Asking—and answering—causal questions using longitudinal data

Despite a growing availability of longitudinal datasets, it can be difficult to select the most appropriate modelling strategy. In particular, there is little guidance regarding which causal questions one can ask using longitudinal data, and what is the best way to answer these questions. This paper distinguishes between three causal quantities that researchers in social and behavioral sciences often want to investigate: the contemporaneous treatment effect, the cumulative treatment effect, and the long-term treatment effect. I provide a definition of these quantities, clarify which causal assumptions are needed to identify these effects, and present statistical models that can be used to estimate these quantities. The aim is to provide an accessible overview of longitudinal models for estimating causal effects using observational data. I illustrate the methods discussed by studying how peer victimization in school affects internalizing behaviors. For this purpose, I used a nationally representative sample of kindergarteners, and focused on peer victimization events from 2nd to 5th grades. Consistent with prior research, the results suggest that peer victimization has short-run, cumulative, and long-term effects on internalizing behaviors.