2019 SA Senate Election – Ballot Paper and Preferences Analysis (Part 1)

This is my latest (and much delayed) post looking at the Senate’s new electoral system, how it worked at 2019 election, how voters completed their ballot papers, and what was the influence of how-to-vote material.

This post on South Australian is my first on a six-member Senate contest. The two previous posts in this series, on the ACT Senate race and the Northern Territory Senate race, dealt with elections for two Senators and were two-party preferred races.

At the start of 2020 I published two other posts with broad overviews of the 2019 result. The first looked at the breakdown of above and below the line voting and the number of preference completed. The second looked at measures of performance, why the new system produced different results to past Senate elections. In particular, it compared the 2019 result with the 2013 half-Senate election, the last conducted using the now abolished group voting tickets.

Part 2 of this post tracks the formal distribution of preferences for the South Australian Senate. The Part 1 post below provides a more detailed analysis of preferences based on examining the electronic ballot papers.

Definition of terms

In this blog I am re-introducing several terms I used in a 2014 post analysing below-the-line preferences in the 2013 WA Senate election.

The first is “concordance“, whose dictionary definition is simply “agreement”. Concordance is also used with that meaning in several areas of mathematics and statistics. I use concordance to describe agreement between how-to-vote (HTV) recommendations and the preferences written on ballot papers by voters. If a high proportion of voters complete preferences that match their first chosen party’s HTV, I say there is a high rate of concordance with the HTV.

The next is “affinity preference“. The dictionary definition of affinity is a natural liking for or attraction to a person, thing or idea. An affinity preference is a second or further preference given by a voter to another party that the voter sees as having an affinity or closeness to the first choice party. Aggregating preference data provides a measure of the affinity voters see between parties. If a high proportion of preferences flow from one party to another, it indicates that many voters for the first party see an affinity with the second party, and so on with further preferences. As we are discussing political parties, affinity is a soft measure of ideological closeness between parties as perceived by voters.

A “proximity preference” is created by ballot paper structure. It is a preference caused by a party’s proximity or closeness to another on the ballot paper. You don’t have to spend much time looking at preference data to note that two parties close to each other on a Senate ballot paper will have stronger preference flows between them than if they were at opposite end of the ballot paper. The most obvious proximity preference is the “donkey vote” flow of preferences from column A to column B, C and so on. It is a weaker and more random preference flow than an affinity preference.

An “effective preference” is a measure of preference destination, ignoring passage through other parties along the way. In the lower house an effective preference is the choice made between the final two candidates in the contest. In the Senate, it is a measure of where a preference finishes between the parties remaining towards the end of the count. It may be that a voter places greater emphasis on their actual 2nd, 3rd etc preferences, but it is the effective preference reached later in the count that is the one that counts.

And I should add a couple of abbreviations. ATL means ‘above-the-line’, a preference vote completed and counted using the party voting squares above the black dividing line on the ballot paper. BTL means ‘below-the-line’, a vote completed using the candidate voting squares below the black dividing line. HTV means ‘how-to-vote’, the generic name for material distributed on-line and outside polling places suggesting an order of preferences to voters. All HTVs suggested ATL preference sequences.

Above and Below the Line Voting

To analyse how voters completed their ballot papers using the Senate’s new electoral system, I have accumulated the AEC’s South Australian Senate ballot paper data into six broad categories. These are –

  • 1-only above-the-line (ATL) votes – the ballot paper instruction for ATL Senate voting 1984-2014. These votes remain formal through savings provisions.
  • 2-5 ATL preferences – a ballot paper with fewer than the instructed minimum six preferences but formal by savings provisions.
  • 6 ATL preferences – the minimum number of ATL preferences stated in the ballot paper instructions.
  • 7-12 ATL preferences – beyond the minimum 6 and up to the 12 preferences stated as instructions for a BTL vote.
  • Greater than 12 ATL – from 13 to the maximum number of ATL squares on the ballot paper, 16 in South Australia.
  • Below-the-line (BTL) – any ballot paper admitted to the count as a formal below-the-line vote.

Table 1 sets out the percentage of ballot papers in each category by ballot paper party/column. The party name and column code are shown along with the party’s first preference percentage. State and national totals for each category are shown in the final two lines. Parties are ordered in descending order of percentage 6-preference ATL ballot papers.

Table 1 – 2019 SA Senate – Ballot Paper Categories

% of ballot papers in ATL/BTL category
Party (Group) % Vote 1 2-5 6 7-12 >12 BTL
Liberal (G) 37.84 3.4 2.3 83.7 3.4 2.6 4.6
Labor (O) 30.39 3.1 3.1 82.0 3.4 3.1 5.3
United Australia (K) 3.03 2.7 2.9 79.1 5.3 4.8 5.2
Animal Justice (P) 1.87 3.3 3.4 76.4 4.5 4.0 8.5
One Nation (E) 4.87 3.4 3.4 76.2 5.5 4.7 6.7
HEMP (I) 2.13 3.6 4.4 74.6 5.2 4.1 8.1
Shooters, Fishers and Farmers (M) 1.10 3.4 3.8 73.5 5.3 4.7 9.3
Greens (J) 10.92 1.3 2.0 72.1 3.6 3.7 17.2
Liberal Democrats (N) 0.67 5.5 4.4 72.0 6.4 4.8 6.8
Centre Alliance (C) 2.60 1.0 1.8 71.4 3.8 3.9 18.0
Sustainable Australia (H) 0.48 3.3 3.0 69.6 5.6 4.5 14.0
Australian Democrats (D) 0.59 3.5 3.4 68.1 5.2 4.7 15.2
Australian Conservatives (L) 1.48 1.9 1.7 64.9 4.2 4.3 23.0
Citizens Electoral Council (F) 0.15 5.9 4.6 63.8 6.5 3.9 15.3
Fraser Anning’s CNP (B) 0.72 1.7 3.0 63.3 5.4 6.4 20.2
Great Australian Party (A) 1.16 4.7 5.0 62.4 7.3 5.3 15.3
Ungrouped 0.09 .. .. .. .. .. 100.0
SA Totals .. 3.0 2.7 79.6 3.8 3.3 7.6
National Totals .. 3.5 3.5 80.0 4.5 1.2 7.3

Note – Calculated by author from AEC formal preference ballot paper files. Note there is no ATL voting square for the Ungrouped column.

The highest rate of 6-preference ATL voting was for the two major parties followed by the United Australia Party then several the larger minor parties. Of the larger parties, the Greens at 72.1% had the lowest rate of 6-preference ATL voting and a much higher rate of BTL voting at 17.2%. The lowest rate of 6-preference ATL voting and highest rates of BTL voting were recorded for smallest parties. There is a clear inverse relationship between the rates of 6-preference ATL voting and BTL voting by party with little drift into other categories of ballot paper completion.

There are only minor differences in other categories of votes. The Greens, Centre Alliance and Australian Conservatives had much lower rates of votes that stopped before 6 ATL preferences, the Liberal Democrats, Citizens Electoral Council and Great Australian Party much higher. There is not much variation by party in tendency to go beyond 6 ATL preferences.

BTL ballot papers completed with the recommended minimum 12 preferences represented 61.9% of BTL votes or 4.7% of all votes (ATL + BTL). Savings provisions allowed votes with between 6 and 11 BTL preferences to remain formal, representing 4.9% of BTL votes or 0.4% of all votes. BTL votes with a full sequence of 42 preferences represented 14.9% of BTL votes and 1.1% of all votes. Ballots with between 13 and 41 preferences represented 18.3% of BTL votes or 1.4% of all votes. The total of ballot papers with a given number of preferences beyond 12 decreased as the number of preferences increased. The trend reversed slightly as the maximum number 42 approached. (See graph below)

There were only minor differences by party in the number of BTL preferences completed. Voters for the Greens and middle-size minor parties were slightly more likely to give more than 12 BTL preferences. 17.0% of Green BTL voters completed all 42 squares compared to 14.4% of Labor BTL voters and 12.4% of Liberal BTL voters. But as a percentage of ATL and BTL votes for each party, the percentages were Greens 3.9%, Labor 0.9% and Liberal 0.7%. It was the decision to use BTL voting that was the most significant difference in votes by party, not how voters completed their ballot paper below the line.

The number of formal ballot papers with a given number of BTL preferences is shown in the graph below. Note that the vertical axis uses a logarithmic scale.

There were 29,732 ATL ballot papers where voters numbered all 16 squares, 2.9% of ATL votes and 2.7% of all votes. By party the lowest rates of all square numbering were the two major parties, Liberal 2.1% and Labor 2.5%. The rate for the Greens was 3.3%, the highest rate 5.3% for Fraser Anning’s Conservative National Party.

There were 229 donkey votes numbering columns A to P in sequence left to right, and 18 reverse donkey votes numbering right to left.

How-to-Vote Concordence

For the 2019 election I was able to locate how-to-votes for 13 South Australian Senate parties. Four recommended only a first preference and left it to voters to determine further preferences. Nine included recommendations for second and further preferences. You can view the how-to votes I use in my analysis at this link.

The table below summarises concordance rates by preference number for ATL votes cast by party. To explain, the first entry for the Liberal Party says 40.0% of all Liberal ATL votes had the same second preference as the Liberal HTV, 30.7% the same 2nd and 3rd preference, 28.4% the same 2nd, 3rd and 4th preference, and so on. The ‘>6’ category is ballot papers that matched the six preferences and then went on to give further preferences.

Table 2 – 2019 SA Senate – HTV Concordance by Party

% of ATL Votes Matching Party HTV by Preference Number
Party (Group) ATL Votes 2 3 4 5 6 >6
Liberal (G) 395,058 40.0 30.7 28.4 27.1 26.7 0.4
Labor (O) 314,876 51.1 27.3 20.4 19.8 17.6 0.2
Greens (J) 98,903 23.0 11.3 8.7 8.2 8.0 0.2
One Nation (E) 49,718 2.1 1.2 1.0 1.0 1.0 0.0
United Australia (K) 31,472 28.4 8.2 5.6 5.1 4.7 0.1
Animal Justice (P) 18,713 32.4 10.6 3.4 3.0 2.6 0.0
Great Australians (A) 10,757 1.4 0.5 0.5 0.4 0.4 0.3
Fraser Anning’s CNP (B) 6,248 10.6 3.6 3.3 2.7 2.6 0.2
Australian Democrats (D) 5,507 6.1 1.1 0.5 0.4 0.3 0.2

Note – Calculated by author from AEC formal preference ballot paper files and using available how-to-votes. Percentages are calculated as a percentage of ATL votes, NOT of all votes.

The two major parties had by far the highest rate of concordance, no doubt due to the two parties distributing the most how-to-votes. 26.7% of Liberal votes had the same 6-preference sequence as the party’s HTV, Labor 17.6%, the next best the Greens on 8.0%.

An obvious point to make is that if a voter does not receive a HTV for their first choice party, the chance of their vote matching the HTV sequence is very low. A voter might guess the correct 2nd or 3rd preference, but the odds of guessing all preferences from 2-6 is low. It is obvious that six preference concordance will be low for parties that distribute few how-to-votes. Yet an interesting feature of Table 2 is that some parties have very low concordance to the second preference.

It is clear from Table 2 that supporters of some parties were not able guess the recommended second preference of their party. Table 3 breaks down the highest rate of ATL second preferences by party. An asterisk (*) indicates the second preference recommendation on each party’s HTV.

Table 3 – Second Preference Choices by Voters – 2019 SA Senate Election

  • Liberal (Group G) – *40.0% United Australia 8.7% Liberal Democrats, 8.3% Labor
  • Labor (O) – *51.1% Greens, 9.1% Liberal
  • Greens (J) – 41.5% Labor, *23.0% Animal Justice
  • One Nation (E) – 19.0% Liberal, 15.7% United Australia, 10.5% Labor, *2.1% Liberal Democrats
  • United Australia (K) – *28.4% Liberal, 22.1% One Nation, 10.8% Labor
  • Centre Alliance (C) – 20.9% Labor, 19.6% Liberal, 19.1% Greens
  • HEMP (I) – 25.8% Greens, 13.8% Animal Justice, 13.3% Shooters, Fishers, Farmers, 12.2% Labor
  • Animal Justice (P) – *32.4% Greens, 16.9% Labor, 11.4% HEMP
  • Australian Conservatives (L) – 47.3% Liberal, 10.0% One Nation
  • Great Australian (A) – 23.6% Fraser Anning’s CNP (donkey vote), 15.9% One Nation, *1.4% Australian Conservatives
  • Shooters, Fishers, Farmers (M) – 15.0% One Nation, 14.4% HEMP, 12,0% Liberal, 11.8% Labor, 11.5% Animal Justice
  • Fraser Annings’s CNP (B) – 44.5% One Nation, *10.6% Great Australians
  • Liberal Democrats (N) – 20.0% Liberal, 18.8% Labor
  • Australian Democrats (D) – 19.1% Greens, 16.8% Liberal, 11.9% Labor, 10.1% One Nation, *6.1% Sustainable Australia
  • Sustainable Austraia (H) – 23.5% Greens, 14.9% HEMP, 11.9% Animal Justice
  • Citizens Electoral Council (F) – 14.4% One Nation, 14.3% Liberal, 11.5% Sustainable Australia

Note – Calculated by author from AEC formal preference ballot paper files. Not all how-to-votes were available and some did not include a second preference recommendation. Percentages are calculated as a percentage of ATL votes, NOT of all votes.

The data in Table 3 suggests higher HTV concordance is best achieved with an obvious sequence of preferences. Liberal, Labor, United Australia and Animal Justice all had second preference recommendations that reflected voter perceptions of affinities between parties. Voters who did not see a HTV could easily guess at least the suggested second preference, boosting HTV concordance.

The Greens distribute more HTVs than parties other than the two majors, yet only 23.0% of voters gave the party’s recommendation of a second preference for Animal Justice. The greatest category of second preferences was 41.5% for Labor. Voters either ignored the HTV recommendation, or the large number of voters who didn’t see the recommendation, chose Labor as their second preference. I suspect both factors played a part. In all work I’ve done on whether voters follow HTVs, the tendency of Green voters to vote 2 for Labor whatever the Green HTV says is one of my most consistent findings.

The data in Table 3 suggests few voters for lower ranked parties saw their chosen party’s HTV. Only 2.1% of One Nation voters gave the recommended second preference to the Liberal Democrats compared to 19.0% for the Liberal Party, 15.7% United Australia and 10.8% for Labor. Only 1.4% of Great Australian Party voters had the recommended second preference for the Australian Conservatives, only 6.1% of Australian Democrat voters for Sustainable Australia. Only 10.6% of Fraser Anning’s Conservative National Party voters gave preferences to The Great Australian Party in the neighbouring column, and a very high 44.5% gave second preferences to One Nation, showing voters thought Anning had some affinity with his former party.

The Australian Conservatives, despite not recommending preferences, produced a a strong 47.3% second preference flow to the Liberal Party. Even without a recommendation, Australian Conservative voters saw an affinity with the Liberal Party and gave second preferences in that direction. The Centre Alliance also did not make a recommendation, and reflecting its more centrist position on the ideological spectrum, produced more even flows of preferences to other large parties. In the case of the Australian Conservatives and Centre Alliance, the preference flows may reflect the affinity of voters to their former party, Conservative voters more clearly derived from one side of politics.

The ‘donkey vote’ as always had an impact on the preference flow from the party in column A. In South Australia this benefited the Great Australian Party. Of the party’s ATL votes, 1,603 (14.9%) numbered 1 to 6 in the first six columns. 225 votes (2.1%) numbered 1 to 16 all the way across the ballot paper. This helped 23.6% of second preferences flow to Fraser Anning’s CNP in column B.

Effective Preferences – Four Party Split

In the five mainland states, only four parties were in the race to win Senate seats – the Liberal Party/LNP/Coalition, the Labor Party, the Greens, and Pauline Hanson’s One Nation. The Jacqui Lambie Network added a fifth party to the contest in Tasmania.

In South Australia, not all four parties were present at the end of the count. Labor polled only 2.1253 quotas and its third candidate was excluded when seven other parties remained in the count. The Green’s Sarah Hanson-Young had her surplus as the fifth elected Senator distributed at a point when four other parties remained in the race for the final seat. At that point it was already clear that the third Liberal candidate, Alex Antic, would win the final seat.

(I explain the final exclusions in the second part of this post summarising the formal distribution of preferences and election of Senators.)

The table below sets out flows of ATL preferences to the final four parties. The first four entries are flows from the final four to the other three parties. The table can be sorted by column. The table only includes ATL preferences, but after carrying out the same analysis on the much smaller number of BTL votes, the preferences flows were roughly the same.

Of the four largest parties, Labor and the Greens had HTV preference recommendations for each other producing strong preference flows. One Nation and the Liberal Party did not swap preferences on HTVs. This generated a weaker flow of One Nation preferences to the Liberal Party, while 41.5% of Liberal voters exhausted their preferences before choosing between One Nation, the Greens and Labor. (Note that these preference flows are based on what voters wrote on ballot papers. In the actual count, Liberal and One Nation candidates competed for the final seat so neither party had their preferences distributed.)

Splits of preferences from other parties make sense if you understand where on the political spectrum parties sit. Centrist parties have preference flows flowing in both directions and a tendency to avoid One Nation. Parties of the right tend to be split on whether preferences reach One Nation or the Liberal Party first.

As already mentioned, only four parties remained after the election of the Green’s Sara Hanson-Young – Liberal, One Nation, United Australia and the Centre Alliance. The Centre Alliance was excluded first, then United Australia, leaving the Liberal Party and One Nation as the last possible destination for an effective preference.

In table 4, the rate of exhausted preferences from the 12 excluded parties was only 7.1% with 26.0% flowing to Liberal, 18.5% Labor, 24.1% Green and 24.3% One Nation. With these four final parties used in Table 4, there is an array of choices for voters arrayed from left to right on the political spectrum. As these were whittled down to two choices, Liberal and One Nation, the rate of exhausted preferences started to rise.

Performing the same Effective Preference analysis but using only Liberal and One Nation, the rate of exhausted preferences rose. Of the 12 excluded parties, preferences flowed 42.3% to Liberal, 32.1% One Nation and 25.6% exhausted. Labor preferences split 34.4% Liberal, 16.1% One Nation and 49.5% exhausted, Green preferences 42.9% to Labor, 7.2% to One Nation and 49.9% exhausted. However, at the end of the count, Labor and Green ballot papers still in the count were at a much reduced transfer value, so the rate of Labor and Green exhaustion was not around half of around 41% of the vote, but around half of 2%.

In the debate on Senate electoral system reform in 2016, there was great concern over how many votes would exhaust preference before the end of the count. In the case of South Australia, only 3.2% of the vote had exhausted by the end of the count. The of exhaustion was much higher at the 2016 double dissolution election.

Analysis of Two-Party Preferred Preferences

Senators for the states are elected by proportional representation, so analysing preferences based on flows to the two major parties is not really appropriate as other parties are in the race to win seats. In addition, the operation of quotas means that Labor and Liberal ballot papers can have their preferences distributed even if the two parties top the poll.

However, as Senate preferences are allowed to exhaust, I thought it worth publishing the Senate two-party preferred preference flows given LNP Senator James McGrath has attepted to start a debate on introducing optional preferential voting for House elections.

As Senate ballot papers recommend filling six-ATL or 12-BTL preferences, and as the overwhelming majority of voters followed the instructions, exhaustion rates under the Senate system will not be the same as for the House where ballot paper would suggest a one-preference minimum.

The flows from third parties to the two major parties are as you would expect for most of the larger parties. For the smaller parties there is a much higher rate of exhaustion. The two parties most clearly aligned at opposite ends of the left-right political spectrum, the Greens and the Australian Conservatives, were the two parties where voters expressed the strongest affinity with the major party on the same side of the political spectrum. Both parties also saw the lowest rates of exhausted preferences.

Table 5 – 2019 SA Senate – Effective 2-Party Preferences

% Preferences of ATL Preference to
Party (Group) ATL Votes Liberal Labor Exhaust
Greens (J) 98,903 14.3 76.4 9.3
Animal Justice (P) 18,713 21.6 49.1 29.3
Centre Alliance (C) 23,294 38.7 47.0 14.2
HEMP (I) 21,375 22.6 43.2 34.3
Sustainable Australia (H) 4,552 25.4 40.9 33.7
Australian Democrats (D) 5,507 39.5 40.4 20.0
Shooters, Fishers and Farmers (M) 10,892 36.5 31.1 32.4
Liberal Democrats (N) 6,843 52.5 29.0 18.5
One Nation (E) 49,718 41.2 27.7 31.1
United Australia (K) 31,472 53.0 26.1 20.9
Citizens Electoral Council (F) 1,365 31.4 22.0 46.7
Great Australian Party (A) 10,757 34.8 21.4 43.7
Fraser Anning’s CNP (B) 6,248 45.6 14.3 40.1
Australian Conservatives (L) 12,437 76.8 10.8 12.4
Total 302,076 32.0 46.7 21.3