This paper identifies a novel form of dynamic inconsistency of stabilization policy in increasing returns models that generate multiple equilibria. We present a two-period version of the Benhabib-Farmer (1994) externalities model and derive closed-form solutions for all endogenous variables in every perfect foresight equilibrium. We provide conditions under which the stabilization policy that maximizes time zero consumer welfare is not time consistent. Furthermore, we characterize the time consistent stabilization policy. Our results cast doubts on the usefulness of government coordination of economic activity when the government lacks a commitment mechanism. Without commitment, a benevolent government can rule out multiplicity only by ensuring that a pareto dominated equilibrium obtains.