Measuring violence and emergent hostility in ongoing civil wars using reimbursed mobile phone surveys
(joined work with Sebastian Schutte and Andrew Linke)
The project aims at measuring and monitoring inter-group hostilities in ongoing civil conflicts to investigate how individual attitudes towards out-groups covary with experienced violence. The research is based on a social-psychological framework and the recent turn in the civil conflict literature toward grievance-based explanations for mobilization. To test observable implications of the theory, we employ a new computer system for conducting reimbursed electronic surveys in low-intensity conflicts in both Kenya and India.
Visualizing and interpreting time-varying effects from duration models
The most commonly used duration models assume that covariate effects remain constant over time. This assumption is often violated in political science analyses with long observation times. While modeling such time-varying effects is easy to implement, the interpretation is not intuitive and prone to severe inferential errors. I study the merits and limitation of existing interpretation and visualization techniques and develop methods and software tools that help applied researchers to interpret estimation results correctly.
- Estimating survival functions after stcox with time-varying coefficient – published by the Stata Journal
- Quantifying change over time: Interpreting time-varying effects in duration analyses – published by Political Analysis
- Bootstrap pointwise confidence intervals for covariate-adjusted survivor functions in the Cox model – revise and resubmit
Disaggregating the relationship between mediation and conflict intensity
My PhD dissertation builds of the observation that conflict management attempts tend to be initiated when conflicts escalate. The dissertation looks into this endogenous process of violence and conflict management and analyzes the relationship between the short-term dynamics of intrastate conflicts and third party mediation attempts. The thesis consists of three analyses and ties together the literature on conflict management and disaggregated conflict research. On a larger scale, the thesis tests the central role of information asymmetries postulated by the bargaining theory of war. Using a combination of empirical work and computational modelling I am able to show that, first, short-term conflict events strongly predict the decisions to initiate and accept mediation onset. Second, these association undermine our ability to estimate mediation effects in existing country-year or conflict-level data. Third, once mediation is initiated and addresses the main conflict incompatibility, mediation is associated with a strong reduction in conflict intensity. Overall, the empirically observed pattern of conflict reduction is substantively similar to the theoretical prediction of the computational model.
- Anticipating mediated talks: Predicting the timing of mediation with disaggregated conflict dynamics – published by the Journal of Peace Research
- Impeding fatal violence through diplomacy: Evaluating the temporal effect of mediation in intrastate conflict
- Simulating the unknown: Illuminating unobserved selection biases in temporally aggregated studies of mediation
Temporal dynamics of one-sided violence (2009-2012)
During my BA and MA studies, I worked as a research assistant with the Konstanz One-Sided Violence Event Data (KOSVED) Project and assisted with coding and data management. The project motivated both my bachelor’s and master’s thesis. The BA thesis examined the effect of various external interventions on the level of violence against civilians using an interrupted time series design. A related paper was eventually published (joint work with Gerald Schneider and Margit Bussmann). My master’s thesis examined our ability to predict one-sided violence. The resulting paper demonstrates that both the number of acts and the intensity correlate with other conflict events and enable fairly accurate predictions.
- The Dynamics of Mass Killings: Testing Time-Series Models of One-Sided Violence in Bosnia (with Gerald Schneider and Margit Bussmann) – published by International Interactions
- Predicting atrocities. Statistically modeling violence against civilians during civil war – NEPS Working Paper