<p>This article offers a methodological critique of the standard practice of modelling party system fragmentation at the national level without accounting for district-level parameters. The central argument is that national party system fragmentation is a near-mechanical product of two underlying quantities, district-level fragmentation and cross-district evenness of party support, measured by the Party System Nationalization score, and their interaction, a relationship that follows necessarily from the mathematical properties of quadratic aggregation. Party system fragmentation is measured with the Herfindahl–Hirschman index, which is mathematically related to the Laakso–Taagepera effective number of parties as its reciprocal under equal-share conditions and inversely related to it in general. The argument is substantiated by formal derivation and tested on a dataset of 568 elections in 70 countries (1946–2021) using random-effects generalised least squares models. Empirically, 97% of elections fall within five HH-index points of the mechanically predicted value, confirming the identity’s precision in real-world data. The analysis further demonstrates that institutional variables affect the two sub-components of national fragmentation in opposing directions, so that their net effect on the national aggregate is attenuated relative to their effects on the components individually. Consistent with theoretical expectations, a boundary case illustrates that the framework’s operationalisation is sensitive to the level at which the unit of electoral competition is defined. The major implication of these findings is that explanatory models of party system fragmentation should focus on district-level fragmentation and party system nationalization as separate dependent variables, rather than on their national aggregate.</p>