The traditional arguments in favour of stabilization policies come either from the existence of liquidity constraints or from the theory of public finance. In the first type of argument, because consumers cannot optimally smooth out fluctuations in disposable income by borrowing and lending on financial markets, the government intervenes to mitigate the effects of changes in GDP on disposable income. In the second type of argument, governments go into debt during contractions and increase public saving during expansions, to smooth over time the distortionary impact of taxation. Countercyclical fiscal policy will be optimal in the first case because of the financial markets' failure, and in the second case because of the presence of exogenous expenditures to be financed by distortionary taxes. For both of these arguments, the welfare gains from stabilization policies are short-term in nature: the policy has no long-term impact since its objective is to smooth the path of consumption.

In Discussion Paper No. 1129, Research Affiliate Philippe Martin and Carol Ann Rogers present a model where there are no public expenditures to be financed and where an externality prevents optimal smoothing by households. The paper analyses, along the transition path and in steady state, the optimal stabilization policy in an economy in which growth is driven by learning by doing. If future benefits of learning by doing are not fully internalized by workers, the optimal fiscal policy is to tax labour during expansions so as to be able to subsidise it during recessions. The long-term impact on output and on human capital of such a policy depends critically on initial conditions: if stabilization is initiated during an expansion its effect on long-term production is positive. The long-term effect is negative when stabilization is initiated during a recession.

Optimal Stabilization Policy in the Presence of Learning by Doing
Philippe Martin and Carol Ann Rogers

Discussion Paper No. 1129, February 1995 (IM)