Blau, Peter M. and Duncan, Otis Dudley. The American Occupational Structure. New York: The
Free Press, 1967.
My bias against this book is based on my awareness of its methods. After the studies of Warner,
Hollingshead, and the others, the whole idea of regression analysis of questionnaire data as a
technique of understanding social class is an absurdity. I cannot say that I gave this book, even
now, a fair reading; this is not the sort of book one can read easily.
The book is frequently hailed as great sociology. Thus, for example, in the recent flap about
Starr's tenure case at Harvard, this work was cited by the quantitativists as a model of the sort of
thing which professionals ought to be doing as rigorous, hardcore, sociology. It won the Sorokin
Award years ago. But what is here in this representative work? For one thing, this is a secondary
analysis. It is a reanalysis of data gathered by the Census Bureau. Blau and Duncan did not gather
the data and they can have no knowledge of the errors of data generation. Furthermore, they are
using data which were gathered for other purposes: the Census Bureau did not envision the
analysis to which Blau and Duncan subjected the data. I have long argued that data produced by
one cannot be used by another. Data are study specific and cannot be reused. Then, too, the
occupational categories used by the Census Bureau are subject to error. I know of no effort by the
Bureau, or by these authors, to estimate the error term in their measures. My guess would be that
the error term is so large as to preclude useful analysis of these data. The question is, then: is this
study worth anything at all?
These authors make it clear that one can disagree with their methods and their interpretations. I
quote from this "rigorous" book: "(The complications of our methods)...require us to ask the
reader to take some things on faith (though his suspicions may well be aroused) and to accept our
pattern of analysis however much he would have preferred another." (p. 116) That hardly seems
to be the sort of thing that "rigorous methodologists" would do: have the reader accept things on
The authors insist that their "rankings" of occupations have remained remarkably stable over
time and it is therefore possible to compare the rankings of fathers' occupations (as reported by
sons) and the sons' occupations. But this assumes a remarkably static social order. I do not doubt
that rankings have remained stable, but only as an artifact of our effort to study prestige. It was
one thing to be a college professor 40 years ago and quite a different thing today - regardless of
the prestige aspect of one's job.
The findings are that one's father's occupation is a reasonably good predictor of one's own
starting point on the social scale but, as one grows older, the relationship diminishes. Let's put it
this way: where else could children start but at the position given them by their parents? With
time, that changes.