Introduction
In October of 2005, Estonia held its first local elections where voters could cast their ballots entirely online.Footnote 1 Two years later, the state made internet voting available in national elections, making Estonia the only country in the world where e-voting was open to all voters at the highest level.Footnote 2 Conservative Prime Minister Mart Laar celebrated the initiative as ‘a tool for creating equal and additional opportunities, for strengthening and developing Estonian democracy’.Footnote 3 Despite legal challenges and vigorous political debate, this assessment seems to have been accurate. In the 2023 national elections, over half of all votes were cast online (Figure 1).Footnote 4 Internationally and at home, e-voting has become one of the central pillars of Estonia’s ‘digital republic’, which its champions typically describe as a set of digital public services that represent ‘the most ambitious project in technological statecraft today’, promising to ‘fundamentally redefine what it means to be a country’.Footnote 5
Figure 1. Online votes cast in municipal, national and European elections from 2005 to 2023. Source: National Electoral Commission (Vabariigi Valimiskomisjon).
‘What it means to be a country’ is, of course, a question with a vibrant historiography. Voting is certainly one of the constitutive elements of modern states, the proof of participation in representative democracies according to political theory, and, in the words of one historian, an act with a ‘solemn character’ that states and citizens treat as a sacred ritual.Footnote 6 The secret ballot, in its present socio-material arrangement, has its roots in mid–nineteenth-century Australia, where it was instituted as part of colonial self-rule.Footnote 7 Though states have tended to emphasise the continuity of voting principles (such as secrecy and uniformity), scholars have increasingly called attention to the struggles that have shaped many innovations in voting technology, from reconfiguring the layout of the ballot to the introduction of voting machines.Footnote 8
Many political scientists have seen Estonia’s digital democracy projects as a natural by-product of technological progress, driven by ‘by the hope to increase voter turnout, attract younger voters, and make voting more convenient’.Footnote 9 These scholars form part of a larger chorus calling for a broader revitalisation of democracy via digitisation, such as Citizens Assemblies, various computer-powered direct democracy initiatives, online public services and so forth.Footnote 10 In this view, digitisation offers solutions to the widespread dissatisfaction recorded in Western democracies, overcoming their ‘deliberative deficits’, the feeling of the public of being excluded from decision-making in spite of increasing numbers of formal mechanisms for participation.Footnote 11 Estonia’s construction of a ‘digital republic’, which includes, in addition to e-voting, services ranging from tax filing to medical aid, is portrayed in this telling as a deliberate strengthening of popular sovereignty.
This view is at odds with the standard historiography on post-Soviet statecraft. Historians generally describe the post-1991 period as one of increasing, if circumscribed, neoliberalisation. In Quinn Slobodian’s account, neoliberalism is characterised by not so much the championing of markets over state power but rather ‘redesigning states, laws, and other institutions to protect the market’, in particular from national sovereignty and mass democracy.Footnote 12 Key to this transformation has been the empowerment of experts and technocrats over electoral politics, a trend that holds true both in Eastern Europe and elsewhere.Footnote 13 Of course, the vision of a world made predictable and safe from the tempers of national publics has always been more of an idealised vision than fully implemented reality. As reformers in Russia, the Baltics and elsewhere discovered, institutions and infrastructures resisted complete recalibrations, and neoliberals had to, as Stephen Collier put it, learn to work through the actually existing state, which retained many of the attributes of Soviet socialism.Footnote 14 Nor was the neoliberal revolution a complete break with the past, since, as Johanna Bockman has shown, post-1989 reformers co-opted ideas about ‘market transitions’ developed by economists on both sides of the Iron Curtain into their practices.Footnote 15
Remarkably, scholars of neoliberalism have had very little to say about voting. The founding fathers of the thought collective certainly had their reservations about universal suffrage: Milton Friedman, John Davenport and William Hutt all supported restrictions on suffrage in South Africa in order to ‘immunize the market against the disruptive effects of an empowered population’.Footnote 16 One could argue that it is precisely because neoliberals have been so successful in constructing supranational institutions to constrain national sovereignty that they have no interest in elections as such – no matter who wins, the ruling parties still have to play by rules made at the WTO or the European Commission. Yet it is surely worth investigating why e-voting in Estonia came into existence during the government of Mart Laar, a politician so committed to neoliberalism that he has repeatedly claimed that the only work of political philosophy he had read when he became prime minister was Milton Friedman’s Free to Choose. Footnote 17
To understand the genesis of e-voting, this article takes seriously the ‘building’ in ‘state-building’ and looks at how the infrastructure required for online voting was developed in the 1990s and early 2000s. This approach is informed by work in science and technology studies that conceptualises infrastructure – the basic technological building blocks that invisibly enable the functioning of contemporary society – as sociotechnical systems, composed of hardware, as well as institutions, laws, cultural practices and maintenance work that make the hardware function, or at least appear to function, smoothly and without friction.Footnote 18 I argue, contrary to the digital republic’s cheerleaders, that, rather than being the culminating step in a grand vision of democratic renewal, e-voting is best understood as a peripheral by-product of a different digital infrastructural project that brought together both public and private interests – the digital ID. Indeed, in the same speech in which Prime Minister Laar touted the benefits of e-voting for equal opportunity, democracy and so forth, he also underscored how basic democratic principles, like the secret ballot or one person, one vote, were safeguarded by ‘the passing of the digital ID law in the Parliament’.Footnote 19 In the context of the history of neoliberalism, my approach argues for understanding the neoliberal turn as one mediated and shaped by the construction and reappropriation of sociotechnical infrastructure, rather than driven principally by individual actors or a ‘neoliberal thought collective’.Footnote 20 The infrastructure in question not only framed and set limits to ideology but also, indeed, created the conditions of possibility for ideological utterances, thus explaining the seeming contradictions of the Estonian ‘digital republic’ project.
Centring infrastructure in the story of online voting helps us better both to appreciate the historical conjuncture that enabled the emergence of this particular project in the first place and to explicate how the specific material and discursive qualities of digital identification shored up support for the online voting project, enabling its success in Estonia while similar projects failed elsewhere. The State Chancellery developed the physical and computational architecture for digital identification in the early 2000s, in a public–private partnership with Estonian banks and telecommunications companies. The latter saw this initiative essentially as an opportunity to outsource the development costs of their own authentication infrastructure for online banking onto the state. Meanwhile, civil servants understood the digital ID programme as a tool for simplifying queries to national registries, which expanded as the state took over functions that had previously been performed in Moscow and tried to comply with EU regulations in advance of accession. Although the state made new digital ID cards compulsory for the entire population, few citizens initially used them. The following discussion argues that for many stakeholders, online voting became a means of revitalising the ID card programme. Voting, in other words, became the problem that digital identification could solve.
The sources for this article include sixty oral history interviews conducted in 2021 with individuals deeply involved in building Estonia’s digital republic. These include civil servants, politicians, lawyers and engineers but also entrepreneurs, CIOs and CEOs of major banks and telecommunication firms. These interviews were commissioned by the Estonian Ministry of Communications and Economic Affairs, and the list of interviewees was developed in consultation with the ministry; however the semi-structured interviews themselves were conducted, together with Raivo Ruusalepp, in private and without oversight. I have complemented these sources with archival material from the State Chancellery and records from the State Information Agency, various ministries and published sources. The article first looks at the genesis of the digital ID programme in various state offices in the 1990s in response to internal and international pressures. I then survey its transformation into a tool for banks and telecoms in the turn of the millennium. Finally, I look at how particular features of the digital ID infrastructure were mobilised to respond to critics of e-voting both at home and abroad.
The Infrastructure Under the Success Story
Understanding why Estonia’s compulsory digital ID programme became so central to the success of online voting requires a look under the hood of the system itself. From the perspective of the voter, the process looks as follows. In local, national and European elections, voters can vote online during the period of early voting using any internet-connected computer with a smart-card reader (Figure 2). Voters download an app, authenticate themselves with the ID card, view the list of candidates running in their district, choose their preferred candidate and confirm their choice with a digital signature authenticated, once again, with the ID card. To protect the secrecy of voting, voters can change their online votes an unlimited number of times until the end of the election period, and a vote cast at a polling station will override a ballot cast online, rendering, as the architects of the system argue, any voter intimidation meaningless (Figure 3).Footnote 21
Figure 2. The Estonian ID card. Front and back. Source: The Police and Border Authority (Politsei Ja Piirivalveamet).
Figure 3. The e-voting GUI. The left column contains party lists in a randomised order. The right column lists the voters’ electoral region and confirms their candidate choice. After choosing their preferred candidate, the voter clicks on ‘choose’, authenticates themselves with the ID card and their PIN and the ballot is encrypted, sealed in a virtual double envelope and transmitted to the NEC’s server.
On the back end, online voting most closely resembles double envelope postal voting. Information about the voter’s choice is encrypted using a public key generated by the central voting system, and it is attached to the ‘outer envelope’, containing the voter’s digital signature. The information is then sent to the central voting system over the internet. At the time of counting, provided that the voter has not voted in person, the ‘outer envelope’ is dropped, the information is decrypted using the server’s private key (stored in a hardware security module) and the vote is added to the overall tally.Footnote 22 Several security and verification mechanisms prevent tampering with the system. Voters can verify that their vote has actually been recorded by the server.Footnote 23 Election monitors and auditors verify that the raw voting data is recorded, in line with regulations, onto DVDs, that the processing of votes is done on computers disconnected from the internet, that different components of the voting system’s private key are sealed, that the seals are unbroken prior to use, that the standard rules of cybersecurity are followed and so on.Footnote 24
Over the years, the system has been overhauled and modified to reduce security risks. Still, as even the developers themselves have admitted, the risks of e-voting can only be minimised and not entirely eliminated. Critics have argued that the e-voting system has ‘serious procedural and architectural weaknesses’ from the vulnerability of voters’ computers, which could potentially be contaminated by malware to the potential of server-side attacks, where third parties could break into the vote-counting servers and alter votes between decryption and tabulation.Footnote 25 Architects of the system have admitted that the security of the computers used for voting is indeed left to the conscience of the voters themselves. They also argue that risks to other components of system, such as tampering with DVDs used for tabulating votes, while ‘theoretically possible’, are as likely as a ‘neutron bomb arriving by train [to the centre of the capital] or the Sun suddenly going extinct’.Footnote 26 Official analyses have found the system to be in compliance with EU rules on ‘traditional’ voting, and some researchers have even argued that e-voting has more risk management built into it than paper-based voting mechanisms.Footnote 27
Still, Estonia remains the only country using e-voting in national elections. E-voting has been used sporadically in Switzerland, Norway, Germany, Canada and elsewhere, usually in local-level elections, and states have often discontinued the practice due to security concerns (Figure 4). It should be distinguished from other types of electronic voting, such as casting ballots in polling stations that use vote-counting machines.Footnote 28 Researchers studying the persistence of Estonia’s e-voting practices have repeatedly returned to the same key difference: the added layer of security provided by the digital ID system, which allows voters to securely authenticate themselves using a smart card and a series of PIN numbers.Footnote 29 Other online voting systems have used passwords delivered through the postal system (Switzerland), IDs tied to mobile phone numbers (Norway), or other more complicated methods.Footnote 30 Estonia has integrated digital identification into various government e-services, further lowering barriers to entry.
Figure 4. Countries that use online voting in regional or national elections. Source: International Institute for Democracy and Electoral Assistance.
Put simply, much of the trust placed in the Estonian e-voting system is based on the country’s unique infrastructure of digital identification, which is universal,compulsory and tied to a physical ID card. The process by which this system of identification came into being, however, was far from meticulously planned. The following discussion reveals how the development of what became touted as a novel experiment in the future of democracy relied on a specifically post-Soviet alignment of business interests, public sector reformers and cyberneticians.
Bureaucrats Face the Information Age
The groundwork for Estonia’s digital ID system was built in the 1990s, when dozens of researchers at the Institute of Cybernetics found themselves relocating from the quintessentially Soviet-looking mass housing district of Mustamäe to the historical buildings in the Old Town of Tallinn, which housed ministries and departments of the once-again sovereign Republic of Estonia. The Institute was founded in 1960 on the order of Gustav Naan, a polymath, tech enthusiast and, by all accounts, relentless womaniser. Naan was known for, among other things, editing the Estonian Encyclopedia, authoring numerous articles on extraterrestrial life and championing the works of Norbert Wiener, including writing an introduction to the Estonian translation of Human Use of Human Beings: Cybernetics and Society. Footnote 31 Cybernetics, which frames interactions between humans, animals and machines as issues of communication and control that could be facilitated via computers, offered an exciting new language for scientists to talk about reforms in the Soviet system.Footnote 32 In Estonia, the Institute of Cybernetics brought together an eclectic group of heterodox thinkers who worked on issues ranging from urban planning to cryptography, often simultaneously.Footnote 33 The Institute was led by Nikolai Alumäe, an engineer working at the Energy Institute of the Academy of Sciences, who saw an increasing need for digital computers in the petrochemical sector of eastern Estonia.Footnote 34 They were also among a small number of experts in computing with experience working in a market society, thanks to a long-standing collaboration with Finnish engineers, who built many of the microelectronics used in the Soviet Union. Some even had experience in politics.Footnote 35 No wonder, then, that both private companies and the state quickly poached many of these researchers once the Institute was forced to downsize after 1991.Footnote 36 Take, for instance, Tarvi Martens, a young engineer whose career began with working on the Institute of Cybernetics’s Juku line of PCs in the late 1980s, before transitioning into cryptography, a long-time interest of the Institute’s chief, Ülo Jaaksoo. When work at the Institute became scarce, Martens shuffled around various private companies for a few years, before being invited to build registries for the state’s borders and customs agency.Footnote 37
Registries are central to the history of the digital state. Nineteenth-century statisticians in the United Kingdom, France, Germany and the United States argued that the aggregation of informational statistics – on births, deaths, marriages, income, ownership, criminality, health and much more – was vital for modern governments. Basic tasks depended on such data. The state might, for instance, have to confirm that citizen A indeed owned the piece of land they claimed to own.Footnote 38 By the 1950s, Western governments collected massive amounts of data on their citizens. By the 1970s, this information was increasingly stored electronically on large mainframe computers held by agencies like the UK Benefits Agency or the Social Security Administration in the United States. In the 1990s, these organisations began to move their data onto personal computers and interlinked registries, promising a future of ever more precise forecasting of demographic and economic trends, as well as cheap and efficient processing of citizen information.Footnote 39
Nothing like this had taken place in Soviet Estonia. Attempts to build a pan-Union system of internetworked data registries had been started in the 1970s but failed.Footnote 40 A development centre founded by the State Chancellery in 1989 studied the information processing capacities of what would soon become independent Estonia and concluded in 1992 that the government of ‘the Republic is completely lacking in centralized registries […] and due to the incredible development of computing power, our information systems are falling even further behind those of Western states’.Footnote 41 Estonia lacked a functioning population registry, a registry of enterprises, a centralised tax board and a customs registry. Other registries existed in the form of punch cards and indexes or were incompatible with Western standards – sometimes both. The State Chancellery was faced with a dual challenge: after the collapse of the Soviet market, the country was in a constant state of near bankruptcy, and there was hardly any cash to pay for pensions, let alone to modernise the state; yet jumpstarting the economy and integrating with Western markets was itself contingent on a functioning and dependable system of registries and economic statistics. For many officials, digitisation was the only way to break this gridlock.Footnote 42
As the number of new electronic registries multiplied, engineers like Martens, with experience in cryptography, began to ask questions about the security and reliability of documents that were now being exchanged increasingly in the form of bits and bytes. How could someone at the tax board, for instance, be certain that a certificate of ownership issued electronically by the land cadastre was actually accurate? How could the authenticity of electronic documents be verified in the same way as a signature and timestamp could confirm the authenticity of a printed document? Answering these questions, of course, also meant more work for the engineers and their colleagues: if Martens could impress the importance of secure authentication on government officials, the Institute of Cybernetics and its offshoots were likely to get the contracts to actually build the system.Footnote 43 The issue was internationally prominent, too. The state of California passed the world’s first law on electronic signatures in 1995, Germany passed similar legislation in 1997, and the European Union, in the mid-nineties, was in the process of drafting directives to regulate digital signatures. In response to growing local and international pressure, a small group of civil servants, including Martens, began work on developing Estonian legislation in early 1998 (Figure 5).Footnote 44
Figure 5. Tarvi Martens. Source: Estonian National Archives, ERA.5446.df.33521.
Martens does not immediately appear charismatic. He speaks quietly, fidgets and, as I discovered during our interview, has a habit of referring to himself in the third person. At the same time, he speaks with confidence and can explain the most complex issues in cryptographies to fairly uninitiated audiences. Nearly everyone we interviewed brought him up unprompted as a key figure in getting the ID card project off the ground, from his first meeting with Migration Agency officials who had a vague idea of introducing some kind of electronic ID once the first set of Estonian passports was due to expire in 2002. At first, the civil servants were interested mostly in simplifying their own workflows. Once most government registries were operating digitally, civil servants imagined a system of data exchange, where governmental and municipal institutions could send and receive documents with timestamps and signatures, speeding up legislative and administrative reviews, rights confirmations and so on. The project soon expanded to encompass interactions with citizens as well, facilitating already existing online service (like e-tax filing) and adding new ones.Footnote 45
The working group converged around the idea of a microchipped ID card, based on the model of public key cryptography, which dated back to British intelligence experiments in the 1970s and has now become the standard model for online encryption.Footnote 46 The card would contain a secure private key that could be used to generate unique stamps on digital documents. A second vital component in this system is a Certification Authority, which associates digital IDs with specific people and holds the repository of public keys used to verify the authenticity of a specific signature.Footnote 47 And the person who led the engineering side of this gambit – none other than Tarvi Martens.
As the project developed, it became clear that neither the microchipped ID card nor the infrastructure for the certification authority would come cheap. As a result, there was little enthusiasm for the project among political elites, who saw it as an expensive and overly technical ‘nice-to-have’. It was not a vote winner.Footnote 48 This tension might have become an insurmountable obstacle for the cash-strapped state – were it not for the interest of the banking sector in the success of the project.
The Bankers’ Gambit
In the first ten years of the young market economy, the financial sector became a considerable force in Estonian society and politics. The first commercial banks were founded in 1988–9, but it was not until the early nineties when banking in Estonia really took off. New enterprises badly needed loans, workers needed safe harbours for their paycheques, uneven information flows created opportunities to make fast cash on foreign exchange markets and a wave of privatisations across the former Soviet bloc offered high returns to smart investors. While Soviet-era enterprises were downsizing or shutting their doors completely, and civil servants were counting cents, banking offered unheard of salaries to young men (indeed, they were predominantly men) willing to put in the hours. In 2001, a banker earned nearly triple the average wage. While Tarvi Martens went to work for the state, his cybernetician colleague Jaan Priisalu ended up at Ühispank, and Aivo Adamson, another course mate, landed at Hansapank.Footnote 49
Hansapank, a self-declared ‘bank for the rich’, was by far the most successful of the new banks. In 2001 its annual profits rivalled those of all other Estonian banks combined.Footnote 50 The bank was also technologically innovative: in 1993, it opened up its first electronic banking service, and in 1997 it started offering internet banking, alongside two other Estonian banks. Indrek Neivelt, the bank’s founder, told me in an interview that at the time, i-banking services were not seen as serious business proposals but were permitted according to ‘the liberal principle to let the IT-guys do what they want’.Footnote 51 After all, in 1997, only a tiny percentage of the Estonian population even owned computers.
But things changed quickly. The popularity of personal computers rose rapidly as they became more affordable, and as the state boosted their proliferation with a variety of computer literacy initiatives, such as Tiger’s Leap, which outfitted all Estonian schools with computing classes.Footnote 52 Meanwhile, economic crises in Asia and Russia hit Estonian banks hard, leading to bankruptcies and mergers. Over the space of just a few years, Scandinavian banks acquired controlling shares in three Estonian banks: Swedbank became the majority shareholder in Hansapank, SEB in Ühispank, and the Finnish bank Sampo acquired Optiva Pank. Competing for new clients became a more urgent task, and i-banking became an increasingly attractive tool for cutting costs and expanding market share.Footnote 53
After some highly publicised security breaches in the mid-nineties, performing sensitive transactions online was no longer as straightforward as it used to be. Early internet banking used password cards to authenticate customers – little sheets of paper with one hundred or so password written on them. These got the job done, but they were cumbersome, easy to lose and not particularly secure. In Scandinavia, banks were developing their own two-step authentication systems, but in Estonia, no single bank, not even the ones under Swedish or Finnish ownership, had sufficient R&D budgets for a task of this magnitude.Footnote 54 Even if that were possible, Estonian customers simply could not afford the password generators that were being distributed to customers in Sweden or Finland.Footnote 55 Like the state, the banks faced an authentication problem.
From the start, Martens had invited cryptography specialists from the private sector to provide input to the ID card project. To him, people like Priisalu or Adamson were not competitors but former colleagues, and the only experts in Estonia who knew the task at hand. After a presentation on the new ID card at the former political education centre of the Estonian Communist Party, where the state revealed it was planning to outsource the critical work of the Certification Authority (CA) to the private sector, in order to save costs, the CIOs of Estonian banks quickly proposed collaborating on a joint CA.Footnote 56 Soon enough, the joint effort had expanded to include not just banks but also telecoms and became the only CA, essentially a government-supported monopoly.Footnote 57
The ‘business case’ for supporting the government’s project was as follows. The banks would offload a large part of the cost of developing an authentication system for their i-banking services. After an initial investment, the CA would become a for-profit institution, charging a small fee for every certification requested by a corporation (while keeping the service free for private citizens). In a major break with past practice, the banks would be allowed to issue and service ID cards, meaning that citizens picking up their new documents would inevitably end up in the orbit of the banks’ tellers and their sales teams. Indeed, to make the proposal even more attractive, the state agreed to subsidise the cost of the ID card by increasing fees on new passports.Footnote 58 Finally, in the long run, increasing the number of i-banking customers would enable banks to close branch offices and reduce brick-and-mortar costs.Footnote 59
The state, meanwhile, was interested in the substantial reduction in costs, and the potential for faster uptake, as people who already used the ID card for their i-banking would be more likely to use it for other public services as well. Critically for both parties, the ID card would be made compulsory to overcome the initial ‘chicken-and-egg problem’, where companies would not develop services because not enough people had the cards, and people would not get cards because there were not enough services for them.Footnote 60
Things did not work out quite so smoothly. The ID card was launched in 2002, with the first cards delivered to the Estonian president and the first lady at a public ceremony. But in the following four years, although the state issued over one million ID cards, only 50,000 certificates were activated. The predicted flood of public and private services that would be developed for the new digital identity did not materialise. Two years into the life of the card, citizens could do little with it other than file taxes, report residency changes and query five or six registries. The state had all kinds of ambitious ideas for how it could be used – from patient–doctor communication to submitting applications for social benefits – but none had yet been realised.Footnote 61 Bank customers preferred to continue using methods they were already familiar with, such as password cards, and the state was slow in developing new online services. Rather than turning a profit, the CA required additional investment.Footnote 62 The card had become a punchline for the media, with one journalist calling it ‘a nearly useless product of banks’ and IT-companies’ lobby that is full of vulnerabilities’.Footnote 63 One popular joke suggested that the only good use for an ID card was scraping ice off a car’s windshield on a cold morning.Footnote 64
To save the project, both the state and the bank–telecom alliance poured money into publicity and the development of new services. The publicity was concentrated into the Look@World Foundation, financed by the same corporations as the CA. The foundation claimed to have trained over 10 per cent of the adult population of Estonia in internet use over two years, with a predominant focus on educating the elderly, blue-collar workers and ethnic minorities. The foundation also tripled the number of public access internet points around the country.Footnote 65 As for services, the Ministry of Economic Affairs directed all other ministries to develop new ways of using the ID card. Among these new services was also a proposal from the Ministry of Justice: to institute online voting as an alternative to paper voting at local elections and, eventually, at national elections.
The governing right-liberal Reform Party (Reformierakond) introduced a draft bill on electronic voting in Parliament in the middle of 2001, half a year before the first ID cards were issued to citizens. Märt Rask, the minister of justice, defended the proposal in Parliament by noting that it would increase turnout, bring more young people to the polls and help retain Estonia’s position as leader in ICT but, critically, also labelled e-voting ‘a side-product of the complete application of the digital signature project’.Footnote 66 Indeed, when the bill was passed in early 2002, the National Electoral Commission (NEC), tasked with implementing the law, had very little idea of how the process would actually work. The first analysis commissioned by the Ministry of Justice judged the idea ‘utopian given current technological capabilities’, due to the difficulty of securing users’ personal computers and ensuring that recorded votes could not be altered by malicious actors.Footnote 67 What was clear to the NEC working group, however, was that the precondition to any secure e-voting system would be the new ID card.Footnote 68 The infrastructure was, in fact, so central to the project that the development of the e-voting system was once again entrusted to Tarvi Martens, who had, by then, moved to work for the bank–telecom-sponsored CA.
Coming up with a system that satisfied both constitutional lawyers and cryptographers took some time.Footnote 69 Conservative Prime Minister Laar had initially envisioned instituting e-voting in the national elections of 2003; in reality, the first online-enabled elections took place in 2005. In 2007, when the first national election with an e-voting option took place, just over 5 per cent of the electorate cast their votes online; by the European elections of 2009, that had risen to 15 per cent. Combined with the growing popularity of other e-services, like tax filing, e-voting had become popular enough to form a core part of Estonia’s new mythology as an innovative digital republic.
One version of the e-voting story depicts it as the brainchild of Laar’s government, known for its love of technology and the exuberant energy of its ‘chief information officer’, Linnar Viik, who tied Estonia’s various e-services together into an ideology of an ‘information society’. In Viik’s view, transforming the Estonian public sector into an online-first ecosystem would prevent the emergence of Soviet-style corruption, educate a critically minded citizenry and leap-frog Estonia ahead of Western welfare states, without requiring similarly generous tax-and-spend policies. Here, e-voting was just one component alongside other e-democracy initiatives. These included e-government, a paperless system of holding cabinet meetings that was more of a PR trick than a substantial innovation, and a participatory democracy portal called Today I Decide (Täna Otsustan Mina, TOM), where citizens could propose new legislation, with the government required to provide feedback on the most upvoted proposals.Footnote 70 There is some truth to this story, which had been elaborated since the state had first started its programme of computerising Estonian schools with Tiger’s Leap in 1997.Footnote 71 But taking an infrastructural view situates e-voting as part of a larger story of civil service reform and the capture of public infrastructure by private actors. Here, e-voting becomes part of a series of initiatives designed to save an ambitious idea – of using the power of the state to lure Estonians to switch to online banking – from an unexpectedly underwhelming execution. It becomes one chapter in a story of engineers with grandiose plans improvising when faced with indifferent politicians. It becomes a problem invented for a pre-existing solution. What it does not do is reinvent democracy.
The E-voting Service
The infrastructure of the e-voting system shaped the discourse over its legitimacy long after implementation. When legislation on e-voting was first introduced in 2002, it encountered principled but light opposition from the parliamentary opposition – hyper-nationalist, left-of-centre and agrarian parties, who opposed it on the grounds of insufficient secrecy (citizens would not be voting in booths, but anywhere with an access to a PC) and security (following a vote from the voter’s computer to the ‘ballot box’ would be complicated).Footnote 72 Citing similar concerns, the Estonian president, Arnold Rüütel, sent the draft law to the Supreme Court for constitutional review, where it received a positive ruling.Footnote 73 After the initial controversy had subsided, the legitimacy of e-voting was rarely contested. Still, two notable exceptions are worth discussing.
The first series of critiques came from local hackers, Paavo Pihelgas and Märt Põder. In 2011, Pihelgas, in an episode of the popular investigative programme Pealtnägija, demonstrated how a virus could prevent the voting software from sending a vote to the NEC server, without leaving any trace of this for the user.Footnote 74 Põder, a long-time internet freedom activist and member of the local Pirate Party, demonstrated in 2015 that it was possible to cast a spoiled ballot for a fictional candidate ‘Tarmo Jüristostein’, a theoretical impossibility, since the e-voting app only allowed voting for candidates on electoral lists.Footnote 75 In the following years, Põder continued to draw attention to what he termed ‘the fundamental unverifiability of the e-vote’.Footnote 76 The NEC responded to these critiques by adding a verification mode to the e-voting app, allowing voters to check whether their vote had been recorded in the NEC server as it was cast. A second update, in 2017, added a data auditor to the vote tabulation process, which could ‘verify the integrity of the voting result without breaking ballot secrecy’.Footnote 77 The Ministry of Economic Affairs even appointed Põder to a working group to review e-voting procedures in 2019 – although the group’s recommendations did not lead to substantial modifications.Footnote 78
A more substantial critique was proposed by a team of American researchers, who observed e-voting at the 2013 local elections at the invitation of the Centre Party (Keskerakond), which had opposed e-voting since its institution in 2005. The highly credentialed observers issued a scathing report, in which they concluded that the system had ‘serious design weaknesses that are exacerbated by weak operational management, […] problems [that] stem from fundamental architectural problems that cannot be resolved with quick fixes or interim steps’ and recommended that ‘to maintain the integrity of the Estonian electoral process, use of the Estonian i-voting system should be immediately discontinued’.Footnote 79 The critique got a lot of play in Estonian media but was widely rejected by local security experts and the NEC. A central claim of the critique was that the system could not guarantee client-side security – users’ computers could be systematically infected by malicious state-level actors, and election results could be irreversibly altered.Footnote 80 NEC experts responded that achieving full operational security with e-voting was impossible, and the user would always hold some responsibility for maintaining computer hygiene. Still, the digital signature created a level of security comparable to paper voting, that, in the view of election officials, set the Estonian e-voting system apart from its international counterparts and more than sufficiently answered the critiques of American observers.Footnote 81
The American report did not limit its critique to issues of user security. The researchers warned that the system could also be attacked on the server side, and that their observation of local elections had revealed major lapses in operational security on the part of the NEC.Footnote 82 Some of their critiques were more convincing than others.Footnote 83 The point here, however, is not to litigate the validity of e-voting but rather to note that the Estonian authorities’ response turned on the physical and rhetorical qualities of its unique digital ID system. It allowed Estonian cryptographers to address specific critiques regarding client-end security, but equally importantly, the digital ID became a kind of cultural totem that authorities used to distinguish the Estonian system from other electronic voting systems, whether those were internet-based or not. Indeed, Switzerland, Norway and the United States did not have a compulsory, encrypted system of authentication like the Estonian ID card. Foreign experts could therefore be credibly accused of transposing legitimate problems with their own systems to Estonia, where the analogies did not apply. In the words of the security expert Anto Veldre, ‘the service offered by the Certification Authority is the “magic stuff” that allows any bank, tax board or small enterprise to verify a person’s identity online’.Footnote 84
Alongside technological critiques (that e-voting is not secure enough) and philosophical critiques (that it does not approximate conditions of physical voting well enough), various actors have, of course, levied political critiques at the programme, suggesting that electronic voting distorts the democratic playing field in favour of certain parties. In the early days of the programme, these accusations most commonly came from the oppositional Centre Party, whose voters tended to skew older and poorer and were therefore less likely to have access to the internet.Footnote 85 In recent years, buoyed by Donald Trump’s rejection of election results in the United States, the far-right Conservative People’s Party has regularly challenged the legitimacy of e-votes, suggesting that these have been manipulated by the liberal bloc.Footnote 86 While such arguments may be useful for mobilising core voters and explaining away election losses, they have not found purchase with courts. Independent analyses of e-voting have shown that although liberal bloc voters were more likely to vote online, there is no evidence that the ability to vote online actually tilted the playing field. In other words, the availability of e-voting did not increase turnout among voters of specific parties; it simply made younger, wealthier and more technologically savvy voters – who tended to vote liberal – switch from paper voting to e-voting.Footnote 87
Conclusion
The e-voting programme was but one of several digitisation initiatives pursued by the Estonian state. Some were more successful than others. Efforts to develop digital democracy often faltered on poor design choices, haphazard implementation and a lack of political support. The Laar government launched two other ventures alongside e-voting: e-government and a participatory democracy forum called Today I Decide (Täna Otsustan Mina). The first can be described as a glorified PR project. Launched at a moment’s notice in 2000, even Linnar Viik, the programme’s godfather, had a hard time describing what e-government was supposed to do. Ultimately, it amounted to little more than an online system for managing the paperwork of government sessions, so ministers could share and approve bills on their laptops.Footnote 88 Today I Decide allowed citizens to propose, discuss and vote on new legislative ideas, which then required a response from relevant officials. The crude online platform was quickly overtaken by trolls, and civil servants had to respond to what they perceived as ludicrous proposals (such as drug legalisation) with no additional resources. In other words, the programme was unpopular among all constituents, and by 2004 it was already considered a failure.Footnote 89 Programmes focused on digital services fared better, particularly if they could be financed from EU structural funds and had buy-in from private actors. Still, many initiatives, such as a government e-health platform developed from 2005 to 2012, were plagued by cost overruns and conflicts between interested parties.Footnote 90 The particular confluence of public and private interests, mediated by the network of cyberneticians and unhampered by legacy systems that made digital ID and e-voting possible, only came around once a century.
International trials of e-voting in Norway, Switzerland, Canada and elsewhere have often resulted in negative assessments or decisions to limit e-voting to ‘lower-stakes’ situations, such as municipal elections or local referenda. In Norway, the opposition of right-wing parties over issues of trust led to the discontinuation of e-voting trials after a change in government in 2013.Footnote 91 In Switzerland, where e-voting is permitted in certain cantons for citizens living abroad, implementation has been slowed by ongoing security concerns and the need to develop technical solutions that respond to these (such as return codes that verify that a vote has been recorded accurately).Footnote 92 In the Netherlands, enthusiasm for e-voting dimmed considerably after activists groups pointed out various ways in which internet voting compromised the secrecy of the vote.Footnote 93 Few other former Soviet bloc countries have pursued digitalisation of the public sector to the same degree as Estonia. Officials in both Latvia and Lithuania have discussed e-voting but rejected it on the grounds that the security of the process and the secrecy of the vote cannot be sufficiently ensured.Footnote 94 Analysts in Ukraine have noted that the minimal preconditions for considering various e-democracy projects – a secure authentication method akin to the ID card and centralised voter registries – were only created in the early 2020s, and thus any serious consideration of e-voting is still years away.Footnote 95 Finally, as a dark reminder that e-voting is not the same as democracy, using the coronavirus pandemic as an excuse, Vladimir Putin’s Russia allowed e-voting in the constitutional referendum of 2020, which removed term limits from the presidency, as well as the 2021 State Duma elections and elections in Ukraine’s occupied regions of Donetsk and Luhansk in 2022.Footnote 96
While there are admittedly few comparative studies on e-voting, two central reasons for limiting its expansion seem to have been political controversy and the lack of an infrastructural base.Footnote 97 While other socialist states, too, benefited from expertise in cybernetics, Estonia’s position on the Baltic made it unique. Nordic banks and telecoms motivated the state to pursue the ID card project, and long-standing ties to the Nordics gave Estonian cyberneticians cross-sectoral expertise that engineers elsewhere lacked. Of course, an authoritarian regime like Putin’s Russia could simply ignore concerns over security and confidentiality and pursue its political agenda. The Estonian case was successful precisely since it was pursued not as a political project but rather as part of a public–private infrastructure project led by civil servants, banks and telecoms. Indeed, political interest in the programme surged years after it had been implemented, once opposition parties began to suspect it might negatively affect their electoral performance. As for infrastructure, state officials and IT engineers, who had bought into the system from the beginning, had anticipated some criticisms and could deflect others by referring back to Estonia’s ‘special’ ID card foundation.
Ideologically, the post-Cold War Estonian state appears deeply contradictory, providing material for narratives of exceptional innovation and depressing typicality alike. Those looking for evidence of Estonian uniqueness can point to its commitment to building a digital democracy, in the form of initiatives such as e-voting or participatory platforms. Those who consider Estonia a textbook case of the neoliberal turn note that the most successful digital initiatives have been vehicles for leveraging state capacity to the service of private enterprises, and that ideologically, the digital republic has been articulated as an alternative to the administratively bulky welfare state. The popularity and longevity of both these narratives are primarily a testament to the salesmen of the digital republic, who have stitched together coherent narratives of contingent, unexpected and haphazard developments, in order to sell the e-state to domestic political and international business audiences.
There is some truth to both these tales, of course. Reconciling them, though, requires an infrastructural approach. Investigating the emergence of the sociotechnical systems that underpin the digital state reveals a story that is much more local, characterised by negotiation and compromise and populated by a very different set of actors than in either the triumphalist or tragic stories. The story of the digital republic is a story of unintended consequences, without a straightforward ideological trajectory, because the actors involved had no singular goal in sight. Much like the story of Latin American ‘mixed economies’, the infrastructure of the digital state generated resources for a host of different purposes.Footnote 98 Some of these goals grew out of Soviet administrative structures and the Institute of Cybernetics, others had to do with European integration and others still with business interests piggybacking off public infrastructure. If Stephen Collier suggested that many of the neoliberal reforms in post-Soviet Russia were constrained by existing socialist infrastructure in the city of Belaya Kalitva, then my analysis suggests that a version of neoliberalism was both made possible as well as constrained by the assemblage of infrastructures whose roots lay, among others, in the Soviet past. Rather than being invented by transnational companies, national politicians or intellectuals, the ‘neoliberal’ digital republic was built around the infrastructure of a universal, compulsory digital ID. The ID card was imagined by Soviet cyberneticians, sponsored by banks and telecoms who offloaded their own R&D work onto the state, but it ultimately facilitated the construction of services that both meaningfully simplified the exercise of democratic power and generated a discourse that emphasised the efficiency of state bureaucracies over all other concerns. The story of the digital state is one of infrastructure that was repurposed and rethought over time – and can be repurposed and rethought today.
Acknowledgements
The author would like to thank Raivo Ruusalepp and Kaidi Ruusalepp, who were both part of the ID card working group, for many informal conversations. Märt Põder offered extensive thoughts on the virtues and vices of online voting. Marysia Jonsson, Linda Kaljundi and Gustav Kalm provided valuable comments on drafts of this article. Versions of this paper were workshopped at a public lecture sponsored by the Institute of Social Studies at Tartu University and the NordAI seminar at KTH.
Funding statement
This project has received support from Estonian Research Council grant PRG346 and from the project ‘Development of the Digital Republic’, financed by the Estonian Ministry of Communications and Economic Affairs and the Information Systems Authority, and executed by Ernst&Young, for which the author was an independent contractor.
Open access
