Western Metro; The flaw in the system ignored in Antony Green's analysis

The method used in calculating the results of the Victorian Legislative Council is flawed. It is not accurate and is semi proportional at best.

If you applied a reiterative count where the ballot is reset and the votes redistributed on every exclusion with a weighted calculation of the surplus transfer value the results are somewhat different to that which Antony Green has presented.

There are 422588 ballot papers

If you recount the vote excluding all candidates but the last six standing (3 ALP, 2 Liberals and 1 Green) and distribute all the votes based on the order of preferences allocated to continuing candidates .

1205 ballot papers exhaust without expressing a valid preference.

The quota based on the adjusted primary vote is 70229

After redistributing preferences and surpluses using a value based formula in calculating the surplus transfer value

The following are elected with quota
PAKULA, Martin 70229
EIDEH, Khalil 70229
FINN, Bernie 70229
ELSBURY, Andrew 70229

Last vacancy
HARTLAND, Colleen 70237
SMITH, Bob 69419

Exhausted with value 811

The gap between Hartland and Smith is 818 votes not 2012 as is derived using the VEC counting method.

Similar distortions were recorded in the 2007 Queensland Senate which resulted in Larissa Waters denied being elected to the sixth Queensland Senate spot.

The system in place does not accurately reflect the voters intention.

The distortion in the count arises from the method of calculating the surplus transfer value and the method of segmentation in redistributing excluded candidates votes.

The system was designed to facilitate a manual count. With the use of electronic counting there is no justification or merit in maintaining the flawed counting system. A reiterative weighted count such as the Wright System or Meek produces a much more accurate reflection of the voters intentions.

Posted by: democracyAtWork | December 18, 2010 at 04:38 PM

Western Metro: BTL Vote Stats

Table showing primary vote and distribution party of second preference

ID	Candidate	Primary	SEX	ALP	FF	DLP	GRN	LIB
1	MURRAY, Joël	1378	881	306	63	28	82	18
2	DAVIS, Merinda	536	277	127	44	15	42	31
3	PAKULA, Martin	2686	109	2073	169	132	157	46
4	EIDEH, Khalil	679	27	503	53	59	23	14
5	SMITH, Bob	557	23	443	37	29	19	6
6	REES, Llewellyn	723	17	551	25	29	72	29
7	SPINNER, Claudine	652	11	499	39	29	62	12
8	MUMBY, Daniel	635	24	65	410	63	37	36
9	GARAY, Lisa	142	6	31	59	21	16	9
10	FARRELL, Mark	631	23	38	39	458	49	24
11	MAZZARELLA, Stephanie	274	7	28	21	179	26	13
12	BYRNE, Georga	121	4	14	7	76	17	3
13	NOTARO, Samuel	89	1	4	4	64	7	9
14	KAVANAGH, Marguerita	113	3	13	3	79	10	5
15	HARTLAND, Colleen	3439	99	272	30	62	2883	93
16	HUMPHREYS, Robert Allan	428	21	39	5	28	312	23
17	INGHAM, Liz	595	23	49	10	22	478	13
18	CRAWFORD, Simon	199	8	12	3	10	155	11
19	GALE, Owen	205	4	12	11	9	161	8
20	FINN, Bernie	2051	38	56	82	79	115	1681
21	ELSBURY, Andrew	386	9	14	16	13	18	316
22	FAIRBAIRN, Victoria	603	12	20	26	20	65	460
23	BARTH, Jeremy	269	11	4	15	7	27	205
24	KENNY, William	309	9	9	9	10	15	257

Table showing number of ballot papers from each primary vote distributed to last remaining party candidates

ID	Candidate	Primary	SEX	ALP	FF	DLP	GRN	LIB	Exhaust
1	MURRAY, Joël	1378	1378
2	DAVIS, Merinda	536	308	45	58	10	36	25	54
3	PAKULA, Martin	2686	127	1958	155	135	204	31	76
4	EIDEH, Khalil	679	27	473	43	29	27	18	62
5	SMITH, Bob	557		557
6	REES, Llewellyn	723	32	471	39	17	111	17	36
7	SPINNER, Claudine	652	19	457	40	16	78	6	36
8	MUMBY, Daniel	635			635
9	GARAY, Lisa	142	10	16	65	16	10	10	15
10	FARRELL, Mark	631				631
11	MAZZARELLA, Stephanie	274	7	12	28	148	23	12	44
12	BYRNE, Georga	121	3	7	12	69	16	2	12
13	NOTARO, Samuel	89	3	9	6	47	8	5	11
14	KAVANAGH, Marguerita	113	8	7	5	69	11	6	7
15	HARTLAND, Colleen	3439					3439
16	HUMPHREYS, Robert Allan	428	26	21	13	10	314	14	30
17	INGHAM, Liz	595	31	14	18	8	493	8	23
18	CRAWFORD, Simon	199	17	12	14	4	127	13	12
19	GALE, Owen	205	13	9	16	7	144	8	8
20	FINN, Bernie	2051	44	18	101	67	101	1676	44
21	ELSBURY, Andrew	386						386
22	FAIRBAIRN, Victoria	603	26	9	42	9	37	438	42
23	BARTH, Jeremy	269	15	7	25	5	12	196	9
24	KENNY, William	309	10	7	16	9	5	250	12

Over 940 data inconsitencies in VEC Western Metro Count

In review and comparison of the two BTL preference data files (Count A and Count B) provided by the Victorian Electoral Commission shows that over 1,000 votes were incorrectly recorded with a disconcerning number being mistches in the batch reference id which indicates poor data management.

There were 15 records added to count B that were not included in Count A

Western-Metro-BTL-2010-B.batch_id
0548-001
0549-001
0549-002
0549-003
0550-001
0551-001
0551-002
0551-003
0551-004
0552-001
0553-001
0554-001
0555-001
0556-001
0557-001

and 8 records in Count A that were not in Count B

Western-Metro-BTL-2010-A.batch_id
0017-050
0026-050
0043-050
0132-048
0144-002
0341-049
0361-049
0426-049

Preliminary Analysis shows that the ALP fell short by one percent of the quota.

In a good example that demonstrates the distortion in the way the VEC counts the vote Preliminary analysis of Western Metropolitan vote shows that had a more accurate and proportional counting system been used Greens candidate Colleen Hartland would have won the seat with three votes above quota and the ALP falling short by one percent of quote just over 700 votes.

This is a stark contrast with the method used by the Victorian Electoral Commission as outlined in the legislation which at best can only be described as semi-proportional.

In our analysis we used the Wright system which included a weighted value method to calculating the surplus transfer value and on every exclusion the ballot was reset and the count restarted and votes redistributed as if the excluded candidates had not nominated. The quota for election was adjusted to accommodate votes that existed throughout the count. The Weight system uses a single transaction without segmentation and only surplus votes are distributed in a single iteration of the count. The process of iteration continues until all vacant positions are elected. ION analysis of Western Metro the vote progressed until there were only six candidates remaining in the final iteration count.

Similar results were shown using the Meek method of counting.

what this demonstrates yet again the need for the state government to abandon the outdated distorted and accurate system currently in use. Had the number of voters not exhausted their ballot papers or Victoria had retained a full preferential voting system the outcome of the election may have been different.

In 2007 the Queensland Senate election which uses the same flawed system of segmentation as exists in Victoria Greens candidate Larissa Waters missed out in being elected. Had a reiterative counting system been used instead Larissa Waters would have been elected.

The current system was designed to facilitate a manual count and with the use of computer based technology is no longer warranted or justified.

Wright System Reiterative Count - Value based Surplus Transfer - Single transaction per candidate Final Iteration  422588 Ballot papers minus 1205 ballots which exhausted with-out value expressing no preference for any continuing candidate
ID	Candidate	Group	01-001 Primary	01-002 Surplus PAKULA, Martin	01-003 Surplus FINN, Bernie	01-004 Surplus EIDEH, Khalil	01-005 Surplus ELSBURY, Andrew	01-006 Elected HARTLAND, Colleen
3	PAKULA, Martin	ALP	196632	70229	70229	70229	70229	70229
4	EIDEH, Khalil	ALP	1242	126653	126653	70229	70229	70229
5	SMITH, Bob	ALP	1411	1736	1761	57929	69419	69419
15	HARTLAND, Colleen	GREENS	69582	69987	70124	70181	70237	70237
20	FINN, Bernie	LIBERAL	151299	151299	70229	70229	70229	70229
21	ELSBURY, Andrew	LIBERAL	1217	1250	82015	82015	70229	70229
	Sum		421383	421154	421011	420812	420572	420572
	Quota		70229
	Exhausted	with value		229	372	571	811	811

Western Metro Greens Hartland elected on below quota vote VEC instructions and optional voting rules under review

The Greens have been elected to the last position in Western Metropolitan Region with 130 votes less than the quota required.

The Australian Labor Party and former Legislative Council President, Bob Smith, more than 2,000 votes behind the Greens' Colleen Hartland primarily due to the optional preferential rules.

Many Liberal Party and DLP supporters who were opposed to the Greens candidature exhausted their preference allocation and in doing so allowed the Greens to be elected by default.

The VEC has also come under criticism for not providing clear instruction to encourage others to fill in all preferences. The instructions on the ballot legislative Council papers encouraged voters to only preference on to 5. The VEC should have been more proactive in encouraging voters to maximise their vote by fulling in all preferences. As a result of the VEC failure many voters were disenfranchised as a result.

As a result of the optional preferential rules that were introduced for the first time this year Labor has lost a third seat in Western Metropolitan.

The conduct of the count was further marred by the fact that the VEC failed to provide scrutineers copies of the below-the-line preference data. Copies of the preference data should have been published on line and in real time as the vote is counted.

Button Day show: Tuesday December 14

Tomorrow the Chief Electoral Commissioner, Steve Tully will press the button to reveal the outcome of the preliminary declaration of the Victorian Legislative Council and the eagerly awaited results for Western Victoria.

Whilst the counting process has been a marked improvement on the events of 2006 where ballot papers went missing between count A and Count B and the Commission failed to maintain backup copies of crucial below the line preference data files for Western Metro.  Over 500 Ballot papers were unaccounted for, double entered or removed from the count. Steve Tully can not afford another stuff-up repeat 

Already the count has been marred by the refusal of the Commission to provide copies of the below the line data files to scrutineers as the cont progresses.  Why this information was not published on line is any one's guess.

"Ladies and Gentlemen watch me pull an election result out of my hat"

Presumably Steve Tully wants to make a show and dance of the event and bask in then lime light of media as the computer pumps out the results as he puts his ego ahead of maintaining a open and transient election. 


Copies of the data-files were available during the 2008 City of Melbourne Municipal election count why weren't available during the State election?  We can monitor the polling booth results  on election night as they are reported why not watch live the tabulation of the below the line preference votes.  Does the commission have something to hide?

In not providing access to the data-files the Commission has denied scrutineers the ability to verify the integrity of the  data file used to tabulate the results. It is also impossible to know if the data files has not been altered in anyway. Had copies of the preference data-files been made available or published on line it would have allowed members of the public to maintain a watching brief to ensure that the data record had not changed without validation or appropriate authority. 

Steve Tully claims that he is not required to make copies of the data available. Whoever he does have an obligation to ensure that the conduct of the election is open and transparent. Whilst there is no specific regulatory statement that says he must make this data available the reality is there is nothing that prevents or phlebitis him from doing so.  In not providing access to this data he has prevented scrutineers from scrutinising the election. 

In the absence of self regulation the State Parliament will have to regulate to restore public confidence and  ensure that this information is readily available during the next round of elections due in 2008.

Steve Tully has promised he will make copies of the data-file including both the preliminary  data-entry and the recheck data-entry verification files immediately after the conclusion of the count.

If as expected Western Metropolitan Region is within 250 votes we can expect a possible recount in which case the Commission must make copies of the the data available as and then data is recorded. already both the green and the ALP have signaled that they will be requesting a recount of the result is close.

Given the history of the VEC under Steve Tully's stewardship anything under 500 votes is worth looking at a second time.

Western Metro Break Even Analysis

In an attempt to determine the likely breakdown of the Western Metropolitan Region we have allocated below the line preferences into two categories.

Full preference flow based on the allocation of preferences of the above the line ticket registration and the optional preferential flow where votes exhaust after five preferences. We believe that you can allocate percentages to these tables to determine the likely data flow. The values we have indicated result in a tied vote between Bob Smith (ALP) and Colleen Hartland (Greens). Both Sex Party and FF have to reference a second Party for their vote to be valid. We have included a most likely what if of FF flowing to the DLP and exhausting.

The Liberal Party Ticket vote will increase in strength as a result of the flaw in the calculation of the Surplus Transfer value which is based on the number of ballot papers not the value of the vote

What is also interesting with this breakdown is that if you applied a reiterative count without segmentation Bob Smith is elected where the method of segmentation gives an unfair advantage to the Greens and other minor parties.

The main player and unknown quantity is where do the Sex Party preferences flow. Many will follow the Sex Party Ticket flow to the Greens where they may exhaust as the instructions given on the ballot paper encourage optional voting as opposed to a full preference distribution. The ALP is expected to pick up a greater share of Sex Party preferences then they did in the August Senate ballot due to the position of the ALP group being next to Sex Party (The donkey vote). There will be some votes that will cross groups but they can be accommodated by selecting and allocating their vote to one of the other records in the table presented depending on which group (ALP,Greens or Liberals) they end up at.  


#1 For clarity sake we have allocated Sex Party BTL Full ticket vote within the Optional Preferential Sex Party to Greens list as it is unlikely to travel beyond the Greens.

On distribution of the below preference allocation the result is within 50 votes between Smith and Hartland

Current Group Vote Tally for Western Metropolitan

		ATL	BTL	TOTAL	%
A	SEX PARTY	18088	1946	20034	4.74%
B	AUSTRALIAN LABOR PARTY	192774	5354	198128	46.86%
C	FAMILY FIRST	14464	825	15289	3.62%
D	D.L.P. - DEMOCRATIC LABOR PARTY	12097	1236	13333	3.15%
E	AUSTRALIAN GREENS	45835	4884	50719	12.00%
F	LIBERAL	121630	3656	125286	29.63%
	Sum		17901

Breakdown allocation of BTL preferences

Below the Line Distribution Allocation
	Full Tickets		Ballots		%
	SEX PARTY #1		0		0.00%
	AUSTRALIAN LABOR PARTY		535		10.00%
	FAMILY FIRST		206		25.00%
	D.L.P. - DEMOCRATIC LABOR PARTY		309		25.00%
	AUSTRALIAN GREENS		1221		25.00%
	LIBERAL		731		20.00%
Optional Preferential (Stop at 5)
	SEX PARTY -> GREENS		778		40.00%
	SEX PARTY -> ALP		681		35.00%
	SEX PARTY -> LIBERALS		487		25.00%
	AUSTRALIAN LABOR PARTY		4819		90.00%
	FAMILY FIRST -> DLP		619		75.00%
	D.L.P. - DEMOCRATIC LABOR PARTY		927		75.00%
	AUSTRALIAN GREENS		3663		75.00%
	LIBERAL		2925		80.00%
	Sum		17901

Segmentation and Surplus Transfer flaws come into play

Flaws in the way the Victorian Upper House is counted are likely to effect the outcome of the Western Metropolitan Region. The problem lies with the method of Segmentation and the calcualton of the Surplus Transfer Value.

Analysis, of what is going to be an extremely close and controversial count, has indicated that if the VEC used a mathematically correct and weighted formula for calculating a candidate's Surplus Transfer Value and  if the results were determined by a single transaction reiterative counting system. such as outlined in the Wright System, the result of the election could differ.  Likewise if Victoria had adopted the Meek method of counting the ballot. Both systems more accurately reflect the voters intentions

The quickest and easiest method of testing to see of the outcome is effected by Segmentation is to recount the ballot excluding all candidates except the last six left standing,  (ALP 3, Liberal 2, and Greens 1) Likewise you could apply the Western Australia method of calculating the Surplus Transfer Value again exlcuding all canidates except the last six standing. Similar flaws had denied Greens Candiat, Larrissa Waters election in the 2007 Queensland Senate count.

Segmentation is the system adopted that distributes an excluded candidates votes based on the value or group of votes allocated.  It was implemented to facilitate a manual count and to minimise the impact of the flaw in calculation of the Surplus Transfer Value.  It is akin  to dealing from the bottom of the deck in a game of cards. With a computerised electronic counting system there is no justification or need for segmentation.

All votes should be transferred in a single transaction.  Better still the ballot should be reset and started from a fresh following every exclusion.  This would allow for a recalculation of the quota taking into account any ballot papers that had exhausted and would allow minor party votes to contribute to a previously elected candidate's surplus value. All ballot papers are treated equally and distributed as if the excluded candidates had not stood.  One vote -one value.

More questions raised about Steve Tully's 2006 Western Metro Count

A number of question have been raised about the data quality and accuracy of Th 2006 Victorian State - Western Metropolitan Region election results.

In 2006 the Victorian Electoral Commission recorded a win for the ALP following the first preliminary count of ballot papers. In the recount there were 500 less ballot papers recorded then the first count and the ALP lost by 127 votes. What was also of interest was that there were 250 ballot papers less than that recorded for the corresponding lower house districts.

The number of total votes should never change between counts. The only reason they could is if there was a mistake in the tabulation of the number of votes or votes went missing between the two counts.

What was of even greater concern is that the Victorian Electoral Commission refused to publish a copy of the below-the-line preference data-file for the primary count.  A copy of the file was requested by the Victorian Parliamentary Electoral Matters Committee to which the Chief Electoral Commissioner, Steve Tully, in giving evidence to the committee responded by stating that the file and been overwritten and there were no copies or backup data files available.  Steve Tully was only able to present a copy of the final count data file.

Without access to the primary count data-file there is no means of independently analysing the validity of the election ot the quality of the count.  A comparison of the two data files would have shown in detail where and what had changed. The fact that the VEC had not back backup copies of the data raised a number of serious concerns as to the professionalism of the VEC's IT support. It costs taxpayers millions of dollars to collate the election results and it is extraordinary that there was no disaster recovery plan in place. - Something that no Professional IT service would miss.

To add to the questions and uncertainty of the 2006 elections result the number of ballot papers recorded for 2010 Western Metropolitan Region election in comparison to its associated lower-house districts is currently less than 50 votes difference. So just how accurate was was the total vote count in 2006?  Has the VEC cooked the books or have they just been more diligent this time around then they were in 2006?  We will never know the answers to these questions, without access to the vote preference data files.

This time round there is no excuses.  Steve Tully has promised to publish both sets of preference data-files (Why the delay? They should have been available in real time and publihsed live as the count progresses adnas was the case on election night) One file for the primary count and one for the recheck. If there is to be a recount in Western Metropolitan (Which is looking most likely) then hopefully the VEC will also publish the final set of data.

What is clear is that Victoria needs to be vigilant and  much more needs to be done to ensure that our electoral system is open and transparent., currently it is not.