World Alignment Map

UN General Assembly Voting — Ideal Point Clusters (2000–2025)

Bailey, Strezhnev & Voeten (2017) · Harvard Dataverse · Louvain clustering on ideal point distance

Methodology

Data Source

Country ideal points derived from United Nations General Assembly roll-call votes (2000–2025), using the dynamic ordinal spatial model from Bailey, Strezhnev & Voeten (2017). This dataset covers all UN member states from 1946 onward.

Ideal Points vs Vote Agreement

The simplest approach — measuring what fraction of votes two countries cast the same way (the "S-score") — produces muddy clusters because roughly 72% of UNGA resolutions pass by consensus. Most countries vote Yes on everything, inflating agreement scores even between adversarial states. Ideal point estimation (Item Response Theory) solves this by modelling each country's latent ideological position from the pattern of votes across all resolutions, producing sharper, more meaningful clusters.

Clustering — Louvain Community Detection

The pairwise distance matrix between country ideal points is fed into Louvain community detection (Blondel et al., 2008), a greedy modularity-optimization algorithm. It identifies groups of countries with similar voting profiles by maximizing the density of edges inside each group relative to a random baseline. The algorithm determines cluster count naturally from the data structure.

2-Group vs 3-Group Modes

The Louvain algorithm naturally partitions the data into 3 clusters, but the ideal point distribution also has a clean binary split at a natural gap (IdealPointScore ≈ 0.30). The 2-group mode shows this split: Western-aligned states vs the rest. The 3-group mode further isolates a "center" group of states (including Russia, Argentina, Chile) that sit between the two poles.

What the Score Means

The IdealPointScore is a 1-dimensional latent estimate of a country's voting position. Positive values indicate voting patterns aligned with the US/Western European bloc; negative values indicate alignment with the Russia/China axis. The magnitude reflects extremity — the further from zero, the more consistently a country votes at that pole.

Known Caveats & Artifacts

  • Serbia (YUG): The Voeten dataset encodes Serbia's historical votes under ISO3 code YUG (Yugoslavia). Serbia inherited Yugoslavia's UN seat and voting record; the YUG code is a legacy artifact of the dataset's panel structure. It maps correctly to Serbia (UN numeric code 688).
  • Crimea: The Natural Earth TopoJSON base map labels Crimea as part of Russia. This reflects the base map's administrative boundary data, and is not a political statement. The map system draws boundaries from the shapefile, not from the voting data.
  • Non-UN members: Kosovo, Palestine (observer state), and other non-member entities are excluded from the dataset and appear gray on the map.

References