Theory 29/Randomised trials settle causality
Trials Can't See Equilibria
The claim
Trials measure partial-equilibrium effects, so general-equilibrium responses can reverse the sign when scaled.
01Mechanism
Prices and wages adjust when a programme is large, changing the conditions the trial held fixed.
02Prediction
Effects should differ between small and large implementations.
03Refutation
Scale-invariant effects would remove the equilibrium response.