On the measurement of electoral volatility

Abstract

Electoral volatility measures the degree of vote switching between political parties in two consecutive elections. Political scientists use this as an indicator of party system (in)stability. Pedersen (1979) states that volatility should increase when the number of parties changes and/or relevant parties experience vote transfer between elections. However, his proposed functional form of measuring volatility does not always respond to these changes. To address these limitations, we introduce a class of additively separable electoral volatility measures which are responsive to changes in both the number of parties and their vote shares. We present a set of axioms that are both necessary and sufficient to characterize the proposed class of indices, making the structure of the indices more transparent. The paper also introduces two quasi orders which can rank party systems in terms of all electoral volatility indices satisfying certain intuitively reasonable axioms. Finally, applications of the proposed class of indices and the quasi orders are provided using data from Indian state elections.