Formatively-measured constructs are increasingly applied in information system research models.
Recent work shows that exogenous formatively-measured constructs suffer from a number of problems that include interpretational confounding and a lack of external consistency. Yet replacement by reflectively-measured constructs can lead to bias if not theoretically appropriate. One solution may be to use a MIMIC construct composed of the formative measures as well as two additional reflective measures. A simulation study indicates that a MIMIC so composed mitigates the problems of interpretational confounding and poor external consistency allowing use broader use in a variety of structural models.
Jiang, James J.; Tsai, Jacob C.; Klein, Gary; and Petter, Stacie
"Fully Mediated Effects of Formative Measures Using MIMIC Constructs,"
Pacific Asia Journal of the Association for Information Systems: Vol. 4
, Article 2.
Available at: http://aisel.aisnet.org/pajais/vol4/iss2/2