Talk:Donkey vote

Great job expanding the article, Capitalistroaster! --Smithfarm 15:00, 21 Mar 2005 (UTC)

Grayndler Greens How-To-Vote

I know for a fact that the Greens did advocate a Donkey Vote in Grayndler in 2004, so I'll update this. Also, Werriwa 2005 provides a good example of use of the Donkey Vote. Braue

intro

Would it be possible for someone familiar with the topic to provide better context in the intro? That is, to more clearly identify the elections (and the voting systems used) in Australia where donkey votes occur. heqs 06:56, 19 May 2006 (UTC)

I'm an American and have never heard the expression till just now. I must say that the article gives a lot of information on it. Steve Dufour 04:06, 30 November 2006 (UTC)

Link for citation of 2% advantage to first place: http://www.vote.caltech.edu/news/RMA_KYCourier-Journal_11-18-06.pdf 71.109.237.121 (talk) 20:00, 22 December 2007 (UTC)

1961 myth

Although in a quote, it's dubious to say that the 1961 election was decided by donkey votes and/or Communist preferences in Moreton:

Those crucial ballots [in the Queensland electoral district of Moreton, in the extremely close 1961 House of Representatives election] turned out to have cast not by Communists but by donkeys, and as [Liberal candidate James] Killen’s name preceded that of the now-forgotten Labor candidate in the alphabet, they flowed largely to the Libs.

Andrew Bartlett posted the actual count figures on his blog a while back, showing that any donkey votes would have transferred in line with ideological consistency and that the crucial factor was the Democratic Labor Party transfers. Sir James Killen: Moreton, Menzies and Mythology Timrollpickering (talk) 20:19, 21 August 2008 (UTC)

"The Senate"

It isn't clear that this refers to the Australian Senate. The previous section's title DID make it clear. Can somebody fix this? —Preceding unsigned comment added by 69.255.193.251 (talk) 06:50, 30 October 2010 (UTC)

I've done that. I was half asleep and mistook it for US senate. I probably wouldn't have if I'd have thought about it for a second but just to make it really clear I've changed it Cls14 (talk) 08:43, 13 October 2012 (UTC)

Proportion of donkey votes?

I was scrutineer at a byelection in 1998 and saw about 50% donkey or near-donkey votes (Labor or Liberal first, the rest in numerical or reverse-numerical order). Do we have anything like a solid number? - David Gerard (talk) 20:01, 27 March 2012 (UTC)

Should donkey votes be officially recorded or measured during the voting process, and taken into account after the voting process. They really are a sort of public vote of no confidence regarding that particular election and the parties/policies involved within it. They provide relevant and important statistical information regarding peoples attitudes towards the the current political climate within a country where voting is compulsory (?) — Preceding unsigned comment added by 124.149.43.106 (talk) 14:23, 10 November 2012 (UTC)

Origin of name

Where did the term "donkey vote" come from? Is there such thing as an elephant vote? (The donkey and elephant are the symbols of the two major political parties in the United States.) --Damian Yerrick (talk) 04:12, 5 November 2018 (UTC)

Victorian Local Government Elections 2008 - 2021

For the 2024 elections refer to the relevant sections below.

When there is a large field of candidates on ballot papers in Victorian Local Government General Elections the results clearly demonstrate the impact of Donkey Vote variations. Donkey Votes are the only reasonable explanation for the greater than 3𝜎 difference between expected and actual election results for the candidates at the top of ballot papers.

Published Election Data 2008 - 2021

The Victorian Electoral Commission publishes the results of Victorian Local Government 2008 to 2021 elections, by-elections and recounts for the 79 Local Government Areas of Victoria (LGAs) which consists of 34 cities, 38 Shires, 6 Rural Cities plus one Borough.

This section of the Talk Page provides an overview of two different techniques for the analysis of the published election results.

Analysis by Deciles

Results For A Subset of Victorian Local Government Elections 2008 - 2020

Candidate Range	Elections	Total Candidates	Decile 1	Decile 2	Decile 3	Decile 4	Decile 5	Decile 6	Decile 7	Decile 8	Decile 9	Decile 10
21-41	32	823	2.41	2.09	1.78	1.45	1.23	1.00	1.70	1.82	1.09	1.57
20-41	42	1023	2.77	1.99	1.82	1.63	1.19	1.00	1.66	1.77	1.16	1.91
19-41	54	1251	2.32	1.95	1.76	1.33	1.23	1.00	1.45	1.60	1.03	1.71
18-41	68	1503	2.15	1.74	1.60	1.26	1.32	1.00	1.45	1.44	1.11	1.75
17-41	83	1758	1.88	1.51	1.48	1.24	1.28	1.00	1.23	1.29	1.10	1.52
16-41	99	1998	1.96	1.68	1.53	1.22	1.37	1.00	1.29	1.22	1.13	1.52
15-41	121	2343	2.06	1.59	1.54	1.27	1.34	1.00	1.32	1.18	1.11	1.54
14-41	150	2749	2.10	1.61	1.71	1.35	1.34	1.00	1.32	1.25	1.08	1.52
13-41	187	3230	1.97	1.42	1.57	1.29	1.29	1.00	1.30	1.23	1.13	1.52
12-41	226	3698	1.90	1.44	1.60	1.26	1.32	1.00	1.28	1.25	1.12	1.45
11-41	272	4204	1.87	1.44	1.61	1.21	1.28	1.00	1.23	1.20	1.08	1.36
10-41	324	4724	1.78	1.38	1.58	1.18	1.27	1.00	1.20	1.14	1.07	1.29

As an example, for a subset of Victorian Local Government elections between 2008 and 2020, with 21 to 41 Candidates standing for election per ward, the published results were:

• 32 elections
• 823 Candidates standing
• 23.1 Candidates elected from the 1st decile (i.e. top 10%) of a ballot paper
• 9.6 Candidates elected from the 6th decile of a ballot paper.

Therefore candidates in the 1st decile were 2.41 times more likely to be elected than the candidates the Victorian Electoral Commission (VEC) had randomly allocated to the 6th decile on the ballot paper. Note: The fractional components are an artifact of the decile calculations and only shown for accuracy.

Analysis of the the Results for the Candidate at the top of Ballot Papers

An alternative method of analyzing the impact of a Donkey Vote is to compare the expected ( E[p] ) and probability ( p ) of election for candidates at the top of the ballot paper. The number of election results available decreases with increasing number of candidates standing for election in a ward therefore Candidate are grouped for this analysis. A simple analysis of all elections (i.e. those with 2 or more candidates standing for election) will tend to mask the impact of Donkey Votes for elections where there is a large candidate field. Therefore results have been grouped by the elections with greater than or equal to the number of candidates standing for election.

For all elections with candidates from >=2 to >=21 the observed probability for the election of a candidate at the top of the ballot paper is always greater than 3𝜎 above the expected probability of election. For >= 5 candidates the result is 7.5𝜎. The 3𝜎 limits shown plotted here could have been increased to 5𝜎. Refer to the 68–95–99.7_rule Wikipedia entry for a further discussion on probability. As election fraud is not a characteristic of Victorian Local Government elections the impact of a Donkey Vote variation is the only explanation for the observed bias in the election results.

Victorian Local Government

Local Government in Australia is effectively a third layer under the National and State government constitutions. Each State regulates its own form of Local Government under their individual Acts and Regulations. In Victoria the relevant legislation is the Local Government Act 2022 plus associated Statutory Rules (Regulations) such as the Local Government Electoral Regulations 2020, etc.

Election Counting Methods

When there is only one vacancy in a Ward a Preferential method is used. For Wards with multiple vacancies a Proportional method applies. Note: Some other States and Territories implement a form of Robson Rotation that minimises the impact of Donkey Votes and the analysis discussed here is not applicable to these jurisdictions.

Victorian Local Government General Elections have unique characteristics that increase or amplify the impact of Donkey Vote variations. Characteristics are:

• Electors are required to allocate preferences for every candidate
  • The counting methods do not permit partial preferences.
  • Failure to provide a preference indication for every candidate means a ballot paper will be discarded as an informal vote.
  • Voters are required to preference candidates even when they may not have received enough information to make an informed choice.
• There can be large candidate fields
  • Any qualifying elector in a Ward can stand for election
  • The largest Wards in 2020 would have more than 35,000 electors.
  • There is no upper limit on the number of candidates standing in a Ward
  • The largest candidate field standing for election in the 2008 -2020 period was 41 candidates
• For the majority of Victorian Local Government elections there is no grouping of candidates or any form of above the line voting.
  • The City of Melbourne elections are an exception as they permit Candidate Teams and Groups therefore all the City of Melbourne election results were excluded from the statistical analysis reported here.
• Minimal direction provided by major political parties
  • Major Australian political parties (e.g. Australian Labor, Liberals, Nationals) typically do not endorse candidates in Victorian Local Govenment elections.
  • How-to-Vote cards are not permitted in the Postal Vote packets sent to electors.

The majority of these Victorian Local Government General Elections have employed a postal vote system where the VEC provides a Candidate Information Sheet with the ballot papers. The candidate information sheet provides an option for each candidate to include a photograph and an upto 300 word candidate statement (an increase from 250 words in 2016). This limited Candidate Statement forms the minimum information supplied to voters. Therefore the preferences cast by an individual maybe based soley on this limited material. The postal vote packages do not include How-to-Vote cards as they are forbidden for Postal Voting under current (2022) legislation. Therefore the major political parties do not direct preferences, influence or provide directions to voters in the material distributed by the VEC.

A Candidate Statement can not refer to another candidate without their written permission. As candidates are competing for a limited number of vacancies this provision is relatively rarely employed. The VEC has not published the relevant data therefore the actual use of this provision is unknown. A submitted Candidate Statement is formatted without paragraph breaks. The largest field of candidates in the 2008 to 2020 period was 41. Worst case for a voter reading a ballot information pack could be 12,300 words (41 x 300 words), a word length for a Novella. Voters in Victorian Local Government elections are required to preference every entry on the ballot paper. A voter is unlikely to personally know every candidate and in a worst case they may only have the information pack to determine their preference order. Typically they will have an informed opinion on a subset of candidates that will receive their highest preferences, maybe some that will be penalised and placed last on the ballot paper. As a voter may not be able to make an informed choice about all candidates on their ballot paper, it is not surprising an individual could cast preferences with the candidates preferenced in one of the variations of a Donkey Vote pattern.

Donkey Vote Classic Definition

The 'classic' definition of a Donkey Vote has been typically restricted to the definition of a preference order straight down (or up) the ballot paper. Previous work has reported Donkey Votes in the order of ~2% for Australian Federal and State elections. The statistics and analysis of Australian Federal and State elections has no relevance to Victortian Local Government General Elections discussed here.

There have been a number of scholarly articles attempting to estimate the impact of a donkey vote in Australia elections, including:

References to Donkey Votes

Estimate	Election	Author	Title	Source
2% to 4%	Representatives	Mackerras	The Donkey Vote for the House of Representatives	APSA Monograph No. 6 (1963), Department of Government, University of Sydney
1% to 2% and 3%	Senate and Representatives	Mackerras	The “Donkey Vote”	The Australian Quarterly, Vol. 40, No. 4 (Dec., 1968), pp. 89-92
~0.65%	Representatives	Peetz	Donkeys, deserters, and targets: causes of swing in electorates in the 1987 federal election	The Australian Quarterly, Vol. 61, No. 4 (Summer, 1989), pp. 468-480
1.3%	Representatives	Kelly, McAllister	Ballot Paper Cues and the Vote in Australia and Britain: Alphabetic Voting, Sex and Title	Public Opinion Quarterly, 48 Summer (1984), pp. 452-466
1%	Representatives	King, Leigh	Are Ballet Order Effects Heterogeneous?	Social Science Quarterly, Volume 90, Issue 1 (2009), pp.71-87
0.34% to 7.53%	Senate and Representatives	Smith, Kildea, Gauja, Keenan	The Challenge of Informed Voting in the 21st Century	Electoral Regulation Research Network Research Report, April 2015

Various dictionaries can include similar definitions for a Donkey Vote while the Australian Electoral Commission (AEC), Glossary definition is a very concise:

Donkey Vote

A ballot paper marked 1, 2, 3, 4 straight down (or up) a ballot paper.

This is not the only form of a Donkey Vote and using this definition underestimates their impact on Victorian, Local Government elections. State and Commonwealth elections are dominated by political parties with group voting, How-To-Vote cards, above the line voting available for Senate elections (where there are a large number of candidates), etc.

The VEC does not release the database of cast Local Government ballots so it is nearly inpossible to estimate this form of Donkey Vote in Victorian Local Government elections.

Victorian Electoral Commission Definition of a Donkey Vote

In relation to the Local Government election the VEC has not published a Donkey Vote definition. A Donkey Vote definition by the VEC was included in the 2010 Victorian State Parliament Election Report, Section 11 Statistical Overview Of The Election, Page 73, Donkey Votes entry:

A donkey vote is a vote cast by a voter who numbers the squares down the ballot paper, without caring about the nature of the candidates on offer. Candidates are pleased if they get the top spot on the ballot paper when the draw for position takes place, because they will have the advantage of the donkey vote. The size of the donkey vote has been a matter for discussion.

The VEC counted the donkey votes as part of its survey of ballot papers. To measure real donkey votes, it is necessary to distinguish them from votes following party advice and from votes that were logically in this order given the nature of the candidates. ...

Figure 64: Proportion of donkey votes minus proportion attributable to random variation on Page 73 demonstrated how minuscule impact of the Donkey Vote had on the election. In Figure 64, the proportion attributable to random variation has been subtracted. ... In the Districts surveyed, the median donkey vote was 1% of the total formal vote. The donkey vote varied widely, from 0.07% in Rodney to 2.59% in Shepparton.

Given the advantage conferred by the Donkey Vote it is not surprising that the VEC states "Candidates are pleased if they get the top spot on the ballot paper". Candidates benefiting from an advantage are the ones least likely to be unhappy. Obviously this analysis is related to the State Election and not directly applicable to Local Government elections. The VEC used such a narrow definition of Donkey Vote that it fails to identify the impact of Donkey Vote variations on Local Government election results.

In the 2020 Victorian Local Government elections the candidate at the top of the South Ward, Mooreland City Council ballot paper did not have a candidate statement or photograph published on the candidate information sheet. Examination of the VEC's Distibution spreadsheet shows 269 first preference votes cast their second preference to the second candidate on the Ballot Paper i.e. the begining of a 'classic' Donkey Vote preference sequence. There were 25,770 valid votes cast in this election therefore these 269 represent a maximum possible 1.004% 'classic' Donkey Vote. The VEC provides a report to the State Parliament after each Local Government General Elections and using their current definition of a Donkey Vote would not consider 1% worthy of comment e.g. refer to the 2020 report to parliament.

1% or 2% 'classic' Donkey Votes could not produce the distortion shown in the election results of Victorian Government General Elections provided in the table above. Given the distortion in the observed results for large candidate fields it is appropriate to examine the effect of alternative forms of a Donkey Vote.

8 Types of Donkey Vote Variation

Shown are examples of eight variations of a Donkey Vote. In this simplistic example of Partial Donkey Votes the first two preferences (1 & 2) are cast with apparent care for candidates 8th then 2nd) then the balance are cast in a Donkey Vote sequence. Circular Donkey Votes are a special case of the Partial Circular Donkey Votes where only the first preference appears to have been cast with apparent care.

Results For Simplistic Donkey Votes

Ballot Position	Linear (Classic)	Reverse Linear	Circular	Reverse Circular	Partial Linear	Partial Reverse Linear	Partial Circular	Partial Reverse Circular
#1	1	10	4	8	3	10	10	3
#2	2	9	5	7	2	2	2	2
#3	3	8	9	6	4	9	3	10
#4	4	7	6	5	5	8	4	9
#5	5	6	7	4	6	7	5	8
#6	6	5	8	3	7	6	6	7
#7	7	4	9	2	8	5	7	6
#8	8	3	1	1	1	1	1	1
#9	9	2	2	10	9	4	8	5
#10	10	1	3	9	10	3	9	4

The following table illustrates preference flows to a candidate that is not Candidate #1. This is not attempting to show all possible combinations, just a simple subset to show the distribution of preferences favours the candidates higher up the ballot paper. Assume the final two candidates will be vying for one councillor position (Candidates #3 and #6) when partial linear donkey votes are distributed. To evenly distribute the preferences each voter allocates their second preference two candidates positions above (circular) their first preference. Shading in the table is used to highlight the distribution of preferences. In this example every voter has cast their first and second votes with care which has exhausted their informed choices for candidates. For their ballots to be valid they must preference every candidate therefore they all resort to a partial linear Donkey Vote to complete a valid ballot. In this simplistic example Candidates #3 and #6 each receive one 1st and one 2nd preference vote while Candidate #3 is elected by the 6 Donkey Votes. The same analysis with a partial circular Donkey Vote provides a similar result.

Simple Example Of Partial Linear Donkey Vote Results

Ballot Position	Elector 1	Elector 2	Elector 3	Elector 4	Elector 5	Elector 6	Elector 7	Elector 8	Elector 9	Elector 10
#1	1	3	2	3	3	3	3	3	3	3
#2	3	1	3	2	4	4	4	4	4	4
#3	4	4	1	4	2	5	5	5	5	5
#4	5	5	4	1	5	2	6	6	6	6
#5	6	6	5	5	1	6	2	7	7	7
#6	7	7	6	6	6	1	7	2	8	8
#7	8	8	7	7	7	7	1	8	2	9
#8	9	9	8	8	8	8	8	1	9	2
#9	2	10	9	9	9	9	9	9	1	10
#10	10	2	10	10	10	10	10	10	10	1

Both candidates (#3 and #6) receive 1 primary vote (Elector #3 for Candidate #3, with Elector #6 for Candidate #6). On distribution of preferences both candidates receive 1 second preference vote (Elector #5 for Candidate #3, with Elector #8 for Candidate #6). Candidate #3 receives all 6 Donkey Vote preferece votes while Candidate #6 receives zero preferences from the distribution of the Donkey Votes.

The Tasmanian Electoral Commission (TEC) investigated Donkey Vote variations cast in two Tasmanian Local Government elections. The results were published in the Robson Rotation Discussion Paper, dated April 2008 and included:

Local government elections

Ballot papers from the 2002 Latrobe and Meander Valley Council elections were examined.

The survey of all formal ballot papers found:

• 1.4% of the ballot papers were full linear votes.
• 0.4% of the ballot papers were full linear votes going in the reverse direction (bottom to top).
• 2.4% of the ballot papers were full circular votes.
• 27.5% of the ballot papers showed only the minimum five preferences and 66.4% of ballot papers showed a preference for all 14 candidates. Only 6.1% voted for an in between number of candidates.
• 27.9% of the ballot papers contained partial linear voting. That is, voters casting their first few preferences with apparent care, and then filling in the remaining boxes in a straight sequence up or down the ballot paper.

In summary linear voting at local government elections did not appear to be a large problem.

There are major problems applying these results or assuming the conclusion applies to Victorian Local Government elections:

• There was no analysis of the wider impact of partial linear voting
• The Tasmanian elections only had 14 candidates. While the Victorian elections have had up to 41 candidates. Analysis has shown the increased number of candidates in Victorian elections is likely to increase the size and impact of a Donkey Vote.
• Insufficient information is available to determine any impact of the partial linear Donkey Vote for such a limited number of Tasmanian Local Government elections. The 2016 Victorian Local Government elections included wards with a much larger field of candidates that could be influenced by a Donkey Vote. The impact of the very high percentage of partial linear voting (27.9%) identified in the Tasmanian council elections is not identified by the VEC in any election reports.

The Tasmanian Electoral Commission report indicated 27.9% of their Local Government ballot papers contained partial linear voting. That is, voters casting their first few preferences with apparent care, and then filling in the remaining boxes in a straight sequence up or down the ballot paper.

As shown by an examination of the Victorian Local Government election results (2008 to 2020), the distortion of results increases with the size of the Candidate field. A Donkey Vote variation can bias the election of more than just the Candidate at the the top of the ballot paper i.e. to unfairly elect candidates from a range of positions at the top of Ballot papers. The effect of Donkey Vote variations is the only reasonable explanation for the observed results. Explanations like human corruption of the results are unresonable. A similar shape to the distortion of results is observed when examining all the subset of elections where preferences cast by voters could impact the results i.e. the subsets of elections with ranges of 3 or more candidates (to 41 candidates).

The Victorian Electoral Commission has not identified the observed distortion of election results in their reports to the Victorian Parliament or in their published research papers. There have not been any peer reviewed or other scholarly published work that could be used as references. Therefore the parliamentarians do not have a reasonable basis to ammend the legislation to eliminate the impact of these Donkey Vote variants. The current (2022) in force Victorian Local Government Act and Regulations do not directly mitigate the impact of Donkey Vote variations. In recent years Victoria changes to legislation has driven a move to a greater number of smaller wards. By this increased division of council areas the average number of candidates per ward is decreased. This reduction in candidate fields reduces the severity of the distortion but has not eliminated the bias. A solution to the problem has been implemented in the ACT and Tasmania with the introduction of different variations of Robson Rotations. This randomizes the position of candidates on ballot papers and eliminates any top of the ballot paper advantage. Refer to the Robson Rotation article for details of this solution.

Victorian Local Government Elections 2024-

For the 2024 General Elections there was the final implementation of multiple, single-member Council Wards to replace most, multi-member Wards. With the increase in the number of Wards there has been a decrease in the average number of Candidates per Ward. This should significantly reduce the impact of Donkey Votes. Given the changes are mostly confined to the urban areas of the capital City of Melbourne it may take some election cycles to provide sufficient data for analysis.

Pdp11.caps11 (talk) 00:56, 29 July 2022 (UTC)

References

— Preceding unsigned comment added by Pdp11.caps11 (talk • contribs) 03:25, 21 May 2022 (UTC)