A Dynamic Separable Network Model with Actor Heterogeneity: An Application to Global Weapons Transfers

TL;DR

This paper extends the separable temporal exponential random graph model to include actor heterogeneity and applies it to analyze international weapons transfer networks, revealing key drivers and country-specific trends over time.

Contribution

It introduces a dynamic model with actor-specific effects for network analysis and applies functional principal component analysis to identify notable country-level changes.

Findings

Security and network effects drive transfer formation.

Market size and military spending influence repeated transfers.

Identified countries with significant changes in weapon import/export tendencies.

Abstract

In this paper we propose to extend the separable temporal exponential random graph model (STERGM) to account for time-varying network- and actor-specific effects. Our application case is the network of international major conventional weapons transfers, based on data from the Stockholm International Peace Research Institute (SIPRI). The application is particularly suitable since it allows to distinguish the potentially differing driving forces for creating new trade relationships and for the endurance of existing ones. In accordance with political economy models we expect security- and network-related covariates to be most important for the formation of transfers, whereas repeated transfers should prevalently be determined by the receivers' market size and military spending. Our proposed modelling approach corroborates the hypothesis and quantifies the corresponding effects. Additionally, we subject the time-varying heterogeneity effects to a functional principal component analysis. This serves as exploratory tool and allows to identify countries that stand out by exceptional increases or decreases of their tendency to import and export weapons.