Dear Vivek,

According to Hair et al. (2010), factor loading estimates should be higher than 0.5, and ideally,  0.7 or higher. So, there is no any obligation that your loading estimates must higher than 0.7. Actually, I think the statement "Hair et al. (2010) require that each item is considered a satisfactory item when item loadings are greater than 0.70. " is not correct. Thus, I think you're right. Send your manuscript to another journal.

Thank you for your suggestions Ali Gohary 

I agree. Factor loadings appear to be decent. Unfortunately it has gone to a wrong reviewer. This happens very frequently. Just target another journal. 

Best wishes 

Hi Mr. Ali Gohary.. can you please give full citation. tq

Hi Dicky,

Here is the reference:

Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., Tatham, R.L., 2009. Multivariate data analysis. (7th, Ed.) (Vol. 7). Pearson Prentice Hall Upper Saddle River, NJ.

Hi colleagues,

To me Hair et al. (2009) are not wrong, but I agree that your manuscript has gone to an inappropriate reviewer (this happens frequently and that's academic life!) . Hair et al. discuss the ideal case for each factor loading, i.e., the common variance should be greater than the unique one (lambda >= 0.72 in order to have a variance >= 0.50), but mainly for the average; that why we use average variance attracted (AVE >=50%). In some cases, especially a new measure, lambda >= 0.5 (AVE > 0.25) can be considered to be acceptable (but we have to address the limitation of this low AVE measure).

THis is very useful thanks

Dear all,

from Dr. Nguyen Tho's comments :

[ In some cases, especially a new measure, lambda >= 0.5 (AVE > 0.25) can be considered to be acceptable (but we have to address the limitation of this low AVE measure).]

I need to know what are the limitations of the low AVE measure (AVE > 0.25) can be acceptable.

Thanks all

what I understand that Factor loading score below 0.70 can be considered if its AVE is above 0.50 like if factor loading is 0.62 but AVE is 0.50 then you are safe to consider this item

Thank you Imran sir

Additionally composite reliability may also be considered.

Rightly said by my fellow researchers Standardized factor loading above 0.5 is totally acceptable..

Rhytheema Dulloo Phd can you give some references to support your argument?

My comment: It is reasonable for requesting a factor loading of above 0.60. When your factor loading is only around 0.5-0.6 or below than 0.5, you calculate the AVE for validity ( AVE: sum up each squared standardized factor loading, divide it by the number of indicators) will resulting in a non-valid conclusion for your construct, as the minimum cut-off value for AVE is 0.50

@ Mohammed Shuhaiber could you please introduce a reference? Through AMOS, I have done SEM. Although there is an acceptable model fit, I am not sure about the correlations.

Mohammed Shuhaiber can you give a reference please?

Dear Vivek Verma ,

There are more than 100 references and literature that support that a factor loading below 0.70 is accepted and satisfies the convergent validity of a latent construct. However, it requires some assumptions to fulfill. The theoretical explanation to have a factor loading above 0.709 is that each item should explain at least 50% variance (i.e., square of 0.709) of a latent variable. That is why many researchers without having a detailed understanding are stringent on these criteria and consider this cut-off value blindly. However, scholars suggest retaining the item having a factor loading above 0.40 in confirmatory factor analysis (CFA), when -

1. An item with a factor loading above 0.40 can be retained if the average variance explained (AVE) is 0.50 or above. Each construct must fulfill this criterion as in a reflective measure scale, on average all items of a construct should explain at least 50% of the variance. If any item individually contributes less variance (less than 0.50) but other items contribute high variance to the construct, this will lead to an AVE more than 0.50.

2. An item with low factor loading can be deleted as long as it increases AVE at least 0.50. One may raise a question - how many items can be deleted from a contract during CFA? The answer is conditional. We cannot delete any item from a latent variable if it has only two items. deleting one item from two will change the property of a latent variable. Even some items may offer very useful information about the construct which may demand to retain in analysis.

3. Also look at composite reliability if it is above 0.70.

Few key references:

1. Hair Jr, J. F., Hult, G. T. M., Ringle, C. M., & Sarstedt, M. (2021). A primer on partial least squares structural equation modeling (PLS-SEM). Sage publications.

2. Hair, J. F., Risher, J. J., Sarstedt, M., & Ringle, C. M. (2019). When to use and how to report the results of PLS-SEM. European business review, 31(1), 2-24.

3. Wieland, A., Durach, C. F., Kembro, J., & Treiblmaier, H. (2017). Statistical and judgmental criteria for scale purification. Supply Chain Management: An International Journal, 22(4), 321-328.

4. Ali, F., Rasoolimanesh, S. M., Sarstedt, M., Ringle, C. M., & Ryu, K. (2018). An assessment of the use of partial least squares structural equation modeling (PLS-SEM) in hospitality research. International Journal of Contemporary Hospitality Management.

Hope, this clarifies your concern well.

M Sarker

It is suggested that indicator loadings should be in the region of .70 and statistically significant at .05 or below (equal to t-statistic of ± 1.96). However, values between .40 and .70 can be justified ( Hair et al., 2019) if acceptable values are obtained on other indices (Internal consistency reliability, AVE, and HTMT) (Hair et al., 2019; Hair and Alamar, 2022).