The debate over measurement models being reflective (where the latent construct is the cause and the indicator variables are the effects) or formative (where the latent construct is the effect and the indicators are the causes) has been addressed in various disciplines of management. However, traditional covariance-based measurement modeling does not offer much in modeling formative constructs. This paper addresses that issue with discussions on Partial Least Squares (PLS), a relatively new methodology in management research. Alongside, the paper also discusses the differences between formative and reflective measurement models and the possible effects of measurement model misspecification.

The use of Structural Equation Modeling (SEM) for the assessment of both measurement and structural models in Management Research has been greatly enhanced by the presence of covariance-based modeling software such as LISREL, AMOS and EQS. However, the increasing use of SEM has also raised additional issues (Shah and Goldstein, 2006). One of these issues is the use of covariance-based SEM to measure almost all latent constructs. Covariance-based SEM highly supports a reflective measurement model. The debate over measurement models being reflective (where the latent construct is the cause and the indicator variables are the effects) or formative (where the latent construct is the effect and the indicators are the causes) has been addressed in various disciplines of management such as psychology (Bollen and Lennox, 1991), marketing (Diamantopolous and Winklhofer, 2001 and Jarvis et al., 2003), organizational behavior (Mackenzie et al., 2001) and strategic management (Hulland, 1999). Other researchers have highlighted the theoretical and statistical issues regarding the differences between reflective and formative measures (Bollen and Lennox, 1991; Edwards and Bagozzi, 2000 and Diamantopolous and Winklhofer, 2001).

One such issue, that of measurement model specification, has not yet been addressed fully in Management Research. Since the measurement model is generally a part of a larger structural model, researchers have also previously evaluated issues regarding estimation of structural paths entering and exiting a latent variable and attempted to determine the impact of misspecification on the structural paths (Mackenzie et al., 2001 and Jarvis et al., 2005). The origin of formative models can be traced back to the work of Blalock (1961), but their use has remained limited, in part due to the unavailability of proper modeling software and the lack of proper testing guidelines. Conversely, reflective models have their foundation in the classical test theory (Bollen and Lennox, 1991), and thus have some well-developed testing criteria.

Measuring Formative Constructs in Management Research: Definitions, Distinctions and Measurement,measurement, formative, models, reflective, management, Bollen, Lennox, effects, indicator, constructs, covariancebased, misspecification, causes, software, researchers, variables, foundation, Goldstein, highlighted, increasing, covariancebased, LISREL, Management, methodology, Modeling, organizational, psychology, research, structural, traditional, evaluated