It's rather specious reasoning at it's surface, as adding additional controls that reduces variance in the model doesn't discount that there is discrimination w

It's rather specious reasoning at it's surface, as adding additional controls that reduces variance in the model doesn't discount that there is discrimination within the labour market. That is to say, the discrimination happens elsewhere in the chain, possibly at at the education level (women pushed out of higher income degree streams), the on the job mentor discrimination (women are less likely to be 'picked' to be mentored), or due to the lack of a comprehensive childcare support system pushes women to lower income jobs or do not have the ability to rise in those jobs. There are no simple answers as "hey, when we control for x, y, z the wage gap disappears", as these are almost inscrutably complex systems.