Sunday 18 August 2024

Northern Territory 2024 - Legislative Assembly

It’s the election for the Northern Territory's only house of government—the Legislative Assembly—this weekend. Normally I’d wait closer to Saturday so I can take advantage of the absolute latest polling available. However, in the Northern Territory there seems to be a collective agreement that we don’t need a whole lot of polling:

This doesn’t give us a whole lot to work with, so what can we do to predict this election?

Powering On

One option is, of course, to use the little data we have. The latest poll available has the Country Liberal Party leading Labor 54:46 on a two party preferred (2PP) basis. This would be a swing of -7.3 percentage points to Labor from the 2020 election result. Applied across the pendulum that would see 4 seats change hands:

SEAT INCUMBENT ALP LEAD
(PRE-ELECTION)
-7.3 PP
SWING
Arafura  ALP 3.6 -3.7
Araluen  IND N/A N/A
Arnhem  ALP 15.9 8.6
Barkly  CLP -0.1 -7.4
Blain  IND N/A N/A
Braitling  CLP -1.3 -8.6
Brennan  CLP -3.0 -10.3
Casuarina  ALP 16.0 8.7
Daly  ALP 6.1 -1.2
Drysdale  ALP 5.4 -1.9
Fannie Bay  ALP 10.9 3.6
Fong Lim  ALP 2.2 -5.1
Goyder  IND N/A N/A
Gwoja  ALP 16.2 8.9
Johnston  ALP 16.0 8.7
Karama  ALP 8.3 1.0
Katherine  CLP -2.5 -9.8
Mulka  IND N/A N/A
Namatjira  CLP -0.3 -7.6
Nelson  CLP N/A N/A
Nightcliff  ALP 24.1 16.8
Port Darwin  ALP 1.9 -5.4
Sanderson  ALP 18.8 11.5
Spillett  CLP -13.5 -20.8
Wanguri  ALP 17.3 10.0

CLP is assumed to retain Nelson, as Labor’s 2020 result was so poor that the 2PP vote there is CLP v IND.

In Blain the incumbent Independent Mark Turner won the seat as a Labor candidate which, after redistribution, effectively holds the seat by a margin on 1.3 on the pendulum. With the narrow margin, large swing indicated in polling, and alleged illegal activity leading to Turner’s removal from the ALP, we can assume also that Blain would pass to the CLP.

In 2020 Araluen’s now Independent (then Territory Alliance) Robyn Lambley only beat the CLP by 0.5% 2PP, so this could easily return to the CLP who have held the seat in every election prior to Lambley’s resignation from that party in 2015. However Lambley is not to be underestimated, and I'd leave this as IND for now.

Mulka was only introduced in the last election, but it largely replaced the Labor-leaning seat of Nhulunbuy. CLP did not contest the seat in 2020, and the ALP is not contesting in 2024. To me this suggests that Independent Yingiya Mark Guyula had middle ground and CLP support in 2020 and will now lose some of the latter but gain the substantial de facto ALP vote which (if comparable to historical Nhulunbuy support) should keep him in the seat.

In Goyder Kezia Purick is retiring, and we may assume that the seat passes to the CLP based on the fact that CLP finished second in 2020 when Labor’s primary vote fell to 12.6 (4th place).

If we assume that CLP retain Neslon, gain Blain and Goyder and that Araluen and Mulka remain Independet, that would give CLP 14 seats, ALP 9 and 2 Independents.

SEAT INCUMBENT SWING BASED PREDICTION
Arafura  ALP CLP
Araluen  IND IND
Arnhem  ALP ALP
Barkly  CLP CLP
Blain  IND CLP
Braitling  CLP CLP
Brennan  CLP CLP
Casuarina  ALP ALP
Daly  ALP CLP
Drysdale  ALP CLP
Fannie Bay  ALP ALP
Fong Lim  ALP CLP
Goyder  IND CLP
Gwoja  ALP ALP
Johnston  ALP ALP
Karama  ALP ALP
Katherine  CLP CLP
Mulka  IND IND
Namatjira  CLP CLP
Nelson  CLP CLP
Nightcliff  ALP ALP
Port Darwin  ALP CLP
Sanderson  ALP ALP
Spillett  CLP CLP
Wanguri  ALP ALP

But That’s Just One Poll from Three Months Ago…

Okay, so how illustrative are the scarce NT Polls? A look at the 2PP data from the last decade shows that what polls have been taken overestimate Labor’s 2PP result, but typically underestimate the primary vote of both major parties—both in terms of data points and when indicating trend.

 

The latter phenomena, where primary votes for the major parties are higher than polling, is likely a result of undecided voters including the less politically engaged, who therefore pick between the major parties on polling day due to limited familiarity with the alternatives. All other observations are probably a phantasm emerging from very limited data.

However, what indications are available are not promising for the ALP, with actual 2PP results potentially even lower than polled. At any rate polls, even those closer to the election that we have available in 2024, can be off by more than 5 percentage points.

So what is the alternative?

In the absence of recent polling, our options are to try to infer voting intentions from other contemporary data, or to take a broader historical look.

Inference from contemporary politic-adjacent data has a number of issues; it will generally prioritise data that is quantitative (e.g. economic measures or crime rates) over that which is qualitative (e.g. labour sentiment or any scandal/outrage) which can only be quantified through the presence of polling; it requires substantial historical voting/polling data to demonstrate relationships which is not currently available; and it assumes that each election is fought on essentially the same issues.

Taking historical voting data is also flawed, again struggling to factor in changes in key issues over time as well as assuming a somewhat fatalistic paradigm where the future is set by recurring patterns that cannot be changed.

But it’s a Sunday afternoon here and I don’t have anything better to spend my time on, so let’s splash around in some data for the fun of it.

The graph below shows each party’s share of the primary vote territory wide. The colour of the background reflects the party in government between elections, and the divider in 1983 marks the Legislative Assembly’s change from 19 to 25 seats.

We can see that there is no a perfect correlation between primary vote and who wins enough seats to for government. Of the five years where Labor formed government, the CLP had a higher primary vote twice. However, Labor has never been ahead in the primary vote and failed to form government. This may indicate something about the nature of the "Other" vote—that it generally breaks in favour of the ALP, meaning the CLP tends to (needs to?) get a higher level of primary vote support to form government.

There is no obvious cycle of government here. CLP governed the Territory from its first election in 1974 until 2001—8 terms over 27 years—and then the NT has been (comparatively) more tumultuous. There is definitely a trend towards a more willingness to support other ("Other" or ALP) parties (with the exception of 1987, support for third parties has generally only picked up in the same time that Labor has been winning elections), but nothing to suggest that political changes are going to become increasingly frequent; indeed, they could scarcely become more rapid than the single term the CLP held in 2012, and if one treated that election as an outlier one could suggest the Territory was effectively CLP throughout the 20th century and Labor throughout the 21st.

Looking at the 2PP vote since 1983 marks a very similar story, though does iron out the anomalously high "Other" primary vote in 1987.

This does hint at a cycle since ~1997 where the ruling party consistently loses popularity from 55-60% 2PP over the course of about 12 years to around 40-45% 2PP, then consistently climbs as the other party suffers the same cycle.

The cycle C plotted here is given by CALP = 0.1 x Sin((Year-3)/2) + 0.5. The reciprocal cycle for the CLP is therefore given by either CCLP = 1-CALP or CCLP = 0.1 x Sin((Year+3)/2) + 0.5. This is not the optimal sin curve fit, just an eyeballed approximation.

With so few data points, no clear underlying mechanism and discarding the first four data points merely for not fitting the pattern, this apparent cycle is highly doubtful. But then, with the data we have to work with, everything is doubtful.

But if we run with it just out of curiosity we see the 2PP prediction for the ALP is 41.1% in 2024, or a swing of -5.6 percentage points, not exactly miles away from the latest polling’s -7.3 pp.

What Do We Do with Sketchy Data?

All of the above is of superficial predictive value, to be sure, but in each scenario things are looking bleak for the ALP. What polling is available suggests a swing large enough to hand power to the CLP; what a glance at past polling shows is that, if anything, the reality may be even crueller than the polling; and if anything can be inferred from past electoral fluctuations Labor’s time is running short.

The historical tide of elections suggests a swing in the order of -5.6 percentage points, polling puts this around -7.3 and historically polling could overestimate the 2PP vote by as much as 6.5 points (i.e. a swing up to -13.8).

We’ve considered a swing of -7.3 above. Let’s now look at an entirely credible range of -5 to -10:

SEAT INCUMBENT ALP LEAD
(PRE-ELECTION)
-5 PP
SWING
-7.3 PP
SWING
-10 PP
SWING
Arafura  ALP 3.6 -1.4 -3.7 -6.4
Araluen  IND N/A N/A N/A N/A
Arnhem  ALP 15.9 10.9 8.6 5.9
Barkly  CLP -0.1 -5.1 -7.4 -10.1
Blain  IND N/A N/A N/A N/A
Braitling  CLP -1.3 -6.3 -8.6 -11.3
Brennan  CLP -3.0 -8.0 -10.3 -13.0
Casuarina  ALP 16.0 11.0 8.7 6.0
Daly  ALP 6.1 1.1 -1.2 -3.9
Drysdale  ALP 5.4 0.4 -1.9 -4.6
Fannie Bay  ALP 10.9 5.9 3.6 0.9
Fong Lim  ALP 2.2 -2.8 -5.1 -7.8
Goyder  IND N/A N/A N/A N/A
Gwoja  ALP 16.2 11.2 8.9 6.2
Johnston  ALP 16.0 11.0 8.7 6.0
Karama  ALP 8.3 3.3 1.0 -1.7
Katherine  CLP -2.5 -7.5 -9.8 -12.5
Mulka  IND N/A N/A N/A N/A
Namatjira  CLP -0.3 -5.3 -7.6 -10.3
Nelson  CLP N/A N/A N/A N/A
Nightcliff  ALP 24.1 19.1 16.8 14.1
Port Darwin  ALP 1.9 -3.1 -5.4 -8.1
Sanderson  ALP 18.8 13.8 11.5 8.8
Spillett  CLP -13.5 -18.5 -20.8 -23.5
Wanguri  ALP 17.3 12.3 10.0 7.3

That gives the CLP a range of 9 to 12 seats, but 12 to 15 when including Blain, Goyder and Nelson, vs the ALP's 11 to 8.

The likely outlook is therefore a CLP victory, and probably a majority government. For the sake of a seat-by-seat final call, I'll restate the earlier prediction:

SEAT INCUMBENT SWING BASED PREDICTION
Arafura  ALP CLP
Araluen  IND IND
Arnhem  ALP ALP
Barkly  CLP CLP
Blain  IND CLP
Braitling  CLP CLP
Brennan  CLP CLP
Casuarina  ALP ALP
Daly  ALP CLP
Drysdale  ALP CLP
Fannie Bay  ALP ALP
Fong Lim  ALP CLP
Goyder  IND CLP
Gwoja  ALP ALP
Johnston  ALP ALP
Karama  ALP ALP
Katherine  CLP CLP
Mulka  IND IND
Namatjira  CLP CLP
Nelson  CLP CLP
Nightcliff  ALP ALP
Port Darwin  ALP CLP
Sanderson  ALP ALP
Spillett  CLP CLP
Wanguri  ALP ALP