Legislative Council of the Hong Kong Special Administrative Region

ISE11/16-17

Subject:

constitutional affairs, electronic voting, electoral arrangement

In the most recent Legislative Council ("LegCo") Election held in September 2016, there were public concerns over issues such as long queues for casting votes at some polling stations after closing time and discrepancies between the cumulative voter turnout and the ballot papers issued at some polling stations. To address these concerns, the Electoral Affairs Commission ("EAC") recommended that "computerisation of the electoral process should be the way forward", including electronic voting ("E-voting").^{1Legend symbol denoting See Constitutional Affairs Bureau (2017).}
Over the past two decades, E-voting has been repeatedly discussed at meetings of the Panel on Constitutional Affairs.^{2Legend symbol denoting Since the 1990s, the Hong Kong Government has stated its commitment to introducing more technology in various elections to facilitate voting and counting. Two feasibility studies on the introduction of E-voting system were commissioned by the Government in 1995 and 2000 respectively. While counting machines have been introduced to the Election Committee Subsector Elections since 1998, there was little progress in the implementation of E-voting. See Minutes of Meeting of the Panel on Constitutional Affairs of the Legislative Council (1996) and Constitutional Affairs Bureau (2000).} After the 2012 LegCo Election, the Registration and Electoral Office ("REO") had also commissioned a consultancy study on the feasibility of implementing E-voting in Hong Kong. At the Panel meeting held on 25 January 2017, the Government, in response to Members' enquiries, said that it was "not yet in a position" to implement E-voting, largely due to concerns over information technology ("IT") facilities and its security, and the time and cost of installing E-voting network and system at a large number of polling stations. The Government considered that more studies should be conducted to assess the use of IT in the electoral process.^{3Legend symbol denoting See Constitutional Affairs Bureau (2017).}
While 30 places around the world have experimented with E-voting, only 15 places have formally adopted it in their elections, according to the most recent stocktaking study on E-voting.^{4Legend symbol denoting See International Foundation for Electoral Systems (2012).} Belgium is one of the pioneers in E-voting, with its application in its elections as early as 1994. This issue of Essentials first reviews the key issues of E-voting in elections, followed by a discussion on its development in Belgium.

Key issues faced by E-voting in elections

Major types of E-voting: By and large, E-voting can be broadly classified into two types, namely (a) E-voting through electronic devices installed at polling stations to facilitate vote casting and counting; and (b) Internet E-voting which allows voters to use generic technology (e.g. Internet or Short Message Service) to cast their votes anywhere. While the latter is apparently more voter-friendly, it is technically more demanding and hence is less commonly applied at this juncture.
Potential benefits of E-voting: There are potential benefits of both types of E-voting. They include: (a) increasing the accuracy of voting by eliminating invalid and incorrectly cast ballots as E-voting software can be configured to ensure only valid ballots are cast and recorded. (b) expediting vote counting and the results could be available soon after the polls close without the lengthy manual counting process; (c) increasing the accessibility of voting, while voting technology would appeal to tech-savvy voters, it also makes voting easier for some voters such as those with disabilities;^{5Legend symbol denoting For example in New South Wales, Australia, E-voting is adopted specifically for voters who are: (a) blind and have low vision; (b) suffered from other disabilities; and (c) living more than 20km from a polling station or overseas to cast their votes from any computers with access to the Internet or telephone. It serves to improve their right to a secret ballot as some of these people were only able to vote through the assistance of a friend or relative in the past. See New South Wales Electoral Commission (2017).} and (d) reducing administrative cost in the long run, despite a high initial set-up cost for installation of the network infrastructure.
Potential risks of E-voting: At the moment, there are views that the technology for E-voting is not entirely mature and proven, posing a host of risks and concerns in the election process. First, there are risks of system reliability in the election process, as disturbances in power supply, system failures or even hacking could occur anytime. Secondly, there are risks of data manipulation and election fraud, as voting through electronic devices or Internet are not entirely transparent, resulting in vulnerability to frauds or manipulation of votes unrecognizable to voters or to polling stations' supervisors. Thirdly, along with the integrity concern of the system, voting behaviour of individual voters might be revealed, violating the principle of anonymity in elections. Fourthly, there are concerns on whether the IT proficiency of certain voters (e.g. the elderly) is good enough to cast the votes through electronic devices themselves.
Recent global developments: In face of these concerns, a few European countries which once experimented with E-voting had reverted to conventional form of paper ballot in recent years, including Ireland (2004), the Netherlands (2007) and Germany (2009).^{6Legend symbol denoting The Netherlands discontinued the use of E-voting machines in 2007 after a group of activists identified a number of security flaws that allowed them to manipulate the results unnoticed. Germany ended E-voting in 2009 following a court case that ruled the E-voting machine unconstitutional as the results from the voting machines were not verifiable. Ireland called off its planned use of E-voting machines in 2004 following the negative review of the E-voting system made by the Commission on Electronic Voting. See International Foundation for Electoral System (2012).} That said, many countries are still considering how to apply technologies in electoral processes, including Finland, Spain, Italy, Greece, Norway, and some developing countries in Asia and Central America.

Earliest form of E-voting in Belgium during 1994-2011

In Belgium, nationwide E-voting in elections was first applied in 1994^{7Legend symbol denoting In Belgium, municipal governments are responsible for the organization of elections and it is up to them to decide which voting method (either E-voting or traditional paper-based voting) to use for their own municipalities. In the most recent elections held in 2014, 209 of the 589 municipalities opted for electronic voting. See Service Public Fédéral Intérieur (2017).} to reduce administrative cost and expedite vote counting.^{8Legend symbol denoting Following an initiative from the Minister of the Interior in 1989, two different E-voting systems were tested in two electoral cantons, namely Waarschot in Flanders Region and Verlaine in Wallonia Region. One of the systems being tested was based on a touch panel, whereas the other one was based on a magnetic card and a voting machine with a light pen, which was chosen at the end. See Gonzalez C.V. (2012).} A dedicated Law on Automated Voting (Loi du 11 avril 1994 organisant le vote automatisé) was enacted in April 1994 to establish the general framework for E-voting in the country. The earliest mode of E-voting was applied in the country for almost two decades till 2011.
Under this earliest mode of E-voting in Belgium, voter would be given a magnetic card for voting at a polling station, after the verification of identity. The voter would then insert the magnetic card into a voting machine to register his/her choice of candidates. The voter would then take out the coded card from the voting machine and insert it into an electronic ballot box which would automatically count the votes at the end of election (Figure 1).

Figure 1 – Voting machine in Belgium in 1994

Source: University of Leuven, Belgium.

Response to E-voting in Belgium was encouraging, with the proportion of voters voting through electronic means rising from 20% in 1994 to 44% in 1998. Apart from cost savings, it could also shorten the time required for counting votes, from six to eight hours under the paper-based system to between two and five hours under E-voting.

The earliest form of E-voting in Belgium also attracted a number of concerns and criticisms, primarily on the public's perception of the opacity of the E-voting process. In response, the Belgian government put in place the following measures:

(a)	Setting up a regulatory body to oversee E-voting: The Law on Automated Voting was amended in 1998 to create a new and independent regulatory body, namely the College of Experts. It comprised IT and legal experts appointed by the federal and regional parliaments. Their responsibilities include controlling the use and functioning of the E-voting system before and on the election day.^{9Legend symbol denoting The 1994 Law provided the College of Experts for access to both the hardware and software 40 days prior to the elections and 15 days after the elections. On the election day, they have access to any polling station. The College of Experts delivers a report within 15 days after each election. See Gonzalez C.V. (2012).}
(b)	Enhancing transparency of the system: To address the transparency concern, starting from 2003, arrangement has been made to appoint an external consultant to certify both the hardware and software of the E-voting system before each election. The source code that powers the E-voting system would also be published on the government website after the election for public scrutiny.
(c)	Enhancing the trust of voters: In 2003, a ticketing system was added to the aforementioned voting machines as a pilot study. This ticketing system produced a paper trail for voters to view behind a glass whether the choices made on the voting machine actually matched with their own. The verified ticket would then be deposited into a box for possible audits and recounts, if needed.

New form of E-voting in Belgium adopted since 2012

Despite the above enhancement measures, there were still concerns over the transparency of the E-voting system in Belgium. An in-depth comparative study on E-voting systems ("the BeVoting Study") was thus commissioned by the Belgian government in 2006. A consortium of seven Belgium universities was appointed to identify the best E-voting system with respect to the European Union ("EU") standards^{10Legend symbol denoting In 2004, the Council of Europe issued its Recommendation on Legal, Operational and Technical Standards for E-voting. As the first international legal instrument that deals with E-voting, the Recommendation is structured into three main parts: (a) common legal standards for E-voting that comply with the principles of universal, free, equal and secret suffrage; (b) operational standards that apply to all phases of the electoral process; and (c) technical requirements for accessibility, interoperability, security and monitoring of voting. See Council of Europe (2008).} and the Belgium electoral law.
After an examination on the latest innovations in E-voting technology and a balance against public concerns, the consortium recommended in 2008 an "improved paper-based voting system". Following the legislative amendment to the Law on Automated Voting in 2011, the new system was put to a large scale pilot test,^{11Legend symbol denoting A large-scale pilot test was conducted across Belgium in October 2011 in order to check the reliability of the new E-voting system. The pilot test involved 6 134 votes cast in 22 different locations with 90 voting machines with polling stations operated in the same manner as real elections by a total of 130 election staff. The pilot was concluded a success with the public finding the system simple and easy to use, despite the occurrence of some minor technical issues. See Service Public Fédéral Intérieur (2011) and Gonzalez C.V. (2012).} after certification by an external consultant.

In 2012, the new E-voting system was formally launched. It comprises five electronic elements, namely (a) electronic ballot boxes; (b) voting machines; (c) smartcards; (d) presiding officer's machines serving administrative functions; and (e) USB pen drives. Below is the new voting procedural sequence (Figure 2):

(a)	The presiding officer of each polling station will activate the E-voting system in the respective polling station by inserting a USB key and a passcode he/she receives a few days prior to the election day in order to boot up the equipment before the poll opens.
(b)	Each voter will be given a smartcard to activate the voting machine in the voting booth, after identity verification by the polling staff. The voter then inserts the smartcard into the voting machine and selects candidates from the lists shown on the touch screen of the voting machine. After completing the voting process, the voting machine will print out a paper ballot containing two parts: a human-readable part and a machine-readable barcode. The voting machine would return the smartcard to the voter, which does not store the voter's choice.
(c)	The voter then goes to the electronic ballot box located in a separate position, scans the machine-readable barcode on the paper ballot with the scanning unit, before depositing it into the ballot box in case a recount is required. The scanning unit is connected to the presiding officer's machine that automatically stores the vote cast in form of encrypted data on two USB pen drives, without the functionality to access vote data.
(d)	Once the polling finishes, the USB pen drives and the electronic ballot boxes would be sealed and transported to the centralized ballot reading center. The ballots would be decrypted under supervision and aggregated for results.

Figure 2 – New E-voting system in Belgium from 2012 onwards

Source: Service Public Fédéral Intérieur.

Under the new E-voting system, as the vote is registered on a paper instead of magnetic card, voters can easily verify the correctness of the vote cast electronically, while the printed paper ballot can be used for recount when required. Moreover, safeguard component was added to prevent double-voting, and mirrored secured storage (i.e. the second USB pen drive) can provide physical and unalterable evidence of the electronic voting process. All of them are meant to address the concerns over transparency, verifiability and security of the earlier system, enhancing voter's trust in E-voting.
In terms of cost, each E-vote costs around €13 (HK$112), as compared to €4.5 (HK$39) per E-vote cast under the earlier system and €1.5 (HK$13) for each paper vote.
In the most recent election held in 2014, some 49% of the voters in Belgium voted electronically, while the remaining 51% used traditional paper ballots. Even with rigorous testing and certification of the system prior to the election, about 10% for the voting machines were not calculating the votes correctly, affecting close to 2 000 votes or 0.06% of overall votes. While the affected votes did not change the election results, they caused a delay in the announcement of the election results by two days. Following the incident, the Belgian government and the College of Experts had conducted a root cause analysis and published a 120-page review report, proposing an array of technical enhancements encompassing different aspects of the E-voting process, including software upgrades, counting tabulation data encryption and decryption methods and procedures, voter identification on smartcard, voting logistics, security and control of the system, and training for polling staff.

Prepared by Phoebe TANG
Research Office
Information Services Division
Legislative Council Secretariat
21 February 2017

Endnotes:

1.	See Constitutional Affairs Bureau (2017).
2.	Since the 1990s, the Hong Kong Government has stated its commitment to introducing more technology in various elections to facilitate voting and counting. Two feasibility studies on the introduction of E-voting system were commissioned by the Government in 1995 and 2000 respectively. While counting machines have been introduced to the Election Committee Subsector Elections since 1998, there was little progress in the implementation of E-voting. See Minutes of Meeting of the Panel on Constitutional Affairs of the Legislative Council (1996) and Constitutional Affairs Bureau (2000).
3.	See Constitutional Affairs Bureau (2017).
4.	See International Foundation for Electoral Systems (2012).
5.	For example in New South Wales, Australia, E-voting is adopted specifically for voters who are: (a) blind and have low vision; (b) suffered from other disabilities; and (c) living more than 20km from a polling station or overseas to cast their votes from any computers with access to the Internet or telephone. It serves to improve their right to a secret ballot as some of these people were only able to vote through the assistance of a friend or relative in the past. See New South Wales Electoral Commission (2017).
6.	The Netherlands discontinued the use of E-voting machines in 2007 after a group of activists identified a number of security flaws that allowed them to manipulate the results unnoticed. Germany ended E-voting in 2009 following a court case that ruled the E-voting machine unconstitutional as the results from the voting machines were not verifiable. Ireland called off its planned use of E-voting machines in 2004 following the negative review of the E-voting system made by the Commission on Electronic Voting. See International Foundation for Electoral System (2012).
7.	In Belgium, municipal governments are responsible for the organization of elections and it is up to them to decide which voting method (either E-voting or traditional paper-based voting) to use for their own municipalities. In the most recent elections held in 2014, 209 of the 589 municipalities opted for electronic voting. See Service Public Fédéral Intérieur (2017).
8.	Following an initiative from the Minister of the Interior in 1989, two different E-voting systems were tested in two electoral cantons, namely Waarschot in Flanders Region and Verlaine in Wallonia Region. One of the systems being tested was based on a touch panel, whereas the other one was based on a magnetic card and a voting machine with a light pen, which was chosen at the end. See Gonzalez C.V. (2012).
9.	The 1994 Law provided the College of Experts for access to both the hardware and software 40 days prior to the elections and 15 days after the elections. On the election day, they have access to any polling station. The College of Experts delivers a report within 15 days after each election. See Gonzalez C.V. (2012).
10.	In 2004, the Council of Europe issued its Recommendation on Legal, Operational and Technical Standards for E-voting. As the first international legal instrument that deals with E-voting, the Recommendation is structured into three main parts: (a) common legal standards for E-voting that comply with the principles of universal, free, equal and secret suffrage; (b) operational standards that apply to all phases of the electoral process; and (c) technical requirements for accessibility, interoperability, security and monitoring of voting. See Council of Europe (2008).
11.	A large-scale pilot test was conducted across Belgium in October 2011 in order to check the reliability of the new E-voting system. The pilot test involved 6 134 votes cast in 22 different locations with 90 voting machines with polling stations operated in the same manner as real elections by a total of 130 election staff. The pilot was concluded a success with the public finding the system simple and easy to use, despite the occurrence of some minor technical issues. See Service Public Fédéral Intérieur (2011) and Gonzalez C.V. (2012).

References:

1.	Chamber des Représentants de Beglique. (2014) Rapport de Collège d'Experts Chargés de Contrôle de Système de Vote et de Dépouillement Automatisés.
2.	Constitutional Affairs Bureau. (2000) Systems of voting and voter registration for LegCo elections.
3.	Constitutional Affairs Bureau. (2017) Electoral Affairs Commission Report on the 2016 Legislative Council General Election.
4.	Council of Europe. (2008) Compliance of the BeVoting Study with the Recommendation (2004) 11 of the Committee of Ministers of the Council of Europe to member states on legal, operational and technical standards for E-voting.
5.	Electoral Affairs Commission. (2017) Report on the 2016 Legislative Council General Election.
6.	European Commission. (2014) Bug in Belgian voting machine should have been found.
7.	Gonzalez, C. V. (2012) New Belgian E-voting System.
8.	International Foundation for Electoral Systems. (2012) International Experience with E-Voting: Norwegian E-Vote Project.
9.	K.U. Leuven et al. (2007) BeVoting Study on Electronic Voting System.
10.	Le Libre. (2010) Vote automatisé: les communes devront "upgrader" à leurs frais. Le Libre. 23 February.
11.	Legislative Council. (2006) Examination of Estimates of Expenditure 2006-07: Controlling officer's reply to initial written question (Reply serial no.: CAB072).
12.	Legislative Council. (2016) Official Record of Proceedings: Wednesday, 2 November 2016.
13.	Malengreau, D. (2014) Vote électronique: voici l'origine du bug informatique. Le Soir. 28 May.
14.	Minutes of Meeting of the Panel on Constitutional Affairs. (1996) 24 June. LC Paper No. CB(2)1841/95-96.
15.	New South Wales Electoral Commission. (2017) iVote^®.
16.	Office for Democratic Institutions and Human Rights, Organization for Security and Co-operation in Europe (2016).
17.	Service Public Fédéral Intérieur. (2011) Test du nouveau système de vote électronique: Plus de 10 000 personnes participant à la grande expérience de vote.
18.	Service Public Fédéral Intérieur. (2017) Le vote électronique.