We examined the prevalence and correlates of both general and workplace-rel
ated drinking measures using data from a telephone survey of 673 workers in
a large municipal bureaucracy and tested the hypothesis that observed diff
erences across job categories can be explained by compositional difference
in terms of demographic variables known to be related to drinking behavior.
Results suggest such factors account for much of the variation in general
drinking measures (prior-28-day quantity, CAGE score, indicating risk for d
ependence), but that significant variation in a workplace-related drinking
measure (times ever drank before, during, or just after work) remains even
after such factors are controlled. Implications of these findings for exist
ing theories of workplace effects on drinking are discussed, along with a c
onsideration of appropriate levels of analysis for future studies.