Underestimation
I'll be honest, I wasn't expecting much from the NT election. To quote Australian political comedy series The Hollowmen, I might have described it as "almost boring". It's unicameral, so if their race isn't close there's not much nuance to consider. There are only 25 seats. There are a lot of large, regional seats which tend to be very predictable. Because territories are more reliant on the federal government, the stakes are a little lower. Last election Labor won in a massive landslide (18 seats out of 25 with the Country Liberals only getting 2), so even with a back-swing it seems a foregone conclusion.
That was before I looked into the madness that's been silently unravelling in the territory while no-one was looking.
In a previous government Country Liberal Chief Minister Terry Mills was replaced during a leadership spill, which was the fashionable thing to do in those days. He won a seat last election as an independent and has since formed the Territory Alliance with two other independents.
Now, if you've been paying attention (and I wouldn't blame you for tuning out--this is the 4th paragraph!) that gives the TA more seats than the CLP, meaning they are bigger than the official opposition. Someone in the parliament noticed this too, because all of the non-government members held a vote to decide who should be the opposition. TA held the most non-government seats, had demonstrated the ability to negotiate their way to this position, had a former Chief Minister at its head was not politically inclined to support the ALP government. The CLP was clinging to official opposition status by refusing to enter a coalition and by virtue of being a branch of the Liberal Party. So naturally the CLP won the vote and remained the official opposition with two seats.
That's all well and good during their term, but now we're in an election and it's not entirely obvious who the opposition to Labor is. It's CDP by tradition and TA on the numbers. This makes the NT a three-corner race, and it's not clear how this newcomer fits in.
Officially it is "ideology-free" which in literal terms should probably place it in the political centre, but is almost always a red flag for a party on the extreme fringe of one side or the other. The party includes a former CLP chief minister, a former member of the far-right RUAP and several other noted hard-line conservatives. On the other hand, it also has former ALP candidates, and preferenced the Greens in a by-election earlier this year. The party is in favour of renewable energy and medicinal cannabis as well as more police recruitment and "swift, impartial justice" for young offenders.
Overestimation
It should be no secret by now that I am not a professional. I speak freely with no real expertise, and lack the professionalism to even follow up on the results of my last prediction (which was a bad prediction, but that's not the reason). I have no idea what I am doing, and that's never stopped me before.
That being said, I'm going to use some statistical methods in this post that are incredibly useful in the right hands and probably butchered by mine. I'm probably not using the student t-test and p values as they should be; I haven't looked at this stuff since high school. But desperate times call for desperate measures, and the lack of polling in the NT causes desperate times.
Estimation
There are 25 seats in the NT. Four are held by independents (Goyder, Gwoja, Mulka and Nelson) and will require special consideration. Two others were races between the CLP and TA at the two-party preferred (2PP) level (Araluen and Fong Lim), which also requires consideration. For the rest of this post, both of these parties will be considered the "opposition" to the ALP. The seat of Spillett is also set aside from the analysis below as it was created just before the last election and doesn't have the requisite history including a meaningful measure of swing from 2012 to 2016.
This leaves 18 seats for analysis in a simple ALP vs CLP binary, plus Blain with is ALP vs TA. Normally, I'd take the polling from this week, calculate the swing from last election, apply it uniformly to these 19 seats and use that as a backbone for my predictions. The issues is that there is no polling available from this week. There is no polling available from this month. There is no polling available from last month. In fact, since last election there have been two polls: September 2019 and June 2020.
That's not a great indicator of current intention, given that the polling predates most of the campaigning.
Given that the federal government plays a greater role in territorial politics than in states, could federal polling be used as a proxy for territorial polling? The short answer is no.
Here's the long answer
There are, to my knowledge, only four polls published for NT voting intention at a territory level. Table 1 shows how the NT ALP performed in these four polls (2PP) and the range of results federal Labor achieved in three federal polls taken in the same months:
Table 1: NT and Aus 2PP polling
| NT Poll | Essential | Newspoll | Roy Morgan | |
| Jun-20 | 53.9* | 49-50* | 49 | 50.5 |
| Sep-19 | 42.6* | - |
49 | - |
| Jul-16 | 64 | 50-52 | - |
- |
| Mar-15 | 61.8 | 52-53 | 51-55 | 53.5-56 |
Plotting Labor's polling visually (X: NT polling; █: Essential range; █: Newspoll range; █: Roy Morgan range) makes it clear there is no meaningful correlation here:
Figure 1: NT and Aus Polling
Let's not be deterred, however. Perhaps there is a correlation between NT Labor's results at elections and federal polls within 7 days?
Table 2: NT electoral results and Aus polling
| NT Vote | Essential | Newspoll | Roy Morgan | |
| 29/08/2016 | 51 | |||
| 28/08/2016 | 50 | |||
| 27/08/2016 | 57.5 | |||
| 22/08/2016 | 51 | |||
| 26/08/2012 | 44 | |||
| 25/08/2012 | 44.2 | |||
| 19/08/2012 | 43 | 47 | 47 | |
| 10/08/2008 | 57 | |||
| 9/08/2008 | 49.3 | |||
| 18/06/2005 | 59.1 | 50 |
Figure 2: NT electoral results and Aus polling
Okay. Fine. Now we can be deterred.
Wild Mathematical Flailings
So, we have no meaningful approximation of a poll. This leaves me in something of an unprecedented situation, where we're relying entirely on the historical data of each seat to predict this election.
Fortunately the NTEC has data back to the 1980 election, meaning (at least for the divisions that were still around back then) we can calculate the 2PP swing to (+ve) or against (-ve) the ALP in every election since 1983 (ALP Margin is calculated as % of voters, and swing as current margin - past margin):
Table 3: Historical electoral data
| 2016 | 2012 | 2008 | 2005 | 2001 | 1997 | 1994 | 1990 | 1987 | 1983 | 1980 | |||||||||||||||||||||||||||||||||
| ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | ALP Swing | ALP | Opp. | ALP Margin | |
| Arafura | 1234 | 1020 | 4.75 | 5.74 | 1488 | 1548 | -0.99 | -14.99 | 1803 | 1014 | 14.00 | -9.58 | 2137 | 767 | 23.59 | 12.09 | 1586 | 993 | 11.50 | 2.49 | 1552 | 1078 | 9.01 | -6.23 | 1481 | 789 | 15.24 | -1.25 | 1431 | 721 | 16.50 | -1.40 | 1286 | 608 | 17.90 | No 2PP data | 1286 | 608 | 17.90 | ||||
| Arnhem | 1777 | 985 | 14.34 | 19.69 | 1102 | 1366 | -5.35 | Uncontested | 2110 | 746 | 23.88 | No 2PP data | 1365 | 998 | 7.77 | -16.73 | 1723 | 590 | 24.49 | No 2PP data | 976 | 795 | 5.11 | -3.22 | 840 | 600 | 8.33 | No 2PP data | |||||||||||||||
| Barkly | 1739 | 1257 | 8.04 | 0.44 | 1849 | 1361 | 7.60 | -8.00 | 1913 | 1003 | 15.60 | -7.43 | 2183 | 806 | 23.03 | No 2PP data | No 2PP data | 1519 | 1423 | 1.63 | 0.96 | 1239 | 1206 | 0.67 | 1.17 | 953 | 972 | -0.49 | 9.81 | 688 | 1045 | -10.30 | No 2PP data | ||||||||||
| Blain | 1849 | 1953 | -1.37 | 11.87 | 1470 | 2529 | -13.24 | -1.68 | 1410 | 2258 | -11.56 | -4.35 | 1594 | 2131 | -7.21 | No 2PP data | 884 | 2491 | -23.81 | ||||||||||||||||||||||||
| Braitling | 2314 | 2287 | 0.29 | 24.14 | 918 | 2592 | -23.85 | -0.39 | 935 | 2588 | -23.46 | IND v CLP | IND v CLP | 947 | 1870 | -16.38 | No 2PP data | No 2PP data | No 2PP data | 1325 | 377 | 27.85 | |||||||||||||||||||||
| Brennan | 2077 | 1869 | 2.64 | 16.88 | 1493 | 2683 | -14.25 | -11.60 | 1807 | 2009 | -2.65 | -4.49 | 1922 | 1785 | 1.85 | No 2PP data | 950 | 2747 | -24.30 | No 2PP data | 1024 | 1541 | -10.08 | ||||||||||||||||||||
| Casuarina | 2688 | 1694 | 11.34 | 2.04 | 2512 | 1724 | 9.30 | -4.86 | 2397 | 1339 | 14.16 | -4.86 | 2466 | 1107 | 19.02 | 15.50 | 1900 | 1650 | 3.52 | No 2PP data | 1328 | 1650 | -5.41 | No 2PP data | 1152 | 1109 | 0.95 | 16.40 | 765 | 1449 | -15.45 | No 2PP data | |||||||||||
| Daly | 1617 | 1772 | -2.29 | 2.35 | 1679 | 2022 | -4.63 | -10.40 | 1832 | 1453 | 5.77 | -9.04 | 2249 | 1221 | 14.81 | No 2PP data | |||||||||||||||||||||||||||
| Drysdale | 1971 | 1597 | 5.24 | 20.92 | 1413 | 2704 | -15.68 | -5.53 | 1457 | 2199 | -10.15 | -11.41 | 1829 | 1739 | 1.26 | 16.65 | 1199 | 2265 | -15.39 | No 2PP data | |||||||||||||||||||||||
| Fannie Bay | 2688 | 1499 | 14.20 | 7.44 | 2263 | 1724 | 6.76 | 5.70 | 1878 | 1800 | 1.06 | -17.56 | 2661 | 1217 | 18.62 | No 2PP data | 1782 | 1582 | 2.97 | 10.88 | 1232 | 1695 | -7.91 | No 2PP data | No 2PP data | No 2PP data | 896 | 861 | 1.00 | ||||||||||||||
| Johnston | 2453 | 1339 | 14.69 | 9.06 | 2224 | 1774 | 5.63 | -2.50 | 2246 | 1618 | 8.13 | -7.44 | 2355 | 1237 | 15.56 | 12.69 | 1858 | 1656 | 2.87 | ||||||||||||||||||||||||
| Karama | 1702 | 1647 | 0.82 | -5.57 | 2219 | 1716 | 6.39 | -4.41 | 2420 | 1560 | 10.80 | -6.18 | 2445 | 1205 | 16.99 | 13.25 | 1990 | 1713 | 3.74 | 5.44 | 1692 | 1811 | -1.70 | No 2PP data | No 2PP data | 1095 | 1423 | -6.51 | |||||||||||||||
| Katherine | 1843 | 1810 | 0.45 | 22.70 | 1146 | 2984 | -22.25 | -13.83 | 1499 | 2106 | -8.42 | -5.77 | 1736 | 1930 | -2.65 | No 2PP data | No 2PP data | 885 | 2127 | -20.62 | No 2PP data | CLP v NAT | |||||||||||||||||||||
| Namatjira/MacDonnell | 1742 | 1235 | 8.52 | 27.12 | 918 | 2006 | -18.60 | Uncontested | 1727 | 1067 | 11.81 | No 2PP data | 1235 | 1383 | -2.83 | -18.74 | 1682 | 870 | 15.91 | No 2PP data | No 2PP data | No 2PP data | 817 | 601 | 7.62 | ||||||||||||||||||
| Nightcliff | 3049 | 918 | 26.86 | 17.75 | 2301 | 1592 | 9.11 | -1.59 | 2293 | 1485 | 10.69 | -4.76 | 2385 | 1259 | 15.45 | No 2PP data | 1656 | 1991 | -4.59 | No 2PP data | No 2PP data | No 2PP data | No 2PP data | No 2PP data | |||||||||||||||||||
| Port Darwin | 1875 | 1676 | 2.80 | 12.46 | 1471 | 2175 | -9.65 | -6.83 | 1569 | 1757 | -2.83 | -3.90 | 1734 | 1661 | 1.08 | 7.99 | 1452 | 1918 | -6.91 | No 2PP data | No 2PP data | No 2PP data | No 2PP data | 690 | 1510 | -18.64 | No 2PP data | ||||||||||||||||
| Sanderson | 2578 | 1680 | 10.54 | 13.65 | 2041 | 2311 | -3.10 | 3.32 | 1723 | 2231 | -6.42 | -14.84 | 2204 | 1569 | 8.42 | 5.43 | 1959 | 1738 | 2.99 | 12.26 | 1369 | 1992 | -9.27 | -2.12 | 1274 | 1699 | -7.15 | No 2PP data | 1007 | 1368 | -7.60 | -2.79 | 973 | 1180 | -4.81 | -5.04 | 1930 | 1912 | 0.23 | ||||
| Wanguri | 3026 | 1302 | 19.92 | 12.92 | 2416 | 1823 | 6.99 | -7.87 | 2407 | 1304 | 14.86 | -5.94 | 2583 | 1065 | 20.81 | 20.81 | No 2PP data | 1649 | 1521 | 2.02 | 0.43 | 1594 | 1496 | 1.59 | -4.66 | 1653 | 1286 | 6.24 | 15.65 | 984 | 1440 | -9.41 | No 2PP data | ||||||||||
For those of you with less than a 12-foot computer display, these swings can be summarised as follows:
Table 4: Known swings to ALP in NT elections
| 2016 | 2012 | 2008 | 2005 | 2001 | 1997 | 1994 | 1990 | 1987 | 1983 | |
| Arafura | 5.74 | -14.99 | -9.58 | 12.09 | 2.49 | -6.23 | -1.25 | -1.40 | ||
| Arnhem | 19.69 | -16.73 | -3.22 | |||||||
| Barkly | 0.44 | -8.00 | -7.43 | 0.96 | 1.17 | 9.81 | ||||
| Blain | 11.87 | -1.68 | -4.35 | -7.21 | ||||||
| Braitling | 24.14 | -0.39 | ||||||||
| Brennan | 16.88 | -11.60 | -4.49 | |||||||
| Casuarina | 2.04 | -4.86 | -4.86 | 15.50 | 16.40 | |||||
| Daly | 2.35 | -10.40 | -9.04 | |||||||
| Drysdale | 20.92 | -5.53 | -11.41 | 16.65 | ||||||
| Fannie Bay | 7.44 | 5.70 | -17.56 | 10.88 | ||||||
| Johnston | 9.06 | -2.50 | -7.44 | 12.69 | ||||||
| Karama | -5.57 | -4.41 | -6.18 | 13.25 | 5.44 | |||||
| Katherine | 22.70 | -13.83 | -5.77 | |||||||
| Namatjira/MacDonnell | 27.12 | -18.74 | ||||||||
| Nightcliff | 17.75 | -1.59 | -4.76 | |||||||
| Port Darwin | 12.46 | -6.83 | -3.90 | 7.99 | ||||||
| Sanderson | 13.65 | 3.32 | -14.84 | 5.43 | 12.26 | -2.12 | -2.79 | -5.04 | ||
| Wanguri | 12.922 | -7.867 | -5.945 | 20.806 | 0.4332 | -4.658 | 15.65 |
Now, we could just take the average swing and see if it's enough to imperil the seat for Labor--the current house is theirs to lose, after all. But that's a little simplistic. Instead, I've opted to use the student t-test, which can be used to determine from a known set of data the likelihood of any given value being exceeded (in short, how likely a seat-toppling swing is, based on past data.) This assumes a normal distribution (i.e. that swings fall on a bell-curve) and is probably not an appropriate use of this test, but that's not going to stop me.
By taking the average swings from history (X̄), their standard deviation (σ), the current margin (μ) and number of historical examples (n), the t-value can be calculated as (X̄-μ)/(σ/√n). Knowing this and the degrees of freedom (DF=n-1), this calculator will give a probability p relating to the probability of getting a swing high enough for the seat to change hands. Unfortunately, the p result never exceeds .5 (or 50%), which means a result of .4 can mean a 40% chance of the swing being large enough or a 60% chance. Which of these is true is determined by whether the t value is +ve or -ve, and is factored into the final column (which also converts to a % probability), the probability of the seat changing hands. (Note: this is based on a comparison of the average swing against the 2016 result, not the current pendulum as all prior swings were calculated election-to-election and did not factor out redistribution).
Table 5: T values and probability of change (from polarised swing data)
| Average | σ | 2016 result | T value | DF | P result | P% of change | |
| Arafura | -1.644 | 8.6408 | 4.7471 | -2.092 | 7 | 0.037 | 3.7 |
| Arnhem | -0.088 | 18.408 | 14.337 | -1.357 | 2 | 0.154 | 15.4 |
| Barkly | -0.51 | 6.5779 | 8.0441 | -3.185 | 5 | 0.012 | 1.2 |
| Blain | -0.342 | 8.4503 | -1.37 | 0.2433 | 3 | 0.412 | 58.8 |
| Braitling | 11.877 | 17.342 | 0.2934 | 0.9446 | 1 | 0.259 | 74.1 |
| Brennan | 0.2626 | 14.826 | 2.6356 | -0.277 | 2 | 0.404 | 40.4 |
| Casuarina | 4.8438 | 10.525 | 11.342 | -1.381 | 4 | 0.12 | 1.2 |
| Daly | -5.7 | 7.0018 | -2.287 | -0.844 | 2 | 0.244 | 24.4 |
| Drysdale | 5.157 | 16.012 | 5.241 | -0.01 | 3 | 0.496 | 49.6 |
| Fannie Bay | 1.6157 | 12.962 | 14.199 | -1.941 | 3 | 0.074 | 7.4 |
| Johnston | 2.9537 | 9.4821 | 14.689 | -2.475 | 3 | 0.045 | 4.5 |
| Karama | 0.5039 | 8.5512 | 0.8211 | -0.083 | 4 | 0.469 | 46.9 |
| Katherine | 1.0325 | 19.195 | 0.4517 | 0.0524 | 2 | 0.482 | 51.8 |
| Namatjira/MacDonnell | 4.1921 | 32.425 | 8.5153 | -0.189 | 1 | 0.441 | 44.1 |
| Nightcliff | 3.803 | 12.185 | 26.859 | -3.277 | 2 | 0.041 | 4.1 |
| Port Darwin | 2.429 | 9.2598 | 2.802 | -0.081 | 3 | 0.47 | 47.0 |
| Sanderson | 1.2323 | 9.4318 | 10.545 | -2.793 | 7 | 0.013 | 1.3 |
| Wanguri | 4.4774 | 11.716 | 19.917 | -3.487 | 6 | 0.007 | 0.7 |
Again, this is probably not an appropriate use of this test, but there is some very rough value here. The least likely to change, with only a 0.7% chance, is Wanguri which had a 2016 margin of almost 20 percentage points and an average swing of less than 5 and the most likely to fall with almost a 3 in 4 chance is Braitling which had a margin of 0.29 and an average of almost 12 percentage points.
On this probably inappropriate analysis, Blain, Braitling and Katherine are likely to fall from Labor's grasp, and Brennan, Drysdale, Karama, Namatjira and Port Darwin are close to a coin flip.
However, this calculation is based on polarised data: an average swing is positive if it historically has favoured Labor and negative if it has favoured the opposition (CLP or more recently TA). This isn't an issue per say, but it assumes that a seat that has experienced a swing one way is volatile in a different way to one experiencing an equal swing the other way. This may be true--a seat may become increasingly conservative or progressively more... progressive, and a swing in those directions is more probable than the reverse. However, the term 'swing' implies that seats tend to follow patterns one way or the other as political fortunes ebb and flow and priorities bubble to the top of societal consciousness. It is assumed that most seats that swing to Labor in one year will, eventually, swing back at another and this ultimately causes the change in government--it is assumed, then, that the swings will tend to even out over time and shouldn't be polarised as above. Removing the negative value of the swings gives this table:
Table 6: T values and probability of change (from unpolarised swing data)
| Average | σ | 2016 result | T value | DF | P result | P of swing | P of against | P of change | |
| Arafura | 6.7219 | 5.11 | 4.7471 | 1.0931 | 7 | 0.155 | 84.5% | 62.5% | 52.8% |
| Arnhem | 13.212 | 8.7761 | 14.337 | -0.222 | 2 | 0.422 | 42.2% | 66.7% | 28.1% |
| Barkly | 4.6346 | 4.2195 | 8.0441 | -1.979 | 5 | 0.052 | 5.2% | 33.3% | 1.7% |
| Blain | 6.2785 | 4.3593 | -1.37 | 3.509 | 3 | 0.362 | 63.8% | 75.0% | 47.9% |
| Braitling | 12.263 | 16.796 | 0.2934 | 1.0078 | 1 | 0.249 | 75.1% | 50.0% | 37.6% |
| Brennan | 10.993 | 6.2169 | 2.6356 | 2.3285 | 2 | 0.072 | 92.8% | 66.7% | 61.9% |
| Casuarina | 8.7303 | 6.6954 | 11.342 | -0.872 | 4 | 0.216 | 21.6% | 40.0% | 8.6% |
| Daly | 7.2645 | 4.3125 | -2.287 | 3.8362 | 2 | 0.031 | 96.9% | 66.7% | 64.6% |
| Drysdale | 13.627 | 6.6527 | 5.241 | 2.5211 | 3 | 0.043 | 95.7% | 50.0% | 47.9% |
| Fannie Bay | 10.394 | 5.2385 | 14.199 | -1.452 | 3 | 0.121 | 12.1% | 25.0% | 3.0% |
| Johnston | 7.9209 | 4.2294 | 14.689 | -3.2 | 3 | 0.025 | 2.5% | 50.0% | 1.3% |
| Karama | 6.97 | 3.5656 | 0.8211 | 3.8561 | 4 | 0.009 | 99.1% | 60.0% | 59.5% |
| Katherine | 14.103 | 8.4685 | 0.4517 | 2.7921 | 2 | 0.054 | 94.6% | 66.7% | 63.1% |
| Namatjira/MacDonnell | 22.928 | 5.9286 | 8.5153 | 3.438 | 1 | 0.090 | 91.0% | 50.0% | 45.5% |
| Nightcliff | 8.0323 | 8.5662 | 26.859 | -3.807 | 2 | 0.031 | 3.1% | 66.7% | 2.1% |
| Port Darwin | 7.7938 | 3.5526 | 2.802 | 2.8102 | 3 | 0.034 | 96.6% | 50.0% | 48.3% |
| Sanderson | 7.4307 | 5.2527 | 10.545 | -1.677 | 7 | 0.069 | 6.9% | 50.0% | 3.5% |
| Wanguri | 9.7543 | 7.0458 | 19.917 | -3.816 | 6 | 0.004 | 0.4% | 42.9% | 0.2% |
The calculation that previously gave us the probability of change has been relabelled as P of swing--that is, the probability of a swing being larger than the margin after 2016. In Barkly the average swing in either direction is 4.6 and the margin is about 8 percentage points. It'd be uncharacteristic for a swing to be large enough in Barkly to topple this--the target is almost a full standard deviation from the average. In fact, the probability of this is a mere 5.2%. However, even if the swing was big enough, it might be a swing to the ALP, making it safer. two thirds of swings we have 2PP data for were to the ALP, leaving us with a 33.3% of a swing against the ALP (P of against). This drops the probability of Labor losing the seat to 1.7%.
This is almost certainly an abuse of the statistics now. It not only assumes a normal distribution of swings, but that the direction of the swing is an independent probability, which it just plain isn't--the highest swings will generally occur when the incumbent has a huge margin as a correction to previous voting patterns.
A Fine Lot of Fine Tuning
Both of these methods are highly suspect, but we are now going to use them to supplement the usual considerations (like not being an typical 2PP race, having an independent incumbent or having a very small margin) to rule seats in for further consideration. If they don't flag something that should be considered, it is either undetectable by any method or should be caught by the usual methods. If it rules in something it shouldn't, on the other hand, there's no harm taking a second look.
On these two tests, then, further consideration is needed for:
Arafura: >50% chance of changing in Table 6 as the average swing exceeds the 2016 margin. Note, though, the probability in table 5 was much lower (3.7%).
Blain: >50% in Table 5, and very close in table 6.
Braitling: >50% in Table 5, and close in table 6.
Brennan: >50% in Table 6, and very close in table 5.
Daly: >50% chance of changing in Table 6 as the average swing exceeds the 2016 margin. Note, though, the probability in table 5 was below 1 in 4.
Drysdale: Very close to being a 50-50 in both tables 5 and 6, though considered more likely to remain ALP than not in both cases.
Karama: Very close to being a 50-50 in both tables 5 and 6, though considered more likely to remain ALP than not in both cases.
Katherine: The only seat considered >50% in both tables 5 and 6.
Namatjira: Very close to being a 50-50 in both tables 5 and 6, though considered more likely to remain ALP than not in both cases.
Port Darwin: Very close to being a 50-50 in both tables 5 and 6, though considered more likely to remain ALP than not in both cases.
Interestingly, with the sole exception of Namatjira, this is identical to the list of seats that would be considered marginal (i.e. with a margin of less than 6 percentage points). Due to redistributions the current pendulum has moved Arafura to "fairly safe" and Karama to "safe". Namatjira, flagged by our system, has become marginal.
On our metrics Arnhem, Barkly, Casuarina, Fannie Bay, Johnston, Nightcliff, Sanderson and Wanguri are safe:
Table 7: Prediction (v1)
| Seat | Incumbent | Prediction |
| Arafura | ALP | |
| Araluen | TA | |
| Arnhem | ALP | ALP |
| Barkly | ALP | ALP |
| Blain | TA | |
| Braitling | ALP | |
| Brennan | ALP | |
| Casuarina | ALP | ALP |
| Daly | CLP | |
| Drysdale | ALP | |
| Fannie Bay | ALP | ALP |
| Fong Lim | TA | |
| Goyder | IND | |
| Gwoja | IND | |
| Johnston | ALP | ALP |
| Karama | ALP | |
| Katherine | ALP | |
| Mulka | IND | |
| Namatjira | ALP | |
| Nelson | IND | |
| Nightcliff | ALP | ALP |
| Port Darwin | ALP | |
| Sanderson | ALP | ALP |
| Spillett | CLP | |
| Wanguri | ALP | ALP |
This is a concerning start for the opposition--be that Liberal or Territory Alliance; Labor only needs to win 13 seats to hold a majority and on our first pass they've locked in eight.
Since the redistribution of boundaries, there are 11 safe seats: the eight set for Labor in table 7, Gwoja where a retiring independent sets the seat up for the ALP, Karama which was only flagged due to a 0.82 margin and is now fairly safe at 12.3, and Spillett for the Liberals. However, there was no TA in 2016 so while it's safely non-ALP it's not impossible the Alliance candidate Vanessa Mounsey could win here. Either way my ultimate assumption is that any hung parliament will see the CLD and TA enter a coalition, so I'm lumping them together as "opposition". This also resolves the seats of Araluen and Fong Lim which are contested between CLP and TA on a 2PP basis
We've discussed Gwoja, considered an easy ALP win. The other independents are in Goyder, where the incumbent has a margin of over 25 percentage points, Mulka and Nelson. The Nelson independent is retiring, which normally would toss this seat into firm CLP territory but a closer look is warranted. CLP is the only party to have held Nelson (one term from 1997) but the rest of its 30-year history has had it in independent hands. This means independent candidate Beverli Ratahi should not be ruled out, and this could be fruitful hunting grounds for the TA looking for a seat that is on the conservative side but not pledged to either the ALP or CLP. This is for sure a seat to watch.
The independents, then, are assigned for Goyder to be retained off the back of a very safe margin, Gwoja to return to the ALP, Nelson to go to the combined CLP-TA "opposition" and Mulka to remain IND. This last seat (formerly Nhulunbuy) is based on the observation that this was a reliable ALP seat until last election when a popular independent took it. Now that he's established as a viable candidate the increased visibility and an option to kick the current government without backing the opposition should see his very narrow margin grow. An ALP win is he only feasible alternative.
Table 8: Prediction (v2)
| Seat | Incumbent | Prediction |
| Arafura | ALP | |
| Araluen | TA | Opp |
| Arnhem | ALP | ALP |
| Barkly | ALP | ALP |
| Blain | TA | |
| Braitling | ALP | |
| Brennan | ALP | |
| Casuarina | ALP | ALP |
| Daly | CLP | |
| Drysdale | ALP | |
| Fannie Bay | ALP | ALP |
| Fong Lim | TA | Opp |
| Goyder | IND | IND |
| Gwoja | IND | ALP |
| Johnston | ALP | ALP |
| Karama | ALP | ALP |
| Katherine | ALP | |
| Mulka | IND | IND |
| Namatjira | ALP | |
| Nelson | IND | Opp |
| Nightcliff | ALP | ALP |
| Port Darwin | ALP | |
| Sanderson | ALP | ALP |
| Spillett | CLP | Opp |
| Wanguri | ALP | ALP |
Labor only needs to win 3 of the outstanding seats to form majority government. This does not seem to be a tall order
Arafura: Arafura is a long-standing ALP seat; from its creation in 1983 it has been ALP except for a brief moment of independence in 2009 and for one term (2012-2016 as CLP, independent, Palmer and then CLP again) by a margin of 0.1. If there is a correction from last year, Arafura must be in the mix. I feel, though that Labor may hold on to this one as last election was itself a correction from the 15-point swing to CLP.
Blain: Held on a narrow 1.4 pp margin by the TA, there are several reasons this should increase: A reaction to the ALP in a traditionally conservative seat, a long history of conservative control, and a high-profile independent candidate who proved not only to be viable enough to become the incumbent but has now formed the Territory Alliance which has some claim to being the main conservative party.
Braitling: This is a traditionally conservative seat, won by a narrow margin in the ALP landslide last election for the first time and destined to return to conservatives.
Brennan: From one of the safest seats to a close race in recent times, Brennan can be won by the CLP if they pay it enough attention, and after the last election's crushing defeat they'd have been foolish not to.
Daly: This seat is certainly close, and historically could go either way, but the CLP held it against the 2016 landslide. That puts them in a good position, but the current incumbent is retiring which puts this potentially back in play. I'm naming this an Opp seat (likely CLP), but also putting it on my watch list.
Drysdale: Drysdale has a mixed history, but was won by the ALP in a landslide when the incumbent transferred to Spillett. Backing incumbency in the absence of anything more meaningful, this could be an ALP hold.
Katherine: This has been a CLP seat since foundation, except for in the landslide last election which brought it to the ALP as its most marginal seat. A return to normal is predicted.
Namatjira: Formerly MacDonnell, there's a mixed history but a CLP lean. It's held by ALP but redistribution makes it nominally CLP before even considering the correction from 2016.
Port Darwin: A strong CLP history, a thin ALP margin and a likely general trend against ALP across the territory. Plug it in the opposition column and print out the results:
Table 9: General Prediction
| Seat | Incumbent | Prediction |
| Arafura | ALP | ALP |
| Araluen | TA | Opp |
| Arnhem | ALP | ALP |
| Barkly | ALP | ALP |
| Blain | TA | Opp |
| Braitling | ALP | Opp |
| Brennan | ALP | Opp |
| Casuarina | ALP | ALP |
| Daly | CLP | Opp |
| Drysdale | ALP | ALP |
| Fannie Bay | ALP | ALP |
| Fong Lim | TA | Opp |
| Goyder | IND | IND |
| Gwoja | IND | ALP |
| Johnston | ALP | ALP |
| Karama | ALP | ALP |
| Katherine | ALP | Opp |
| Mulka | IND | IND |
| Namatjira | ALP | Opp |
| Nelson | IND | Opp |
| Nightcliff | ALP | ALP |
| Port Darwin | ALP | Opp |
| Sanderson | ALP | ALP |
| Spillett | CLP | Opp |
| Wanguri | ALP | ALP |
That makes ALP - 12, Opposition - 11 or a minority government. However, the majority of less certain seats have ALP incumbents and opposition gains; additionally, the NT has done well (comparatively) through coronavirus and that's likely to play well for the government. So, despite the seat-by-seat predictions here, my sum total expectation is a Labor majority. If that's not correct, given the division of CLP and Territory Alliance, it seems likely the ALP will have the advantage of securing a minority government before a conservative coalition.
Seats to watch
I haven't done tossups on this blog for a while now, but these are the seats I think are least predictable and how these flop will probably determine whether we're looking at a Labor majority or a hung parliament.
Arafura - Predicted as an ALP hold on a very tiny margin this could easily fall.
Daly - The fact that Daly held out in 2016 and now likely has a state-wide swing holding it in line may not mean much if the candidate that secured the 2016 win retires.
Drysdale - This one's a coinflip. The ALP call is based solely on incumbency and doesn't mean much.
Nelson - Interesting only to see if it goes to the CLP or TA, or at an outside shot to the independent candidate.
Predictions
For once I'm not putting my seat-by-seat predictions forward as my main result. There're a lot of unknowns, particularly without any reliable polling to work from. While I'll still check those results, my main predictions are as follows:
1) ALP majority government - in table 8 there are 10 safe ALP seats; just securing half of the undecided slots in that table would be a majority government.
2) Swing against ALP - the landslide in 2016 will face a correction, and there's certainly not much room for the ALP to increase its lead at a seat-level.
3) CLP regains upper hand as opposition - Most of the seats given to the opposition will go to CLP rather than TA. The recent stint as the largest opposition party is not enough to shift the two-party psyche. In particular, Braitling, Brennan, Daly, Katherine, Namatjira, Port Darwin and Spillett are picked for CLP, while Araluen, Blain and Fong Lim are likely TA seats. Nelson will be one of the two in my mind, but it's not clear which:
Table 10: Final prediction
| Seat | Incumbent | Prediction |
| Arafura |
ALP | ALP |
| Araluen |
TA | TA |
| Arnhem | ALP | ALP |
| Barkly | ALP | ALP |
| Blain | TA | TA |
| Braitling | ALP | CLP |
| Brennan | ALP | CLP |
| Casuarina | ALP | ALP |
| Daly | CLP | CLP |
| Drysdale | ALP | ALP |
| Fannie Bay | ALP | ALP |
| Fong Lim | TA | TA |
| Goyder | IND | IND |
| Gwoja | IND | ALP |
| Johnston | ALP | ALP |
| Karama | ALP | ALP |
| Katherine | ALP | CLP |
| Mulka | IND | IND |
| Namatjira | ALP | CLP |
| Nelson | IND | Opp |
| Nightcliff | ALP | ALP |
| Port Darwin | ALP | CLP |
| Sanderson | ALP | ALP |
| Spillett | CLP | CLP |
| Wanguri | ALP | ALP |
