When preparing for an election coverage, my first step is always to compile a set of last election results as history for the ABC’s election website and election computer.
The data includes historic votes by candidate and party for each electorate. To enable more complex analysis on election night, the same historic data is collected for each polling place.
When there has been a redistribution, I also re-calculate last election results to match the new electoral boundaries. Those calculations become the stored history for use in analysis.
Another important job is to give all electorates an initial party classification. On election night, party classification is combined with results to label seats as either “gains/losses” or “retains”.
Normally I classify seats based on last election votes. But sometimes I depart from this practice when by-elections, candidate defections and redistributions complicate analysis. Relying too strictly on past votes can make it hard to explain a result on election night.
A prize example is Novembers Victorian election. A major redistribution since the 2018 election has left a number of seats as unclear on how best to allocate an initial party classification.
Recently the Victorian Electoral Commission (VEC) released estimated margins for all seats. You can find the estimates on the VEC’s research and publications page at the link named “Methodology of estimating 2018 election results on new electoral boundaries”.
The table below shows the seats won by party at the 2018 Victorian election on the first line. The second line shows the seat holdings if you strictly apply the VEC’s estimated margins. The third line is the seat holdings I have advised the ABC to use for the Victorian election site (coming soon!) and in the ABC’s election computer.
| ALP | LIB | NAT | GRN | IND | |
|---|---|---|---|---|---|
| 2018 Election result | 55 | 21 | 6 | 3 | 3 |
| VEC party seat totals | 58 | 19 | 7 | 3 | 1 |
| ABC party seat totals | 56 | 21 | 6 | 3 | 2 |
A two-page pdf with a new electoral pendulum and alphabetic list of seat margins can be found via this link.
The VEC totals explain why the Labor Party has been reported as having gained two seats through the redistribution. In my view that is not the case. Labor does have two more safe seats than it did on the old boundaries, but at the expense of some marginal Labor seats. The two possible extra Labor seats come about through VEC estimated margins that flip two Liberal seats on fewer than 70 votes.
Whether these two seats are classed as Labor or Liberal held, it has no impact on the swing of around 9.6% needed for the Labor Party to lose its majority
The ABC will be using the same margins as the VEC in most seats. The VEC’s estimates have only minor variations from the estimated margins I published last year. (My original estimates can be found in this post).
The ABC and VEC party totals differ because the ABC is choosing to classify ultra-marginal Caulfield and Hastings as Liberal held (the VEC suggests Labor held) and to keep Mildura as Independent held (the VEC suggests National held). How these seats are classified has no impact on the swing needed for Labor to lose its majority.
There are also four seats where the ABC and VEC agree that party status has changed because of the redistribution.
Where the ABC and VEC Seat Classifications Differ
Redistribution calculations are not predictions of who will win a seat at the next election. Rather they are backward projections, estimated margins for the last election assuming the election had been contested on the new boundaries.
In three seats the ABC will be leaving seats with their 2018 holding party rather than flip party holdings based on tiny estimated new margins.
Caufield (old margin LIB 0.3%) has undergone some minor boundary adjustments. The VEC estimates that Caufield now has a Labor majority of 0.16%, representing about 64 votes. The ABC will classify the new Caufield (LIB 0.04%) as Liberal held. This simplifies analysing Caufield on election night given Liberal Deputy Leader David Southwick represents the seat.
The point of assigning an initial classification is so that results can be listed as “retain” or “gain”. If Caulfield is classed as Liberal held, victory for David Southwick will display as “Liberal retain” rather than “Liberal gain from Labor”. Also better to have defeat for Southwick by an Independent displayed as “Independent gain from Liberal” rather than “Independent gain from Labor”.
Hastings (old margin LIB 1.1%) has also undergone boundary changes and the VEC calculates the re-drawn seat has a notional Labor margin of 0.01%. That’s about nine votes. While Liberal MP Neale Burgess is retiring, the ABC has retained Hastings (LIB 0.0%) as a Liberal seat with no margin. It makes more sense for a Labor victory in Hastings to display as “Labor gain” than “Labor retain”.
In 2018, Independent Ali Cupper won Mildura (old margin IND 0.3% v NAT), but the new boundaries have added an area around Donald she did not contest in 2018. The VEC has calculated a new margin for the seat that would put the National Party ahead. However, such calculations ignores how the Independent could poll in the new area. The ABC has chosen to categorise Mildura as Independent held with no margin. Given there is an Independent MLA, this makes more sense than to rely on 2018 votes in an area Cupper did not contest.
Seats Changing Party Classification
There are four seats where the ABC and VEC agree on re-classifying seats – Bass, Bayswater, Morwell and Ripon. These four seats will be displayed on the ABC site with an asterisk ‘(*)’ after the name to signifiy a change in the seat’s status.
Ferntree Gully (LIB 1.6%) has been abolished and merged with old Bayswater (ALP 0.4%) creating a notionally Liberal seat in the new Bayswater (LIB 0.6%). The new Bayswater will be contested by the sitting MPs from both former seats, Labor’s Jackson Taylor from Bayswater and Liberal Nick Wakeling from Ferntree Gully. With two sitting MPs contesting, it makes more sense to use votes in classifying the seat. (Summary – Labor minus one seat)
On Melbourne’s outer south-east fringe, three seats have been formed from the abolished seat of old Gembrook (LIB 0.8%) and surplus votes from the over-quota old Bass (ALP 2.5%). The new seats are new Berwick (LIB 1.3%) and new Pakenham (ALP 2.2%), while the new boundaries transform new Bass (LIB 0.7%) into a notional Liberal seat. Gembrook Liberal MP Brad Battin will contest Berwick, Bass Labor MP Jordan Crugnale will re-contest Bass, while Pakenham will have no sitting MP. (Summary – Liberal plus one, Labor minus one)
In western Victoria, a redistribution has caused old Ripon (LIB 0.02%) to flip allegiances, the new Ripon (ALP 2.7%) having a notional Labor majority. Liberal MP Louise Staley will re-contest her seat. Flipping the classification avoids confusion on election night if there is a swing to the Liberal Party but Staley is still defeated. (Summary – Labor plus one, Liberal minus one)
In Eastern Victoria, old Morwell (IND held) now includes Moe, significantly increasing the Labor vote. Given Independent MP Russell Northe is retiring, the new Morwell (ALP 4.0%) has been classed as Labor held using alternative two-party preferred results from 2018. (Summary – Labor plus one, Independent minus one).
Other Significant changes
The inner south-east seat of Keysborough (ALP 14.9%) has been abolished and divided between surrounding electorates. Labor MP Martin Pakula will retire. (Summary – Labor minus one seat)
West of the Yarra, the very safe Labor seat of Yuroke (ALP 20.3%) on Melbourne’s north-west fringe has been abolished. It has been divided into equally safe Labor seats, Greenvale (ALP 22.0%) and Kalkallo (ALP 20.9%). A new seat called Laverton (ALP 23.4%) has also been created straddling Wyndham and Brimbank councils. The electorate of Altona (ALP 14.6%) has shifted south and been re-named Point Cook (ALP 12.8%). (Summary – Labor plus two seats)
In eastern Melbourne, Burwood (ALP 3.3%) and Mount Waverley (ALP 1.8%) have been abolished and largely replaced by the new seat of Ashwood (ALP 2.0%). Ashwood will be contested by Labor’s MP for Mount Waverley Matt Fregon, while Burwood’s Will Fowles is contesting Ringwood in place of the retiring Dustin Halse. (Summary – Labor minus one seat)
Also in Melbourne’s east, Forest Hill (LIB 1.2%) has been merged with parts of abolished Mount Waverley (see Ashwood above) to create the new seat of Glen Waverley (LIB 0.9%). Forest Hill Liberal MP Neil Angus will contest Glen Waverley. (Summary – no change)
Extraordinary insight, as always Antony. I look forward to more analysis of the Victorian Election.
Interesting that redistribution report differs with your analysis here – it stated that Greenvale is a new seat and Kalkallo is the new name for the Yuroke electorate, referring to “Proposed new Greenvale district” and “Proposed Kalkallo (Yuroke) District” and stating “The EBC proposed to rename the district Kalkallo, after the locality in the centre of the district.”
COMMENT: The Redistribution reports have a habit of following the numbered order in which they draw electorates. Yuroke was split 32,525 voters to Greenvale and 34,814 to Kalkallo. This even split is why I describe Yuroke has been abolished and Kalkallo and Greenvale created. Like Yuroke, Kalkallo is the last seat before the country but that’s the only real way you can describe it as the successor.
The Redistribution report also says Gembrook is replaced by Berwick, but the VEC’s redistribution calculations say Gembrook has been replaced by Pakenham. What’s really happened is that Gembrook and surplus voters from Bass have been used to create two seats, Berwick and Pakenham. I don’t think either seat is a good enough fit to say it replaces Gembrook.
As a returning candidate but in a different seat, I’m obviously interested in the Point Cook margin. With the high numbers of early voters and not being able to tell which polling booth they should be from, how do you adjust for that?
COMMENT: The best you can do is split them according to the proportion of votes transferred, and then weight them based on party vote share in the area weighted against party vote share in the old electorate as a whole.