The Negative Consequences of Measurement Model Misspecification: A Response to Aguirre-Urreta and Marakas

It has been more than 40 years since Blalock (1964) noted the distinction between what he called “cause” (formative) and “effect” (reflective) indicators of latent variables, and three decades since the academic literature recognized that some SEM measurement models don’t fit classical test theory’s assumptions about the direction of causality of the relationships between constructs and their indicators (Bagozzi 1981; Fornell and Bookstein 1982). Since that time, interest in modeling constructs with formative indicators has significantly increased in the business, social science, and information systems literatures (e.g., Bollen 2007; Cenfetelli and Bassellier 2009; Diamantopoulos and Papadopoulos 2010; Diamantopoulos et al. 2008; Diamantopoulos and Winklhofer 2001; Edwards and Bagozzi, 2000; Jarvis et. al. 2003; Law and Wong 1999; MacKenzie et al. 2005; Marakas et al. 2007; Petter et al. 2007), and an increasing number of academic studies are incorporating these types of measurement models in their substantive investigations. However, recently researchers in a variety of disciplines (Franke et al. 2008; Howell et al. 2007; Kim et al. 2010) have raised questions surrounding the correct conceptualization and operationalization of formative indicator measurement models. 1