Page 1 of 1

Endogeneous Formative Constructs

Posted: Sat Aug 08, 2015 10:46 am
by RaffaeleP
Hi there

I have a general questions about formative constructs, especially endogeneous ones!

Is it critical to use formative constructs at the endogeneous level?
I recently read about general problems with endogeneous formative variables.
Has anybody heard about that and if there is a "real" problem, is there any feasible solution to that?

Should I perform the analysis as described in this text?: ... p14005.pdf

For a better understanding I attached my research model.
Basically, I have 2 endogeneous (formative) variables (Diversification and Risk) but I'd like to measure even the influence of Diversification on risk.
The measurment of the remaining constructs is mixed, some are formative, some are refelctive.

Thank you very much
Best regards

Re: Endogeneous Formative Constructs

Posted: Wed Dec 09, 2015 7:30 pm
by Beraza
Hi Raffaele,

I have just seen your message and your model. Have you found a solution for your model? I have a model with 3 endogenous formative constructs and I have been recommended to read the same article. The problem is that I didn´t include reflective items in my cuestionaire. Is it absolutly necessary to include reflective items to get a solution to this kind of models?

Best reggards,

Ana Beraza

Re: Endogeneous Formative Constructs

Posted: Mon Dec 21, 2015 8:27 am
by MeiPeng
Dear Raffaele,

Have you found an answer for your question? I noticed that it was posted in Aug 2015, and now is Dec 2015.

Based on my understanding of reading the Book and some workshops I attended, it is okay to have formative endogenous construct. Just that the model assessment will be different. That's also mean, you cant run "Predictive relevance Q square".

In my point of view, your model is fine.

mei peng

Re: Endogeneous Formative Constructs

Posted: Sun Apr 10, 2016 11:04 am
by RaffaeleP
Hi there,

this question isn't resolved yet. There is no comment in literture on the publication of Temme et al.
And I got no answer from the authors regarding this point.

But I think that they are wrong with their "absolute" reccomendation of CSA because in my opinion it is possible (and quite easy)
to use a second order construct to circumvent the addressed problem.

They say:
"PLS requires at least one (additional) reflective indicator if an FMC is endogenous"
but when I use second order constructs properly it should be no problem.
This idea is also mentioned in "Improper use of endogenous formative variables" Journal of Business Research 66 (2013) 233–241
"A Critical Review of Construct Indicators and Measurement Model Misspecification in Marketing and Consumer Research" JOURNAL OF CONSUMER RESEARCH, Inc. ● Vol. 30 ● September 2003

And when you have no reflective items it is still no problem I think, because if you measure every single antecendent construct with a single item
it should work.