Résumé
This paper revisits that of Barrett, published in Food Po licy in 1997. Even if Barrett resorted to the use of the appropriate methodology, ARCH - M models, it would appear that he misinterpreted its philosophy, just as he misspecified the variance equation of the ARCH. This led him to a spurious modeling at severa l levels, methodological and theoretical. Using food price time series (as Barrett did), we first explain the inconsistencies in Barrett (1997) and then we propose three modeling procedures: that of Engle, Lilien and Robins (1987) and then two others simil ar to Barrett's, but differing by the fact that one is carried out with good identification of the time series data generating process, and the second without. These three models allowed us to apprehend the gap between results of a "correct" ARCH - M model a nd those of Barrett. The interest of this comment on is to highlight few irregularities of drawn conclusions that were subject to economic recommendations at that time, and also that are being largely replicated in agricultural economic papers. (Résumé d'auteur)