“Big Data”. Source: betakit.com
In June 2006, I was a doctoral student at the University of Colorado-Boulder in the final stages of writing my dissertation on why terrorism occurs in democratic countries. I was spending my last summer teaching undergraduate courses in Boulder before heading off to Cambridge, Massachusetts, where I was due to take up a predoctoral fellowship at Harvard University’s Belfer Center for Science and International Affairs in the fall.
Earlier that summer, I received an email from my colleague Victor Asal, forwarding an invitation he had received to a workshop called “People Power and Pedagogy.” “The other side of the coin – thought you might be interested,” Victor’s email said. Sponsored by the International Center on Nonviolent Conflict and hosted at Colorado College, this workshop was meant to introduce social scientists to the literature on nonviolent or “civil” resistance—a concept with which I was not familiar. In all of my time researching social movements, political power, and organized violence, I had never come across the work of ICNC’s co-founders, Peter Ackerman and Jack Duvall, nor had I heard of Gene Sharp or any other academic powerhouses whose work had established the field of civil resistance studies.
Skeptical but curious—and more than a little enticed by the free food and books—I applied to the workshop. Soon thereafter, I received an acceptance letter from ICNC followed by a FedEx package full of books and papers on civil resistance, its theoretical and strategic dimensions, and the ways people power movements had accomplished in many cases what violent rebellion could not. These works made the claim—sometimes implicit, other times explicit—that civil resistance was as effective or even more effective than armed struggle in achieving major political concessions. They based these claims on cases like Serbia, where the Otpor movement had initiated the downfall of Slobodan Milosevic; Poland, where the Solidarity movement had successfully challenged the entrenched Communist Party; the Philippines, where the People Power movement had removed Ferdinand Marcos from power; and the U.S. Civil Rights movement, where lunch counter sit-ins, boycotts, and marches had initiated the desegregation of many southern cities and created the base for a broader campaign.
I must say I attended the workshop as a cynic. Although I found the various cases of successful civil resistance interesting, I thought they were exceptional. For every case like Serbia, Poland, or the Philippines, I could recall a case like Tiananmen Square, Hungary 1956, or Burma 1988. After all, even Mahatma Gandhi’s dubious victory in expelling the British from India had ushered in a period of violent turmoil, punctuated by India’s bloody partition with Pakistan. Even more persistent was my suspicion that successful cases could be explained by other factors—weak states incapable of suppressing unarmed actors, international actors willing to patronize them, moderately democratic institutions that accommodated them, social, economic, or demographic characteristics that predisposed some populations to embrace nonviolent action where others would turn to violence, or plain old government incompetence.
I voiced these concerns repeatedly throughout the workshop and expressed my desire for more and better data on the phenomenon. Maria Stephan, then academic outreach coordinator for ICNC, was receptive to my concerns but was also eager to challenge me: “Well, how would we know if civil resistance was as effective as armed struggle? How would one develop a study that tested your hypotheses about structural factors explaining the outcomes?” Maria had just completed her own doctoral thesis on the topic of “extending the nonviolent battlefield,” or how nonviolent self-determination movements could solicit third-party support to increase their leverage.
In response, I developed what I thought would be a reasonable research design that might actually assess—systematically and empirically—both the relative rates of success of nonviolent and violent mass movements as well as the underlying causes for these successes. The study, I ventured, would have to be much broader in scope that the limited number of cases examined by much extant literature on nonviolent resistance. It would have to include all known cases—both successes and failures—rather than making inferences from a relatively small number of cases. It would have to have wide geographic and temporal scope so as to increase the number of observations in a way that would provide more statistical power to the study. With Maria’s recent fieldwork, familiarity with different cases, and vast knowledge of the canonical literature, we decided to put our heads and our skill sets together and conduct a study on whether it was indeed the case that nonviolent resistance was a relatively effective form of political struggle.
To ICNC’s credit, the organization’s leadership encouraged us to do this. I remember telling Maria that we might find out the opposite—that violence was actually more effective—and asking her whether Peter Ackerman understood that this was a real possibility. She assured me that if we discovered that civil resistance was ineffective, then he and everyone else at ICNC would accept this outcome. After all, Peter had a Ph.D. from the Fletcher School and fully understood and respected the scientific process. Looking back, I can see what a gamble this was for ICNC. But Maria and I agreed that if it turned out that civil resistance was not very effective, then we had better find that out sooner rather than later.
My own reservations about the effectiveness of nonviolent action directed me to take a particularly skeptical approach to the data collection and analysis, applying a “hard test” to nonviolent action wherever possible. This is reflected somewhat in the sample population itself. I decided we would only include cases where the insurgents were seeking the removal of the incumbent regime, territorial self-determination, or the expulsion of a foreign military occupation—I saw these as “maximalist” goals that fundamentally altered the shape of the state and would therefore be more difficult to achieve. Data coverage would be global and cover the time period 1900-2006. Whether it was nonviolent or violent, we would only code a case as “successful” if it achieved the full removal of the incumbent leader, de jure and de facto secession, or the expulsion of a foreign military. The campaign had to achieved this objective within a year of its peak, and it had to have had a discernable impact on this outcome. Moreover, we could collect data on a variety of features of the campaign’s structural environment—including features of the regime, socio-economic trends, ethno-linguistic characteristics of the population, international support for the campaign or the regime, region, and time period—and we would account for these factors when identifying the correlates of success.
To obtain the cases of violent insurgency, I started by looking at all of the cases of intrastate conflict from the Correlates of War dataset, which features cases of organized armed actors using violence against one another that results in at least 1,000 battle deaths. I then turned to collecting data on analogous nonviolent campaigns. Drawing on thousands of source materials—including encyclopedias, bibliographies, case studies, historical documents, news reports, and other scholars’ published lists of popular revolutions—I began to assemble a list of cases of nonviolent mass mobilization featuring at least 1,000 observed participants seeking maximalist goals from 1900-2006. My impending dissertation defense and fellowship obligations at Harvard introduced a variety of distractions at this point, so ICNC gave me a stipend to keep the data collection process on pace. By spring of 2007, I had a preliminary list of campaigns of nonviolent resistance, which I circulated to experts of nonviolent action—including some of ICNC’s staff and academic advisors, as well as subject experts like Doug Bond—to ask for any additions and to get their sense of whether we had properly characterized the outcomes of these campaigns (e.g. successes, partial successes, or failures). At the end of this often tedious process, we had produced the Nonviolent and Violent Campaigns and Outcomes (NAVCO) data set (version 1.0)—the first dataset of 323 maximalist campaigns over 106 years with global coverage.
Then I ran the numbers. I was shocked. More than 50% of the nonviolent campaigns succeeded whereas about 26% of the violent ones did. Moreover, when I ran a variety of regression models that included features of the regime as control variables, I could find no systematic statistical association between structural features of the country and the outcomes of the campaigns. Generally speaking, nonviolent campaigns were succeeding more often than violent campaigns despite a variety of structural factors that we typically associate with predetermining such outcomes. Other studies have since reached similar conclusions.
Maria and I wrote up these preliminary results and submitted them to International Security, which published the paper in its Summer 2008 issue. At the same time, we suggested to ICNC that we ought to write a book together on this topic. Given the striking nature of our findings and their potential impact, they agreed to provide me with a stipend to work on the book manuscript the following year while a postdoctoral fellow at Harvard. In 2011, Columbia University Press published our book Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict. It hit the shelves during the height of the so-called Arab Spring, a series of upheavals that leant greater urgency to the need for even more systematic, empirical research on the topic.
ICNC has been ahead of the curve of this effort. In 2009, they gave me a five-year contract to further develop the NAVCO Data Project to provide greater detail on the characteristics and tactics of different campaigns. Thanks to this funding, in May 2013, Orion Lewis and I published a release of the NAVCO 2.0 data set, which contains annual data on a variety of organizational characteristics of 250 nonviolent and violent campaigns from 1946-2006. Additionally, ICNC’s support will allow me to release NAVCO 3.0, which contains data on the tactical sequences of nonviolent and violent campaigns in 30 countries from 1987-2012, in late 2015. At this point, the NAVCO data set is the signature resource on the comparative characteristics and outcomes of nonviolent and violent campaigns, and its release has catalyzed a number of new studies on civil resistance.
ICNC has also set aside funds to support anyone interested in adding to empirical understandings of civil resistance. They have developed a generous $10,000 stipend for PhD students writing promising dissertations on the topic. Their monograph award is meant to catalyze new research on this topic by established scholars, while their curriculum development award is meant to promote innovation among scholars wishing to apply their pedagogical insights to the field.
So the story of the NAVCO Data Project is one of skepticism, hypothesis-testing, and discovery—basic principles of the scientific method. I am grateful to ICNC for providing me with the intellectual provocation to begin this research as well as the support to carry it out. In this case, as it turned out, the intuitions of many advocates of nonviolent resistance were correct, and the assumptions of many cynics—including myself—were dead wrong.
—-This post has been updated to reflect the role of lunch counter sit-ins in the broader U.S. civil rights movement, rather than in desegregating Nashville alone.