Dresden 2026 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 4: Poster Session
SOE 4.2: Poster
Montag, 9. März 2026, 18:00–21:00, P4
Validating diffusion-based dimensionality reduction of political regimes: conflict and child mortality benchmark — •Paula Pirker-Diaz1, Radost Waszkiewicz1, Matthew Wilson2, and Karoline Wiesner1 — 1University of Potsdam — 2University of South Carolina
Political regime indices such as the Electoral Democracy Index (EDI) are widely used for measurement and policy analysis, yet their additive aggregation rules can obscure nonlinear relationships among regime indicators. In earlier work, we showed that a subset of EDI components lies on a low-dimensional nonlinear manifold recovered with the Diffusion Map method [1]. Here we introduce a new one-dimensional regime index by fitting this manifold, identifying the latent dimension along which regime features co-vary most strongly while preserving their nonlinear structure. The resulting index parallels the EDI's characterization of democracies but produces a different ordering among autocracies. By varying the smoothness penalty, we identify the indicators whose contributions to the latent structure remain most stable, providing a robustness-based measure of their relevance. The new index improves the resolution of intra-regime conflict onset probability and child mortality relative to the EDI, with the largest gains among autocratic regimes that the manifold distinguishes more effectively. These findings reinforce that political regimes lie on a nonlinear low-dimensional manifold identified by diffusion methods, which capture social and political variation more effectively than linear aggregation rules. [1] Pirker-Diaz et al., RSOS (2025)
Keywords: Diffusion; Political science; Risk-of-conflict; Child mortality; Regime types