Preference Flows at the 2018 South Australian Election and the Influence of How-to-Votes

The 2018 South Australian election saw a record vote for minor parties. This was largely due to the campaign by Nick Xenophon and his SA-Best party, polling 14.2% in the House of Assembly, a creditable 18.4% in the 26 seats it contested. The failure of the party to poll as strongly as published polling well out from the election suggested, or to elect a member to the Assembly, saw its campaign labelled a failure by political commentators.

In the end the party was used by voters as a conduit for preferences to the Liberal and Labor Parties. As you would expect for a party viewed by voters as sitting in the political centre, the party’s preferences split evenly, 51.6% to the Liberal Party and 48.4% to labor.

The release of preference flow data by the SA Electoral Commission provides an opportunity to analyse preference flows against party preference recommendations. Several unique features in the conduct of SA House of Assembly elections allows the comparison of preference flows with how-to-votes lodged by candidates and displayed in voting compartments.

The Display of How-to-votes on South Australian Voting Screens

Across Australia, (Tasmania and the ACT excepted), how-to-vote material is distributed outside polling places by volunteers working for candidates and parties. The material is designed to attract first preferences, to help ensure ballot papers are completed correctly, and in some instances to influence preferences. (See my previous post on why candidates distribute how-to-vote material.)

While voters are permitted to bring the material into the polling place as an aid to voting, effort is made by polling place staff to ensure the material is not left in the polling place. Voters are encouraged to put the material in bins or return it to volunteers outside of the polling place for re-use. Any how-to-votes left in voting compartments are removed by polling place staff.

Something different occurs in South Australia. The distributed how-to-vote material is still removed, but a special provision of the SA Electoral Act allows candidates to register how-to-votes for display in the voting compartments. The how-to-votes are displayed on the screen in front of voters, as shown below.

HTVsDisplayed

The HTVs for the district of Adelaide at the 2018 South Australian election are shown below.

Screen-HTVs

How-to-votes (HTVs) are displayed left to right and in ballot position order, in rows if there are more than five candidates. Candidates and parties are limited to basic text and colour with pictures not allowed. Candidates can lodge 1-only HTVs, or split HTVs, as long as the how-to-vote indicates that all boxes must be numbered. Candidates can choose to leave party affiliations off the list of candidates. Candidates that don’t lodge a HTV have their name shown in ballot position order with the message “no how-to-vote lodged”.

Every voter at a voting centre within their home district will see the registered how-to-votes for their district arrayed in front of them when voting. Absent voters, postal voters, and out of district pre-poll voters will not see a HTV for their district, but all polling places and pre-poll voting centres have available a booklet containing the HTVs for each district.

This widespread access to HTVs makes South Australia a  useful test case in assessing the impact of HTVs. In other jurisdictions, without knowledge of where HTVs were distributed, it is difficult to assess their effectiveness.

Preference Distributions versus Preference Flows

Preference data from Australian elections comes in two forms.

Preference Distributions show the transfers of ballot papers from excluded to continuing candidates as each candidate is excluded in ascending order of vote. This produces an inverse tree diagram as preferences flow from their first preference source to their two-candidate preferred destination.

The table below shows the formal distribution of preferences for Adelaide at the 2018 South Australian election.

Adelaide-Distribution-2018

Source: Electoral Commission SA Statistical Return.

From this table, we know that of the 946 votes for Price (Dignity Party), 192 had second preference for Chapley (Labor), 205 for Sanderson (Liberal), and 549 for Simms (Green). What we don’t know is the third preference of those 549 ballot papers transferred to Simms, that preference being the final destination choice of Price voters between Chapley and Sanderson.

Similarly, the 3,288 votes for Simms distributed at the second exclusion includes 549 second preferences from Price. It is not possible from the above table to work out the destination preference of the 2,739 first preference votes for Simms because of the inclusion of Price preferences after the first exclusion.

What we need is preference flow data, counts of preferences that are tallied by their source, that is first preferences, and their destination, which of the final two candidates received the preferences.

Thankfully, South Australia is one of the few states where this information can be obtained. The preference flow data is not part of the SA Electoral Commission’s formal results reporting process, but can be accessed well after the election for research purposes.

An example of this data is shown in the table below. The data is derived from the same Adelaide ballot papers that produced the distribution of preferences, but the tallying is by pairings of candidate first preference and candidate destination preference.Adelaide-Flows-2018

Source: Calculations by Antony Green based on raw data released by the Electoral Commission South Australia.

The preference distribution table did not allow the Price and Simms preference data to be disentangled. The preference flow data above does.

From the flow data, we know that 63.2% of Price’s votes flowed to Labor, and 36.8% to Liberal. We can also work out that of the 549 Price votes with second preference for Simms, 406 gave a third preference to Chapley, and 143 to Sanderson.

Similarly, the flow data allows us to know that of the 2,739 first preference votes for Simms, 82.9% flowed to Labor as preferences, and 17.1% to the Liberal candidate.

In the flow table I’ve also noted that the “donkey vote” from Price at the top of the ballot paper favoured Labor. In the “HTV” column I’ve also noted that for both candidates their HTVs recommended preferences for Labor.

In this post I’m going to compare the preference flow data with the HTVs lodged by each party. As we know most voters had access to these HTVs on the screen in front of them, the flow data gives us a measure of the influence of HTV preference recommendations.

Measuring How-to-Vote Concordance

I define HTV concordance as where a  completed ballot paper exactly matches the HTV for the candidate that received the ballot paper’s first preference.

The preference flow data used in this post does not measure concordance. All it measures is pairings of first preference source and final preference destination, not the path between the two. Preference flow data measures whether a destination preference was influenced by a HTV recommendation, not whether a voter copied the HTV sequence.

The SA Electoral Commission conducts research on concordance. In a survey of 17 districts at the 2018 election, the Commission found 37.7% of ballot papers matched lodged HTVs. This was compared to survey figures of  42.8% in 2014 and 42.1% in 2010. The Commission did not break down the data by party in 2018.

The Commission did publish concordance detail by party in 2010 and 2014. The data for the four main parties contesting those two elections is shown in Table 1.

Table 1 – Ballot Paper Concordance with HTVs – 2010 and 2014 SA Elections

% Concordance with HTV
Party 2010 2014
Labor Party 41.4 41.1
Liberal Party 51.1 48.2
The Greens 14.4 26.9
Family First 25.9 29.7

Source – Electoral Commission SA, State Election report 2014, pp62-65

The above data disproves claims that voters blindly follow how-to-votes. The overwhelming majority of voters at both elections had how-to-vote recommendations in front of them, but less than half of voters followed them.

That still leaves open the question of whether voters were influenced by the how-to-votes. For that, let me turn to analysing the 2018 preference flow data.

Overall Preference Flows in 2011

Two-party preference flow data counting out preferences between the Labor and Liberal candidates was collected for all 47 districts, though data for the rural electorate of Narungga was incomplete. There are minor discrepancies between the flow data and the official preference counts, but this is due to the flow data being collected separately from the official count.

Flow data was collected for another 12 two-candidate contests, seven between Liberal and SA Best, two between Labor and SA Best, two Liberal versus Independent contests and one Labor versus Independent contest. Two candidate flows were not collected for Labor-SA Bests contests in Elizabeth, Port Adelaide or Ramsay.

The table below shows the total of two-party preferred preference flows by party.

Table 2 – House of Assembly Preference Flows by Party

Party Candi-
dates
First
Pref %
% Prefs
to labor
% Prefs
to Liberal
Nick Xenophon’s SA Best 36 14.1 48.4 51.6
The Greens 47 6.7 74.5 25.5
Australian Conservatives 33 3.0 27.6 72.4
Dignity 30 1.5 57.0 43.0
Animal Justice 4 0.3 58.8 41.2
Others/Independents 20 3.6 44.0 56.0
Total Preferences 170 29.2 52.3 47.7

Source: Calculations by Antony Green based on raw data released by the Electoral Commission South Australia.

Nick Xenophon’s SA Best lodged HTVs recommending a first preference for the party’s candidate, but made no recommendation for further preferences. When counted out, preferences favoured the Liberal Party in 26 districts with a strongest flow of 59.0%, and in 10 they favoured Labor with a strongest flow of 57.0%.

The Greens lodged 36 HTVs with a preference recommendation for Labor over Liberal, six with no recommendation and in five districts did not lodge a HTV.

The Australian Conservatives lodged 26 HTVs recommending preferences for the Liberal candidate, five with recommended preferences to Labor and two with split HTVs.

Dignity lodged HTVs in 18 districts with a preference recommendation for Labor, and in another 12 districts recommended preferences to Liberal.

Green Preference Flows

Table 3 – House of Assembly – Green Preference Flows

How-to-vote Type (Districts) First Pref % % Prefs to labor % Prefs to Liberal
Recommended to Labor (36) 6.9 76.2 23.8
1-only, no recommendation (6) 5.1 67.6 32.4
None lodged (5) 6.1 67.5 32.5
Total (47) 6.6 74.5 25.5

Source: Calculations by Antony Green based on raw data released by the Electoral Commission South Australia.

Preference flows to Labor were 76.2% in the 26 districts with a preference recommendation for Labor, and 67.6% in 11 districts where there was no recommendation.

This suggests there was a difference created by the HTV recommendation, but some of the difference is accounted for by the electorates with no HTV recommendation generally having a lower Green first preference vote. The data is weighted towards contests where the Greens polled strongly and where there was a preference recommendation for Labor. (At some point I hope to conduct some regression analysis to clarify this observation.)

Without an example of the Greens recommending preferences to the Liberal party, it is difficult to know whether the Greens are able to induce a strategic as opposed to ideological preference flow. Green preferences across Australia seem to be very much based on ideology and Green voters tend to direct preferences to Labor irrespective of the Green HTV. The February 2020 Johnston by-election in Darwin was a rare example of Green preferences being directed against the Labor Party.

Australian Conservatives Preference Flows

Table 4 – House of Assembly – Australian Conservatives Preference Flows

How-to-vote Type (Districts) First Pref % % Prefs to labor % Prefs to Liberal
Recommended to Liberal (26) 4.0 22.8 77.2
Recommended to Labor (5) 4.0 48.5 51.5
Split how-to-vote (2) 5.6 32.9 67.1
Total (33) 4.1 27.6 72.4

As the successor party to Family First, the Australian Conservatives inherited its predecessor’s habit of directing preferences to Labor in a small number of SA seats, usually to socially conservative members of Labor’s Right faction.

Preferences were recommended to Labor in Florey, Lee, Playford, Port Adelaide and Ramsay, and split in Light and Newland.

The recommendations clearly made a difference, only 22.8% of preferences flowing to Labor where the recommendation was to Liberal, flowing 32.9% to Labor on split tickets, and 48.5% to Labor where recommended to Labor.

Dignity Preference Flows

Table 4 – House of Assembly – Dignity Preference Flows

How-to-vote Type (Districts) First Pref % % Prefs to labor % Prefs to Liberal
Recommended to Labor (18) 2.4 59.5 40.5
Recommended to Liberal (12) 2.2 53.0 47.0
Total (30) 2.3 57.0 43.0

With a supporter base less driven by ideological fervour than the Greens or Australian Conservatives, and with a lower first preference vote, random factors and voter’s own choice produced weaker flows of preferences for the Dignity Party. The gap in flows was only 6.5% between seats with a Labor recommendation and those with Liberal recommendation.

In only five districts did Dignity preferences favour the Liberal Party, and in two of those the recommendation had been to Labor.

Labor Preference Flows.

There were nine districts where Labor preferences were distributed. Seven were Liberal-SA Best contests in rural seats (Chaffey, Finniss, Hammond, Heysen, Kavel, Mackillop, Narungga) and two were Liberal-Independent contests in regional seats (Frome, Mount Gambier).

Labor recommended preferences to SA Best in seven contests producing preferences flows between 65.8% in Chaffey and 79.9% in Finniss. Labor recommended preferences to Independent Geoff Brock in Frome and 82.3% of Labor voters followed the recommendation.

With criminal charges over Mount Gambier Independent Troy Bell, Labor recommended preferences for the Liberal candidate and 53.2% of Labor voters followed that recommendation.

Liberal Preferences

Preference flow data for excluded Liberal candidates was captured in three districts. Two were Labor-SA Best contests, Giles and Taylor, and the third was the Labor-Independent contest in Florey. The Liberal Party recommended preferences against Labor in all three contests.

Liberal preferences flowed 74.3% to SA Best in Giles and 69.8% in Taylor. Liberal preferences flowed 80.1% to Independent Frances Bedford in Florey.

Green Preferences in SA Best Contests

There were five electorates where the Greens made preference recommendations in Liberal-SA Bests contests.

In four the Greens recommended preferences to the Liberal Party ahead of SA Best. Only in Chaffey did a majority of voters follow the recommendation with 50.7% of preferences flowing to the Liberal candidate. In the other three contests, preferences flowed to SA Best 57.3% in Finniss, 64.0% in Heysen and 59.9% in Kavel.

The Greens recommended preferences to SA Best in MacKillop and 63.2% followed the recommendation. In Mount Gambier, 64.5% of Green preferences flowed to Independent Troy Bell despite no preference recommendation being made. In Frome, again with no recommendation but with a donkey vote favouring the Independent, preferences flowed 72.5% to Independent Geoff Brock.

Interestingly the Green preference flows to Labor from the same ballot papers were Chaffey 64.4%, Finniss 73.8%, Heysen 82.6%, Kavel 77.5% and Mackillop 66.9%. This suggests Green voters might have cared more about their preferences reaching Labor than any choosing between other parties.

Interestingly, only 55.6% of Green preferences reached Labor first in Frome, much lower than reached Brock. Two-factors were at play here. First, the Greens issued a 1-only how-to-vote. Second, the Liberals benefited from the donkey vote. Perhaps Green voters cared little after preferencing Brock, knowing full well he would be in the final two-candidate contest rather than Labor.

In Conclusion

Overall it appears that how-to-votes had only limited impact on how voters filled in their ballot papers, and mixed influence on how voters completed their preferences.

The concordance data shows clearly that voters do not blindly follow how to vote cards. Less than half of voters transcribed the sample HTV recommendations from the voting screen on to their ballot papers.

For the Greens and Dignity voters, HTV recommendations had only marginal impact on how voters completed their ballot papers. Australian Conservative voters seem to have been swayed more by preference recommendations in the limited number of districts where preferences were recommended for Labor.

For Green voters, there were much stronger flows of preferences to Labor in the two-party contest against the Liberal Party than was produced by the choice between Liberal-SA Best candidates. That, and the flows in Frome, suggest that maybe Green voters cared more about higher and more ideologically defined choices than they did about preferences further down the ballot paper..