times, figuring out what data “actually speak for” is less than straightforward.
The literatures on collective mobilization and political violence suggest why this
might be the case: first, sources come with their own biases. They provide a
skewed picture of violent events, reflecting their own preferences (which seems obvious in the case of government documents). Second, aggregate accounts typically
mask actors, methods and nature of violence. The end result is that sources and
the information they present can effectively constrain what can
reasonably be argued with data at hand. A multi-source approach and a
disaggregation effort have been offered to overcome these issues. In
this page, I discuss briefly several examples from my own work to demonstrate
the usefulness of this approach to answer questions related to violence.
In this book, we consulted government documents, insurgent publications
and national newspapers to compile datasets on insurgency, counterinsurgency
and protest. A multi-source approach that included cross-checking events from alternative sources and
recognizing their distinct reporting biases led to more reliable and
extensive information. Furthermore, a disaggregation
effort that involved identifying targets, repertoires and intensity of violence
across space and time revealed an interesting feature of violence. When viewed
from a long-term perspective, violence is sticky and hard to export it
elsewhere in the same form and intensity. Please read the Appendix of the book for an extensive discussion and check out box 2 to get a sense of rebel methods in the
This article discusses the impact of non-state actor violence on
electoral success of an ethnic party. A multi-source approach was instrumental
for identifying perpetrators and victims since most killings remained legally
unsolved. The key was figuring out the specific language used in each source base for
reporting this particular set of events, as these killings were taking place in
a larger context of violence. An insurgent publication, reports by human rights
organizations, national newspapers and legal claims by relatives of victims
were carefully surveyed to understand who did what to whom. At the end, the
distribution of killings strongly showed that violence concentrated in spots where ethnic politics initially enjoyed strongest support
and violence eventually backfired at the ballot box. See Figure 1 for some of
the properties of these killings and victim demographics.
In this manuscript, the major benefit of a disaggregation effort was
identifying the mechanism through which emergency zone practices translated
into electoral support for a pro-insurgent, ethnic party. By distinguishing
military encounters from detentions and further unpacking detentions based on accusations leveled against detainees, the manuscript shows that civilian
detentions (situations where civilians were accused of having rebel sympathies
with no organizational ties to insurgency) brought fresh support to the ethnic
platform (Map 2). A similar political impact did not materialize with military
encounters or incidents involving other types of detentions.
This manuscript examines violence employed against civilians in early
phases of an insurgency by consulting the main rebel publication and several
national newspapers. It seeks to answer which villages were targeted by
insurgents in order to test major theoretical frameworks for understanding violent
events. The disaggregation effort that included data on roughly 5000 villages
demonstrated that opportunity and grievance are not necessarily competing
explanations but complementary mechanisms that foster violence in different
ways. Insurgents attacked villages which were difficult to reach by security
forces, close to rebel camps/border and also had a political history of siding
with political parties that voiced their grievances. The manuscript also
reveals another mechanism at play: signaling. Violence was used as an
instrument to sanction future behavior of in-group
The return of insurgent violence to Turkey during mid-2015 puzzled many.
While pundits were busy discussing why it happened, the common wisdom in the media
was that the insurgent group was stronger than ever simply because it was able
to strike everywhere. Disaggregation of insurgent attacks across time and space
revealed another story. The article showed that insurgent control remains
almost fixed in the long-history of war, concentrating in 30 districts that we
previously named Zone 1, the heartland of insurgent activity (Figure 1).
This article demonstrated the varied impact of violence on electoral
outcomes. A disaggregation effort that located violence into districts revealed interesting facts.
First, it showed how the intensity of violence affected the incumbent
(AKP) and pro-insurgent ethnic party (HDP) differently. Second, it demonstrated that the
ethnic party suffered minimal losses when insurgent violence targeted security forces in
urban areas. Viewed together, these results underscored the importance of identity of targets and urban/rural
divide in the translation of violence into electoral politics (Figures 1A-1B).
This article shows that the incumbent used a counterinsurgency
campaign to secure electoral outcomes. Specifically, the government imposed
curfews to suppress the turnout in ethnic party strongholds and turned a blind
eye to insurgent violence where it had high electoral expectations. Once again,
a two-way disaggregation effort that divided the region into districts based on
electoral support for the ethnic party and the number of voters in the prior election
were instrumental to find out whether a curfew was imposed in a particular
district. We found that the government was careful not to alienate voters in
vote-rich districts by imposing curfews whereas the opposite was true in ethnic
party strongholds (Map 1).