Laurence Broze

SOLUTIONS OF LINEAR RATIONAL EXPECTATIONS MODELS

Linear rational expectations models generally have a large number of solutions. It is thus important to describe them exhaustively in order to study their properties and subsequently estimate which solution best fits the data. In this paper, a global approach is suggested allowing a simultaneous treatment of all possible cases. The fundamental concepts are the revision processes appearing in the procedure of updating expectations. It isfound that the set of solutions is completely described by using a limitednumber of these processes. We show how the method may be applied to determine the set of stationary solutions admitting an infinite moving-average representation. We give a natural parametrization of this set and discuss the exact number of independent parameters.

Quasi Indirect Inference for Diffusion Processes

We present an estimation procedure for continuous time models based on discrete sampled data with a fixed unit of time between two observations. Since in general the conditional likelihood of the model cannot be derived an indirect inference procedure based on simulations of a discretized model is proposed. Asymptotic properties of this quasi indirect estimator are given and some particular cases are examined. Finite sample properties are also considered through Monte Carlo experiments. We also examine the problem of the choice of an appropriate simulation step.

Solutions of Multivariate Rational-expectations Models

The aim of this paper is the study of the path solutions of a multivariate rational expectations model. We describe several procedures for solving such dynamic systems based on either the adjoint operator method or the Smith form. As a by-product, we derive the dimension of the set of solutions in terms of martingale differences and the dimension of the set of linear stationary solutions when we restrict ourselves to the linear case. These dimensions are functions of the number of equations in the system, of the maximum lead, and of the orders of some eigenvalues of the characteristic equation associated with the system.

Reduced Forms of Rational Expectations Models

A comprehensive exposition of rational expectations models is provided here, working up from simple univariate models to more sophisticated multivariate and non-linear models.

Non redundancy of high order moment conditions for efficient GMM estimation of weak AR processes

This paper considers GMM estimation of autoregressive processes. It is shown that, contrary to the case where the noise is independent (see Kim, Qian and Schmidt (1999)), using high-order moments can provide substantial efficiency gains for estimating the AR(p) model when the noise is only uncorrelated.