Micropolling results of Presidential Elections 2011

Update 24 Sept : One more result in from Fuhua Secondary (Jurong GRC).  TCB did very well at this counting centre with 46.8%, almost an outright majority. Again, entirely consistent with media reports, and this data point brings TCB’s average up to 34.5% in this sample, closer to his actual vote share of 34.8% overall. This data set now has results from 21 counting centres, covering 278,878 voters (13.0% of votes cast).

Update 11 Sept 11:00 pm : Thanks to Donaldson, BK, Betty, Paul, Wei Ming, Justicia, Yingru, Dexter , Jacqui, WF, Randi and others who prefer to remain unnamed 🙂 We now have results from 20 out of 162 counting centres, representing 268,865 voters or 12.5% of the 2,153,037 votes cast in Singapore. Breakdown of votes for each candidate and spoilt votes by polling district are available in 11 out of the 20 counting centres. In the other cases, the Counting Agents who responded were only able to provide totals for the counting centre, or I was not able to match the counting tables to a specific polling district. For polling district boundaries, see this gazette notification on ELD’s website. Not very user-friendly, unfortunately, but I don’t know of an easy way to convert the text descriptions to graphic maps. Ideally, we would map the polling districts to interesting characteristics, e.g. income, age, education etc, but that’s beyond my capabilities at the moment. Are there any readers who could help ? Having four choices instead of a binary choice also complicates the statistical analysis. I would be very happy to receive any suggestions as to how to analyze the data rigorously. Eyeballing the data, this sample confirms media reports that TCB dominated the Western part of Singapore but failed to overcome TT’s smaller but more spread out advantage in other parts of the island. Raw data is in http://bit.ly/ocqvUB (Alternate URL http://db.tt/2WiQ39I ) and for avoidance of doubt, I will state that this post and the compilation of election results is released into the public domain. Attribution would be appreciated but is not required.

Update 31 Aug 11:30pm : We’re up to 15 Counting Centres, representing almost 200,000 votes or 10% of the votes cast. Keep the data coming 🙂 As always, latest data in http://bit.ly/ocqvUB

Update 30 Aug 8:45pm: Several other counting agents have responded and I now have results from a total of 11 counting centres. If you wish to download the data, the spreadsheet at http://bit.ly/ocqvUB will be kept up to date, through there may be a lag before I can update the chart in this post.


Five Eighteen other counting agents responded to my call so I am now able to post results for six twenty-one counting centers, nine eleven of which are broken down by polling district. This is essentially raw data, but I am posting it early so that other people would be able to make use of this data for their own analysis. I am still hoping to get more data. If you volunteered as a Counting Agent and still have the vote tally from your location, I would be extremely grateful if you could forward the results to me so that I can add it to the publicly available information on voting patterns in Singapore.

Micro-polling results from selected Counting Centres - Presidential Election 2011

Tan Cheng Bock did very well in Jurong West, Choa Chu Kang and Bukit Panjang in the Western part of Singapore,  and also edged ahead of Tony Tan in Sengkang and Punggol East. At Fuhua Secondary in Jurong West, TCB received close to an outright majority, with 46.8% of the vote, ahead of TT by almost 18%. Tony Tan’s best performance was in the Orchard-Tanglin-Farrer Road-Sixth Avenue area (SCGS counting centre). In one polling district, in fact, TT almost achieved an outright majority with 49.7% of the vote. TCB still managed to achieve 36.9%, higher than his national average, at the SCGS counting centre, but TJS clearly rattled the voters in this area, with as low as 13.9% of the vote in one of the polling districts. As I mentioned before, if you were a counting agent, know someone who was, or better yet, know someone inside the campaigns who would be willing to share their data, I would really like to hear from you at PE2011@ngiam.net. An example of the results from one counting centre are shown below, and the complete data table can be downloaded at http://bit.ly/ocqvUB. Please note that this is NOT OFFICIAL DATA and was reported by volunteers after a very late night. As the Election Department’s favorite saying goes, “the decision of the Returning Officer is final”. See my earlier post for more background on counting procedures. And of course, a big thank you to the counting assistants, facilities, security and other personnel who worked through the night to ensure an orderly election despite the recount. Kudos especially to the AROs and ELD officers who worked over 24 hours straight, from 5 am on Saturday to 7 am on Sunday.

Constituency Marine Parade GRC
Counting Centre Nanyang Junior College
Counting Place GK01 GK02 GK03 GK04 GK05 GK06
Polling District MA27 MA24 MA23 MA22 MA26 MA25
Polling Station Nanyang JC Braddell Heights CC (B) Braddell Heights CC (A) 419 Serangoon Central 305 Serangoon Ave 2 240 Serangoon Ave 2
Valid votes for
Tan Cheng Bock 1,104 1,000 1,006 1,150 986 1,124
Tan Jee Say 861 880 917 921 745 851
Tony Tan Keng Yam 1,099 1,014 1,113 1,266 1,071 1,257
Tan Kin Lian 172 180 162 202 144 202
Percentage of Valid votes
Tan Cheng Bock 34.1% 32.5% 31.5% 32.5% 33.5% 32.7%
Tan Jee Say 26.6% 28.6% 28.7% 26.0% 25.3% 24.8%
Tony Tan Keng Yam 34.0% 33.0% 34.8% 35.8% 36.4% 36.6%
Tan Kin Lian 5.3% 5.9% 5.1% 5.7% 4.9% 5.9%
Rejected ballots 61 52 77 52 41 59
Total Valid Votes 3,236 3,074 3,198 3,539 2,946 3,434
Total votes cast 3,297 3,126 3,275 3,591 2,987 3,493

See http://bit.ly/ocqvUB for complete data set.

Advertisements

Call for volunteers: Micro-polling results for Presidential Elections 2011

UPDATE: Results by polling district for six counting centers are posted at
https://stngiam.wordpress.com/2011/08/29/flash-results-micropolling-results-of-presidential-elections-2011/

I’m looking for volunteers who are able to help me increase the transparency of the electoral process in Singapore by volunteering as counting agents in the upcoming Presidential Elections and sharing with me detailed election results by polling district. Even though there is no legal prohibition against revealing precint-level elections results, the Elections Department only publishes aggregated results for entire constituencies – which in the case of GRCs can comprise over 100,000 voters. For more details, see my earlier post, GE2011: Micro-polling results from one counting centre.

In the case of the upcoming Presidential Elections, I expect that they will only publish figures for the whole of Singapore. This is unfortunate, because detailed results would be a very valuable resource for political scientists and other researchers trying to understand voting behaviour in Singapore. (OK, yes, it would be very valuable for politicians, too, but right now, only one party has the resources to be able to collate electoral data across Singapore. And if you believe them, they are not endorsing any presidential candidate so would not have any counting agents in the counting centres to report results to them. OK, this also requires me to assume that there is a separation between party and government, but lets’s not get into that for now.)

Please contact me at PE2011@ngiam.net if you are interested in participating in this project. What you will need to do:

  1. Sign up as a Counting Agent for any candidate. Doesn’t matter which one. This is a non-partisan project, so ideally there will be a spread of participants across the candidates.
  2. Print out the Excel template in this link (http://db.tt/90PR8OT) before you go for counting. Things will get rushed when the results start coming out so best to prepare in advance.
  3. There will be 4-6 counting tables at each counting centre. Be sure to record the counting place number, polling district number and the name of polling station for each of the tables.
  4. Record the number of votes for each candidate for each of the counting places when the results are announced. I expect the number of spoilt votes to be higher than in the last GE, so be sure to also record the number of rejected ballots.
  5. Send the results back to me at PE2011@ngiam.net. For the moment, email will do but if response is overwhelming I will think of something else.
Finally, if anyone has any connections to the candidates’ campaigns and is able to help me obtain the list of counting centres and polling districts, I would really appreciate it so that I can perform some error checking.
Please send this link to anyone you know who will be acting as a counting agent and email me at PE2011@ngiam.net if you have any questions.

GE2011: Micro-polling results from one counting centre

In an interview after the May 2011 General Election, a retiring PAP Minister alluded to the oaths of secrecy taken by candidates and advised reporters not to publish suggestions that any particular division in a Group Representation Constituency (GRC) had lower support for the ruling party because that could not be verified and might be wrong.[1]

The Minister was only partially correct. The oath of secrecy that she was referring to is a reminder to candidates and their agents that individuals’ votes are secret. Regardless of whether an oath is taken or not, the Parliamentary Elections Act prohibits anyone from attempting to determine who any particular voter voted for [2]. This is one of the ways in which the law backs up the principle “Your vote is secret.”

However, while the law prohibits any attempts to determine how a particular person voted, it does not prohibit disclosure of the total vote count for a Polling District. The Assistant Returning Officer is specifically required to announce results over the table at the conclusion of counting of the votes for a polling district, in the presence of any of the candidates or counting agents who are there. Again, this is one of the means by which we ensure transparency of the vote count. It is the responsibility of candidates to total up the reports from all their counting agents and compare that with the final results announced by the Returning Officer.

As PM Goh Chok Tong put it in 1997, “Each party- Workers’ Party, People’s Action Party – will have counting agents at the counting station. It is not something secret, which the PAP knows and the other side does not. It is transparent, it is clear as daylight. If a counting agent is alert when votes are counted, they will know how each precinct votes.”[3]

Prior to the 1997 general election, all ballot papers for a constituency were mixed together prior to counting. By 1997, however, GRCs had grown so large that it was clearly impractical to mix together over 100,000 ballot papers. Vote-counting was therefore decentralized but with the safeguard that the smallest unit at which results could be resolved would be the polling district. Polling districts appear to have shrunk over time[4] but each polling district still consists of 2,000-4,000 voters. While this is adequate to ensure the secrecy of any individual’s or household’s vote, the government has since 1996 linked upgrading of HDB neighbourhoods to the level of that neighbourhood’s support for the PAP. In the absence of public domain data on voting patterns by polling district, however, it is impossible for outside observers to verify the extent to which the government has followed through on this threat.

More recently, when electoral boundaries were redrawn prior to the 2011 election, seven polling districts were transferred from Aljunied to Ang Mo Kio GRC, and two more were transferred to Pasir Ris-Punggol GRC. The Workers’ Party (WP) accused the government of gerrymandering and stated that they had “significant support” in the polling districts that were transferred out of Aljunied [6]. Since they had contested that constituency before, however, they could have strengthened their case by revealing the actual vote percentages of those polling districts compared to other parts of the GRC in the 2006 election.

Can we infer from their silence that in fact the polling districts concerned did not have a higher vote for the WP, and that the government was not guilty of gerrymandering ? Not necessarily. It is possible that they did not reveal actual figures because they, like PAP Minister Lim Hwee Hua, were under the misconception that revealing district-level results was illegal. It is also possible that they did not have the figures because they had neither enough counting agents nor sufficient management capacity to be able to collate those results in 2006. Unfortunately, the present policy of the Elections Department is to only publish in the Government Gazette results at single-member constituency (SMC) or GRC level. There is no evidence that the Elections Department directly provides precinct-level results to any outside party, but the PAP almost certainly has sufficient resources and administrative capacity to compile their own lists of electoral results by polling district. I am not able to say whether the WP was able to compile those results on its own in 2006. Certainly, the media and other researchers do not have access to that information.

Thumbnail precis of different election officials

I worked in the civil service for several years and was arrowed to serve as an assistant to a Group Assistant Returning Officer (GARO) during the 2006 general elections. Had I not left the service I would have served as an Assistant Returning Officer (ARO) in Tampines GRC in the recent elections. The ARO is usually a graduate officer or first-level manager, and is overall-in-charge of one polling station from opening of polls early in the morning all the way through counting of ballots at night and deposition of ballot papers in the Supreme Court vault in the wee hours of the next day. Presiding officers (POs) are the nice people you meet guiding voters and handing out ballot papers at polling stations while Counting Assistants (CAs) perform the actual counting of ballots. At counting centres, the Senior Assistant Returning Officer (SARO) oversees the counting of votes from 4-6 polling districts. The SAROs, who are typically director-level officers, report in turn to Group Assistant Returning Officers (GAROs) who are Deputy Secretaries or equivalent. GAROs would be in charge of a GRC and one or two SMCs. And the GAROs report to the Returning Officer (RO) and YouTube sensation Yam Ah Mee. Like Yam Ah Mee, whose day job is Chief Executive Director of the People’s Association, all 16,000 or so of the above civil servants perform elections duty on top of their normal jobs. In addition several thousand other police and SCDF officers are also deployed during an election. There are only about 20 full-time staff in the Elections Department itself [7].

Microanalysis of results from Zhonghua Secondary School counting centre

I volunteered as a counting agent for the National Solidarity Party (NSP) at Zhonghua Secondary School in Marine Parade GRC. Zhonghua Secondary was the counting centre for six of the polling districts in the Braddell Heights division of Marine Parade GRC (Figure 1). While MA30, MA31 and MA32 contain a mix of HDB and private housing, each of MA28, MA29 and MA21 consist entirely of landed and non-landed private properties. While all of the polling districts counted at this centre had a lower proportion of votes for PAP than the rest of Marine Parade GRC, one of the districts (MA28) was not significantly different from Marine Parade GRC as a whole (see Table 1). Interestingly, while some commentators have suggested that private property residents would be more likely to vote for the opposition because they have fewer bread-and-butter worries, in this counting centre, the highest support for the PAP was in MA28 which is made up entirely of private properties (landed and non-landed). But the lowest support for PAP was also from an entirely non-HDB polling district, MA29, and that district is immediately adjacent to MA28. The second highest vote for the PAP was from a PD, MA30, which is a mix of private and public housing with HDB-dwellers in the majority.

Figure 1 – Polling districts counted at Zhonghua Secondary School Counting Centre

Table 1 – Polling results by polling district

Votes

Significance

Polling Station

Polling District

Housing type

NSP

PAP

p

Zhonghua Sec A (AZ01)

MA28

Private

1,134

(44.8%)

1,398

(55.2%)

0.149

Zhonghua Sec B (AZ02)

MA29

Private

957

(48.4%)

1,019

(51.6%)

0.000

St Gabriel’s Sec A (AZ04)

MA30

Mixed (more HDB)

1,419

(45.1%)

1,725

(54.9%)

0.044

St Gabriel’s Sec B (AZ03)

MA31

Mixed (more private)

1,611

(46.2%)

1,878

(53.8%)

0.001

Maris Stella High School (AZ05)

MA32

Mixed (more HDB)

989

(48.0%)

1,071

(52.0%)

0.000

Maris Stella High Schol (AZ06)

MA21

Private

1,235

(46.1%)

1,442

(53.9%)

0.004

Marine Parade GRC (Total)

59,926

(43.4%)

78,286

(56.6%)

The last column refers to the level of significance of Fisher’s two-tailed exact test for the proportion of PAP votes in that PD compared to Marine Parade GRC as a whole.

Spoilt votes

The proportion of rejected ballots (spoilt votes) was consistent across all the polling districts and were not significantly different from that in Marine Parade GRC as a whole (Table 2). In Tables 1 & 2, I have described MA30-MA32 as “mixed”, with both private and HDB housing present. In principle, it would be possible to calculate the exact number of electors residing in HDB vs non-HDB properties based on their addresses shown in the electoral register. However, I do not have access to the electoral register, and more importantly, the Parliamentary Elections Act specifically limits the allowable uses of information from the register [8]. Without information on the number of potential voters in each polling district, it is also impossible to calculate the turnout by polling district even though we know that the overall turnout for the GRC was 91.5%.

Table 2 – Rejected votes by polling district

Polling Station

Polling District

Housing type

Rejected ballots

As a proportion of total ballots cast

Zhonghua Sec A (AZ01)

MA28

Private

55

2.10%

Zhonghua Sec B (AZ02)

MA29

Private

40

2.00%

St Gabriel’s Sec A (AZ04)

MA30

Mixed (more HDB)

61

1.90%

St Gabriel’s Sec B (AZ03)

MA31

Mixed (more private)

71

2.00%

Maris Stella High Schol (AZ06)

MA21

Private

55

2.00%

Marine Parade GRC (Total)

3,082

2.20%

Note: I failed to record the number of rejected ballots in MA32

Non-governmental organisation Maruah has produced a video on the election process and several volunteers have also described their experiences as polling agents and counting agents [10, 11]. These are positive steps in demystifying the electoral process and reassuring Singaporeans of the integrity of the system. Unfortunately, I think disproportionate attention has been placed on the issue of spoilt votes, even in the mainstream media [12]. To be frank, some of the more egregious cases mentioned by commenters (not Alex Au himself) on Yawning Bread sound to me to be more like hearsay than reliable reports. The overwhelming majority of rejected ballots I saw had clearly been deliberately spoilt by voters. The most common method was simply marking both boxes, but I was amazed at how accurately some people could draw a cross freehand across the ballot paper with the centre of the cross exactly intersecting the line dividing the boxes for the two parties. Even in the few doubtful cases, I would not fault the decision of the ARO. Realistically, civil servants are sensitive to accusations of bias, and if anything, where the guidelines allow them to do so, they are more likely to sympathize with the underdog, not just in election administration but also in other areas of their day-to-day work. I am therefore very disappointed with the typically Singaporean clamour for “clear guidelines” in interpreting doubtful ballots. The narrower more specific the guidelines the more likely that nonsensical outcomes (e.g. interpreting any mark, even profanities, as a mark for a candidate) will result. The ARO of course has to act within workable parameters set by the Elections Department, but within those parameters, and after hearing the views of the counting agents present, the ARO on the ground has to be empowered to use her common sense and just decide.

Crowdsourcing precinct-level election results

As PM Goh Chok Tong put in in 1997, all candidates have the opportunity to observe voting patterns at the precinct level [3]. However, this opportunity can only be made use of by a party which has sufficient resources to put a counting agent in each counting place and is disciplined enough to compile the results after the excitement of the hustings have passed. In any case, even if an opposition party has reached the level of organisational capability to compile results for the constituencies that it contested, it would not know the results from other areas. In principle, opposition parties could pool their data and share polling results but given the present state of the opposition in Singapore, I doubt that this will happen any time soon. Even if the political parties were to get their act together, however, the public would still be left in the dark. This makes it impossible for independent observers to assess whether there has been any gerrymandering or the degree to which public monies are allocated according to voter support for the ruling party.

During the 2011 elections, I came to know of two attempts to crowd-source precinct-level results by collecting results directly from volunteer counting agents. Unfortunately both attempts came to naught. One of the challenges is that while there is no legal prohibition on releasing the results, political parties may want to treat the data as proprietary and prohibit their counting agents from releasing the information.

It is not clear yet if there will be a contested presidential election this year, but that would be another opportunity to test out the feasibility of an independent non-partisan analysis of micro-level voting patterns in Singapore. Given that candidates are required to resign their membership of any political party, I would hope that the candidates would embrace transparency and non-partisanship and allow their counting agents to participate in such a study or perhaps even to release the data that they collect themselves during the course of counting. Better yet, we can hope that the Elections Department changes its policy and publishes polling district level results on its own accord !

References

  1. ‘Noting that candidates were under an oath of secrecy, she told reporters “My only advice is to tell you not to continue to publish things that you can’t verify because they could be wrong.” Teo Xuan Wei, Today, 12 May 2011, “It is a surprise for us that the resentment is so deep”.  http://www.todayonline.com/Singapore/EDC110512-0000330/It-is-a-surprise-for-us-that-the-resentment-is-so-deep
  2. Parliamentary Elections Act, Section 56 (5) Every officer, clerk, interpreter, candidate and agent in attendance at the counting of the votes shall maintain, and aid in maintaining, the secrecy of the voting, and shall not —
    (a) attempt to ascertain at the counting the number on the back of any ballot paper; or
    (b) communicate any information obtained at the counting as to the candidate or group of candidates, for whom any vote is given by any particular ballot paper.
  3. Chua Mui Hoong, Straits Times, 1 Jan 1997, page 1, “PM: Precincts with greater support get upgraded first”
  4. The average number of voters in each polling district in Cheng San GRC in 1997 was 5,200. Straits Times, 1 Jan 1997,”GE 97: On the drawing board”
  5. Chua Mui Hoong, Straits Times, 12 Jan 1997, page 3, “Let’s weigh the balance of tying voters’ support to programmes”
  6. http://wp.sg/2011/02/workers’-party’s-response-to-the-electoral-boundaries-review-committee-report/
  7. http://www.elections.gov.sg/about.html
  8. Parliamentary Elections Act, Section 21A (1) Any candidate or his election agent who or any political party which, on payment of any fee, acquires from the Registration Officer any copy (whether in printed or electronic form) of any register of electors —
    (a) shall use any information recorded in the register only for communicating with electors;
    (b) shall not use any information recorded in the register for commercial purposes; and
    (c) may disclose any information recorded in the register to others only after obtaining their written acknowledgment that they are bound by the restrictions in this subsection.
  9. http://maruah.org/2011/05/03/votewithoutfear/
  10. http://yawningbread.wordpress.com/2011/05/08/counting-agent-me/
  11. http://yawningbread.wordpress.com/2011/05/10/counting-agent-me-too/
  12. Anna Teo, Business Times, 14 May 2011, “In S’pore’s GE, ‘X’ marks the spot, right ?”
Revision history
8 June 2011 – The significance level in the last column of Table 1 had been mistakenly calculated with spoilt votes included in the denominator. The p-value for MA30 was thus originally reported as 0.071 when it should be 0.044 if only valid votes are included in the total. While this pushes MA30 below the customary 0.05 threshold for significance, it does not change any of my observations regarding the relative strength of support for the PAP among the six PDs.