The Limits of TERGM for Analyzing Regional Alliance Networks: A Case Study of Post-Soviet Alliances

The article examines methodological aspects of applying dynamic network analysis in contemporary international political science. The study is motivated by the limitations of standard statistical methods, particularly logistic regression, which ignores network autocorrelation and endogenous interdependence of observations, leading to biased estimates in the analysis of international processes. The paper aims to demonstrate the analytical potential and limits of applicability of Temporal Exponential Random Graph Models (TERGM) for studying the evolution of regional security subsystems. Using the example of modeling the dynamics of military alliance networks in the post-Soviet space, a methodology for constructing and specifying TERGM models is presented. The procedure for incorporating both endogenous structural configurations (inertia, triadic closure) and exogenous covariates (trade flows, sanctions status, influence of external powers) into the analysis is described. The analysis was conducted in the R programming environment using the «btergm» package. The practical challenges of interpreting stochastic modeling results in small-sample conditions are demonstrated. In particular, it is shown that the inclusion of deterministic covariates can lead to statistical distortions requiring specific interpretation. The limitations of the method in assessing the influence of node attributes, specifically sanctions status, are identified in cases where these attributes are present in only a small number of network participants. It is concluded that TERGM is an effective tool for measuring structural effects; however, when analyzing small regional network structures, its application requires rigorous preliminary data diagnostics and combination with alternative verification methods.