Amid heightened attention to issues of causality throughout the social sciences, many researchers point to the need for not only causal research but also high-quality descriptive work. But what does high-quality description look like in a context where (correlated) measurement error and multicollinearity are prevalent, as when studying relationships among attitudes? And to what extent can challenges associated with observational studies of attitudes be addressed through use of structural equation modeling (SEM) approaches coming out of psychology? This study aims to address these questions through a detailed comparison of analytical approaches, with examples offered through reanalysis of two datasets used in prior studies. I argue that in many cases, econometric and SEM tools yield similar substantive conclusions, though they also tend to be associated with workflows emphasizing somewhat different concerns. SEM also encourages researchers to think carefully about both mediation and measurement, and it offers greater flexibility in correcting for measurement error. At the same time, measurement error in the real world is typically even more complicated than what SEM can account for. Econometric tools offer more options for dealing with measurement error than many researchers realize, and econometric models offer more flexibility in terms of the number and type of variables that can be realistically analyzed. The study also provides a practical set of guidelines for analyzing hard-to-measure constructs—guidelines which can be used by researchers regardless of which disciplinary tradition of statistical modeling they draw on.
Nathan Favero (Tue,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: