Establishing causality using observational panel data: models and applications

While much empirical research in the area of travel behaviour is based on cross-sectional data, it is well known that models based on such data are not well suited to satisfy all criteria for establishing causal effects (most notably the criteria of non-spuriousness and time precedence). Randomised controlled trials offer the gold standard for establishing causal effects, but such experiments are often unfeasible and/or unethical. In this contribution we first review several relationships in travel behaviour research where establishing causal effects is crucial (from a scientific and policy perspective). Next, we explore the benefits of (observational) panel data to satisfy the criteria for establishing causation by reviewing the various statistical models that can be used to handle panel data. In addition, we also show that -with the use of panel data- the distinction between “within-person” and “between-person” effects becomes crucial; a distinction that is (as of yet) not well-embedded in theories of travel behaviour. Finally, we review several recent empirical applications of panel data models in travel behaviour research, illustrating the novel behavioural insights that can be gained.