Reading public evidence

1967

Bandler’s Campaign: Two Referendum Questions

Separate two ballot questions and test two invented cross-tables to see why aggregate approval figures do not reveal how each person voted.

Two invented cross-tables share the same margins but different joint votes.
3
Evidence distinctions

The short answer

A shared polling day does not make two ballot questions the same proposal, nor does it reveal how individuals paired their answers.

Record and inference

Keep the questions separate

Each input establishes a different kind of evidence.

A limited input cannot establish every proposed conclusion.
InputWhat followsWhat does not follow
Recorded marginsEach question’s published aggregate result.Individual paired choices are unknown.
Original cross-tablesTwo joint patterns fit the same margins.Neither is recovered polling data.
Campaign recollectionAn organiser’s retrospective account.It is not a cross-tabulation of ballots.

Chronology

A timeline of the turning points

  1. Separate ballot outcomes

    Parliament proposal failed; Aboriginals proposal passed.

Two questions, two recorded outcomes

The AEC table gives 27 May 1967: Parliament, 40.25 per cent Yes, not carried, a majority in New South Wales; Aboriginals, 90.77 per cent Yes, carried, majorities in every state.

The constitutional changes in the latter question concerned the race power and population reckoning. The federal voting change preceded it, in 1962.

The two questions must remain separate before any explanation of campaigning or results begins. “The referendum won” is too imprecise when a poll contains different proposals.

Sources for this section: [1] · [2]

An invented common-electorate exercise

Use an invented group of 100 people who each answer both questions: 40 vote Yes on A and 70 vote Yes on B. These are deliberately different from the historical percentages. Construct one table with 10 Yes on both and another with 40 Yes on both.

For each, find A-only Yes, B-only Yes and neither. What can the shared marginal totals tell you about whether the same people supported both questions?

Sources for this section: [1]

Reference answer

With 10 Yes on both, the four cells are: both 10, A only 30, B only 60, neither 0. With 40 Yes on both, they are: both 40, A only 0, B only 30, neither 30. Both tables total 100 and preserve A=40 and B=70.

The margins therefore do not determine the overlap. The historical questions may also have different valid-vote denominators, so these common-100 tables are not reconstructions or rounded versions of their results.

Sources for this section: [1]

Use testimony without changing the results

The supplied ABC transcript records Bandler’s memories of the campaign.

A retrospective account can explain how an organiser understood the work. It cannot, on its own, establish each voter’s paired choices. We read the supplied transcript, not the video or a complete separately catalogued interview. The original exercise concerns aggregate inference, rather than repeating the constitutional double-majority calculation.

Sources for this section: [3]

Continue the chronology

Biography context

Explore Australia through connected lives

Connected biographies

The related public career

Quick answers

Common questions

Are 40 and 70 rounded historical results?
No. They are stipulated counts for a different, invented common electorate.

Primary and institutional sources

How this history was checked

Selected original records, retrospective testimony and invented exercises are identified separately. This article was reviewed on 31 October 2026.

  1. Australian Electoral CommissionReferendum dates and results: 1967 rows
  2. National Archives of AustraliaThe 1967 Referendum: constitutional text and explanation
  3. ABC EducationTimeframe: Australia’s 1967 Referendum: supplied transcript