The effective space of party competition
A conventional triangle display recommended by major authors has begun to be widely used for studying the patterning of party scores at constituency level across general elections. We show that it is systematically misleading in several ways and should not be further employed. To replace it we introduce the improved concept of ‘effective competition space’ (ECS) defined as the space of all feasible combinations of party vote shares, given the ‘number of observable parties’ (Nop) (defined as those with 1 per cent support or more). We employ a two-dimensional operationalization of the ECS concept to analyse changes across four Indian general elections, demonstrating that the space defined by the median constituency Nop captures well the variation of results. We also show that the effective competition space can be segmented in ways that respond to the context of each election but also facilitate meaningful comparison across them. We briefly show also how multi-dimensional versions of ECS operate. We conclude that greater attention should be paid to the factors driving changes in the number of observable parties, the key variable determining the size and patterning of effective competition space. More generally, we demonstrate the value of a relativized approach to measuring and comparing patterns across elections. Whenever an index score changes from one election to another, the frame within which the index scores are determined also often changes.