Aggregation choice only matters when constituent units are heterogeneous.

Evidence

Although substitutability is a property of considerable theoretical interest, in practice its significance depends on the degree of heterogeneity among constituent units. Obviously all combinatorial algorithms yield exactly the same result under conditions of perfect homogeneity [e.g., a = b in equation (5.2)]. Even under conditions of near homogeneity the measurement procedure may be of little consequence since aggregate measures are apt to be highly correlated. But as the homogeneity of individual units decreases, the choice of aggregation strategy is increasingly likely to have significant empirical implications. — Gary Goertz, Social Science Concepts, 896

Categorization

• XREF: Same logic as any composite index (HDI, credit ratings, aggregated noise metrics): the aggregation rule is inert under homogeneity and decisive under spread.
• SEED: Angle: ‘The tyranny of the average’ — why every composite index (school rankings, ESG scores, credit ratings) encodes a hidden substitutability assumption nobody defends.