The same questions, asked of every relationship on earth
Bilateral analysis has traditionally rested on impression and proximity. Our method replaces it with a fixed procedure applied identically to all 18,528 country pairs, so that a comparison between two neighbors and a comparison between two countries with no history together are produced to the same standard.
Tens of thousands of indicators, per country
The World Bank catalog, which aggregates country data across the UN system, serves as the primary source at roughly 1,400 active indicators. We extend it with specialized databases until each country is described by tens of thousands of unique measures. Every indicator is selected on two tests: whether it is recognized by the general public, and whether it carries weight with professionals in international relations. Sources are cited in the tables themselves and listed in full at the back of every volume.
Curation, not compilation
Each table is built in four stages. Data is compiled across the sectors relevant to a bilateral relationship. It is then filtered and sorted into the thematic categories that give the volume its structure. Algorithms narrow millions of candidate data points to the indicators that actually distinguish the two countries in question. What remains is arranged for reading, with paired circles showing the scale of difference in each row and a three-year trend line beside it.
Position, not magnitude
Indicators arrive in incompatible units, so we compare the rank each country holds among the 193 UN member states rather than the raw value. Canada has many times the forest area of Sweden, but both sit near the top of the world in forest coverage, and that is the more meaningful fact. Every indicator carries equal weight in the final calculation, on the principle that each one contributes a distinct view of a country. Where two countries rank identically, GDP per capita breaks the tie.
Weighted Jaccard similarity, normalized
Similarity is calculated with the generalized Jaccard coefficient applied to rank differences. A country’s weight on any indicator is its rank subtracted from 193, so first place carries a weight of 192 and last place carries zero. Summed across indicators, this produces a similarity score for each domain.
Raw scores skew differently from one domain to the next, which would let geography quietly outweigh demographics. To prevent this, each distribution is transposed onto a standard three-sigma bell curve. Normalization changes the scale but preserves the order, so every domain enters the final index with equal influence and a median of 50 percent.
One number for the whole relationship
The Bilateral Similarity Index averages the normalized similarities across every chapter of a volume. Worldwide, the highest score is 81 percent, between Belgium and the Netherlands. The lowest is 29 percent, between North Korea and Nauru. A score of 68 percent, as between Canada and Sweden, indicates substantial common ground despite distance, language, and separate histories.
What the method cannot yet do
Data coverage is uneven, and some countries are documented far more thoroughly than others. Indicators are not refreshed on a common schedule, so a comparison may draw on figures from different years. And parts of a bilateral relationship resist measurement entirely, which is why every volume also carries the testimony of the ambassadors, consuls, and experts who live the relationship in practice.
Enter the Library