Doing a Number on Violators


As Patrick Ball scrawled equations across a conference-room white board, his talk was of regression analysis, matching methodologies and capture probabilities.

His numbers were the equivalent of blood spatters at a crime scene.

For three years, Ball traveled back and forth to Kosovo, systematically culling data on civilian deaths from refugee reports, exhumations and witness accounts. Building on that evidence, he and his colleagues compiled a database documenting the ebb and flow of “ethnic cleansing” of ethnic Albanians during the spring of 1999 in Kosovo, a province of Serbia, Yugoslavia’s main republic.


The statistical portrait of the displaced, missing and killed reveals the timing and ferocity of fatal blows that fell across an entire province. This numerical pattern of death and panic exonerates some people; it points toward others.

Now, the statistics that Ball calculated on a Boston white board have become evidence in a war crimes trial. On Wednesday, in an international courtroom in The Hague, Ball confronted the man he believes is responsible for the deaths--former Yugoslav President Slobodan Milosevic.

The evidence Ball laid out as an expert witness for the prosecution represents the newest infusion of technical expertise into the human rights movement--an effort to harness information science to track the beatings, rapes, killings and mass executions of systematic political violence.

Milosevic has argued strenuously in his own defense that NATO airstrikes or the ethnic Albanian guerrillas of the Kosovo Liberation Army, or KLA, could have been responsible for the estimated 11,000 civilians killed between March and June 1999. The statistical analysis, Ball testified, demonstrates that neither is possible.

The numbers, he said, establish a clear pattern: The culprit was an organized campaign of “ethnic cleansing” by Yugoslav military and paramilitary forces under the command of Milosevic.

“When we looked systematically and really carefully at the killing data, I found this pattern,” Ball said during an interview before his testimony. “My jaw dropped through the floor. It blew me away.”

Negotiating a Minefield of Unreliable Data

Ball, the deputy director of the science and human rights program at the American Assn. for the Advancement of Science in Washington, has spent a decade perfecting the use of computer technology in the service of human rights.

To arrive at their conclusions about Kosovo, he and his team had to negotiate a minefield of technical uncertainties and unreliable data.

In all, more than 800,000 ethnic Albanians fled Kosovo in 1999, in the largest mass expulsion of people in Europe since the 1940s.

No one knows exactly how many died. International prosecutors and human rights activists allege that as many as 11,000 men, women and children were killed in a campaign of terror by the Yugoslav government designed to trigger panic and cause people to flee their homes. Atrocities in more than 500 towns and villages have been documented.

The worst of the killing took place during the 11 weeks of NATO airstrikes in Yugoslavia, from March into June 1999, when there were few outside observers in Kosovo who could independently document human rights violations.

To complicate any subsequent investigation, military and paramilitary forces routinely confiscated identity papers from the fleeing refugees. Sensitive government records were purged. Houses were burned and belongings scattered. Mass graves were obliterated. Bodies disappeared.

Beginning with a knapsack full of registries he rescued from the rubble of an Albanian border crossing, Ball gathered as much information as possible about the number of people leaving Kosovo, where they came from and when they reached the border. The one knapsack contained files on 272,000 people.

From that set of records and other reports, Ball could map the timing and source of the surges of refugees. Next, he sought to estimate how many civilians had been killed, and where and when they may have died.

Working independently of the U.S. government, Ball and his colleagues drew information from 15,000 interviews and exhumation reports conducted by four different humanitarian groups. Thousands of duplicate or misspelled names had to be culled.

Ball’s team reliably identified 4,400 people who had been killed. They then used a standard population sampling technique to estimate the total dead--10,356 Kosovo Albanians, with an error margin of several hundred people more or less.

By comparing the refugee movements against the death records, Ball discovered the numbers rose and fell in the same pattern in the same parts of the country, suggesting that they shared a common cause.

Then Ball compared records of the flight of the refugees to daily military action reports and tallies of the dead. To ensure fairness, he used the records from Milosevic’s government of NATO airstrikes and KLA ground actions.

He lined all the patterns of behavior up against one another and created graphs of the results.

The peaks of refugee flight consistently occurred during and after intense activity by Serbian forces, he found. Actions by the North Atlantic Treaty Organization and the KLA generally happened after the surges in refugees and killings, not before them.

“I find the data are consistent with the explanation that Yugoslavian forces conducted a systematic campaign of ‘ethnic cleansing.’ The data reject the hypothesis that KLA or NATO activity was responsible,” Ball said.

Several experts called the statistical analysis “particularly innovative.”

UC Berkeley demographer Ronald Lee, who reviewed the work, said: “There are very subtle and difficult statistical questions about how to draw these conclusions. I thought what they did was very impressive, very valid . . . persuasive.”

A Scientific Application ‘Born of Desperation’

Ball’s work is an application of science “born of desperation,” said independent human rights scholar Louise Spirer in Stamford, Conn. “The reason for the push is that we have to come up with a means to show what happened when there aren’t any records.”

In World War II, Nazi officials who orchestrated the death camps meticulously documented their work. Allied armies captured almost 14 tons of such files, which were then used as evidence at war crimes trials involving 90,000 cases.

Attempts at concealment were half-hearted at best, said Richard Pierre Claude, an expert on human rights and government at Princeton University, who is author of the forthcoming book “Science in the Service of Human Rights.”

Today, by contrast, those who use terror and mass murder as tools of statecraft take pains to cover their tracks “to preserve plausible deniability,” said Yale University international law expert Harold Hongju Koh, who was U.S. assistant secretary of State for democracy, human rights and labor from 1998 to 2001.

So often too, both sides in a conflict can muster victims of atrocities. Statistical analysis holds out the hope of a reliable way to determine relative guilt.

“What we can do with statistical analysis that you can’t do with anecdotes is make overall scientifically valid estimates,” said human rights activist Herbert Spirer, an international law and information management expert at Columbia University and husband of Louise Spirer. “We don’t have to sit and look at a single mass grave and try to decide how many people died in an entire country.”

During truth commission proceedings in Guatemala, which were convened to investigate abuses from more than 35 years of civil war, human rights statisticians analyzed 7,500 cases compiled from 11,000 depositions documenting 24,910 killings.

In rural areas, the number crunchers proved, native people were killed by government death squads at rates five to eight times greater than rates among other ethnic groups, offering a statistical hint of genocide.

In South Africa, researchers digested interviews with more than 21,000 witnesses covering 49,000 incidents during the latter decades of apartheid. By comparing rates of death among groups of people in different parts of the country, they developed statistics demonstrating that police were responsible for the overwhelming majority of the killings and that most of the victims were young black men.

By matching a database of 9,000 witness accounts of beatings and killings in El Salvador against comprehensive career records of military and police officials, statisticians showed how units became more violent when certain officers were placed in charge. The analysis helped get those officers banned from government service.

But in Kosovo, the high technology of human rights may face its severest challenge yet.

Rarely has a government worked so effectively to mask its operations against civilians, several experts said. Never has so much high technology been marshaled in the effort to uncover evidence of sustained human rights violations.

Said Koh, who helped lay the groundwork for the technical evidence being presented against Milosevic: “Kosovo is what I consider the state of the art.

“You want an undeniable scientific account of what happened,” Koh said. “Getting out the historical record is as important as holding someone accountable. These advances help make it impossible to erase history.”

A ‘Hacktivist’ Who Thinks in Code

Ball, 36, is a “hacktivist,” employing his programming skills in the service of human rights. A sociologist by training, he has been writing computer software since high school.

“I think in code,” he said.

He worked his way through graduate school at the University of Michigan by writing computer databases. His passion for dBase, Fox and Paradox code was more than matched by a sense of political outrage. He wrote his dissertation on human rights movements in Ethiopia, Pakistan and El Salvador.

When Ball found himself in El Salvador as that country’s civil war was ending in 1991, he heard that a local human rights group “needed someone who could hack a database.”

He volunteered.

In the years since, every truth commission or major human rights investigation in the world has in some way drawn on his computer skills.

“It is an odd line of work,” Ball said.

For a decade, he has been pioneering the use of computer databases and statistical analysis to document human rights abuses. With his colleagues at the American Assn. for the Advancement of Science’s science and human rights program, Ball recently published a hacktivist manual. It teaches activists around the world how to design computer entry screens, questionnaires and databases that can be used to track human rights abuses.

Helped initially by the American Statistical Assn., a 162-year-old professional group that advocates the use of statistics in science, a global network of documentation specialists has spread information management techniques to human rights activists in 150 countries.

With solar-powered laptops, activists log testimony of abuses from survivors hiding far from any power outlet. They use the precision navigation capacities of the Global Positioning System to pinpoint the coordinates of mass graves they find. They combine databases and integrated mapping software to chart the geography of terror.

To protect the integrity of the information they are collecting, they code it with encryption software and secure it behind Internet firewalls. They create virtual havens for witnesses on Web sites.

To further evade surveillance, they communicate with each other through anonymous e-mailing techniques that disguise the location and identity of the sender.

In the computerized pursuit of justice, analysts like Ball employ the same census and data-mining techniques used by marketing experts to analyze purchasing habits, naturalists to estimate wildlife populations and medical experts to document epidemics.

Instead of consumer profiles, herd counts or public health warnings, however, the product of Ball’s work is compelling circumstantial evidence of official brutality.

Just as pathologists can use forensic DNA techniques to restore the identities of those exhumed from anonymous mass graves, human rights statisticians can reveal much of what is hidden to the individual eye.

What they are creating, said Spirer at Columbia, is “the epidemiology of horror.”

Traditional Activists Uneasy With Statistics

Other more traditional human rights activists are made queasy by so many formulas. They recognize the power of statistics as an analytical tool, but they are uncomfortable with it because abstract mathematical modeling can rob the victims of their humanity.

“You turn this into something you quantify--a smear of refugees--and you have to be careful about dehumanizing it,” said Eric Stover, director of the Human Rights Center at UC Berkeley. “The real testimony is the body with the single gunshot to the head and the hands tied behind the back.”

For Ball, however, the anonymity of numbers is a saving grace.

One afternoon during the data entry phase, Ball found one of his co-workers Net surfing. She was building a bookmark file in the database of Internet links to online photographs of the dead. He was dismayed.

“Stop. Don’t look,” Ball told her.

“She was useless for the next two days,” he recalled. “All she could do was cry.

“This is data. Numbers. This is a technical problem. Otherwise, you’re done. Burned out. Gone.