A Comparative Assessment of Electronic Voting
Part V: Kiosk and Telephone Voting Methods
The above sections have focussed primarily on remote Internet voting methods primarily involving use of computers not under the direct control of the electoral authorities. Here we examine more closely two other remote voting options, namely remote Internet kiosk voting and telephone voting methods.
Next to remote Internet voting, remote kiosk Internet voting seems to have the highest potential to positively impact the electoral process, given its ability to enhance accessibility for electors. Unfortunately however, there has been little written on the subject of kiosk Internet voting and few trials of the technology. Although remote Internet kiosks can make the voting process more convenient for electors, they do not have the same potential to do so as remote Internet voting given that the latter enables so many more points of access for electors. And while they have lower potential to enhance accessibility, kiosks generally raise the same concerns with respect to security and secrecy since in most cases voting kiosks are unsupervised. The need to secure data, lack of a paper trail for recount and auditing purposes, and the susceptibility to machine malfunctions are also concerns. Notably, the kiosk fallibility caused Ireland and the Netherlands to end the use of the voting machines which were to be trialled as remote Internet kiosks (Evans, 2001).46 Controversy over voting machines and negative experiences with electronic voting machine trials in the USA have also tarnished the reputation of voting machines in general and kiosks by association. The UK, Ireland, the Netherlands and Finland are four countries that have conducted, or attempted to commence, remote kiosk voting trials. In all instances the projects were terminated.
As outlined above, the UK offered touch screen Internet kiosks in councils that elected to use the service. In total, 808 kiosks were tested, some in traditional polling places and others in remote locations, such as supermarkets and other places within the city centre. However, as explained, the project was cancelled because of security concerns and the fact that the extension of electronic voting methods did not have a significant positive effect on voter turnout. Survey research conducted by the UK Electoral Commission also revealed that respondents were more positive toward remote Internet and telephone voting than toward kiosk voting (BeVoting, 2007). For example, 87 percent of telephone voters and 81 percent of remote Internet voters reported feeling positively about the secrecy of the vote compared with 57 percent of Internet kiosk voters. Furthermore, 87 percent of telephone voters reported being positive about the security of the vote, whereas remote Internet voters (59 percent) and remote kiosk Internet voters (60 percent) reported similar assessments. Electors who voted electronically reported finding remote Internet (93 percent) and telephone voting easy to use (88 percent) in slightly greater proportions than those who used kiosks (84 percent) (BeVoting, 2007).
In 2000, the Netherlands implemented remote Internet kiosks in the 2003 Provincial Council Elections as part of a program to enable non-place-dependent voting. With a budget of 8.1 million euros, the goal of the Remote (Electronic) Voting project was to allow electors the option of voting from remote electronic kiosks installed in public areas other than polling places. In 2002 however, two government ministers informed the House of Representatives they felt the original program would not be feasible and decided to pursue an alternate model. So, while the remote electronic voting project continued, it deviated from its original intention, instead focusing on allowing internal electors the option of voting at any polling place and external Dutch electors the possibility of remote Internet and telephone voting.
While the specific concerns that prevented the introduction of remote Internet kiosks in the Netherlands are not entirely clear, media reports indicate that the chosen Nedap machines produced radio emissions which could allow remote detection of how a vote was cast. While the government also had security concerns regarding remote Internet voting and telephone voting, they felt these issues were manageable enough to proceed with implementation (BeVoting, 2007; Kitcat, 2007). In fact, the Netherlands' experience with kiosk Internet voting was so negative that a 2007 government commission concluded that it was not feasible to introduce kiosk Internet voting and suggested that paper and pencil voting remained the best method for complying with election requirements, including transparency, controllability and integrity (Dutch News, 2007; Slashdot, 2007).
The impetus for electronic voting in Ireland was unrelated to turnout, having the purpose to "improve the speed, efficiency, accuracy and user friendliness of the Irish elections" (BeVoting, 2007:29). Since turnout was not a consideration, the government selected the Nedap voting machines. A substantial amount of money (51 million euros) was invested in the purchase of the machines that were to be used in polling places as well as remote locations. They were tested in 3 and then 7 constituencies in 2002 and although they were supposed to be tested in the local and European elections in 2004, the electronic voting project was abandoned just prior to these pilots. More recently, in 2009, the electronic voting system was officially cancelled.
Skeptical and unfavourable reports came from the Commission on Electronic Voting concerning issues with the software (they recommended it be open source code). Protest groups, such as the Irish Citizens for Trustworthy eVoting, also created considerable controversy concerning the lack of a paper-based audit trail, the initial cost of the machines, the 700,000 euros it cost to store them annually and the additional 20 million euros Nedap speculated would be necessary to adequately develop the machines. These efforts resulted in the termination of the project (BeVoting, 2007; Electronic Voting in Ireland, 2009). The failure of the Dutch experience has also been an important consideration for not pursing machine development in Ireland, but the government is not ruling out the possibility of trying to introduce electronic voting again (BeVoting, 2007; MacCárthaigh, 2009).
Finland also sought to incorporate remote Internet kiosks into its electoral process by placing the machines in three small municipalities in its 2008 local elections. Approximately 30 kiosks were used and although they were not all remote, they were all supervised by election officials. For example, one supervised remote Internet kiosk was located in a mobile library (on a bus) that drove around the municipality. Electors could use the kiosks both on advance voting days and on election day. The system attempted to integrate software provided by a Spanish company (Scytl) with the standard Finnish Election Information System. Overall however, the trial was considered a failure due to flaws in the machines. A total of 232 votes were not recorded because the voting was interrupted and so the ballots did not register into the electronic ballot box. Finnish officials also report no noticeable effect on voter turnout. The government is currently discussing whether it will use electronic methods of voting in the future (Aaltonen, October 7, 2009).
These examples illustrate that more research and thorough testing and development are needed prior to the implementation of (remote) kiosk Internet voting, at least with respect to the machines that have been trialled in Europe. Though Internet kiosk voting can enhance accessibility while still giving election officials a moderate degree of control, cost is a serious consideration and there appear to be more technological problems associated with machine voting in general (whether polling place Internet kiosks or machines or remote Internet kiosks) than with remote Internet or telephone voting options.
Telephone voting holds considerable promise for improving the operation of elections, most of all by increasing accessibility. It not only enables remote accessibility and allows electors the option of voting at any time, but the presence of telephones in Canadian households is also nearly universal. The existence of cell phones makes this method even more mobile than Internet voting and creates many more access points from which electors could cast a ballot. Furthermore, many electors, especially older cohorts of potential voters, are much more accustomed to telephone technology than the Internet, and so the opportunity cost of voting by telephone is lower because electors would not have to familiarize themselves with new technology. It is also much less costly than voting machines.
Telephone voting has the same benefits for election administration as other types of electronic voting in terms of efficiency and accuracy of electoral results, but does pose risks regarding security and ballot secrecy, especially since election officials have little or no control over the process. Generally however, concerns relating to security and secrecy of the ballot can be minimized through system design, as shown in Halifax's experience with telephone voting. Other issues however, such as the impossibility of a traditional recount (because there is no paper trail), and the possibility of a service interruption, are aspects which should be given consideration. Despite its potential, there have been few trials of telephone voting in Europe with the exception of the UK (detailed above) and the Netherlands. In both cases the development and use of telephone voting has been stopped. There have been multiple instances where it has been used in Canada at the municipal level in small townships.
The Netherlands offered remote Internet and telephone voting to citizens living abroad. To permit these remote types of voting in the 2004 European Parliament election, an on-line voting experiments act was passed in December 2003. Electors were required to pre-register by mail and were then sent a vote code or login and a list of candidates, each with their own candidate code. Of a potential 600,000 electors (20 to 30 thousand of whom would normally vote by postal ballot), 5,351 votes were cast electronically, 480 by telephone and 4,871 over the Internet. Most of the concerns raised in government reports related to the security of voting over the Internet, although a lack of transparency about the software programs that operated both systems, provided by LogicaCMG, was also criticized. Public attitude surveys showed that electors reported using the services because they were simple and fast. Sixteen percent of voters who cast their ballots electronically reported only having voted because these methods were available. The majority of positive feedback however centered on remote Internet voting, and in 2006, when the government decided to pursue another trial, telephone voting was removed as an option and only remote Internet voting was available (BeVoting, 2007).
Overall, while there has not been much testing using telephone voting, the potential benefits are worthy of further exploration and discussion. Notably, that telephone voting appears to be a useful complement to remote Internet voting given electors' familiarity with the telephone and its widespread presence in households. As the case of Halifax highlights, offering telephone voting in conjunction with remote Internet voting is one solution to accessibility problems associated with a remote Internet option. At the same time, rates of use of telephone voting are much lower than remote Internet voting for the data that is available and so, while it is more accessible, it may not have the same effect with respect to engagement as the Internet.
European Trials and Developing a Model for Canada
The preceding European examples impart some salient lessons for the development of Internet voting models elsewhere. In particular, there are eight important considerations that should be taken away from the experiences of Estonia, Geneva and the UK.
First, prior to model development it is essential to clearly establish the principles that the electoral process must live up to as well as lay out any constitutional or legal requirements. It is imperative that the characteristics of the selected model fulfill these obligations and requirements. Where applicable, the criteria used to evaluate an electronic voting system should be the same as those benchmarks used to assess the effectiveness of the traditional paper ballot voting system. The principles and criteria will be different for every country and jurisdiction, given variations in expectations, legal structures and political, cultural and contextual features. Therefore, while we can learn about model development from other cases and experiences, a Canadian application must be customized to the benchmarks and features required by the Canadian electoral process to ensure system success and maintain electoral integrity. The success of remote Internet voting systems in Estonia and Geneva can be attributed to this approach to development.
Second, while there may be a desire to test multiple methods of electronic voting, it is more useful to pick one or two methods and work on developing and integrating those before considering others. Further, whatever types of electronic voting are selected, each should be based on a unified prototype. Adopting a single approach is likely to produce a system that is more thoroughly developed and will make it easier to identify which elements do and do not work well. This is the major lesson imparted from the UK experience. Evidence from Estonia and Geneva suggests that a unified model is a superior approach in terms of administrative and technical considerations.
Third, a step-by-step approach in developing a model is important, and was one of the primary reasons for the success of remote Internet voting in Geneva. Each successive trial reached more electors and occurred in elections of greater prominence. Gradually increasing the salience of the vote across these two dimensions allowed for optimal risk management and enabled officials to measure and understand the effects of the extension of electronic voting. This approach allows election officials to systematically build on established success before progressing. Part of the failure of the UK trials was due to its overly hasty and ambitious approach. The time constraints imposed as a result of rushed implementation made adequate testing impossible, resulted in technical malfunctions and likely undermined support for the project as a whole.
Fourth, the engagement of key stakeholders seems to be a useful mechanism to ensure all affected parties are satisfied with proposed changes or additions to the electoral process. While there was not much mention of these with respect to Estonia and the UK – with the exception of government officials, election administrators and public attitudes – the approach taken by Geneva focused on regular communications with representatives from all the political parties, feedback from all levels of government, and an emphasis on what the public wanted. This is considered to be one of the major contributing factors to the success of remote Internet voting in Geneva (Chevallier et al., 2006).
Fifth, it may be useful for the system to be made available in multiple languages, in order to maximize accessibility. A large Russian-speaking minority in Estonia is hindered from voting remotely by Internet because the interface is only available in Estonian, the only official language in the country.
Sixth, the Estonian and Swiss examples suggest that it is possible to adequately address concerns pertaining to security and secrecy. While the USA and UK cases do raise valid issues regarding the security of Internet voting, their significance should be placed in perspective. The USA has not attempted formal testing in official polls. The UK did encounter technical issues, but mostly with regard to Internet machines and kiosks rather than remote Internet voting. On the other hand, all successful trials examined in this report illustrate that, with careful model development, potential security and secrecy problems can be minimized. The feature of the Estonian system that allows electors the option of recasting the ballot is a good example of how the system can be adapted to prevent the threat of vote buying. The intricate envelope schemes used in both Estonia and Geneva are examples of how concerns about voter anonymity can be addressed. These examples show that it is possible to develop model features that not only meet the requirements of the political and legal framework, but can also effectively address important concerns.
Seventh, while some digital divides appear to be of less concern, there are others which should be taken into account. Available evidence from the European trials illustrates that a digital gap is not necessarily a concern in terms of gender, education or geographical location. However, there appears to be a distinct gap with respect to age, as younger electors are more likely to make use of Internet voting than older electors. Furthermore, there is a gap in use among those who have better computer knowledge than others. The more knowledgeable are naturally more likely to use on-line voting. The Estonian case suggests some program options that could be instituted to educate older electors and increase their comfort with using the technology. The age and knowledge disparities are important considerations in ensuring equality of access and potential for engagement.
Finally, although broad conclusions regarding the potential of Internet voting to positively affect voter turnout are not possible, there is evidence (particularly in the cases of Estonia and Geneva) that the extension of remote Internet voting can engage some groups of electors who do not currently vote at high levels. There are two trends in particular. One is that, where data is available, we see electors voting on-line who previously classified themselves as abstainers or said they would not have been likely to vote otherwise. Second, there is evidence of a 'faithfulness effect', whereby electors who opt to vote on-line will continue to do so in subsequent elections if given the opportunity. It appears that making voting accessible through the Internet has the potential to engage additional electors in the voting process, and these electors will faithfully continue to use the service.
46 Machine security can be increased by employing certain measures, such as a centralized account system, whereby voters must register before being allowed to vote, or by issuing electors a personalized card that must be inserted into the machine (Evans, 2001:3).