Panel data in sociology: the fixed effects paradigm and empirical practice in panel regression and event-history analysis

In sociology there is a growing body of literature discussing the ability of quantitative research methods to test causal inferences. This discourse introduces panel data as an important instrument to defend the interpretation of coefficients as effects: by focusing on what is going on within individuals, panel data allow for longitudinal empirical modelling, which reflects the inherent semantic of scientific hypotheses. As a consequence, unobserved heterogenity is absorbed and alternative explanations to the presumed causal mechanisms are largely ruled out. However, a within-, or rather Fixed Effects-estimator is needed to realize this potential of panel data. Consequently, such within-estimators are well established as standard in the methodological discourse on panel data. We explain this standard with references to the mechanics of within-estimators in panel-regression and event-history analysis. Finally, inspection of contributions to the two major German language journals for sociology shows that within-estimators are rarely used in empirical practice. We conclude that the potential of panel data to control for heterogeneity is largely underused in empirical practice. At the same time, our inspection of studies illustrates that the benefits of panel data are multidimensional and go beyond statistical virtues.