This article explores how aggregate level data may be used to make inferences about individual level behaviour. A common strategy in the past was to assume that the relations evident in aggregated data are also present in individual data. Analysis of datasets where there is both individual and aggregated information demonstrates that this assumption is most often incorrect. This means that the relationships observed between variables at an aggregated level are unlikely to be observed in individual level data. This is a problem because quite often social scientists only have aggregated data for exploring individual level behaviour. A key question explored in this article is how is it possible to validly and reliably use aggregated datasets to make inferences about relationships between variables at the individual level. An example analysis is given using electoral data from the Czech Republic., Pat Lyons., 4 obrázky, 4 tabulky, Obsahuje bibliografii, and Anglické resumé
This study has two key aims. First, it explores the two main methods used in the Czech Republic to operationalize the concept of party identification. Second this study demonstrates the merits of both methods; and on the basis of this research proposes one of these party identification measures for use in future studies. This study builds on the classical conceptualization of party identification developed by the Michigan School, but also utilizes (1) reformulations proposed by the Revisionist School; (2) Macropartisanship theory; (3) the Social Identity approach; and (4) experience of using the party identification measure in Europe. Using the Czech Election Study of 2002 this research shows that estimates of partisanship are influenced by how the survey question is formulated. Thereafter, the relationship between the two main variants used in the Czech Republic is presented. This work reveals that use of a (prior) vote intention item leads to an ‘over estimation’ of partisanship when compared to the other partisan measure. Comparing estimates from the Czech Election Studies of 1996, 2002 and 2006 this article reveals that there is no simple association between party identification and voter turnout. In fact, it is better to think of party identification as having two components: (a) probability to vote; and (b) probability to support a party that elicits a sense of identity from voters., Lukáš Linek., and Obsahuje bibliografii a bibliografické odkazy