Experiments are a major tool in the toolbox of behavioral economists. Whether laboratory- or field-based, experiments provide controlled environments to detect causal relationships between the variables of interest. However, like all methods, they have strengths and weaknesses. In the recently published paper by BCCP Fellow Levent Neyse and his coauthors, in the Journal of the Economic Science Association, we discuss what large-scale household panel studies, like the German Socio-Economic Panel Innovation Sample (SOEP-IS), can offer to behavioral economists and how they can complement experiments.
First, panel studies are rich in terms of number of respondents and control variables. This is an important feature that laboratory experiments usually lack: a large majority of laboratory studies rely on student samples. Second, panel studies provide information about the history of the respondents as they regularly collect information from the participants over many years. Third, in examples like SOEP-IS, researchers can introduce novel experiments and questions into the panel, which can be matched with the existing data. The existing data does not just include socio-economic information or psychological variables like risk attitudes, patience, life satisfaction, but also information based on real actions and outcomes: insurance decisions, taxes, diagnosed illnesses, or church attendance.
Although the paper is based on the experience with SOEP-IS, its content can be generalized to other panel studies around the world, including Understanding Society from the UK, The Panel Study of Oncome Dynamics from the US (PSID), or the Korean Labor & Income Panel Study (KLIPS).
Through panel studies, behavioral economists can extend the diversity of their research questions to more applied and policy relevant topics. Some examples are household-level decisions, impacts of public and economic policies on behaviors, personality effects on job selection, and the role of natural disasters on insurance demand. Such topics are particularly difficult to tackle in the laboratory setting and in the absence of a panel structure.
Open-access article can be found at the journal’s website: link.springer.com/article/10.1007/s40881-023-00150-6