The rational expectation hypothesis, time-varying parameters and adaptive control. A promising combination? (Q5970337): Difference between revisions

This book attempts to justify adaptive control methods in economic models. This area of research started thirty years ago and encountered serious difficulties when it was realized that a government policy has an influence on the expectations of economic agents which, expressed in the equations, meant that the parameters of the model depended on the control, an example of gross mismodelling. The author tries here to circumvent this obstacle by representing the impact of the agents' expectations in the equations, as time-varying quantities which have to be estimated iteratively. The first part of this book is devoted to the the adaptive control methodology. The author uses an ``error-in-variable'' model with time-varying parameters. The average, covariance and transition matrix of the parameters are first supposed to be known, and such hypotheses are then progressively relaxed to a situation with an unknown mean and finally to an unknown mean and unknown covariance and transition matrices. Some Monte-Carlo results address several learning procedures including a ``certainty-equivalence'' method, which the author developed. Nonconvexity of an associated ``cost-to-go'' is considered too. The second part of the book deals with a formulation of the ``rational expectation hypothesis'' for a linear ``error-in-variable'' model with time-varying parameters. Such models are then cast in state-space form leading to convenient Kalman filtering solutions. Identification of such models is exposed next. Several estimation and hypothesis testing methods are considered. The book is not well written; the style lacks fluidity and ideas too. This is accentuated by an assumed known jargon, and many references to other works along the development. But the average reader will not have read these publications. Moreover, a desirable index for the terminology is missing, unfortunately. An author index and bibliographical references are included at the end, however. That time-varying parameters constitute ``the right word'' for expressing unknown rational expectations of agents in an economic model, is far from convincing, and the author himself writes that the proposed parametric modelling could cover other cases. But a correct modelling might involve structural aspects, and not parametric aspects only. Adaptive control makes sense for linear systems with time-varying parametric uncertainty, but stumbles when unknown nonlinearities are present, and they do in most practical situations. Here, the models are supposed to be linear without such perturbations; quite convenient. But even if accepting this, some of the author's hypotheses on parameter variation can be contested. Overall, the author was right to write the subtitle in the interrogative form.