Conferences

Thanks to the increasing availability of international survey data, research involving cross-national comparisons is burgeoning in various domains of social science. A rapidly increasing number of studies deals with the simultaneous impact of individual and country characteristics on individual-level attributes, such as opinions, attitudes or behaviour. Very often, multi-level techniques are considered as the state-of-the-art analytical tool for this type of research.

At first sight, the choice for multi-level models in cross-national research is far from illogical. Because of its statistical properties, multi-level modelling is appropriate to take the hierarchical structure of cross-national data—citizens are nested within countries rather than independent observations—into account. Furthermore, multi-level models render it possible to study the interplay between variables that are situated at different levels (here: country and individual level).

Yet, despite these desirable characteristics, the application of multi-level models in the domain of cross-national research might not be as straightforward as it seems. Several methodological issues challenge the choice for multi-level techniques. First, the number of participating countries in cross-national surveys does usually not exceed 25 or 30. Simulation studies have shown that similar numbers of higher-order units are too small to guarantee unbiased multi-level estimates (the so-called small-N problem). Second, whether or not a country participates in an international survey is generally not a question of random sampling. Instead, participation depends on factors such as funding possibilities and research interests of the local scientific community. Third, in times of globalisation it can hardly be sustained that countries are units that develop independently, free from external influences. Especially in the case of Europe, the independence that multi-level models assume at the country-level is highly questionable (the Galton problem). For these reasons, it is uncertain to what extent the statistical inference at the country level implied by multi-level models is meaningful.

This session explores the usefulness of multi-level techniques for the analysis of cross-national survey data. The session welcomes papers that deal with the above-mentioned or other problems of multi-level modelling in cross-national research, papers that propose possible solutions to these issues, as well as applications of multi-level in cross-national research.