2026 – Farrer By-election – Final Count Data Released

This morning the AEC finalised counting for the Farrer by-election with the release of detailed preference data.

It answers what I thought was a key question in the result – whether there was a difference between Liberal and National preference flows to One Nation. It was not possible to answer this question using the previously released full distribution of preferences.

Preference flow data tallies destination preferences from the primary votes for each excluded candidates to one of the final two candidates. In Farrer this data tallies preferences from ten excluded candidates to One Nation’s David Farley and Independent Michelle Milthorpe, the final pairing of candidates. Preference flow data ignores the intermediary stages published as part of the formal distribution of preferences.

The data released today reveals a difference in preference flows by supporters of the two Coalition parties . Liberal voter preference flows to One Nation were 59.0% compared to 69.1% for Nationals voters.

The flow data also reveals that One Nation’s David Farley did not need Liberal preferences to win. In the full distribution of preferences, Farley did not pass 50% until the Liberal candidate was the last excluded. Using the flow data where National preferences do not first flow through the Liberal candidate, Farley reached 50.2% on the flow of National preferences without the need for any Liberal preferences.

The question is whether the different flows tell us anything about how Liberal and National preferences would flow at a general election. In Farrer there was no Labor candidate in the final pairing. I suspect both National and Liberal preference flows to One Nation would have been higher in a final contest involving Labor rather than an Independent.

But it does suggest that Liberal preferences to One Nation will be weaker than those of National candidates. That was the case two decades ago in Queensland before the formation of the LNP. It in part reflects the different appeal of One Nation in rural electorates contested by Nationals versus the flows in urban electorates contested by Liberal candidates. One Nation polled a high enough first preference vote in Farrer for this difference not to matter. But a lower primary vote and weaker Liberal preferences would hamper One Nation hopes of winning urban seats.

Against this view, that the Liberal candidate drew top spot on the ballot paper with Milthorpe in second position suggests that the donkey vote may have played a party in the weaker Liberal preference flows to One Nation.

And it has to be remembered that preference flows from voters on the right of politics have always been weaker than from voters on the left. One Nation and the two Coalition parties rarely achieve the sorts of flows produced by the Greens in Farrer. Despite polling only 2.3% of the vote, Greens preferences flowed 91.8% to Independent Michelle Milthorpe. At general elections, Green voter preferences to Labor are consistently above 85%, a figure that the Liberal and National parties struggle to achieve from their supporters in three cornered contests.

I haven’t had time to pull apart the data by polling place, but the swing to One Nation was smaller in Albury, by far the largest regional centre in the seat. It was also in Albury where Liberal and Milthorpe support was strongest. Farley was better known in the west of the electorate where National support was also strongest. It may be the different preference flows reflect this familiarity with local candidates and the weaker One Nation support in Albury.

There’s more data on the preferences inside the post, along with an effort to graph the preferences data.

Final Composition of the Two-Candidate Preferred

The stacked bar chart below shows the composition by source of the final two-candidate preferred count for One Nation’s David Farley and Independent Michelle Milthorpe.

I think the chart illustrates that it was the first preference gap between Farley and Milthorpe that largely determined the result. National and Liberal preferences boosted Farley’s margin, but he was far enough ahead that even a different decision on Coalition preference recommendations would have been unlikely to change the result.

Charting Preference Distributions and Flows with Step Line Charts

The chart below plots the progressive preference count based on exclusions during the distribution of preferences. By this measure, One Nation did not reach a majority until the Liberal candidate was excluded. The chart’s horizontal axis shows the total votes distributed in total up to each exclusion. The votes for the four highest polling candidates nudge upwards with each exclusion as the total for Other dips down before the National and Liberal candidates are finally excluded. I’ve opted to use a step chart rather than a traditional line chart as I think it shows the preference distribution more clearly.

Based on the flow data you get a different picture. Again I’ve used four separate candidates, Milthorpe, One Nation, Liberal and National with the other eight candidates shown as ‘Others’.

This chart shows that based on FLOWS, One Nation’s David Farley reached 50.2% before the flows of the Liberal candidate are added to the count. For those who have been saying that it was Liberal preferences that elected Farley, this is technically not correct. National preference flows elected Farley, preferences from the Liberal candidate were not required.

In the end preferences merely confirmed a victory for David Farley which was set up by him leading the primary count with 39.6%, well ahead of his Independent challenger Michelle Milthorpe who began the count on 28.1%.

Charting Preference Distributions and Flows with Stacked Column Charts

Another way of representing the preference data is by using stacked column charts. The first version below is based on the formal distribution of preferences.

I think stacked column charts like this work better when a second or lower placed candidate wins. See my version for the recent Stafford by-election and also for the fourth placed victory in Finniss at the South Australian election.

Here’s the same chart style using preference flows. I don’t think this flow based version works as well as the distribution of preferences version above.

I have seen these results illustrated with ‘alluvial’ charts and ‘sankey’ charts, but they also run into problems with trying to show both preference totals and preference flows in the one chart. Different graph types highlight either counts or flows, but showing both is not easy.

Preference Flows

The previous charts highlight the vote for candidates and parties at each stage of the count, but gives less attention to the flows involved.

The graph below is commonly used to show the percentage of preference flows completed by supporters of parties and candidates. All bars in the chart add to 100%. The chart highlights what first preference voters for each excluded candidates think of the final two candidates. But it doesn’t explain which preferences were important. Only 8.2% of Greens preferences flowed to One Nation compared to 69.1% from the Nationals, but the Nationals having four times the primary vote made its preferences more important in determining the winner.

The version below shows the data differently and breaks down the final two-candidate preferred percentage (2CP%) by source. The first line is the 2CP%, then the primary vote percentages, total percentage preferences received and a break down of the percentage received by each candidate. This chart is better at showing the composition of the final result by putting less emphasis on the percentage flow by party. It’s a work in progress I might use in future.

Finally, below is another version showing the lead of Farley with One Nation’s gain or loss on the exclusion of each candidate. I suspect this chart would have been more interesting in a contest where the leading candidate had a smaller initial lead.