Confounding caused by causal-effect covariability

Abstract: Confounding seriously impairs our ability to learn about causal relations from observational data. Confounding can be defined as a statistical association between two variables due to inputs from a common source (the confounder). For example, if $Z\rightarrow Y$ and $Z\rightarrow X$, then $X$ and $Y$ will be statistically dependent, even if there are no causal connections between the two. There are several approaches available to adjust for confounding, i.e. to remove, or reduce, the association between two variables due to the confounder. Common adjustment techniques include stratifying the analysis on the confounder, and including confounders as covariates in regression models. Most adjustments rely on the assumption that the causal effects of confounders, on different variables, do not co-vary. For example, if the causal effect of $Z$ on $X$ and the causal effect of $Z$ on $Y$ co-vary between observational units, a confounding effect remains after adjustment for $Z$. This causal-effect covariability and its consequences is the topic of this paper. Causal-effect covariability is first explicated using the framework of structural causal models. Using this framework it is easy to show that causal-effect covariability generally leads to confounding that cannot be adjusted for by standard methods. Evidence from data indicates that the confounding introduced by causal-effect covariability might be a real concern in applied work.