The authors empirically explore whether and to what extent the content and thematic structure of television news influence the perceived importance of an issue among the population. They present several methodological and conceptual innovations to traditional agenda-setting research. First, they utilise a data set that is somewhat unique in the Czech scientific community. It was constructed by combining data from repeated public opinion surveys and from content analysis of television news; both segments were collected over a period of more than four years. Unlike in most research, the data are not aggregated before the analysis. Second, the authors' choice of issue overcomes the problem of endogeneity that is ubiquitous in agenda-setting research. Third, the authors employ hierarchical linear models to represent adequately the multi-level and clustered data structure and to obtain unbiased estimates of model parameters. Finally, in addition to using standard measures of the intensity of media exposure, they utilise a rather unique set of explanatory variables that represent the homogeneity of media coverage and the relative salience of the issue. They conclude that the media salience of an issue does indeed increase its perceived importance among the population, but media homogeneity and relative salience induce no effect on perceived importance. Owing to a number of methodological improvements, the authors' results are more robust than those produced in previous research.
The article briefly describes multilevel models and presents their simplest applications. After the methodological and statistical need for this procedure is explained, real data are used to demonstrate how a hierarchical linear model is constructed. The article presents models with a random intercept, models with random slopes, and models with explanatory variables measured at higher levels. In the conclusion, other possible applications of multilevel analysis are discussed, and the basic readings on multilevel analysis are presented.