Recent Developments and Current Approaches to the Analysis of Panel Data

Establishing causal relationships is arguably the most important task of the social sciences. The randomized experiment is the gold standard in causal analysis. However, many interesting questions cannot be examined with experiments. Feasibility and ethics limit the use of randomized experiments in some situations, and retrospective questions require a different approach, as well.

Panel analyses offer a promising way to examine causal questions with non-experimental, i.e., observational data. Panel analyses refer to studies of the same observational unit over time, and are effective at addressing unobserved heterogeneity in the form of time-invariant confounders.

However, panel data are not a silver bullet – many difficult questions and issues plague the investigation of causal effects with panel data. Some of these issues include:

• heterogeneous effects,
• the choice between discrete and continuous time models,
• obtaining the optimal lags between treatment and outcome measurement,
• the choice between looking at contemporaneous and lagged effects, and
• the inclusion of lagged dependent variables, the assessment of bidirectional effects

to name a few.

We encourage paper presentations looking at these and other issues related to the use of panel data for causal analysis. Papers can be methodological in nature, or focus on novel applications. We welcome papers looking at traditional (e.g., survey) and intensive (diary studies) longitudinal studies, as well as those focusing on either case studies (e.g., synthetic control) or natural experiments (differences-in-differences). Papers comparing various approaches and modelling strategies are also encouraged.