Technical Issues with Electronic Registration System for Polling that Resulted in Removal of Voter Names from Registers of Electors

Mr Pritam Singh asked the Prime Minister in view of the Elections Department’s press statement dated 24 August 2023 regarding the wrongful removal of names from the Registers of Electors after General Election 2020 (a) how many times has this happened in the aftermath of past elections or by-elections; and (b) what steps will be taken to prevent a re-occurrence.

Ms He Ting Ru asked the Prime Minister regarding voters who were removed from the electoral register due to an error (a) what was the total number of voters affected; (b) whether there was any individual notification exercise to expunged voters in reminding them to add their name back to the register since the last such exercise in 2019; and (c) why was a tally of the number of NRICs scanned and the number of votes collected after the last general election not conducted as a first instance check.

Mr Liang Eng Hwa asked the Prime Minister (a) what are the reasons resulting in the names of about 200 Singaporeans not being reflected in the Registers of Electors eligible to vote at Presidential Election 2023 despite voting in General Election 2020; and (b) what are measures taken to prevent such recurrence.

Thank you, Mr Speaker. Let me first set out the Elections Department's (ELD) key executional objectives at every election, which is to ensure: firstly, that only bona fide voters are issued with ballot papers; secondly that no extra ones were put in; and thirdly that all votes cast are counted, so that the election result is determined by the eligible voters for that election.

To achieve this, we have put in place a set of procedures that need to be complied with. Let me explain.

Upon the voter’s arrival at the polling station, the election official will verify his identity and eligibility to vote at the polling station, and then register the voter’s attendance via a hardcopy polling station register or the electronic registration system. Thereafter, the voter is directed to the ballot paper issuance counter. The election official issuing the ballot paper is required, by law, to read out the name and voter serial number of a voter before the voter is issued a ballot paper. This allows the polling agents to mark their copy of the register and provide an additional check to ensure that only eligible voters in the register for that polling station are individually issued with a ballot paper. As a result, when the votes are tallied for an election, we can be confident that the outcome is the one decided by the eligible voters for that election.

After the end of each election, ELD compiles the list of voters registered to have voted on Polling Day and compares them against the list of voters in the certified Registers of Electors for the election. Those who were not registered as having voted on Polling Day are included in the list of non-voters, which is published after each election for inspection. The non-voter list contains the names of people who have not turned up to vote. The vast majority of them did not vote either because they were not in Singapore or were ill on Polling Day. For these cases, their names were not crossed out in the hardcopy polling registers as they did not turn up at the polling stations.

After each election, there would be a few voters who claimed that they had voted even though upon checking, their names were not crossed out in the hardcopy polling registers. Thus, in reply to Mr Pritam Singh’s question, prior to the introduction of electronic registration at General Election 2020 (GE2020), ELD encountered only a few cases of voters who were on the non-voter list who claimed that they had voted.

For GE2020, GE2015 and GE2011, ELD recorded about 111,000, 155,000 and 147,000 non-voters respectively. As can be seen, the non-voter list for GE2020 was much smaller, rather than bigger, than the two previous GEs.

As per every election, ELD published the non-voter list for voters to verify about three months after GE2020. This was publicised via a press release on 5 October 2020, which was covered by the mainstream media.

Ms He asked why a tally of the number of NRICs scanned and the number of votes cast after the last general election was not conducted as a first instance check. Given the much lower number of non-voters for GE2020 compared to the two general elections before, ELD did not suspect that an unusual number of voters had not been recorded as registered. Checking the registration figures at the end of Polling Day may also not be instructive – and this is because the process to ensure voting secrecy means that we will not be able to identify individual voters, including those who had not been registered accurately and would therefore be listed as non-voters in due course. Furthermore, every voter receives only one ballot paper and the ballot boxes which contain all of the cast ballot papers are always in the custody of election officials, properly sealed, with polling and counting agents able to witness this process, and secured while in transit.

Ahead of Presidential Election 2023 (PE2023), ELD publicised that the revised Registers were available for inspection from 15 June 2023 and that the certified Registers which contains updated information from the inspection period were available for inspection from 21 July 2023. These exercises were carried out to enable Singaporeans to check their details and restore their names to the Registers ahead of the election.

From October 2020 to the day of the Writ in August this year, some 32,000 voters checked and restored their names. Similar to previous exercises, there were very few who requested ELD to restore their names due to inaccuracy in the recording of their voting status in GE2020. As such, ELD did not suspect that there were significant inaccuracies in the recording.

The list of non-voters had been opened for inspection shortly after the GE2020 and the public were reminded that the registers were available for inspection in June and July 2023 before the Writ for PE2023 was issued. Voters need not wait for these periods, but are able to check their status in the registers all year round either through Voter Services on ELD’s website, at any of the over 100 community centres or clubs or by calling ELD.

However, there remained some voters who did not check that their names were on the Registers. After the Writ was issued for PE2023 and up to Polling Day, 1,093 Singaporeans informed the ELD that they did not receive their poll card even though they said that they had voted in GE2020. For context, this number averages one voter per polling station, which had an average of 2,400 registered voters in GE2020. So, this is about one voter per polling station with an average of 2,400 registered voters in GE2020. Further, ELD’s analysis of the distribution of the affected voters showed that there is no pattern across the polling stations and that the 1,093 cases were distributed across more than 500 polling stations.

ELD has investigated this and concluded that the most probable cause of the error was that the NRIC details of these affected voters had not been captured properly by the electronic registration system, which was only introduced from GE2020. This is because the proper registration procedure may not have been followed for the affected voters, especially in some of the more crowded stations.

Let me explain. A voter’s NRIC and poll card are first checked by election officials at the registration counter before the NRIC is scanned to register the voter. When the NRIC of a voter is scanned by the barcode reader, an alert or "beep" sound will be emitted and a digital screen will flash an "OK" button, indicating that the voter data has been successfully captured in the system. At this point, the election official should tap the "OK" button to return to the registration screen so that the NRIC of the next voter can be scanned.

If the NRIC of the next voter is scanned when the screen has not returned to the registration screen, the alert sound will still be emitted, as this indicates that the NRIC has been scanned. However, the record of this voter would not have been captured. It is likely that in their effort to clear the queues at some polling stations quickly in GE2020, some election officials might have missed out this step. As a result, the registration of the affected voters was not captured.

For PE2023, we implemented a new electronic registration system with a simplified process. Election officials need not press any button to return to the registration screen to register the next voter. Therefore, the issue of non-captured registrations as a result of not pressing a button would no longer arise.

Dr Tan asked what is being done to ensure that all who voted were registered, and Dr Tan, Mr Singh and Mr Liang also asked what will be done to prevent a recurrence where those who had voted were inadvertently removed from the Registers. Going forward, besides making a public call for voters to check their status when the non-voter list is published for verification, ELD will institute an additional measure to individually inform Singaporeans on the non-voter list via mail and through SG Notify in Singpass to verify their status, and to restore their names to the Registers if need be. We thank Ms Ng Ling Ling for her suggestion.

Dr Tan asked about the number of polling stations that experienced technical issues with the electronic registration system during the recent PE2023. Based on ELD’s preliminary investigations, there was a 30% average loss in device connectivity in the first hour of polling and this reduced to 16% by 10.00 am. There was no evidence that this was caused by cyber-attacks. Instead, this could have been partly contributed by the surge in the volume of transactions during the morning peak, where about 52% of the total number of voters had already voted in the first four hours, compared to 32% in GE2020.

For PE2023, ELD had endeavoured to reduce queues and waiting times for the convenience of voters.

The number of polling stations was increased by 15% to 1,264 stations, from the 1,097 stations in GE2020. This reduced the average number of voters assigned to a polling station from about 2,400 voters in GE2020 to 2,150 voters in PE2023.

Each polling station also had more e-Registration devices than before, from between two and four devices each, proportionate to the number of voters in the polling stations, with an additional spare device each for contingency. For reference, a polling station with only two e-Registration devices can process at least six voters per minute on average, which is a 50% higher throughput compared to manual registration which does so at the rate of an average of four voters per minute.

I would like to also take this opportunity to thank our election officials who adapted quickly and kicked in the contingency measure as planned to register voters using the hard copy registers, which they had been trained to do. As a result, by about 9.55 am, ELD had reported that queues had improved at most stations. We would also like to thank voters who had to queue longer during the morning of peak hours Polling Day for their patience and understanding.

The electronic Registration system seeks to reduce waiting times and automate attendance taking for the benefit of voters. ELD will continue to enhance the system so that the registration data are captured swiftly and accurately.

Mr Speaker, I thank the Minister for his answer. Let me first declare that I have Clementi residents who were affected and shared their concerns with me as their Member of Parliament. I declare that I have a Clementi resident who remembered voting in 2020, but in 2023, found she was not on the Registers belatedly and she regrets not having had the chance to vote and choose her President at the ballot box that day. I also have Clementi residents who saw how hard our election officials, how hard our public officers worked, and the heroic efforts made by our public officers when technical difficulties arose.

Can I ask the Minister supplementary questions on what I will call the "3Rs": reliability, resilience, responsibility?

Firstly, on reliability. When an electronic registration system is tested, is it benchmarked for 100% accuracy in the tender requirements, at least as good as paper, if not better?

On resilience, can I ask ahead of the Presidential Election, was there a proper full-dress rehearsal, incorporating set-up, log-in, getting the terminals up and ready, and then, comprehensive load-testing beyond what would be the traditional expected loads, to stress test the system? This is because I have residents who have concerns about whether someday, there may be cyber interference and we know what advanced persistent threats are out there.

Lastly, the third "R", on responsibility. I note that the ELD website mentions "Your vote decides the future of Singapore". How we design these systems also shapes the voting process as well. Can the Minister assure us that ELD will do its level best to learn from this and continue improving, in particular, contractor specifications, tender requirements, so that the situations like these can be avoided in future?

Thank you, Mr Speaker, Sir. Let me respond to the three supplementary questions.

I will start with the third one. Yes, definitely, ELD will continue to improve just as per every election. We will take in the debrief and feedback from all members of the public and also Members in this Chamber, and ELD will continue to seek to strive better, be it process wise or the technical issues that we face. So, that is the third question that Dr Tan asked.

The second one, on the issue of reliability, yes, indeed, ELD has specifications on the reliability standards. And let me perhaps take this question a bit further to explain what we mean by "reliability standards". The reliability of a system depends on a few parts and to answer the second question, all these were also stress tested. First, we have to make sure that the hardware is functioning well. Second, we have to make sure that the connectivity between the over 4,500 systems all around the island, is connected well and has certain redundancy. And third, we have to make sure that the software is also working fine.

Sir, as I have explained, the first is that, for this particular PE, we have ruled out the issue of a cyber-attack. The second is that we have also checked that the hardware has worked well and it was not the main contributing factor to the slowness in the first few hours, the slowness that started at 8.30 am that lasted until about 10.00 am during Polling Day. The third thing is about connectivity. For all the polling stations, we had double redundancy to make sure that we did not rely on any particular network, and this was also stress tested during the run-up. In fact, ELD undertook the technical tests. Prior to Polling Day itself, they conducted full-dress rehearsals with all the devices on, to check the connectivity and hardware issues; and they were satisfied that they were all in working condition.

ELD, in the initial findings, has pretty much narrowed this down to a technical issue that has to do with the software. And let me explain it in layman's terms as follows. In ELD's stress tests, they stress test a certain volume of transactions within a certain amount of time. So, this election, notwithstanding the fact that within the first four hours, 52% of the people have voted compared to 32% in the last GE, we had stress tested that scenario.

But there was one particular feature that perhaps, in hindsight, the technical officers could have looked deeper into. And that is when all the devices were sending messages to the central server and when the inordinate amount of surge happened, some of the messages did not get through. And when the packets of messages did not get through, the system kept trying. And because as the system kept trying, it continued to pile up a backlog of packets that have to be cleared.

When ELD technical officers realised this, after 8.30 am when this thing happened, they quickly increased the memory and computing power capacity, and this helped us to quickly resolve the issue. So, if you like, some messages did not get through, they continued to pile up, creating a backlog and it caused a bit of a cascade effect. But with the quick discovery of this problem, the ELD officers were able to quickly increase the computing power and also the memory in order to clear the backlog. But once the backlog had set in, it took us about an hour plus to clear the backlog and return the clearance speed to normal.

But I must say, even with assurance about the redundancy, ELD's planning norms are such that in every polling station, even if the electronic registration systems failed, we will be able to clear the queues with the manual processes, as we had done in the past. So, the baseline is that manual processes will allow us to comfortably clear all the voters in that polling station. On top of that, if this is supplemented by the electronic system, it will help us to speed up the clearance even faster. Which is why, for this particular PE, even though there were 52% of the voters voting in the first four hours, by about 10.00 am in the morning, we had pretty much managed to clear all the queues and clear 52% of the voters within the first four hours.

Sir, between 11 and 24 August, ELD's press release of 24 August stated that about 200 Singaporeans had informed ELD that their names were not in the Registers of Electors even though they had voted in GE2020. The Minister said in his reply that this number is now 1,093. Based on the Minister's reply, will the ELD be sending letters and notifications to all non-voters in GE2020 who have not applied to put themselves back on the Registers of Electors to ensure that no voter has been struck off, due to the additional steps in the electronic registration system process that was employed in GE2020?

Mr Speaker, Sir. I thank Mr Pritam Singh for his supplementary questions. First, let me assure Mr Pritam Singh all the 1,093 voters who were left out or not registered properly for PE2023, all their names have been restored to the Registers. And now ELD is doing the reconciliation for PE2023. We will be sending all those in the non-voters list for PE2023 personal notifications, as I have explained, to encourage them to verify their status.

Thank you, Sir. Two further questions for the Minister. Firstly, given that voting is compulsory in Singapore, why does ELD not automatically update all the citizens' names on the Registers of Electors after each election? That will minimise some of these problems that we have just talked about.

My second question is what is the purpose of this administrative step for those who did not vote in the last election, to apply for their names to be restored to the Registers? Is this necessary or is the purpose still meaningful today, given that it can open up to potential operational slips and also additional administrative work for ELD and the voters as well, since they have to go and put their names back? So, is the purpose still a meaningful one?

Mr Speaker, Sir, we thank Mr Liang Eng Hwa for his feedback. Precisely because voting is compulsory, the Election Acts requires that the names of the non-voters be expunged from the Registers if they did not vote. This is in the Election Acts. And a penalty of $50 must be paid for one's name to be restored if he or she does not provide a valid reason for not voting. The penalty is waived if the voter is able to provide a valid reason such as being overseas on Polling Day. Otherwise, he or she will no longer be able to vote in future elections, or to stand nominated as a candidate for elections. In other words, the penalty for not voting and the process of expunging the names of non-voters, give effect to the notion of compulsory voting. Starting on a clean slate with no penalty will undermine this.

Thank you, Mr Speaker. I have two supplementary questions for the Minister. The first being, leaving aside secrecy, which of course we agree is very important, does ELD have data about how many total identity cards were scanned during GE2020 versus how many votes were eventually recorded as being cast during the election? That would actually give us a first instance indication of whether there is any discrepancy beyond the 1,097 voters that have contacted ELD to say that they voted but their names were erroneously removed from the Register. That is my first question.

My second question is, I note that in a reply to one of my Parliamentary Questions in 2020, the Minister stated then that, in 2019 and 2016, one-off exercises were done to remind voters whose names were not on the Registers to ask them to take action and to re-apply to go back onto the Registers. I am just wondering whether we can do a one-off sweep to all voters, including voters that were removed from the Registers prior to GE2020 and PE2023, to ask them to take action to put their names back on. And also, in that vein, whether or not we can make it standard practice after every election that ELD would then send out a notice to voters who did not vote, asking them to apply, so that you do not just rely on press releases?

Mr Speaker, Sir, let me answer Ms He Ting Ru's first question. At the end of every Polling Day, there will be a combination of electronic records in the current system plus some manual records. Because in instances where the electronic system malfunctions or fails for whatever reason, we will have to rely on the manual record. So, at the end of each Polling Day, we would not immediately have the numbers because there is the electronic dataset, which we have to reconcile with the manual set. So, that is the answer to the first question. And even if we can reconcile it immediately, it will not be able to tell us who are the non-voters accurately.

On the second question, indeed, it is not so much about a one-time sweep or when the voters can check their names. As I have explained in my reply, once the reconciliation is done, from that period until the next election when the Writ is issued, at any point in time, the voters can check their status. And we agree with you; we encourage all the voters to check their status anytime. It is not just only during the period when ELD publicises this. ELD will of course still make the publications, but the voters can go to any of the community centres (CCs) or the ELD website to check their status.

On top of that, this time round, we are specifically adding this extra step whereby those who are in the non-voters list will be encouraged to go and check their status way ahead of the next election.

Thank you, Speaker. To the Minister, just on my earlier supplementary question, the 200 and later on the number was revised to 1,093, these are individuals who got in touch with ELD to inform ELD that, "I am not on the list although I did vote in 2020".

My question is, again, based on your reply, the Minister said, following GE2020, there were 111,000 non-voters. And then, from October 2020 to August 2023, 32,000 individuals applied to restore their names. So, this 32,000 would have to be a part of this 111,000 who were non-voters. My point is there is possibly another 80,000 voters who, as of PE2023, could also be non-voters but who voted in GE2020?

I just want to confirm whether ELD is also going to reach out to everyone to ensure that they have not been erroneously struck off the voting list. I hope the Minister got the subject matter of my first supplementary question.

Let me explain the numbers. There are 111,000 non-voters. The vast bulk of these non-voters, be it in GE2011, GE2015 or GE2020, were either overseas or some of them could be very old and very frail and so, did not vote. They are the vast bulk of these. For the 32,000 who asked to be restored, chances are that most of them were perhaps overseas or ill on that day and they have asked to be restored.

I do not think you will have 111,000 minus the 32,000 and, therefore, you end up with about 79,000 persons who have voted. That would not happen because, from the numbers, you will know that this is not the case. Also, from the previous experience, we know that the number of people who are so-called registered wrongly is not in that order of magnitude.

Perhaps, I should explain, just to put aside any doubt that this has in any way affected the accuracy or validity of the results. Of course, the first thing I must say is this – ELD will endeavour to make sure that we minimise the error to as close to zero as possible. It is very difficult because, whether it is the manual or electronic system, sometimes, there will be an error. But we will endeavour to make sure that this error is the minimum possible.

But just to give a bit of context as to what the error rate is like. Just now, I have explained that a typical polling station will have about, say, 2,500 voters. And the error, as what we have described just now, on average, is about one per polling station, which means about one in 2,500. That is the margin of error.

If we look at our recent GE results, the closest contest, if I remember correctly, should be Potong Pasir in 2011. It is about 100-plus votes difference out of about 20,000 voters. So, that margin is about one in 200 or thereabouts. That is the winning margin, the closest ever. But our error, the closest ever, is more like one in 2,000 – or rather, less than one in 2,000.

So, I just want to assure the public that ELD has gone through the records and, while there is a margin of error, this margin of error is nowhere to the extent that it will call into question any of the election results that we had in recent memories.

Yes, indeed. So, 111,000 after the 32,000 – we will do the reconciliation again this time – exactly like what I said, we will send all those on the non-voters' list an individual reminder henceforth to remind them to make the check. So, we do not rule out that there could be a few more who may have voted but may have recorded wrongly.

And going forward, we will always do this process. That means, all those on the non-voters' list, for whatever reason they are on the non-voters' list, we will send them an individual reminder, on top of the public service announcements that we have made.