Dr. Detmar Straub, SCIS Visiting Scholar, Research Presentation
| Date/Time: | Friday, 15 Nov 2013 from 10:30 am to 12:00 pm |
|---|---|
| Location: | 3164 Gerdin Business Building |
| Cost: | free |
| Contact: | Stacey Ross, |
| Phone: | 515-294-3659 |
| Channel: | Ivy College of Business |
| Categories: | Lectures |
| Actions: | Download iCal/vCal | Email Reminder |
"Handling Formatively-Measured Constructs in the Organizational Sciences"
Abstract
While researchers go to great lengths to justify and prove theoretical links between constructs, the relationship between measurement items and constructs is often given second shrift. The relationship between construct and items is often assumed to be reflective, meaning that the measurement items are a reflection of the construct, highly correlated, and interchangeable. Many times, though, the construct is not being measured through reflective scales, but rather through formative measures. Formatively-measured constructs (i.e., "formative constructs") occur when the scale items describe and define the construct rather than vice versa. Formative elements may be differentially weighted in "forming" the construct and are not interchangeable.
In this research, we examine whether formative constructs are indeed being mistaken for reflective constructs by organizational researchers. By examining complete volumes of MIS Quarterly and Information Systems Research over 3 years, we discovered that a significant number of articles have indeed misspecified formative constructs. These results were very similar to results found in marketing by Jarvis and colleagues.
For scientific results to be valid, we argue that researchers must properly specify formative constructs. This paper discusses the implications of different patterns of common misspecifications of formative constructs on both Type I and Type II errors. To avoid these errors, the paper provides a roadmap to researchers to properly specify formative constructs. We also discuss how to address formative constructs within a research model after they are specified. Choice of SEM modeling (PLS or covariance-based SEM) of formatively-measured constructs is also discussed. A quick review of the many articles that have come out since the publication of this work in 2007 will wrap up the session.