Public Opinion in Policy Studies: Lessons from the Logic of Structural Equation Modeling
Policy studies scholars regularly investigate linkages between public opinion and policy. The use of public opinion as a variable in empirical research poses special challenges. In this article I suggest that the logic and methods inherent in the art of structural equation modeling provide opportunities to overcome some of these challenges. I describe this type of logic as it pertains to measurement error, context effects and endogeneity. Using General Social Survey data for the United States and taking thermostatic feedback models as an example, I demonstrate why it is important for policy scholars to attend to (a) measurement modeling and cross-level isomorphism, (b) shocks that might bias survey response patterns and (c) endogeneity implied by the theoretically reciprocal nature of opinion and policy feedback. These examples come with discussions of why scholars should pay attention to model specification so that theory and empirics are in unison and how to perform model fitting and testing to better develop theories and models of policy processes.