Compare Arbitration Decisions Without Erasing Their Differences

A list of cases is not yet comparative analysis. Decisions that use the same treaty article or arbitration term can arise from different instruments, enactments, rule editions, procedural stages, factual records, and judicial functions. A synthesis must preserve those differences long enough to explain why the sources can or cannot be compared.

Build the matrix before writing a trend statement.

Define one proposition

Write the narrow proposition that the comparison will test. Avoid a heading such as “public policy cases.” A usable question identifies the decision maker, instrument, procedural task, and point of interpretation.

For example, a researcher can compare how specified enforcement courts describe one element of a convention ground. That does not automatically include tribunal merits reasoning or set-aside decisions under a national statute.

Create one row per document

Include exact document identifier, body, jurisdiction, date, posture, requested relief, disposition, instrument and version, provision, material facts stated by the body, quoted or pin-cited passage, source layer, language, and later history. Keep related documents connected but separate.

The UNCITRAL New York Convention page provides the official convention context and access to the text, status, CLOUT abstracts, and related official resources. Use a common official instrument text as a reference point, then verify how the relevant court connected it with domestic law and the record before it.

Mark comparability before direction

Add a field for exact, partial, or weak comparability. Explain the choice. Decisions can be partially comparable when they address a similar phrase but differ in procedural posture or enacted text. Weakly comparable decisions should not be counted as support on the same basis as direct decisions.

Only after this step should the researcher record whether a decision supports, limits, distinguishes, or does not reach the proposition.

Avoid vote-counting

Five decisions with thin or different reasoning do not mechanically outweigh two decisions with directly relevant reasoning. Court hierarchy, jurisdiction, date, later treatment, procedural role, and source completeness matter. A database count can show search coverage, not the legal weight of a conclusion.

Describe clusters and differences. State when the set is shaped by publication availability, language access, database coverage, or a limited date range.

Write from the matrix back to the record

Every synthesis sentence should be traceable to one or more rows and exact passages. State the common feature, then the material differences and exceptions. Use an as-of date and a source-coverage statement.

The matrix is successful when a reader can see why the decisions were placed together and where the comparison stops. It should make disagreement and missing evidence visible, not force every document into a simple line of authority.