It was 2010. I had come back to Pakistan after completing my studies on a Fulbright scholarship from an Ivy league business school in the USA. With a background in computer and communications engineering, and now specialization in technology management, the aim was to contribute by developing products and solutions that would bridge the gap between the users and technology. I started working in Islamabad with the research and development lab at COMSATS Institute of Information Technology as marketing manager and the first big project that I helped conceive, design and launch was Electronic Voting Machine for Election Commission of Pakistan (ECP).
I presented the first prototype to a diverse group of politicians, ECP officials and diplomats at a showcase event at Marriot hotel in March 2011, organized by the Election Commission. There were six other presenters, including local and foreign solution providers. Our solution was unequivocally applauded as the best. We were the only solution providers to receive coverage in the next morning’s national newspapers, which was yet another proof of its excellence and acceptance.
Our mantra was: “Intikhaab Pakistan, a solution made in Pakistan, made for Pakistan.” This was an end-to-end solution, starting from biometric voter identification, electronic voting machines with paper audit trails, instant and secure results transmission to ECP headquarters, and even a voting solution for overseas Pakistanis. We were ahead of the game and we believed that we had solved the problem using an elegant and reliable solution by combining a clear understanding of Pakistan’s localized election landscape and applying state-of-the-art technologies that best fit unique requirements of our democracy. We had developed a panacea for the electoral problems of Pakistan, or so we believed.
We spent the next three years presenting the solution to numerous audiences, including political parties, media personnel and even foreign embassies, all in the interest of digitizing their countries’ electoral processes as well. We even participated in field test in a by-election in Multan to validate the system in an applied setting Unfortunately, as with most such initiatives, gradually the interest in EVMs faded and soon the prototypes we had developed with great aspirations, became nothing more than a display piece for dignitaries who would visit COMSATs university. As time progressed, the staff used these machines to show how the university was trying to solve real-world problems and developing industry applications, in hopes of getting recognition, and perhaps much needed funding.
Fast forward to 2021, Fawad Chaudhary is now the Minister of Science and Technology and through the prerogative of his portfolio, chancellor of COMSATS University. The prime minister showed interest in voting machines and the next thing we knew, the same EVM that was now celebrating its 10th anniversary of collecting dust, was being viewed with thorough interest, awe, and media fanfare at Bani Gala. EVMs these days are being hailed as the savior of our democracy and also being positioned as the cure-all of all ills of Pakistan’s electoral process. Promises that EVMs will allow for every eligible vote to be casted and counted are prevalent in media talk-shows. The president of Pakistan has stepped forward, leading the charge in support of e-voting, swiftly promulgating a presidential ordinance to procure and install these machines. The president’s quote on May 5, 2021: “So if the matter is decided with everyone’s input 4-5 months down the line, we will need to produce 1,000 machines a day to meet the target of 325,000 machines in total.”
It has been 10 years since I started working on the EVM project and it has been five years since I ceased working on it in an official capacity. This multi-dimensional perspective has given me an opportunity to analyse electronic voting implementation in Pakistan both from an insider’s perspective as well as be able to look at it objectively.
ECP, whose current headcount is around 2,500, will need to upskill a team of 735,000 across the country, imparting technical skills and capability to set-up and operate EVMs.
EVMs: How do they work?
There are three concepts about e-voting that we need to understand. First, electronic voting is not about a product but the process. It is not the 325,000 machines but more importantly the infrastructure around them, including pre-election and post-election mechanisms, that need to be in place to make sure the system works. Leave one end loose, and the whole system becomes disputable. Second, every feature, be it necessary or desirable, comes at a cost which needs to be realized and measures need to be taken to ensure the cost remains bearable. And third, the risks associated with a monumental transition such as this, need to be evaluated from all angles within the society.
Let us begin with voter identification. A voter walks into a polling station. The presiding officer places the voter’s finger on EVM’s Identification module. The module is either centrally connected to NADRA’s database or stores local voters’ data. If there is a match, a green light flash informs the staff and polling agents of the voter’s validity, ID card information which is automatically captured, and the voter is allowed to cast her vote. The solution is simple and easy to follow. However, there is a problem. Even though Pakistani citizens facing frustrations while verifying themselves at banks or purchasing SIM cards may choose to differ, NADRA claims that at best it can recognise thumbprints of 82 percent voters. That is 18 percent unrecognisable fingerprints. According to the latest population figures, the 260 National Assembly constituencies (excluding 12 for former FATA from the total of 272) should have an average of 780,000 inhabitants and about 415,000 voters, 18 percent of which is 74,750 voters. Let me rephrase this – on average, even at best, NADRA’s database will not verify 74,750 voters in every constituency.
In the 2018 general elections, 32 percent of National Assembly constituencies had a victory margin under 10,000, 43 percent had a margin of under 15,000 and 52 percent had a margin of under 20,000 votes. So, the number of unrecognizable votes will be four to eight times the victory margins. Also, we must remember the fact that NADRA has now been collecting digital fingerprints for about two decades and claims to be using the latest machines with improved technologies and self-proclaims to maintain one of the most rigorous citizens’ biometrics databases in the world. Like mentioned earlier, e-voting is a process, not a product. And the process begins with accurately capturing and maintaining a biometrics database.
Now, let us move on to the Ballot Casting module. This module is the heart of EVM solutions. It has at least three units but could be more. First, it has a control unit. This is the counter that keeps record of votes casted and is managed by the presiding officer through a smart key. The control unit connects with one to eight other balloting units using cables. The ballot units are large tablets with images of candidates next to push buttons. The ballot units are easy to operate and resemble an actual balloting paper. There are no touch screens to ensure inclusion of population with low digital literacy skills. The balloting units are placed behind secrecy screens to ensure voter privacy. There is no wireless connection between the control unit and the ballot unit to ensure data security. All activity from the start till end is time and date stamped. Multiple ballot units could be required, depending on the number of candidates in a constituency. A single ballot unit can accommodate 16 to 24 candidates. Furthermore, based on ECP’s requirement for a paper audit trail, the module also consists of a printer, affixed in a ballot box. Once a voter pushes the button against the candidate of her choice, a copy of her vote is printed for the voter to view and verify and then drops into the ballot box. When the election ends, the presiding officer, using a set of digital keys, locks the electoral process and pushes a few buttons to begin the display of results. The printer prints Form 45 with the tallied results on it. With slight variation, almost all EVM solutions currently under consideration by ECP follow the same principles explained above.
The Indian EVM solution is similar, only it does not contain the paper audit trail. Therefore, the Pakistani solution is in fact an upgrade from its Indian counterpart due to its auditability. However, this auditability comes at a cost.
Using EVMs in general elections
First, let us talk about scale. About 85,000 polling stations and 242,000 polling booths were set up in the 2018 elections. A nationwide EVM-based election would require an estimated 85,000 voter identification machines, 170,000 control units, 340,000 ballot units, and 340,000 printer ballot boxes. In effect, we are talking about one million machines as opposed to 350,000.
Even if the designs of various components of the system are approved and tested and full-scale manufacturing exercise is carried out, a timeline of one year would require non-stop production of 3,000 units a day. How do we do this? One solution is to move production overseas. However, outsourcing such sensitive operation would result in a kink in the national security armor and potentially the sovereignty of the country. It is not unreasonable to imagine various actors, national and international, making attempts to embed software and hardware backdoors to the systems only to activate them for their nefarious agenda. This has happened in recent past. The Philippines government, after receiving military intelligence reports that China might sabotage their election, had to order transfer of EVMs’ production to Taiwan. Although China denied any such allegations calling it “sheer fabrication”, it did strain diplomatic ties between the countries.
A nationwide EVM-based election would require an estimated 85,000 voter identification machines, 170,000 control units, 340,000 ballot units, and 340,000 printer ballot boxes. In effect, we are talking about one million machines, as opposed to 350,000
In that case, let us discuss onshore production. Do we have capacity to manufacture, test, and package 3,000 electronic devices a day for three hundred and sixty-five days straight, without fail? Unfortunately, it seems unlikely that any government or trustworthy private organization will be up to the task. Indian EVMs are manufactured by two Indian companies; Bharat Electronics and Bangalore and Electronics Corporation of India. Bharat electronics has about 10,000 employees and an annual revenue of USD 2 billion. The company is state owned and was established with the aim of meeting national needs onshore. National Institute of Electronics was established in Pakistan back in 1979 with a similar purpose. Unfortunately, 40 years later, we are yet to see any meaningful product roll out of their facilities.
One potential solution could be to let ‘strategic’ institutions like NRTC manufacture the EVMs. But by getting directly involved in the election process, these institutions run the risk of facing serious allegations of election engineering and results manufacturing. The 2018 general elections were no exception. Moreover, it will be a direct conflict of interest for these institutions to be getting involved in the process of electing their future administrations.
Let us talk about costs now. In 2013, ECP spent Rs4 billion on elections. In 2018, the election bill inflated to Rs21 billion. A conservative estimate for the cost of EVM machines is USD500 million, so about Rs75 billion for just the machines. The packaging, transportation and installation involves costs. Likewise, voter education is a cost. Training polling staff is a cost. Before we know, this becomes a billion-dollar initiative. On training, about 735,000 polling staff were engaged in the 2018 elections. This means training three quarters of a million people (BPS 16-21 officers) to make sure that the biometric unit, control units, balloting units and printer boxes are accurately connected, appropriately placed, and operating properly on election day. ECP, whose current headcount is around 2,500, will need to upskill a team of 735,000 across the country, imparting technical skills and capability to set-up and operate EVMs. This involves a comprehensive training regime, and teams of technical trainers. More processes, more costs!
Speaking of technical skills, we also need to address tech-support. Tech-support by field engineers is a mandatory and critical requirement of this endeavor. A nationwide EVM rollout will require election day tech-support in all 130 districts of the country. A certain number of EVMs will fail in the field no matter how rigorously they are tested. Similarly, polling staff will invariably commit errors and run into technical/training issues which would need to be managed, within eight hours of election day activity. As soon as EVM malfunctions and technical assistance is required, the polling staff should be able to reach out to the tech-team for telephonic support. If the issues remain unresolved, an expert technician or an engineer from the tech-support team needs to reach the polling station within one hour and attempt to fix the machine. If the issue still remains unresolved, the faulty device needs to be retrieved and a replacement needs to be installed. Moreover, the faulty EVM needs to be sealed as integrity of the election results thus far, need to be preserved and later collated with results from the replaced device. These are all very realistic scenarios and would require ECP to set-up a tech support infrastructure including scenario planning, SOPs, technical teams, call centers and dispatch units. More costs, most processes, more risk!
Let us now talk about process auditability, which is critical in the election process, thus ECP’s decision to use printers in every EVM. However, every feature comes at a cost. The provision of paper audit trail brings in added complexity. Printer is a power-hungry device. It cannot operate on commonly available alkaline batteries. Due to uncertainty of power infrastructure, EVMs could not be operated through the national power grid either. Durable and reliable 12-volt batteries are needed to make sure the printers are able to provide eight to ten hours of operation. Keeping in view the hot and humid climatic conditions in Pakistan, these batteries need to be properly stored before election.
And this brings us to storage. EVMs’ application is peculiar in the sense that they are used only once in five years and for the remainder of the four years and eleven months, they need to be stored under suitable environments in tamper-free warehouses, waiting for the next elections. The storage facilities need to account for hot and humid weather conditions as well as natural calamities such as floods, earthquake, and other possible disasters. ECP may opt for a central warehouse, which would mean bringing all one million devices from all across the country into one location after each election cycle. This may be considered a risky proposition – to have all EVMs in one warehouse. Another option is to have regional warehouses across the country, perhaps a few dozen. ECP would need to build the warehouses and also recruit staff for the management of these warehouses. This adds multiple layers of complexity and costs. One of the main reasons Ireland gave up on EVMs was storage costs, logistics and corresponding security concerns. Back in 2007, annual storage costs for just 7,500 machines swelled up to €140,000 (About Rs3,500 per EVM, per year – or Rs3.5 billion per annum for 1 million units). In April 2009, a decision was taken to scrap the EVMs in Ireland because it would cost too much to maintain them and were termed “a scandalous waste of public money.”
We also need to discuss EVMs’ surveillance when in storage in these purpose-built warehouses. While the EVMs would hibernate in store rooms, they would need security as well, otherwise this provides a perfect opportunity to carefully and discreetly “manage” the future elections. Once out of their slumber every five years, all one million EVMs units would need black box and white box testing, to ensure not only that they are operational but also to make sure that they have not been tampered with. The Netherlands, one of the first countries to implement electronic voting, eliminated EVMs in 2007 after it was discovered that it was possible to get complete and undetected control of the election results if one had access to the devices before the elections, even for a brief period. So ECP would need adequate measures and investment to make sure that all stakeholders would have full confidence that no security breach took place in five years, even for a brief period of time, in any of the warehouses, with any of the machines. Have these costs and risks been factored in?
For any electoral process to be acceptable under internationally recognised standards, the voter’s choice needs to be protected through anonymity, yet the process needs to be verifiable, secure, and auditable. This presents a conundrum. To provide anonymity, the solution is to de-link identification from voting, hence two separate units for identification and balloting. However, this is only pseudo-anonymity. Since all activity in identification and balloting units needs to be time and date stamped, collating data of the two machines later would easily show who voted for whom. Another solution could be staggered storage of information, like is currently being done in the banking sector. But this also results in more complexity. The more complex we make our electoral process, the more obscure it becomes. The more obscure it becomes, the more distant an ordinary citizen gets from it.
The hallmark of an electoral process is its inclusiveness and transparency. Every citizen should be able to contest in an election, vote for a candidate of his or her choice, and also be able to monitor the process. Technology does bring the promise of making the process efficient and secure, but the decision makers need to fully comprehend the monumental challenges this endeavor would entail. This task is not insurmountable for sure but it is definitely not also a matter of merely procuring a few hundred thousand devices and pushing a few buttons. No matter how much we may desire it, presently there is no off-the-shelf quick solution available in the market that could help hold nationwide elections electronically. Many other countries around the world are experimenting with it and learning from their mistakes. Perhaps we could also start small but start smart. We need to learn to crawl before we run. Are we willing to take a comprehensive and sustainable approach towards the solution? Are we ready to appreciate the challenges that technology will bring with it? Or are we looking for a quick band-aid solution? Does EVM produce undisputable results or does it pave the way to rigged and even more fraudulent, difficult-to-trace results? Is EVM a panacea to our democracy or is it Frankenstein’s One-Billion-dollar Creation? These questions are for readers to ponder and stakeholders to decide.