What happens when we vote?

“You don’t need the bullet if you got the ballot” (George Clinton, Chocolate City)

At last, on 9 October 2004, the people of Afghanistan have an opportunity to choose their leaders. They will exercise the franchise under intense pressure and scrutiny. Under Nato protection and with the support of the United Nations, many will brave threats of violence within insecure territories beyond Kabul.

Despite all the preparations and precautions, both voters and candidates find their lives in peril in the days leading up to the vote. And questions remain about the legitimacy of the voter registration efforts themselves. Will every Afghani be able to vote? Will some places have more votes cast than eligible voters?

Amid such familiar doubts and concerns, everyone who wants Afghanistan to have a peaceful and just future hopes this election attains some level of legitimacy in the eyes of the world and the Afghan people. If it works, the Afghan experience will serve as the model for relatively quick transitions to democracy. If it fails, for whatever reason, the cause of democracy in unstable places could suffer deeply.

We justifiably celebrate the rapid spread of democracy to most corners of the globe – from Indonesia to Georgia - over the past twenty years. We often pay close attention to questions of access, registration, personal security, and ballot-counting procedures. Yet we rarely examine the actual act of voting - the most direct form of communication that exists between citizens and their state.

Voting is subtle, simple, important - but also dangerously susceptible to manipulation. In the world’s most comfortable democracies we’ve come to take voting so much for granted, that it’s easy to forget that it rests upon a lightly-spun web of trust.

The question of “how we vote” is essentially one about the nuts and bolts (or papers and pencils, circuits and memory chips) of democracy. Confidence in democracy is based on the transparency and trustworthiness of the voting system.

It’s a timely question. Governments around the world are being lured into purchasing new and automated systems for their elections. But at what cost and at what risk? After all, new is not always better.

In some countries - like India, the Netherlands and Brazil - new machines and techniques seem to be working rather well, augmenting faith in the democratic process and strengthening voices of citizens across cultural, linguistic, ethnic and political boundaries.

Elsewhere, and particularly in the United States, officials seem to have unlearned the lessons of several centuries in their zeal for new technologies sold by wily corporations. As we approach the most important election in a generation, Americans are anxious to see whether the next person sworn in as president will really be able to claim democratic legitimacy.

One thing is certain: citizens lulled by the regular rhythms of electoral pageants must step up and hold their leaders accountable and ensure their ballots are countable (and recountable) before bad technology causes undue damage to the foundation of democracy.

America, the imperfect

The United States secretary of defense Donald Rumsfeld declared on 23 September 2004 that even if huge sections of Iraq could not participate in the country’s planned January 2005 elections, the vote would still stand as valid and legitimate.

"Let's say you tried to have an election and you could have it in three-quarters or four-fifths of the country. But in some places you couldn't because the violence was too great," Rumsfeld said at a Senate Armed Services Committee hearing. "Well, so be it. Nothing's perfect in life, so you have an election that's not quite perfect. Is it better than not having an election? You bet."

Put aside for a moment all the turmoil over the justification for the Iraq war and the contradictory messages we hear about the potential for peace and security. Is it true that an imperfect election is better than no election at all?

Well, there is imperfect. And then there is invalid. The American electoral system is grossly imperfect. Routinely, the states and cities in the United States hold elections in which at least 2% of voters find their ballots uncounted. That’s considered standard and an acceptable loss.

However, if one-quarter of the potential electorate is unable or unwilling to brave grave threats of violence in order to vote, it’s hard to claim that an election is valid and the resultant government is legitimate.

So where does the line fall? Somewhere between 2% and 25%? Why should we accept any vote loss? And if we agree that minimal vote loss is healthy for democracies, what methods and technologies should we employ?

As the most powerful nation on earth prepares for its quadrennial national elections it still can’t answer these questions.

While Iraqis will have to brave car bombs and drive-by shootings to express their wishes for their nation, Americans will drive peacefully to the polls realising that there is a one-in-fifty chance that their efforts to vote will be for naught. Iraqis will be disenfranchised by insecurity generated by political arrogance. Americans will be disenfranchised by technological ignorance.

A leap of faith

This year, with the widespread use of unverifiable, un-recountable, top-secret proprietary electronic voting systems, Americans are wondering if their system is getting more or less accountable.

The incumbent president assumed office four years ago despite losing the popular vote by more than half a million votes. He claimed victory by winning the electoral college votes of Florida, a state in which 174,000 people who thought they had voted for president had not had their votes counted.

More than 3% of voting Floridians (disproportionately African-American and Democratic) did not have a say in whom their electors would choose. George W Bush was able to convince the US Supreme Court to stop a manual recount of ballots while he still held a lead of 537 votes. Is this validity? Is this democracy? Just what are we exporting with such fervour?

When long-time civil rights activist, Reverend Al Sharpton, addressed the Democratic National Convention in Boston in July 2004, he proclaimed the value of the vote to those who deeply and passionately believe in democracy: “Our vote was soaked in the blood of martyrs,” Sharpton told Democrats. “This vote is sacred to us.”

Sharpton was right. Voting is a quasi-religious experience. It demands a leap of faith.

Voting is essentially irrational. It’s a common example of “Sorites paradox”. One individual vote means little or nothing. Opting out costs nothing in the immediate term. There is no rational reason to believe that a single vote will ever make a difference even in a tiny local election.

Yet billions vote worldwide. Even in the United States, where democracy is so taken for granted that the result of a presidential plebiscite is occasionally ignored, more than half of the population are likely to show up at the polls on 2 November. They justify their actions as “civic duty,” and remind themselves that there is satisfaction in being part of a collective process, even one so obviously flawed and easily corrupted.

Voting depends on trust in one’s community, one’s officials, and in the communicative technology that facilitates the delivery of a simple message across long distances.

Voting as communication

Voting is a communicative act. It’s odd to think of voting this way. We think of communication as having a more complex nature – more data sent over thicker bandwidth. But even a thumbprint on a piece of paper can be one of the most powerful messages there is. Most public forms of communication call citizens to shout loudly, proclaim their distinct voices, and attract attention. But voting is only effective if the particulars are transmitted secretly, and as part of a collective will. A lone, eccentric vote for Mickey Mouse is a failed or wasted vote. An anonymous, one-in-a-hundred-million vote for Lula to be president of Brazil is a successful communicative act.

But no other form of communication allows citizens to complete a circuit so easily. No other form allows a poor citizen to issue a signal with the exact same amplitude as a rich citizen. Yet no other regular communicative act is so fragile. Because the message itself can be a single point of data (to remove or retain Venezuela’s president, Hugo Chavez, for instance), it’s easy for that datum to be altered or halted along the circuit from voter to counting machine to certifying body to public declaration of results.

Voters everywhere need to believe that their vote will be considered sacred. And they demand that the entire system be dependable. This need may be more acute in places where democracy is new, dangerous, and not yet taken for granted. It’s safe to say that if the practice of voting were less trustworthy, fewer people would freely choose to vote. And what remained of democratic practice would soon degrade into empty rituals of voting without meaningful interaction between states and citizens.

Voting has been less trustworthy, of course. There have been more fraudulent, coerced, and conjured elections in the last half-century that we can begin to count. And no one pretends that the few countries that held elections in the 19th century came anywhere close to generating governments that represented the will of the governed. Yet every time we seem closer to agreeing on the values that might enrich and maintain democracy – trust, accountability, openness, and universality – we see how hard it is to engage those values in practice.

Computers and humans

There are five basic voting technologies that dominate electoral practices around the world. The oldest and most common is pencil on paper, read and counted by human hands and eyes. More recent systems include computer-read punch-cards (as were tragically common in Florida in 2000), computer-read optically scanned paper ballots (the best system by far), and newer direct-voting computer units that allow for instant counting and prevent recounting. Some isolated corners of the world, such as New York state, use rusty and unreliable old cabinet machines that require voters to push a lever next to the candidate’s name. These machines are disasters waiting for a close election to happen.

What’s wrong with pencil and paper? In much of the world people express their will by checking a box printed next to the name or symbol of a candidate or political party. The voter then places the paper ballot in a padlocked tin box. At the end of the voting day, deputised election officials tally the votes, make their best guess about the intent of the voter (not all marks are clear), and submit the total to a central counting office.

This simple method does not risk blackout or battery failure, distant hacking, software glitches, paper jams, or centralised corruption. It’s relatively easy for illiterate people to participate. And the common, familiar nature of this technology does not alienate those who lack exposure to more complicated interfaces.

To steal an election in which everyone votes on paper with pencil, and the counting is distributed throughout the region or state, the corrupting party must intimidate or bribe thousands of people or prevent the boxes themselves from reaching a central depository where they can be audited. This is difficult, but not impossible. Only a bold or desperate government as unconcerned about global public opinion as it is of local opinion would stoop to such coercive methods. Still, it happens everywhere. It always has.

The chief problem with paper is that it takes too long to count thousands or millions of ballots. We are just not patient enough for such levels of accuracy and accountability. In a close American presidential election, the final results might not be available for many weeks. We get pizza delivered to our homes in thirty minutes. We get video on demand. We can’t wait any longer to see who will lead us for the next four years, even if we get the wrong answer.

Old-fashioned paper-and-pencil voting systems are also best when the number of electoral questions is small. It’s great for up-or-down votes, as in Venezuela. It’s good for a short election of six city council seats. But when the ballot stretches for eight pages and contains sixty candidates for twenty offices, paper ballots and distributed local counting can become unwieldy.

Also, distributed counting by local election staff risks error. Eyes tire. Tempers shorten. And training and expertise range from veteran arrogance to rookie ignorance. Over an entire state, such errors may cancel themselves out and are correctable upon recount. But the price may still be high.

So why not impose computers between voters and officials? For almost a century we have been discovering new, powerful ways of manipulating and tabulating complex and lengthy data sets in mere seconds. Computers are very fast and don’t tire. The boxes in which they come are lighter and cheaper than ever before. And we increasingly trust them in other essential areas of life, such as banking, education, and reading articles on openDemocracy.net.

Computers have one other tremendous advantage over humans: they appear neutral. This can generate trust. If voters do not trust their neighbours or the soldiers waiting outside to interrogate them, they may assume that the impersonality of the computer interface – whether a punch-card, scannable sheet of paper, keyboard, or touch-screen – grants fairness and accountability to the process.

For those who trust computers, this impersonality can be valuable when taking the leap of faith. Of course, computers and the software that runs on them are not neutral. They embody the biases and values of their designers. For this reason, software should be open to scrutiny and correction. In the United States and many other countries, such common sense public oversight and quality control are beyond the bounds of political imagination.

In America, the earliest widespread effort to install the aura of computer accuracy and efficiency to ballots during the first election of President Lyndon Johnson in 1964. Ironically, Johnson had built his entire political career in Texas on winning elections fraudulently. Texas lore is filled with stories of Johnson cronies stuffing, swapping, and stealing ballot boxes to ensure their man won.

When Johnson ran for president, many states adopted an IBM-developed technology called punch-card balloting. Almost immediately, it became clear that the process confused voters. If a voter mistakenly punched more than one candidate for office (as happened frequently) the entire ballot would be rejected – an “overvote.” If the voter failed to completely punch through the thick paper, leaving the “chad” attached, that particular vote would not count, regardless of the clear intent of the voter. This is known as an “undervote.”

Despite more than thirty years of documented problems with both overvotes and undervotes, many states retained these punch-card voting systems through the 2000 election. Everyone in the election administration community knew these machines were bad. But few thought massive disenfranchisement of millions of American voters was an important enough problem to replace these machines.

In 2000, it became clear that when the election is close and voter turnout is high with first time voters, these systems are destined to fail spectacularly. And so they did.

Technofundamentalism

US officials scrambled to fix the problems revealed by the Florida debacle of 2000. Democrats hoped to install systems nationwide that would simplify the voter experience so that voter intent would be perfectly clear. Republicans wished for a system that would standardize voting and counting and limit the possibility for messy recounts. Both parties looked toward a collection of private companies that had been developing proprietary, closed systems to record and tally votes.

In 2002, Congress passed and President Bush signed the Help America Vote Act, which authorised $3.9 billion to states so they could purchase new voting systems. Without blinking, many states jumped right into contracts with these voting companies. After all, officials figured, the next technology would necessarily be better than the last technology.

As in 1964, it became immediately clear that these machines created more problems than they solved. While they did improve access for the visually impaired, they offered few benefits to the rest of the electorate. The first generation of direct voting machines did not offer paper audit trails so that voters and officials could re-check anomalous results. Many machines were to be connected to central tabulation computers (in the control of the companies, not the government) by insecure telephone lines. Many of the software systems were ridiculously easy to hack. And some just plain failed to work properly.

Conspiracy theories abounded. The 2002 Georgia races for senate and the governor’s office both resulted in upsets, with Republicans defeating popular incumbent Democrats after the state adopted electronic voting machines. Later, reporters discovered that the company that supplied the machines had sent software updates to the machines after they had been certified by the state. In other words, the Georgia voters were using uncertified software secretly supplied by a private company. Compounding suspicions, the CEO of Diebold, the company that supplied the Georgia machines, openly declared his support for President Bush in the 2004 elections. Mitigating the conspiracy theories was the fact that the chief election official and most county officials in Georgia were Democrats.

Nonetheless, corporate secrecy compounded doubts about accountability. Public criticism mounted. The state of California decertified its new machines and demanded paper trails throughout the state. The companies fought back with public-relations campaigns and legal bullying.

While the United States prepares for another close presidential election with high turnout, it retains its patchwork system of electoral technologies. Some states leaped head first into the era of secret, private computer code, putting their futures in the hands of private companies with poor records of quality and openness; others have done nothing to improve the voter experience or install trust in the system.

One size fits all?

While the world’s oldest democracy creaks and chokes its way to electoral chaos, the world’s largest democracy adopted computer-voting systems relatively smoothly. In June, India employed millions of electronic voting machines in its national elections. Complaints were few and the election ran smoothly considering widespread religious violence in recent years. And to everyone’s surprise, the opposition Congress party emerged as the strongest single party and was able to form a governing coalition.

In 2002, Brazilians used a different electronic voting system (one that originally raised considerable criticism for lacking a verifiable paper audit trail). But the opposition party triumphed there as well. In both India and Brazil the elections were not close, so any errors and systematic problems may have been hidden. Since that election, almost all the machines in Brazil have had paper receipt system installed. Despite Diebold’s involvement in Brazil’s programme, the system has attracted more praise than scorn or suspicion.

In both the Brazilian and Indian cases, the lack of obvious tampering does not indicate immunity to hacking or disruption. These systems are still potentially vulnerable. But in both countries, the battery-operated machines never send data over insecure channels – a vast improvement over the insecure American model. Still, the very idea of using proprietary software for public elections raises suspicions among voting rights advocates and computer scientists in these vibrant, growing democracies.

Perhaps the greatest success story for direct electronic voting has been in the Netherlands. Some version of computer voting has worked in the Netherlands since the mid-1970s. And in recent years, the state has been posting the source code for the counting machines publicly on the internet. Once open and posted, coders may point out insecurities and errors, thus improving the system constantly. Dutch officials are also forging ahead in offering on-line e-voting to expatriates.

None of these democracies use the same system. And many other countries – Ireland, Canada, South Korea, for example – have been considering electronic voting machines for some time.

There is no such thing as a perfect voting method that should be applied everywhere in all circumstances and environments. Every technology enfranchises some and disenfranchises others. Some work better in highly-wired small countries. Others work better in vast nations with multiple languages and levels of literacy. But there are general lessons available from the experiences of India, the Netherlands, and Brazil which show the four practical values democracies worldwide should embrace when they build their voting systems: trust, accountability, openness, and universality.

• First, the code, from pencil to screen, must be open to public scrutiny. The state must control the process and dictate its needs and policies to the contractor – not the other way around
• Second, the interface must be simple and clear, so that people with different language skills and abilities may use the system
• Third, if machines are used they must have a verifiable paper trail so that glitches can be addressed immediately and recounts can proceed without controversy
• Fourth, the data should be secure (never connected to a phone line or the internet; data should be stored in encrypted form on a solid medium like a disk).

These are principles and lessons the United States (the home of “vote early and vote often”) still has to learn and apply.