In our research on authentic leadership and conflict management we're using a number of constructs such as authentic leadership ALQ, Thomas Kilman conflict mode instrument CPP , and conflict management style based on the the Rahim organizational conflict inventory. We are trying to figure out if these three constructs are formative and therefore do not need reporting of reliability such as composite reliability. The data analysis technique we are using is smart PLS. I can make arguments both ways as to whether they are reflective or formative constructs. Is a construct designated reflective or formative or does it depend on how we use it in the research. Does anyone have any guidance on this?
As a rule of thumb: If items "reflect", i.e. depict a construct, it is a reflective measurement. If items "form", i.e. build a construct, it is a formative measurement.
There are some methodological papers with regard to the formative vs. reflective discussion that I would recommend for reading and citing. For example:
Finn, A., & Wang, L. (2014). Formative vs. reflective measures: Facets of variation. Journal of Business Research, 67(1), 2821-2826.
Diamantopoulos, A., & Siguaw, J. A. (2006). Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration. British Journal of Management, 17(4), 263-282.
Edwards, J. R., & Bagozzi, R. P. (2000). On the nature and direction of relationships between constructs and measures. Psychological methods, 5(2), 155-174.
As a rule of thumb: If items "reflect", i.e. depict a construct, it is a reflective measurement. If items "form", i.e. build a construct, it is a formative measurement.
There are some methodological papers with regard to the formative vs. reflective discussion that I would recommend for reading and citing. For example:
Finn, A., & Wang, L. (2014). Formative vs. reflective measures: Facets of variation. Journal of Business Research, 67(1), 2821-2826.
Diamantopoulos, A., & Siguaw, J. A. (2006). Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration. British Journal of Management, 17(4), 263-282.
Edwards, J. R., & Bagozzi, R. P. (2000). On the nature and direction of relationships between constructs and measures. Psychological methods, 5(2), 155-174.
Louise, actually you might want to look more detailed into the differences between reflective or formative measurement, especially the respective errors related to the chosen approach and (from the conceptual starting-point) the casualty between the latent variables and its indicators. Choosing between those two approaches should not be a coin flip. Further, it is not enough to state that your measure is formative, their validity still need to be tested.
Maybe those can help you finding the more convincing argument:
Diamantopoulos, A., & Siguaw, J. A. (2006). Formative versus reflective indicators in organizational measure development: A comparison and empirical illustration. British Journal of Management, 17(4), 263-282.
Kim, S. (2011). Testing a revised measure of public service motivation: Reflective versus formative specification. Journal of Public Administration Research and Theory, 21(3), 521-546.
Coltman, T., Devinney, T. M., Midgley, D. F., & Venaik, S. (2008). Formative versus reflective measurement models: Two applications of formative measurement. Journal of Business Research, 61(12), 1250-1262.
http://www.sciencedirect.com/science/article/pii/S0148296312002147
Hi Louise,
Here are just my thoughts on your question.
You are talking about measurement theory and this theory describes how you will measure a latent variable (construct). There is always a directional relationship between the indicator variables and the construct.
Using a reflective measurement model (arrows from the latent variable to the indicator variables) suggests the assumption that the latent variable causes the measurement of the indicator variables. This means that a change in the latent variable will cause a change in the indicator variables. Furthermore, changes in the indicator variables will not cause changes in the latent variable (since it is a directional relationship).
On the other hand, using a formative measurement model (arrows from the indicator variables to the latent variable) suggests the assumption that the indicator variables cause the measurement of the latent variable. This means that a change in at least one indicator variable will cause a change in the latent variable.
The question is now, what is the inferred directional relationship between the indicator variables and the latent variable? Does a change in the latent variable causes a change in the indicator variables or vice versa?
For example, ‘the Authentic Leadership Questionnaire (ALQ) is designed to measure the components that comprise Authentic Leadership.’ (http://www.mindgarden.com/69-authentic-leadership-questionnaire). This directional relationship could well be interpreted as to be formative, since a change in ‘Self-Awareness’ will cause a change in Authentic Leadership.
References:
Coltman et al. (2008). Formative versus reflective measurement models: Two applications of formative measurement (http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1693&context=infopapers)
Hair et al. (2014). A Primer on Partial Least Squares Structural Equations Modeling (PLS-SEM
http://www.mindgarden.com/69-authentic-leadership-questionnaire
As noted in the other posts, I would suggest it is very important to understand the assumptions you are making when choosing a measurement model.
I found this paper very helpful in understanding these assumptions:
Edwards, J. R. (2011). The fallacy of formative measurement. Organizational Research Methods, 14(2), 370-388.
Morning,
It is always about the theory and a common science. Think like this: if I delete one factor, will the results still be complete? If the items are reflexive than even deleting one or more items will make sense. If a measure is formative- not.
One more comment - composite reliability is tsimilar to cronbach's alpha. So it is needed only for reflective items. You need a Variance Inflation Test and some other tests. You can find them in papers on reflective measures.
I hope it helps.
Best,
Eugene
- Badges
- Science topic