Speaking Stats to Justice: Expert Testimony in a Guatemalan Human Rights Trial Based on Statistical Sampling

In October 2010, I testified as an expert witness in a legal case that set a historic precedent for human rights in Guatemala. The Guatemalan Attorney General’s office summoned me to present evidence in the trial of two former Guatemalan National Police agents accused of forcibly disappearing 26-year-old student and union leader Edgar Fernando García. A married man and the father to one young daughter, García disappeared in 1984 after being detained by police on his way to work one morning.

My testimony to the court was based on my analysis of random samples drawn from the millions of documents in the Guatemalan National Police Archive. The archive was found by chance in an explosives storehouse in 2005 and held what archivists estimate to be 8 kilometers, or approximately 80 million sheets, of paper. Many of the police documents were created during the country’s internal armed conflict from 1960 to 1996, during which tens of thousands of Guatemalans disappeared. After four years of intensely analyzing the archive documents, I stood before three judges to defend my statistical findings, which supported the prosecutor’s case against the police officers. Statistical data are seldom used as evidence in court cases in Guatemala, and defense attorneys were attempting to discredit my testimony.

Some content is only viewable by Chance Subscribers

Investigators find approximately 80 million sheets of paper inside the Guatemalan National Police Archive.

Investigators find approximately 80 million sheets of paper inside the Guatemalan National Police Archive.

The judges’ assistant handed me the judges’ version of the report I had submitted to the Guatemalan attorney general a few months earlier. “Is this the report you wrote?” he asked. I looked at him slightly perplexed, wondering if he really expected me to just accept the version he gave me as unaltered. He suggested I have a look, so I quickly flipped through pages of words, tables, and graphs and recognized my work. “It looks like it,” I responded. One of the judges spoke to me: “Raise your right hand.” He asked me if I “promised” to tell the truth before the court of law and the people of Guatemala. “Keep in mind that failing to tell the truth could result in a prison sentence or the death penalty.” Did I hear that correctly? Did he really say death penalty? With a knot in my stomach, I answered “yes.”

To my left, sat the two former police officers, Héctor Roderico Ramírez Ríos and Abraham Lancerio Gómez, who were accused of forcibly disappearing García. Behind me, the courtroom was packed with members of Guatemalan civil society organizations and diplomats, including the U.S. ambassador and members of the press. Also present in the courtroom was García’s family, which had never given up its search for him. Many of the other spectators in the courtroom lived in Guatemala City during those years of intense political violence in the late seventies and eighties.

One week after I presented my statistical evidence, the judges found the two former police officers guilty of forced disappearance and sentenced each to 40 years in prison. The judges also ruled that government prosecutors should begin investigating higher-ranking officers for their possible role in the disappearance.

Beyond the García case, the verdict set the legal precedent that established forced disappearance as a crime in the Guatemalan judicial system. The swift ruling was a great triumph for justice in Guatemala and will help pave the way for judges to trust the validity of the archive documents—and accept statistical evidence in human rights cases.

As a statistician with Benetech’s Human Rights Data Analysis Group (HRDAG), my team and I work with partners around the world to analyze large collections of human rights data. Since the statistical evidence presented in the García case set a precedent for this type of analysis in court, I would like to explain how we worked with the archive staff and prosecutors to present the data.

Archive workers receive police records from throughout Guatemala.

Archive workers receive police records from throughout Guatemala.

Prelude to the Case

The prosecutors in the trial of Ríos and Gómez, who were led by the Guatemalan attorney general’s office, assembled evidence to demonstrate that the former officers were criminally responsible for García’s disappearance. They built their case by carefully piecing together information from experts, witnesses, family members, and the rescued documents from the Historical Archive of the Guatemalan National Police, which was disbanded after the 1996 peace accords between the Guatemalan government and the country’s guerilla movement.

Despite the role of former Guatemalan police forces in human rights violations, no one was aware of the Archive of the National Police until it was discovered by chance in 2005. The Guatemalan Human Rights Ombudsman’s Office (the PDH, in Spanish), concerned with munitions stored in warehouses of the current-day police, asked to inspect the buildings. Upon inspection, the PDH team was suprised to find a stockpile of concealed information in millions of moldy, rotting pages of paper infested with bats, bugs, and rats. The National Police had created many of these documents to guide their urban social control and counter-insurgency campaign. The documents were placed under the jurisdiction of the PDH, which knew they could be important evidence in the search for justice.

Overwhelmed with the magnitude of documents in the archive and hoping to establish a rigorous research design of its contents, the leadership of the archive restoration project called on HRDAG. Patrick Ball, the director of the group, was invited to the archive to provide expert advice on the analysis of the records. Based on his previous experience with Guatemalan non-governmental organizations (NGOs) and Historical Clarification Commission in Guatemala, Ball knew that data about the role of police was missing. Perhaps even more important, any findings about police abuses based on such a massive and mixed population of documents could face political opposition.

In response to these two challenges, Patrick recommended random sampling of the documents. Probability sampling would allow researchers to quickly gain a general understanding of the contents of the archive. Sampling would give analysts of the archive the ability to make inferences about the contents of the whole population of documents. Furthermore, it would give the archive leadership a defensible methodology for studying a subset of the documents. Claims about the archive’s content from a random sample could better withstand political attacks of bias than other research strategies.

Shortly thereafter, I traveled to Guatemala with my colleagues Romesh Silva and Tamy Guberek to design the sampling plan and coding scheme to study the archive contents. From the first day, we worked closely with the Guatemalan sampling team, led by Jorge Villagran. Our goal was to design a plan that would respond to the multiple research objectives of the archive team and not disrupt the careful efforts by archivists to restore the fragile documents.

Determining an appropriate sample design for the archive was a significant challenge. We were faced with the unusual situation of having to draw a sample from 8 kilometers of paper (what archivists estimated to equal approximately 80 million pages) stored across five warehouses with no apparent order or inventory that would allow a straightforward sampling design. From the beginning (and for the four years since), my colleagues and I have been fortunate to have the generous support and advice of Paul Zador and Gary Shapiro from the American Statistical Association’s Volunteer Group.

Archivists use tape measures to draw samples from the fragile paper records in the archive.

Archivists use tape measures to draw samples from the fragile paper records in the archive.

The documents contained in the archive were heterogeneous along any of the dimensions one could imagine using for a reference frame. This includes the size and type of repository that held the documents (what we call “containers”), type of documents, size of documents, dates of creation, physical state of documents, etc. Some bundles of paper in the archive had paper labels suggesting particular content, such as the police unit of provenance (i.e., “Headquarters”) or years relating to some aspect of the documents. But after launching a pilot study, we quickly learned that those markers were inconsistent and could sometimes be deceiving.

In some places, the paper records were scattered on the floor. In other places, documents were tied in bundles, or pages were stuffed in bags. An additional variable in the sample design was that the project timeframe and resources were unknown. We also had to consider that the ongoing cleaning and preservation efforts would require moving the documents from one place to another throughout the premises. Under these circumstances, we were obliged to design a sample plan that was flexible.

Probability Sampling

Any method of sampling that uses random selection such that each element in the population has a probability greater than zero to be selected and this probability can be accurately calculated. This kind of sampling makes it possible to calculate unbiased estimates for the entire population by weighing each observation according to its selection probability.

Multi-Stage Sampling

A complex sample design in which two or more levels of sample units are nested. At first, a sample of primary units is selected, then, in each of those selected units, a sample of secondary units is selected, and so on. This type of sampling is often used to reduce cost or when no direct sample frame of the elements in the population exists.

Our first version of the design included three main components to overcome the challenges we faced. The samples were to be small and iterative. Each sample would serve as a kind of snapshot of the archive contents collected in a short period of time. The sample selection would be multi-staged, selecting from a full enumeration of the archives’ containers (a table, a bookshelf, a drawer set, a floor space, etc.) of paper. The team examining the selected containers would then get a sample of its content by selecting three-dimensional coordinates inside the container, and then random points within bundles or drawers.

Beyond the sampling challenges, the physical environment inside the warehouses was trying. The air was permeated with mold and mildew. Despite these conditions, the coders worked with unwavering commitment, masks covering their faces and clipboards and tape measures in hand. Throughout the years of working together, I have developed a great respect for the Guatemalan team. They have maintained the clarity of vision that random selection will allow them to defend analytic conclusions based on the content of the archive. While the rigor and patience required by scientific methods are not always immediately apparent to human rights groups, this team understood the potential of the methods from the start.

The coding team took seriously the implementation of the sampling. Each time an unexpected challenge arose that they suspected might affect the calculation of the sampling probabilities, they stopped and consulted with me and my colleagues. One of the most dramatic events took place in February 2007. Without warning, a large sinkhole suddenly swallowed a full street block less than a mile from the archive. Fearing the documents could be lost, the archive staff began packing all the records into cardboard boxes that could be moved quickly in an emergency. This prompted us to reconsider the assumption that the containers would remain relatively still for 3–4 weeks (one wave) of each sampling period. After wave nine, we eliminated the container stage, which had the added benefit of making the sampling more efficient.

A sinkhole opens in Guatemala City, less than a mile from the archive.

A sinkhole opens in Guatemala City, less than a mile from the archive.

Our ASA advisors and I later spent months poring over the data and building the sample weights. Although the sample was carried out with great care and attention, there were nevertheless difficulties in calculating the weights. For example, errors in small measurements, such as centimeters, had a potentially large impact on the weights and variance. While I was deep in the technical details of making rigorous estimates based on sample data, I hardly imagined I would be defending my work in a court of law. I was more concerned about presenting and defending these findings before a different kind of jury—the peer reviewers and statistical colleagues at the Joint Statistical Meetings. In effect, together with my colleagues, I presented papers on the sample design, weights calculation, and preliminary estimates of the contents of the archive at JSM in August 2009.

Building the Case

My role in the García case involved presenting comparative analysis on two sets of documents. The first set was composed of the 667 documents submitted by the historians, archivists, and human rights defenders who are permanently conducting research at the National Police Archive. These documents dealt specifically with the role of the police in García’s disappearance. The second set of documents came from the random samples drawn over a year of the archive’s fragile content. (I will call these the “comparison set.”) The comparison set only included sampled documents that the police produced from 1960–1996, roughly the period of Guatemala’s internal armed conflict. In essence, I was asked to characterize in what sense the 667 evidence documents were typical or atypical of documents in the archive more generally in the comparison set.

Sample Weights

Adjustment factors applied to each element in the sample to adjust for differences in probability of selection in a sample, either due to design or happenstance. Many times, samples are selected with unequal probability to expand the number of cases available (and hence reduce sample variability) for certain areas or subgroups for which statistics are needed. Weights need to be applied when estimates are calculated to produce the proper representation.

Estimates and Confidence Intervals

To calculate the estimates we used the traditional approach (described in all survey sampling textbooks), with base weights being equal to the inverse of the probability of selection. All the proportions are sample-based estimates with 95% confidence intervals.

The 667 archive documents provided evidence that the National Police had a policy to hunt, “control, and clean” Guatemalans they considered “subversives.” For example, documents told of one such operation that took place the day García disappeared. Other documents show the director general and coordination units of the National Police were informed of that specific operation. Documents outline the police units in charge of the “control and clean” operations by neighborhood. These documents state that the 4th Corps of the National Police was in charge on the day of García’s disappearance. Still other documents mention a proposal to reward the four police officers who successfully completed an operation against two “subversives” (García and his colleague, Danilo Chinchilla) on the day, place, and time of García’s disappearance. Other documents of interest were those that labeled García an “enemy of the state” and records of relatives denouncing his disappearance and requesting “proof of person.” These requests to know García’s whereabouts were never answered by police. Specifically, my report responded to seven questions to which the Guatemalan General Attorney requested written answers. The questions were intended to strengthen the arguments and evidence the prosecutor would present in court.

To respond to these questions, my colleagues and I divided the answers into three types of information. First, we showed that the 667 evidence documents have similar structural characteristics, such author recipient and creation date, compared to the documents in the comparison set. In other words, the 667 evidence documents presented by the prosecutor have “typical” characteristic of documents that belongs to the archive. Second, we compared the proportion of all the documents in the comparison set that the police leadership had knowledge of to how informed they were of the 667 key documents concerning García. Third, we presented general estimates of the magnitude and physical condition of all the archive’s documents using the comparison set.

My expert testimony report showed that the overwhelming majority of police documents circulated within urban settings, as opposed to rural places where the military was the dominant security force. I also estimated the percent of documents “known by” a police unit either as authors or direct or indirect recipients. I presented our estimate of the total number of documents in the archive, which documented disappearances, killings, and other abuses.

To me, the most fascinating finding is that in nearly all the structural aspects of the documents, the documents related to García are almost exactly like the broader population of documents in the comparison set. In Table 1, we can see that most pairs of percentages are close. Some categories of records, such as “completely readable documents” and “documents without authors or recipients,” are particularly similar.

Table 1—A Comparison of Basic Information in Both Sets of Documents (Legibility, Dates, Authors, and Recipients)

Table 1—A Comparison of Basic Information in Both Sets of Documents (Legibility, Dates, Authors, and Recipients)

* All estimated proportions and 95% confidence intervals are weight-based calculated.

Table 2—A Comparison of the Level of Knowledge of Both Sets of Documents by Specific Police Units

Table 2—A Comparison of the Level of Knowledge of Both Sets of Documents by Specific Police Units

* All estimated proportions and 95% confidence intervals are weight-based calculated.

The key variation between the broader patterns of documents in the comparison set and the specific patterns of the documents related to the García case concerns informed the police Directorate and Coordination Units and military-police Joint Operation Unit were about the content in the documents. This information helped support the arguments put forward by prosecutors. While these units were informed of approximately 30% of all documents in the comparison set, they had knowledge of 70% of all the documents in the García case. This finding may not seem surprising given that the documents concerning García were selected based on their content (a non-random sample), but the comparison strengthens the case that relatively high-ranking officers of the National Police were highly aware of this specific “control and clean” operation.

The Trial

When the Guatemalan attorney general’s office invited me to submit expert testimony, I wrote my report with much support from the sampling and coding team at the archive. A volunteer lawyer from the Guatemalan NGO Center for Forensic Analysis and Applied Sciences (CAFCA) coached me so I would be prepared to take the stand. The coaching sessions were grueling and focused on three tactics the defense might use to attempt to undermine my testimony. These included personal attacks on my experience, age, and professional degrees; discredit by association, citing earlier work by members of Benetech and the National Police Archive Project; and insufficient bureaucratic knowledge about the legal case. After the many rounds of preparation, I felt sufficiently confident that I could make my points clearly in court and avoid getting tangled in questions that were irrelevant or inappropriate to my testimony. Even with this kind of preparation, however, I did not anticipate the personal and emotional process of testifying in a Guatemalan court of law for such an atrocious crime.

On the first day of the trial—October 18, 2010—I arrived with my archive colleagues early to the tribunals building in Guatemala City. The 15-floor tower contains a series of courtrooms for criminal and civil proceedings. The courtroom for the García case was on the 14th floor. In the lobby, a disorderly and extremely long line of people was waiting to take the elevator. Impatient to get to the trial, my colleagues and I decided to take the stairs. Sweaty and more anxious after the climb, we arrived to find the courtroom crowded with spectators. I took a seat with my colleagues and listened to the judge take roll call of the participants in the trial. Then, the witness and experts were asked to leave the room while the judges summarized the contents of the case.

Waiting outside in the hall for my turn to testify was difficult. It was still unclear whether I would be asked to give my testimony on that day. There was a bit of bustle about whether the judges would allow those testifying to use a video projector. No one was overly emotional, but I could feel the tension and anxiousness in the air. García’s widow, Nineth de Montenegro, who today serves on the Guatemalan Congress and is the founder of a leading Guatemalan victims organization, GAM, was present. She was accompanied by Garcia’s mother, who has been seeking justice for her son’s disappearance for more than two decades. Also present was Iduvina Hernandez—the partner of Chinchilla, who was with García on the day of his disappearance and who was subsequently disappeared some time afterward. I mostly spoke with Velia Muralles, the archive’s expert witness.

As we waited, there was a constant flow of handcuffed prisoners and members of their families walking by, escorted by heavily armed policemen. The whole scene gave me a tight knot in my stomach. For a fleeting moment, it seemed easy for anyone to take one of the huge guns held by the police and claim their own retributive justice. I also had ambivalent feelings about the police officers on trial. Although we had sufficient evidence to believe the two men had illegally captured García, their role in subsequent disappearances was still unknown. While contemplating all this, I was overly conscious that the bathroom was located in the basement, and we were on the 14th floor with almost no possibility of using the elevator.

While the rest of us continued to wait outside, Kate Doyle—a staff member of the Washington, DC,-based nonprofit the National Security Archives—was called as the first expert witness. I was called in second. I traversed the spectators’ gallery and took the stand. The accused and their defense lawyers were to the right, the prosecution to the left, and a panel of three judges in front. To my relief, I faced the judges and had my back to the spectators. I think it would have been harder to see my colleagues and García’s family as I read a relatively cold-sounding, technical analysis of documents’ characteristics. While the benefit of my testimony was the objective, scientific nature of my findings, it contrasted harshly with the real-life experience of losing a husband, father, and son.

I brought with me a copy of my report, but the first thing the judges said was that I was not allowed to use any external materials. After putting my copy away, the judges’ assistant handed me their version of my submitted report. The actual testimony was much simpler than expected. I was asked to read the report’s introduction and conclusions. This felt slightly silly, since the report was written to be read as a whole, not in excerpts and not for a public listener. These two parts of the report mentioned points to reinforce the body of the report, rather than a complete summary. I wondered if I would have written the report differently knowing I would be reading it in court. After the reading, the prosecution asked me questions. Did my analysis offer information about the urban/rural divide in the police documents? Was there any evidence about the specific police units’ knowledge of the archive’s documents? I gave straightforward, affirmative answers based on the analysis in my report.
Then it was the defense attorney’s moment to question me. One of the questions concerned whether the documents proved the criminal responsibility of the two ex-police officers on trial. Immediately, the prosecution interjected and the judges overruled the question. Every other question they asked me was also inappropriate, raising objections from the prosecution and overruled by the judge. After being spared from answering any questions from the defense, I was asked to step down and told I could stay in the room if I wished.

In retrospect, my experience on the stand was not particularly rough. But, at the time, I was exhausted from the tension. I took comfort in staying with my colleagues in the courtroom for the remainder of that day and returning the following day to watch the next witnesses. We all had some hope that the verdict might actually be announced that Tuesday. Not surprisingly, however, the trial did not end so quickly. Wednesday, October 20, was a national holiday—a memorial day for the Guatemalan revolution in 1944, during which longtime dictator Jorge Ubico was ousted and a period known as the “10 years of spring” began. It was a time known for relative political freedoms and democracy in Guatemala. I thought to myself how appropriate it was to remember a time of plurality and openness in Guatemala’s history while we were seeking justice for those who violently took some of those freedoms away.

The Verdict and Its Long-Term Implications

On October 28, only 10 days after the start of the trial, the two former policemen, Héctor Roderico Ramírez Ríos and Abraham Lancerio Gómez, were found guilty of the crime of forced disappearance. The men were absolved of responsibility for the crime of illegal detention due to the expiration of the statute of limitations for this crime. Both defendants received the maximum sentence of 40 years in prison.

Daniel Guzmán testifies in the case of Edgar Fernando García.

Daniel Guzmán testifies in the case of Edgar Fernando García.

There are two other fundamental victories that resulted from this case. During the reading of the verdict, the judges announced that the public prosecutor’s office must investigate responsibility for García’s disappearance further up the National Police hierarchy. This mandate gives further hope to the García family and other victims that there may be more wholesale recognition and potential justice for their decades of suffering. In addition, since this is the third conviction for a case of forced disappearance in the Guatemalan legal system, this crime will now be recognized under Guatemalan law. Future prosecutors have a stronger legal basis to seek justice for this tragic crime in other cases.
In regard to my statistical testimony, judges said after the trial that they understood the information and considered it relevant to building the case. In particular, they highlighted its use in corroborating information about the disproportionate urban focus of the National Police and estimating the number of documents in the archive related to disappearances. The high proportion of documents known by the police leadership in the García case compared to the general proportions in the archive also prompted the judges to investigate the chain of command. Finally, the verdict in the García case provides enormous credibility to the use of historical archives as admissible evidence in Guatemalan courts. After six years of hard work preserving, classifying, and scanning millions of the National Police documents, this was incredibly rewarding for the more than 100 staff members dedicated to the archive’s rescue.

It is satisfying to use the tools of statistical analysis to help Guatemalans find closure for these crimes and to help families like that of Edgar Fernando García find some measure of justice. It is rewarding to know that, as we hoped from the start, the probability sample of documents gave credibility and support to the findings from the archive’s reserves. The new samples will shed further light on the chain of command and communication flow within the National Police. I am confident the benefits of archival evidence and statistical analysis will continue to reinforce each other in the campaign for truth and justice in Guatemala.

Back to Top

Further Reading

Guzmán, D., T. Guberek, G. Shapiro, and P. Zador. 2009. Studying millions of rescued documents: sample plan at the Guatemalan National Police archive. In JSM Proceedings, Section on Survey Research Methods. Alexandria, VA: American Statistical Association. www.hrdag.org/resources/publications/JSM-GT-sample-design.pdf

Price, M., T. Guberek, D. Guzmán, P. Zador, and G. Shapiro. 2009. A statistical analysis of the Guatemalan National Police archive: searching for documentation of human rights abuses. In JSM Proceedings, Section on Survey Research Methods. Alexandria, VA: American Statistical Association. www.hrdag.org/resources/publications/JSM-GT-estimates.pdf

Shapiro, G., D. Guzmán, P. Zador, and T. Guberek. 2009. Weighting for the Guatemalan National Police archive sample: unusual challenges and problems. In JSM Proceedings, Section on Survey Research Methods. Alexandria, VA: American Statistical Association. www.hrdag.org/resources/publications/JSM-GT-weights-paper.pdf

Tagged as: , , ,

1 Comment


  1. Guatemalan police archive goes online – SUBMITTERATOR