Sensitivity of optimal programs with respect to changes in target stocks: The case of irreversible investment

Consider an aggregative model of economic growth with changing technology and tastes, in which investment is irreversible. It is shown that initial decisions in finite-horizon optimal programs are insensitive to changes in terminal stocks, provided the horizon is long enough. This generalizes Brock's result, which was proved assuming investment to be reversible. The irreversibility constraint does not allow one to follow Brock's method of proof, using the dual (Shadow Price) properties of optimal programs. An alternative method of proof is developed, using a primal approach, and exploiting dynamic programming arguments.