There is a subtle bias in controlling for variables in a world with equilibrium. Ie suppose some observable variable (house price) trades off with an unobserved one (say crime) in equilibrium as compensation. “Controlling” for price now means crime is an unobserved confounder