Studies of the social bases of environmental concern over the past 30 years have produced somewhat inconsistent results regarding the effects of sociodemographic variables, such as gender, income, and place of residence. The authors argue that model specification errors resulting from violation of two statistical assumptions (interval-level measurement and parallel regression) may contribute to such inconsistency. Using data from an April 2000 Gallup poll commonly known as the Earth Day Survey, they found that violation of the former assumption has little effect on analysis results, whereas violation of the latter assumption leads to inconsistent effects of sociodemographic variables. (Contains 5 tables.)