VIC22 – 2-Party Preferred Results and Swings by District

Inside this post I am publishing corrected two-party preferred (2PP) results, state-wide and by district, for the 2022 Victorian election.

The post is based on the Victorian Electoral Commission’s (VEC’s) published two-party preferred totals. At the moment the difference in my table is that I include corrected the two-party preferred totals for Brighton and Werribee.

These two corrected totals have been calculated from the data entered ballot paper files for both seats. The VEC did not publish a completed preference distribution for either seat but the correct 2PP can be calculated from the data files. I recently analysed the preference flows for seven districts where data entry was available, including for Brighton and Werribee.

In February the VEC intends to undertake formal preference distributions in districts where a distribution was not required. I will update the table and this post as the new figures become available.

As well as publishing two-party preferred totals, the post explains the VEC’s counting procedures that are responsible for me having a different state-wide 2PP, and also why there will be further changes when the VEC conducts the additional distributions in February.

Narracan Currently Excluded

The current published totals only include 87 of the 88 Legislative Assembly districts. There was no election held in Narracan owing to the death of a candidate. A supplementary election will be held for Narracan on 28 January. As is normal practice, for instance Victoria in 1999, NSW 2003, and Federal elections in 1993 and 1998, supplementary election results are included as part of general election totals.

Labor is not nominating a candidate for the supplementary election. Once the election is held, I will use the Labor-Greens final count (if that is the final pairing) as a pseudo-2PP count to calculate state-wide 2PP totals. I suspect the VEC will not include a Greens-based Narracan 2PP in state-wide totals.

Note – I am publishing background on the Narracan supplementary election using the Narracan page of the ABC’s Victorian Election website. I hope to publish the election results on the same page.

Explaining Victorian Count Procedures

Figures in the VEC’s two-party preferred table are derived from three sources.

  • In 38 districts, 40 in my table, a full distribution of preference was conducted that finished with a final two-party preferred count between Labor and Coalition candidates.
  • Another 12 districts finished as two-candidate preferred (2CP) counts, that is the final pairing of candidates was not Labor versus Coalition. In these districts a separate two-party preferred count was conducted.
  • In the other 37 districts, 35 in my table, preferences were either not distributed or the distribution stopped with more then two candidates at the point when one candidate passed 50%. In these districts, the VEC used the totals of indicative preference counts by polling place and vote type as the 2PP or 2CP for each district.

The problem with this final category of seats is that indicative preference counts by polling place and vote type are done on the first count. All votes are then check counted and any errors corrected in the first preference tallies, but the indicative preference counts are not altered to correct errors identified in the check count..

In seats where a preference distribution is required, the VEC first undertakes an amalgamation count. Ballot papers initially bundled by candidate by polling place/vote type are amalgamated and re-bundled by candidate by electorate. The preference distribution is done on these new bundles. After the amalgamation count, it is not possible to cross-reference ballot papers back to where they were cast or counted.

The preference distribution is therefore carried out on the check count, so the final distribution will not match the total of indicative counts by polling place/vote type. In some cases these differences are significant.

In Pakenham, a significant number of first count errors were identified when check counting pre-poll votes. Based on the check count, the distribution of preferences delivered Labor’s Emma Vulin victory over Liberal David Farrelly 19,587 votes (50.4%) to 19,280 (49.6%). The uncorrected indicative preference counts had Farrelly ahead 19,558 (50.1%) to Vulin 19,468 (49.9%).

We can spot these errors in seats with a full distribution of preferences or an alternate 2PP count. All these seats have a correct preference count based on the check count, but they do not match the indicative preference count.

Other districts have errors in the indicative preference count that can’t be identified, but will be spotted when the VEC undertakes further preference distributions in February.

In my tables below, I have included the corrected two-party preferred count for Brighton and Werribee derived from the data entered ballot papers. The VEC’s table uses the indicative count totals. The corrected figures lift the Liberal two-party result in Brighton from 54.2% to 55.1%, and the Labor result in Werribee from 60.5% to 60.9%.

These discrepancies do not occur in states with different counting procedures.

The Australian Electoral Commission (AEC), and commissions in most other states, no longer undertake amalgamation counts. Rather the distribution of preferences is undertaken using the bundles by polling place and count centre. A step of the distribution excluding one candidate is done across all polling places before moving on the next candidate. A computer program is used to keep track of these distributions and determine the next candidate to be excluded. At the end of a count using this method, the 2PP/2CP totals by polling place and count centre always add to the 2PP/2CP for the electorate.

The VEC’s problem is the current count procedures have been implemented into its election management system, used to conduct the count. On its website and in the results feed provided to the media, the VEC can publish the first count totals for a district and by polling place, or the check count totals. In Queensland both sets of figures are published, while other states display first count or check count by polling place and vote type, so the district totals are an amalgamation of the two sets of totals.

The VEC’s computer system also enforces a rule that the indicative preference count must match the first count formal vote. Even if the VEC check counted the indicative preference count, this can’t be entered into the election management system because it insists on checking against the first count formal total, not the check count formal total.

This inconsistency between the storage and display of first counts, check counts and preference counts, is a significant problem in VEC count management. In particular it makes it harder for scrutineers to have a reliable source for the current state of the count.

The other big problem is Pre-Poll totals being provided as a single total rather than the largest Pre-Poll centres being separately identified in the count. A solution to that problem goes back to the way pre-poll centres are conducted and requires pre-poll centres to be declared as counting centres for more than one district.

These procedural problems have arisen largely because the VEC’s counting systems were put in place before the last decade’s explosion in pre-poll voting. The procedures can be fixed, but they require changes to the election management system software. The pre-poll count centre problem might also require changes to legislation.

The Two-Party Preferred Table for Victoria

The table is set out below. (+) indicates districts that were not final two-party preferred counts. I have more information on those counts below the table.

The column labelled ‘Source’ explains where the 2PP has been derived.

  • Actual – result from full distribution of preferences. Category includes corrected values for Brighton and Werribee.
  • 2PP – result from alternate two-party preferrd count.
  • Bths – VEC total from polling place and vote type indicative preference counts.

The table can be sorted by Labor %, Coalition % and Swing. Sorting by either % figure gives you a single column 2PP electoral pendulum.

Two-Candidate Preferred Finishes

Benambra - Liberal 20,956 (50.9%) versus Independent 20,179 (49.1%), a 1.6% swing to Independent
Brunswick - Green 27,699 (63.5%) versus Labor 15,948 (36.5%), an 11.5% swing to Greens
Footscray - Labor 21,880 (54.2%) versus Greens 18,460 (45.8%), an estimated 13.9% swing to Greens
Melbourne - Greens 25,593 (60.2%) versus Labor 16,926 (39.8%), an 8.5% swing to Greens
Mildura - Nationals 19,520 (51.2%) versus Independent 18,620 (48.8%), a 0.9% swing to Nationals
Mornington - Liberal 21,326 (50.7%) versus Independent 20,736 (49.3%), no comparable swing
Northcote - Labor 21,413 (50.2%) versus Greens 21,229 (49.8%), a 1.5% swing to Greens
Pascoe Vale - Labor 20,950 (52.0%) versus Greens 19,323 (48.0%), no comparable swing
Prahran - Greens 24,334 (62.0%) versus Liberal 14,909 (38.0%), a 3% swing to Greens
Preston - Labor 20,761 (52.1%) versus Greens 19,096 (47.9%), a 19.2% swing to Greens
Richmond - Greens 22,772 (57.3%) versus Labor 16,959 (42.7%), a 14.1% swing to Greens
Shepparton - Nationals 23,455 (56.8%) versus Independent 17,846 (43.2%), a 12.1% swing against Independent

Note that the swings against Labor in Labor-Green contests are magnified by the Liberal Party's decision to change its preference recommendation. In 2022 the recommendation was towards the Greens, in 2018 it was to Labor.

Leave a Reply