Kazuma Mizukoshi (University College London)
Abstract: “Science is impossible without an evolving network of stable measures” (Wright 1997: 33), but to what extent should measures be stable? Though measures seem still stable as long as they return scores systematically (Carmines and Zeller 1979: 14-5), how can we distinguish systematic and random errors? This is a question that has been placed, but never answered comprehensively, at the heart of an inquiry into technically called party system nationalization. The concept, party system nationalization, has been developed along with two paradigms. One of them has seen it as an extension of party nationalization, initially projected by Schattschneider (1960) and Stokes (1965), which can be defined as the equal shares of electoral support for political parties across electoral districts in an election. And, so they did for the measure of party system nationalization simply as the summation of party nationalization scores obtained by individual parties (Jones and Mainwaring 2003: 143). On the other hand, standing on the classic of Duverger (1954), some scholars have re-defined party system nationalization as the inflation and dispersion of district-level party systems. In this sense, for them, it would not be surprising to see the effective number of electoral parties as the foundation of alternative measures (Kasuya and Moenius 2008: 129-131). Acknowledging the advantage of each measure, I will argue that they still have serious drawbacks respectively and develop two hypotheses. First, upon defining party system nationalization, the interaction (not simple sum) of competing parties should be considered (Sartori 1976: 39). Second, following the variance-components tradition of Stokes, a new indicator should be able to distinguish systemic and random errors appropriately (Morgenstern et al. 2014; Mustillo and Jung 2016). The new indicator, as well as previous measures, will be tested with a hypothetical dataset initially to clarify the advantages and drawbacks of each measure, and then, investigated with the actual district-level electoral results of 54 political parties in 14 Latin American countries between 1980 and 2015.