Duygu Sili

with Özgür Yılmaz and Ömer Faruk Şahin

We develop a dynamic model of multi-hospital kidney exchange incorporating inter-hospital coordination costs, stochastic patient–donor arrivals, and compatibility constraints. Formulated as a Markov decision process, a centralized clearinghouse decides on each arrival whether to match within a hospital, across hospitals, or defer to thicken the pool. To handle the four-dimensional state space we analyze two benchmarks — Complementary Pools (reciprocal types matched only across hospitals) and Substitutable Pools (same types competing for over-demanded pairs) — proving the optimal value function is componentwise concave and superconcave, supermodular in the complementary case and submodular in the substitutable case, with optimal policies characterized as monotone threshold rules. Inter-hospital exchanges concentrate in complementary states and shut down as coordination costs rise: hospitals operate as a single pool when costs vanish and revert to local matching when costs are high. The results point to transport reimbursement and incentive design as central policy levers.

★ This is my primary job market paper.

with Ömer Faruk Şahin, M. Utku Ünver and Özgür Yılmaz

We develop a new methodology that characterizes and computes dynamically optimal mechanisms for bilateral matching across arbitrary state spaces, provided that compatibility between agent types follows a linear spatial structure. The approach leverages second-order properties of the value function, extending recent advances in Markov Decision Processes and queueing systems to characterize optimal dynamic kidney exchange mechanisms and other spatial matching problems.

with Ali Nalbant and Barış Yıldız

An independent market operator allocates last-mile parcel deliveries among competing providers to maximize public value. We establish existence of Nash equilibria and propose a Mixed Integer Programming formulation that scales to 2,500+ customers. Welfare gaps across equilibria exceed 20%, and centralized assignment typically lies on the environmental–quality efficient frontier.

with Antonio Miralles, Pietro Salmaso and Claudia Meo

Italian public schools have legal autonomy to allocate students across classrooms, but competitive pressure from demographic decline renders this autonomy largely illusory. In under-demanded catchment areas, schools sort students by economic and sociocultural status (ESCS) as a dominant strategy resembling a prisoner's dilemma. Using INVALSI data and municipality-averaged dissimilarity and distance indices, fixed-effects regressions show that a one-percent greater five-year decline in the local school-entrant population is associated with increases of 0.041 (dissimilarity) and 0.058 (distance) in within-school segregation.

with Claudia Meo, Antonio Miralles and Pietro Salmaso

We study coalition formation in a purely hedonic setting where the number of coalitions and their sizes are fixed ex ante. We establish existence of swap-stable partitions when the utility agent i derives from sharing a coalition with agent j is the sum of a symmetric "friendship" term and an agent-specific "quality effect" — beneficial or disruptive — that j exerts on the coalition. The result generalizes Bilò et al. (2022) to non-symmetric preferences.

with Antonio Romero Medina

A graph-based framework for fair classroom assignment and coalition formation, showing that complex group reallocations can be represented through simple cycles, yielding efficient algorithms with guaranteed convergence. Under symmetric friendship preferences, fair and welfare-maximizing assignments are stable against group deviations and implementable through a strategy-proof mechanism; the paper also identifies when fairness and stability become incompatible under asymmetric preferences.