Are structural equation models theories and does it matter?

Abstract

ABSTRACT In their introductory marketing, management, and social psychology courses, undergraduates learn that correlation coefficients provide weak evidence for causal conclusions. Nonetheless, researchers conclude causally from correlation coefficients by drawing causal arrows in their structural equation models (SEMs). Although most researchers avoid describing their findings in causal language, obligatory recommendations for applying those findings insert causation. Researchers’ standard rejoinder to validity challenges is “the critics have ignored theory’s role in rendering our SEM internally and externally valid”. To evaluate this rejoinder, we explore SEMs based on comprehensive underlying theories and as stand-alone and testable context-specific theories that blend previously published hypotheses and findings. In most cases, the rejoinder is unconvincing.