Abstract 

Do partners influence each other?
An application of dynamic panel estimators to dyadic influences

*Matthias Collischon ^1,2 , Andreas Eberl ² & Ansgar Hudde ³
1 Institut für Arbeitsmarkt- und Berufsforschung
2 FAU Erlangen-Nürnberg
3 Universität zu Köln

Romantic relationships are at the core social environment for many adults and understanding dynamics within these relationships is therefore essential for the social sciences. A consistent finding is that partners are more similar than random mating would predict concerning diverse traits that include behaviors, values, and attitudes. However, less is known about the dynamics of these traits in with relationships. In particular, there is a lack of statistically solid identification answering whether and how partners influence each other. This paper uses dynamic panel estimation methods (System-GMM) to identify the causal effects of partners on each other’s attitudes. Using this method, we can account for biases that cast doubt on previous research results, which only used cross-sectional data or did not account for potential biases in longitudinal analysis, e.g. reverse causality. With simulated data, we compare our results to classical methods in the literature. Our results show that dynamic panel estimations provide consistent unbiased results in cases in which classical methods (e.g. APIMs) are biased. Overall, we demonstrate that these dynamic panel estimation methods, which have previously proven useful in fields like labor economics can also solve problems of statistical identification to study mutual influences in dyads, such as romantic partnerships.