formative or reflective?

Research topics can be discussed in this area.
Post Reply
Jiraporn
PLS Junior User
Posts: 3
Joined: Thu Oct 04, 2012 8:06 pm
Real name and title:

formative or reflective?

Post by Jiraporn » Mon Feb 25, 2013 11:52 pm

Hi,
I'm a new user of SmartPLS. I have read through reflective and formative measurement model. Base on the decision rules from Javis et al., (2003), two constructs of my research should be explained as "formative" constructs. Because indicators defines different characteristics of the constructs. However, some previous literature were analyse them in a reflective way.
Are my indicators now formative or reflective?
Construct A:
Compared to competitors, our firm with supply chain partner....
…offer a variety of products/services efficiently.
… innovate in products/services which are offered.
…offer high quality products/services to our customers
…are be able to offer lower prices.

Construct B:
Since our firm engaged with its supply chain partners, the firm…
...has improved its sale growth.
...has improved its return on investment (ROI).
...has improved its profit margin on sales.
...has improved its overall competitive position.
...has improved its overall customer service levels.

Thank you in advance

User avatar
Hengkov
PLS Super-Expert
Posts: 1619
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov » Thu Feb 28, 2013 10:08 pm

Hi,

Check correlation among indicators. If not high, your construct is Mode B (formative).

Regards,
Hengky

Jiraporn
PLS Junior User
Posts: 3
Joined: Thu Oct 04, 2012 8:06 pm
Real name and title:

Post by Jiraporn » Fri Mar 01, 2013 11:47 am

Thank you very much Hengky.

Twan
PLS User
Posts: 16
Joined: Sat Apr 06, 2013 5:29 pm
Real name and title:

Post by Twan » Sun Apr 07, 2013 8:57 pm

Hengkov wrote:Hi,

Check correlation among indicators. If not high, your construct is Mode B (formative).

Regards,
Hengky
I don't think you're right. If you have formative items, and the correlation among those items is high, it doesn't mean that they are reflective. In my opinion there are 2 options:
1. Items are formed in bad way
2. Answers on those items are just simple high, for example

Image

F1. I have a balanced diet
F2. I exercise regularly
F3. I get suffiecent sleep each night.

That you have a balanced diet doesn't mean that you exercise regularly. However, to be healthy you should have a balanced diet and your should exercise regularly. If you send a survey among fitness (or body builders) persons the chance is high that F1. F2. and F3. are highly correlated. However, they still are formative items. Measuring those items reflective is a bad choice.

When you send this survey to people that don't have a job. The change is high that F3 doesn't correlate with F1. They have enough time to sleep, but don't have the money to buy a balanced diet. So the correlation is low, does that mean that the items are formative? No, they are formative because a balanced diet isn't the same as exercising regularly.

Hence, think twice before you considering your items reflective. I've read a lot of articles were researchers just used a reflective validation method. But when I asked why they used a reflective validation method, because their items are just not reflective to me, they didn't answer anymore, or they gave a very vague answer.
Last edited by Twan on Mon Apr 08, 2013 9:59 am, edited 1 time in total.

User avatar
Hengkov
PLS Super-Expert
Posts: 1619
Joined: Sun Apr 24, 2011 10:13 am
Real name and title: Hengky Latan
Location: AMQ, Indonesia
Contact:

Post by Hengkov » Mon Apr 08, 2013 3:38 am

A rejoinder to Twan Peters:

I surprised and very glad with your comment above and some opinion about choice reflective or formative indicators. This is a important and special issue in some journal since four years ago. Until now, our discussion of construct and the measurement model has assumed reflective measurement theory because formative measurement have conceptually ambiguous (Edwards, 2011, p. 373). For health example above you can see that. Coltman et al. (2008), Jarvis et al. (2003), Mackenzie 2003, MacKenzie et al. (2005) and other agree formative model not have “common cause” for items in the construct, so among items uncorrelated (see detail explain some article above). This might be the case where a composite latent construct (formative model) is represented by mutually exclusive types of behavior (Jarvis et al. 2003, p. 202). If the items in formative model have high correlation, it’s indicate potential vertical collinearity (see Kock and Lynn 2012, p. 549). Nature of collinearity first introduce by Ragnar Frisch (Herman Wold Advisor together with Harlad Cramer). Collinearity problem made estimate unstable with large standar error (see Cohen et al. 2003, p. 420). PLS regression develop for handle it and also recent procedure for handle formative model using CTA-PLS suggest by Gudergan et al. (2008, p. 1241).

References:
Cohen, J., Cohen, P., West, S. G., and Aiken, L. S. 2003. Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Coltman, T., Devinney, T. M., Midgley, D. F., and Venaik, S. 2008. “Formative versus reflective measurement models: two applications of formative measurement,” Journal of Business Research (61:12), 1250-1262.
Edwards, J. R. 2011. “The Fallacy of Formative Measurement,” Organizational Research Methods (14:2), pp. 370-388.
Gudergan, S.P., Ringle, C.M., Wende, S., and Will, A. 2008. “Confirmatory Tetrad Analysis in PLS Path Modeling,” Jounal of Business Research (61:1), pp. 1238-1249.
Jarvis, C. B., Mackenzie, S. B.,and Podsakoff, P. M. 2003. “A critical review of construct indicators and measurement model misspecification in marketing and consumer research,” Journal of Consumer Research, (30:2), pp. 199–218.
Kock, N., and Lynn, G.S. 2012. “Lateral Collinearity and Misleading Results in Variance-Based SEM: An Illustration and Recommendations,” Journal of the Association for Information Systems (13:7), pp. 546-580.
MacKenzie, S.B. 2003. “The Dengers of Poor Construct Conceptualization,” Journal of the Academy of Marketing Science (31:3), pp. 323-326.
MacKenzie, S.B., Podsakoff, P.M., and Jarvis, C.B. 2005. “The Problem of Measurement Model Misspecification in Behavioral and Organizational Research and Some Recommended Solutions,” Journal of Applied Psychology (90:4), pp. 710-730.

Twan
PLS User
Posts: 16
Joined: Sat Apr 06, 2013 5:29 pm
Real name and title:

Post by Twan » Mon Apr 08, 2013 9:56 am

Hengkov wrote:A rejoinder to Twan Peters:

I surprised and very glad with your comment above and some opinion about choice reflective or formative indicators. This is a important and special issue in some journal since four years ago. Until now, our discussion of construct and the measurement model has assumed reflective measurement theory because formative measurement have conceptually ambiguous (Edwards, 2011, p. 373). For health example above you can see that. Coltman et al. (2008), Jarvis et al. (2003), Mackenzie 2003, MacKenzie et al. (2005) and other agree formative model not have “common cause” for items in the construct, so among items uncorrelated (see detail explain some article above). This might be the case where a composite latent construct (formative model) is represented by mutually exclusive types of behavior (Jarvis et al. 2003, p. 202). If the items in formative model have high correlation, it’s indicate potential vertical collinearity (see Kock and Lynn 2012, p. 549). Nature of collinearity first introduce by Ragnar Frisch (Herman Wold Advisor together with Harlad Cramer). Collinearity problem made estimate unstable with large standar error (see Cohen et al. 2003, p. 420). PLS regression develop for handle it and also recent procedure for handle formative model using CTA-PLS suggest by Gudergan et al. (2008, p. 1241).

References:
Cohen, J., Cohen, P., West, S. G., and Aiken, L. S. 2003. Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Hillsdale, New Jersey: Lawrence Erlbaum Associates.
Coltman, T., Devinney, T. M., Midgley, D. F., and Venaik, S. 2008. “Formative versus reflective measurement models: two applications of formative measurement,” Journal of Business Research (61:12), 1250-1262.
Edwards, J. R. 2011. “The Fallacy of Formative Measurement,” Organizational Research Methods (14:2), pp. 370-388.
Gudergan, S.P., Ringle, C.M., Wende, S., and Will, A. 2008. “Confirmatory Tetrad Analysis in PLS Path Modeling,” Jounal of Business Research (61:1), pp. 1238-1249.
Jarvis, C. B., Mackenzie, S. B.,and Podsakoff, P. M. 2003. “A critical review of construct indicators and measurement model misspecification in marketing and consumer research,” Journal of Consumer Research, (30:2), pp. 199–218.
Kock, N., and Lynn, G.S. 2012. “Lateral Collinearity and Misleading Results in Variance-Based SEM: An Illustration and Recommendations,” Journal of the Association for Information Systems (13:7), pp. 546-580.
MacKenzie, S.B. 2003. “The Dengers of Poor Construct Conceptualization,” Journal of the Academy of Marketing Science (31:3), pp. 323-326.
MacKenzie, S.B., Podsakoff, P.M., and Jarvis, C.B. 2005. “The Problem of Measurement Model Misspecification in Behavioral and Organizational Research and Some Recommended Solutions,” Journal of Applied Psychology (90:4), pp. 710-730.
Thanks, what a great reply :)
Indeed, unfortunally, at the present moment there aren't usable and acceptable validity and reliability measures for a formative construct. Using collinearity isn't enough. Edwards (2011) even argues that you shouldn't use formative measures at all.

Hence, I think that the main conclusion is; just don't use formative items. It makes the results of your research doubtful. And it also attracts the attention from your results to your validation methods. Hence, conducting a research, and in the end hearing that your conclusions are very doubtfull is a nightmare to every researcher.

@Hengkov, I assume that you are very interested in this subject, as I'm. I'm interested because currently I'm writing my master thesis, and I've used in my researchmodel a construct + it's measures from another study who used reflective validation methods. However, I'm doubting if using AVE was the right choice for that construct, because the used items are very suspicious for me.

If you're interested, I can send this recently published study to you. I'm very curious what your thoughts are about his model. I'm doubting to report that study, but that will in the end destroy some literature background of my study :).

srahagloria
PLS Junior User
Posts: 1
Joined: Sat Aug 30, 2014 4:50 am
Real name and title: Gloria Sraha, PhD Candidate Victoria University, School of Marketing and International Business. 23 Lambton Quay, Rutherford House Wellington 6011

Re: formative or reflective?

Post by srahagloria » Sat Aug 30, 2014 5:23 am

Hello,
This is a very interesting discussion on formative and reflective construsts. I am also a new PLS user. I have the same dilema on whether to use reflective or formative measurements for my structural model as Hair (2013) indicated there are no particular criteria but it depends on the researcher's discretion.I have the same problem with whether my constructs (Export Marketing Strategy 4Ps) are formative or reflective. When l check the sources I adapted the consructs they used reflective measurements. Is this justified? Any suggestions will be welome please.

Thank you.

Post Reply