Public health data do not simply exist: the variables included or excluded from any given data set reflect the choices of individuals and institutions with the power to make these decisions. Their judgement typically is guided by prevailing theories of disease causation, which in turn usually resonate with their society's predominant political, economic, and ideological characteristics. This essay examines the making of public health data as a social process, both historically and in the present, and critiques the routine omission of social class data from US public health data bases, the treatment of "race" and "sex" as primarily biological variables, and their conflation with ethnicity and gender. Overcoming these problems will require developing social theories of disease causation and ending the pervasive silence about the health consequences of class, race, and gender inequalities.