Within the Information Systems
literature, there has been an emerging interest in the use of
formative measurement in structural equation modeling (SEM).
This interest is exemplified by descriptions of the nature of
formative measurement (e.g., Chin 1998), and more recently the
proper specification of formatively measured constructs (Petter et
al. 2007) as well as application of such constructs (e.g., Barki et
al. 2007). Formative measurement is a useful alternative to
reflective measurement. However, there has been little guidance on
interpreting the results when formative measures are employed. Our
goal is to provide guidance relevant to the interpretation of
formative measurement results through the examination of the
following six issues: multicollinearity; the number of indicators
specified for a formatively measured construct; the possible
co-occurrence of negative and positive indicator weights; the
absolute versus relative contributions made by a formative
indicator; nomological network effects; and the possible effects of
using partial least squares (PLS) versus covariance-based SEM
techniques. We provide prescriptions for researchers to consider
when interpreting the results of formative measures as well as an
example to illustrate these prescriptions..
Keywords:
Structural equation modeling, formative measurement, formative
indicators, measurement theory management, information management practices