoll, the states ordered from strongest to weakest support to YES are:

Vic          NSW      Tas          SA           Qld         WA

For Roy Morgan, the states ordered from strongest to weakest support to YES are:

Vic          NSW      Tas          WA         SA           Qld

 

Compiling this data, the most winnable states for YES look to be Victoria, New South Wales and Tasmania, with Queensland the hardest to win. YES will need all three plus either SA or WA.

PREDICTION: The states will resolve in the following order of support:

Vic:         YES (Strongest YES vote)

NSW:     NO

Tas:        NO

SA:         NO

WA:       NO

Qld:       NO (Strongest NO vote)

Summary of Predictions

The YES campaign will fail decisively on both the popular and state counts.

The NO vote nationally will sit at around 57%.

Victoria will be the strongest YES voting state, and Qld the least.

NSW Legislative Council

The NSW Legislative Council has 21 seats up for election through single electorate proportional representation. That means that there are 21 other seats that are carrying over:

➎ Liberal
➌ National
➐ Labor
➋ Green
➋ One Nation
➊ Shooters, Farmers and Fishers
➊ Animal Justice.

◼◼◼◼◼◼◼◼◼◼◼⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚⬚◼◼◼◼◼◼◼◼◼◼

There is no direct polling of Legislative Council voting intention, but assuming this mirrors the Legislative Assembly, The primary vote would elect:

7.98 Coalition candidates (38% of 21 seats)

7.98 Labor candidates (38% of 21 seats)

1.68 Greens candidates (8% of 21 seats)

It is very likely this will convert after preferences to at least 8 seats each for the Coalition and Labor:

◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼⬚⬚⬚⬚◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼

With four unknowns, and a long push for Labor or Coalition to gain a 9th seat, it seems likely that minor parties will fill the remainder. SFF have not polled over 2% (0.42 LegCo seats) since the start of 2022, but a recent(ish) Roy Morgan poll puts One Nation at 8.5% (1.79 seats). With a remainder of  0.68 seats, it's in the realm of possibility for the greens to pick up another, as may One Nation:

◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼⬚◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼

The final is anyone's guess, depending of what minor party can scrape enough preferences together to float to the top of the pack, which is entirely speculative even without NSW's option to not number every candidate. For colour-thematic reasons I'll propose Independent, but really this is a total tossup and more likely to be a party like Animal Justice, SFF or Legalise Canabis.

◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼◼

There's no motivation to form a permanent coalition in the Legislative Council, but assuming the Coalition has minority control of the Legislative Assembly (see previous post), they will need the support of six crossbench councilors to pass controversial legislation. This is unlikely to be the Greens for ideological reasons, requiring support from all ONP and two of the remaining three crossbenchers, adding to the complexity of the next government. Conversely, to block Coalition bills or support their own legislation a Labor-Greens alliance would only require 2 or 3 crossbenchers respectively.

Hanging out in NSW

The NSW election is coming up this weekend, and the polling superficially appears to be a concern for the Coalition. The last poll with a two part preferred (TPP) vote where the Coalition led Labor was in 2021. The reality is a somewhat closer race that this might suggest, however.

In the 2019 election the Coalition won 52.0% of the TPP vote. on 28 February Roy Morgan had this down to 47.5% - a drop of 4.5 percentage points. Taking Antony Green's electoral pendulum and applying a 4.5 percentage point swing to Labor (ignoring all minor parties) takes the Coalition's legislative assembly seats from 51 to 46 or, to put it another wat, gives Labor a majority of 1 seat.

* Incumbent is nominal only - new seat following 2021 redistribution of seats
† Alternate margin by Antony Green (where pendulum did not reflect TPP contest between Coalition and Labor) 
‡ 2019 TPP result used (where pendulum did not reflect TPP contest between Coalition and Labor, and no alternative was provided by Antony Green)

However, this obviously ignores the nuances in a parliament that currently has three Greens, three Shooters, Farmers and/or Fishers and five Independents.

Taking instead the expected races in each seat, we can retain the above results for Liberal v Labor and Nationals v Labor races, but factor in the following alternative swings based on primary vote:

Total represents the sum of compared primary votes. The % of Total is the primary vote as a proportion of this Total. Primary vote swing is the difference in percentage points between the % of Total from the 2019 election and latest 2023 polling. 

Thus while the Coalition is broadly expected to lose seats to Labor, in some instances they may pick up seats from the Greens or, more likely, Shooters, Fishers and Farmers to offset this.

Races involving an independent (where the same independent who ran in 2019 is running again in 2023) are assessed on a case by case basis as follows:

Coffs Harbour - Nationals held a 10.3 percentage point lead over Sally Townley in 2019, which is normally classified as 'safe'. It is not expected that Sally Townley will be able to surmount this, even with the general 4.5pp swing away from the Coalition potentially flowing to her. Purely for the calculations, we will use a swing of 0.0%--this is not intended to accurately represent a possible swing but simply to return a result of no change to the seat.

Lake Macquarie - Independent Greg Piper has a 23.2pp margin over Labor. This is considered 'very safe'. Again a nominal 0.0% swing is adopted to represent no expected change.

Sydney - A 'safe' margin of 11.8pp has been calculated by Antony Green for incumbent Alex Greenwich. The TPP race is very close, creating a very favourable situation for the independent to retain the seat. Nominal swing of 0.0%

Tamworth - Nationals won in 2019 with a 20.7pp 'very safe' margin over Independent Mark Rodda. No change anticipated, nominal swing 0.0%.

Wagga Wagga - Incumbent Joe McGirr has a 'safe' 15.5pp margin. No change anticipated, nominal swing 0.0%.

Willoughby - In the 2022 by-election, Liberals has a 3.3pp 'marginal' margin over Independent Larissa Penn. With the drop in Coalition primary vote, it is anticipated that the Independent who made it to second place once might win this. To reflect this a swing of 3.4pp is nominally adopted to reflect this possible change.

Woolondilly - On the border of 'marginal' and 'fairly safe' at 6.0%, the Liberal lead is technically larger than the swing to Labor which may insulate them from change, but with a viable Independent candidate, this could see a stronger swing. Nominal 6.1% swing adopted to reflect a possible change.

* Incumbent is nominal only - new seat following 2021 redistribution of seats
† Alternate margin by Antony Green (where pendulum did not reflect TPP contest between Coalition and Labor) 

‡ 2019 TPP result used (where pendulum did not reflect TPP contest between Coalition and Labor, and no alternative was provided by Antony Green)
! Antony Green assessment of likely contest adopted
!! Incumbent updated to reflect likely contest (i.e. where Coalition candidate won in 2019, became independent, and it is expected that the Coalition will win the seat again rather than be held by the independent defector)

This would give a hung parliament result of:

Coalition 43 (30 Liberal, 13 National)
Labor 41
Greens 3
Shooters, Fishers and Farmers 1
Independent 5

Assuming SFF supports the Coalition and the Greens support the ALP this is 44 seats apiece, with support from 3 out of 5 independents requited for majority. Given four of these (Lake Macquarie as the exception) are in contest with Coalition parties there is likely going to be pressure within their electorates, and from the voters who prioritised them over Labor, for the independents to align with the Coalition. If Willoughby and/or Woolondilly don't flip this further strengthens the Coalition's position to form minority government.

The ultimate result in the Legislative Assembly, therefore, will come down to the will of a handful of Independents. I'm tentatively suggesting a Coalition minority government against the simplified TPP polling which is tipping ALP to win. This would be touted as a significant achievement for Perrottet, likely to be framed by supporters as snatching victory from the jaws of defeat, but the practical implications of such a broad crossbench means a tough 4 years ahead for whoever ultimately forms government.