SenateDB Senate ballot analysis

SenateDB

SenateDB is a tool for performing analysis of formal ballots cast for the Senate during the 2016 Australian federal election. The source code and instructions for running it are available on GitHub.

General notes

The raw data

These reports are composed by the SenateDB tool from raw data made available on the AEC’s website. In particular, they rely on the large csv files on that website containing the preferences on every formal ballot paper at the 2016 federal election.

Formal ballots

The tables in these reports are constructed based on the formal ballot papers included in the AEC data. The 567,806 informal ballots cast are not available in the AEC data, and so are not included in this analysis.

Formality and savings provisions

For each ballot, the AEC data includes the number (or tick or cross) written in each square on the ballot paper. We need to apply the rules in the Electoral Act in order to retrieve the actual candidate order used in the count.

This process is performed by the BallotNormaliser class. Broadly, it applies 5 processes:

  1. Converts ticks or crosses to a ‘1’ in the marked square, as per sections 269(1A)(a) and 268A(2)(a) of the Act.
  2. Truncates preferences expressed above the line at any counting errors after the first preference, as per section 269(1A)(b) of the Act.
  3. Truncates preferences expressed below the line at any counting errors after the sixth preference, as per section 268A(2)(b) of the Act.
  4. Distributes preferences expressed above the line to the candidates below the line in the preferenced groups, as per section 272 of the Act
  5. If, after the above steps, the ballot is formal both above and below the line, we use the preferences expressed below the line, as per section 269(2) of the Act. Otherwise we use the preferences expressed above the line.

Tabulating by first-preference

Many of the reports tabulate by ballots’ first-preferences. Doing this at a state level is relatively simple: we simply tabulate by the group that received the first preference (either above or below the line).

Aggregating first-preferences nationally requires doing so by party rather than by group. This requires some judgement calls to ensure the results both accurately reflect the ballots, and are not overshadowed by corner-cases.

SenateDB uses the following process:

  1. Above-the-line ballots use the party associated with the group that received the first preference. If the group is not associated with a party (eg group B in Victoria at the 2016 election), the first preference is marked as “Independent”.
  2. Below-the-line ballots use the party associated with the candidate that received the first preference. Note that this is not always the same as the party of the associated group. For example, group E in Victoria at the 2016 election was associated with the party “SCIENCE PARTY / CYCLISTS PARTY”, but Luke James (the first candidate in the group) was associated with the “Science Party”. Again, if the candidate is not associated with a party, the first preference is marked as “Independent”.
  3. State-specific versions of national parties are rolled into their national equivalents (eg “Australian Labor Party (Northern Territory) Branch”) ballots are rolled into the “Australian Labor Party” count.

The Coalition

The above process leaves a lot of ambiguity regarding how votes for the coalition in different states should be aggregated. In their aggregate tables, the AEC seems to have decided to keep the counts for the different constituent parties separated. This principle is even applied to the Country Liberals(NT), who are listed separately. On the other hand, the WA Greens are added to the national Greens count.

When aggregating nationally, SenateDB combines all votes for Coalition parties into the “Liberal Party of Australia” count, except where those votes were specifically cast for The Nationals. For example, a vote whose first preference was above the line in group AF (Liberal / The Nationals) in Victoria at the 2016 federal election would be included in the “Liberal Party of Australia” count. A vote whose first preference was below the line for Bridget McKenzie (in the same group) would count toward “The Nationals”, as that is the party listed next to her name on the ballot paper.

Findings for the 2016 Election

Check out the full reports for more data and notes on methodology.

Donkey votes

Full report

Nationally, donkey votes were a tiny fraction of the total.

  Donkey votes %
Total 20,553 0.15%

The only candidate elected to the Senate from group A was Derryn Hinch in Victoria. Only 7.74% of his primary vote came from donkey votes.

How-to-vote card usage

Full report

There were large variations in how-to-vote card usage by party. In particular, a high proportion of Liberal/Coalition voters used a how-to-vote card.

Party Ballots matching an HTV card Total formal ballots for party %
Liberal Party of Australia 1,322,391 4,821,314 27.43%
Australian Labor Party 522,915 4,123,084 12.68%
The Greens 112,670 1,197,657 9.41%
Pauline Hanson’s One Nation 17,617 593,013 2.97%
  Ballots matching an HTV card %
Total 2,010,114 14.53%

Ballots that marked only ‘1’ ATL

Full report

2% of the population still voted ‘1’ above the line, despite new rules in 2016 requiring them to mark at least 6 squares.

  Ballots with only ‘1’ above the line %
Total 290,758 2.10%

Ballots saved from being informal by savings provisions

Full report

More than a million ballots were saved from being informal by savings provisions in the Electoral Act.

Savings provision Ballots saved by provision %
Insufficient squares numbered above-the-line 913,730 6.60%
Counting error above-the-line 442,132 3.19%
Insufficient squares numbered below-the-line 66,895 0.48%
Counting error below-the-line 44,915 0.32%
Used tick for first preference 28,247 0.20%
Used cross for first preference 23,415 0.17%
Total 1,046,837 7.56%

Votes above the line

Full report

Despite changes to the Electoral Act making it easier to vote below the line, the vast majority of Australians continued to vote above the line, preferencing parties rather than particular candidates.

  Votes above the line %
Total 12,940,784 93.51%

Bucking the national trend, Tasmanians were much more likely to vote below the line. No doubt this is related to the extraordinary election of Lisa Singh, who was elected on below-the-line votes despite being the last candidate on the ALP ticket.

State Votes above the line Total formal ballots for party %
TAS 243,942 339,159 71.93%
ACT 216,278 254,767 84.89%
NT 93,307 102,027 91.45%
SA 971,322 1,061,165 91.53%
QLD 2,556,482 2,723,166 93.88%
WA 1,291,224 1,366,182 94.51%
NSW 4,252,904 4,492,197 94.67%
VIC 3,315,325 3,500,237 94.72%
Total 12,940,784 13,838,900 93.51%

Votes above and below the line

Full report

A very small number of Australians voted both above and below the line.

  Ballots formal both above and below the line %
Total 163,141 1.18%

Exhausted votes

Full report

There were more than 1 million exhausted votes in the 2016 Senate election. Votes whose first preference was for a minor party contributed the vast majority of these.

Party type Exhausted votes Total formal ballots %
Minor parties 906,319 3,626,476 24.99%
Major parties 124,754 10,188,987 1.22%
Independents 11,060 23,437 47.19%
Total 1,042,132 13,838,900 7.53%

Exhausted ballots

Full report

Nearly 50% of ballot papers eventually exhausted. Many of these exhausted after electing a candidate, and so contributed very little to the above tally of exhausted votes. Take a look at the exhausted votes report for an explanation of this distinction.

  Exhausted ballots Total formal ballots %
Total 6,865,733 13,838,900 49.61%

Feedback

If you have any feedback or suggestions, don’t hesitate to contact @tmccarthy_ on Twitter, or raise a ticket on GitHub. Pull requests are welcome.