Research
How can we evaluate a policy when the policy changes the system through which its effects spread?
My work sits at the intersection of causal inference, network analysis, and comparative politics. I am interested in settings where an intervention does not simply act on a fixed structure, but reshapes it — and in what credible evaluation looks like once that is true.
Most methods for estimating treatment effects assume that the relationships through which effects propagate remain stable. My dissertation takes the opposite case as the object of study: an intervention changes behavior and changes the network of relationships through which behavior spreads.
The aim is to make credible evaluation possible when interventions, relationships, and behavior all move together over time — estimating effects along a trajectory rather than at a single fixed point.
Three essays:
The question. When a country democratizes, do the effects stop at its borders — or do they reshape party politics in neighboring states?
The question. Does democratization itself make party systems more polarized, or does polarization simply continue a trend already underway?
The approach. A fuzzy regression discontinuity design isolates the causal effect of transition from the pre-existing trajectory — a quasi-experimental strategy for a question that cannot be randomized.
Why secessionist factions come to dominate some nationalist movements while autonomist elites prevail in others — studied through three decades of recorded interaction among minority nationalist elites. With Başak Taraktaş.
Developing a measure for the effect of ethnic and religious affinity on the likelihood of international disputes. With Abdullah Kabaoğlu.
An original relational dataset covering 2,019 documented interactions among 606 political elites across three late-Ottoman nationalist movements — Albanian, Kurdish, and Macedonian/Bulgarian — spanning thirty years.
No comparable dataset existed, so the records were assembled and coded by hand from primary and secondary archival sources in five languages: Turkish, Ottoman Turkish, English, Albanian, Bulgarian, and French.
