Policy Simulations
What Do They Expect?

The 'rational expectations revolution' has had a profound if controversial impact on macroeconomics and the analysis of economic policy-making, particularly in open economies. In a recent Discussion Paper, CEPR Programme Director Willem Buiter surveys some recent developments in analysing dynamic economic models which incorporate rational expectations by private agents and can be represented formally by systems of simultaneous linear differential equations. While much of the literature he surveys is technical in nature, the techniques he discusses are now commonly used in simulations of economic policy.

Until recently the dynamics of these models were restricted to only two or three dimensions because of the difficulties of obtaining exact analytical solutions. This is a considerable drawback, since many interesting policy questions can only be answered by studying the simultaneous and interdependent adjustment of various asset stocks (money, domestic debt, foreign assets, the capital stock, etc.) and several prices (domestic wages and prices, the exchange rate, bond prices, etc.). The methods Buiter surveys can be implemented numerically even for large dynamic systems with simple and efficient computer algorithms. The consequences of alternative policy rules and of any combination and sequence of anticipated or unanticipated, current or future, and permanent or transitory exogenous shocks can be evaluated.

These methods can also be applied to the problem of optimal policy design when the model is linear and the objective function is quadratic. 'Optimal' is defined with reference to the policy maker's objectives, whatever these may be.

A frequently encountered problem with policy design in rational expectations models is that optimal policies tend to be 'time- inconsistent'. Even if no new information arrives concerning the exogenous environment of the policy maker, there will be a tendency for the latter to change previously announced plans and policies, once these announcements have had the desired effect on the private economy. Time-inconsistent policies are therefore not credible, unless the government can somehow precommit itself not to change them. Without precommitment, a credible, but inferior 'time-consistent' policy can be derived which again fits into the general solution format outlined by Buiter. Recent work by Marcus Miller and Mark Salmon (not surveyed in the paper) shows that decentralised policy design with many policy makers in an interdependent world economy (or within a single national economy) can be analysed within the same formal framework.

The paper concludes with an illustration involving a numerical example of optimal and time-consistent anti-inflationary policy design in a model with long-term nominal contracts.


Policy Evaluation and Design for Continuous Time Linear Rational Expectations Models: Some Recent Developments
Willem H Buiter

Discussion Paper no.15, May 1984 (IM)