Berlin 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 20: Focus Session: Statistical Physics of Political Systems
SOE 20.4: Talk
Thursday, March 21, 2024, 16:10–16:30, MA 001
Modelling dynamics of political regime types in the 20th century — •Paula Pirker-Diaz1, Sönke Beier1, Matthew Wilson2, and Karoline Wiesner1 — 1Universität Potsdam, Potsdam, Deutschland — 2University of South Carolina, Columbia, USA
Can the evolution of political systems be predictable? Is there a way to define their state with a maximum of two state variables?
To answer these questions we are analysing part of the V-Dem Research Project dataset, the world*s most detailed democracy ratings. It consists of hundreds of indicators quantifying different aspects that define the democracy level of more than 170 countries, which reveals the complexity of political systems and their definition. (Wilson, Matthew et al., The Hidden Dimension in Democracy (2023). V-Dem Working Paper 2023:137)
We apply the dimensionality reduction method called diffusion map and identify a two dimensional manifold containing all datapoints corresponding to all countries and years available, between 1900 and 2021. Being the first coordinate strongly aligned to the Electoral Democracy Index, this novel representation distinguishes autocracies from democracies in a surprisingly clear way. We also identify regions in the manifold that are related to specific regime types and additionally, suffrage types. By fitting a Gaussian Mixture model to the temporal evolution, we detect the collective response of the system to historical events. Based on our results, we suggest a predictive model for the evolution of political systems.
Keywords: political systems; nonlinear dimensionality reduction; diffusion maps; gaussian mixture models; dynamics