Reply to Hauser and Sewell

Hauser and Sewell's criticism of our paper is wide-ranging. They find fault with the specification of our models and with the quality and character of our data (i.e., "problems of coverage, nonresponse, comparability, and validity"). It is because of these presumed failings, we suppose, that they apparently find the main points of our inquiry unworthy of their consideration: first, that the Wisconsin recursive model of social-psychological school achievement processes contains many arbitrary assumptions regarding permissible influence flows; and, second, that this arbitrariness makes insecure many of the substantive conclusions from this literature regarding the patterning and magnitude of significant others' influences in such models. The purpose of our paper, it will be recalled, was to get such matters of conventional practice on the table for reflection and discussion, not to produce new, improved estimates of how one variable affects another. Our paper was quite explicit on this, and we even are quoted to this effect in Hauser and Sewell's comment. In our opinion, the limited purposes of our paper make many of their specific criticisms irrelevant, although we will address them nevertheless. We think it especially unfortunate, though, that their preoccupation with technical failings distracted them from these broader concerns, which have merit quite apart from the details of our empirical exercises. Still, we felt initially that our analysis was serviceable for its intended purpose, and nothing in the Hauser-Sewell critique convinces us otherwise. We first address their concerns regarding the Growth Study design and data, then the matter of our model specification.