Variables binaires et tests prédictifs contre les changements structurels

Abstract

The main purpose of the paper is to illustrate the use of a dummy variable interpretation of the predictive Chow test against structural change. After describing how the predictive Chow test against structural change in linear regression models can be viewed as a test on the coefficients of a set of dummy variables, it is shown that these can provide useful additional information on the importance and timing of structural changes. Then, the approach is illustrated by applying it to a version of the St. Louis equation (in rate-of-change form) estimated over the period 1953/I-1976/IV: we detect some instability in the 1970's but find it is rather localized, being linked mainly to two quarters (1973/IV and 1975/III).