Why is gender and sex critical when doing causal inference, and how critical they are?

Sex and gender, two distinct concepts, are critical for causal inference practice in social sciences, epidemiology and public health. Dichotomized sex and gender indicators, often messed up with each other, have long been practiced as stratifying variables, effect measure modifiers, potential confounders, or often-ill-defined treatment/exposures. As two distinct, co-constituted, multidimensional, bio-socio-political, dynamical and culture-dependent constructs, sex and gender’s dimensionality and complexity is almost always overlooked. The common practice, using dichotomized indicators and collapsing/discarding minor groups, mismeasures underlying mechanisms, providing bad and malfunctioning proxies. Once sex and gender have any roles in a causal question and mechanisms, these bad proxies can make causal assumptions less plausible and introduce bias. Sometimes the magnitude of the bias could be argued as trivial, but there exist many situations where bias is not negligible and may lead to wrong conclusions and do harm.