Space and Time in Comparative Political Research. Pooled Time-series Cross-section Analysis and Multilevel Designs Compared

Abstract

The combination of cross-section and time dimension is a central issue in current comparative political research. The state-of-the-art procedure in this context is pooled time-series cross-section analysis (PTSCS), which is en vogue in today's relevant literature but not uncontested. An interesting option are multilevel designs, which allow the combination of time and space by considering observations over time nested within country-specific contexts. The purpose of this paper is to illustrate the advantages of multilevel designs in comparative political research, which mainly concern the modeling of time-invariant variables, the possible distinction between cross-sectional and time related variance in the data, and the possibility to model heterogeneity instead of just correcting it. Using the example of an analysis of public education expenditure in the 26 Swiss cantons between 1978 and 2003, it can be shown that multilevel analysis – mainly due to its statistical and conceptual advantages –is indeed a promising alternative to PTSCS.