Measurement Error Bias In Causal Inference

Measurement error stems from inaccuracies in the measurement or categorisation of variables. The integrity of measurement is paramount to every science, yet measurement errors are nearly inevitable. Here, I use causal diagrams (directed acyclic graphs or DAGs) to elucidate five distinct pathways of threat to causal inference. Illustrations from comparative cultural research demonstrate limitations of psychometric validations. A structural approach in causal inference approach reveals the need for (1) meticulous attention to measurement conceptualisation informed by local knowledge and domain expertise; (2) careful attention to measurement error bias at every stage of research; (3) sensitivity analyses that quantitatively evaluate the robustness of causal inferences to measurement error bias.