If the original authors say not to combine the scales, that's the starting point - though they could turn out to be incorrect. A few thoughts:

You suggest that even Kim & Leung say there is an "overall fairness" dimension. I'm not familiar with their work, but it sounds like they're saying that fairness is organized hierarchically, with a general factor at the apex and three lower-level factors, each semi-independent of the others, going to make it up. If the lower-level factors (distributive, interactional, procedural) are entirely independent of one another, then they ought not to be combined either in measurement or by the theory (i.e., there is no general factor). If they do correlate with one another, then it makes sense to try combining them. My guess is that this is the case, and Kim & Leung are mostly arguing that we gain predictive power by measuring each component as opposed to mushing them all together. Of course these are really empirical questions. Are the domains orthogonal or correlated? Do we gain by measuring three facets as opposed to one global factor?

The simplest way of answering the question is to administer the scales in question to a decent-sized sample, then examine the factor structure (a principal components analysis followed by one of the standard rotations, such as varimax). If everything loads on one factor, then that's all there is. You could also look at the internal consistency of the scales and the item-total correlations.

The longer-term, more complex way is to collect not only scale data but also data on various criterion variables, then running regression analyses. How much variance can we account for if we use all three scales? And does each contribute to the prediction? How much variance can we account for with just an overall score? Do we really make significant gains by using the three separate scales?

All this assumes, of course, that the scales in question are an adequate operationalization of the theory...

Hello Stephen

Thank you so much for your informative response. I will do as you suggested.

Best regards

Sam

Hello Dr. Davies:

Unfortunately, I cannot answer your question with a definitive “yes” or “no.” My sense is that our field continues to sort thought these psychometric issues. Pooling justice faucets into a single measure has been called “composite” justice. As you point out, composite justice is distinct from Ambrose and Schminke’s (2009) measure of overall justice, and it’s important not to conflate the two. My best guess is that overall justice would be the preferred approach, though a case could be made for each.

I would encourage you to see the following chapter, which compares composite and overall justice:

Ambrose, M. A., Wo, D. X. H., & Griffith, M. D. (2015). Overall justice: Past, present, and future. In R. S. Cropanzano & M. A. Ambrose (Eds.), The Oxford Handbook of Justice in the Workplace (pp. 109-137). Oxford, UK: Oxford University Press.

Ambrose and her colleagues (2015) provide a table on pp. 115-116 that reviews the different measures employed. You’ll find a few articles, though not many, that assessed composite justice.

In short, there are some studies that have used measures of composite justice (as opposed to overall justice). However, doing so is likely to be regarded with skepticism.

Best of luck with your work,

Russell

Hi Russell,

Thank you so much, this is exactly what I needed to proceed.

I will check out the references you kindly provided...and perhaps proceed riskily with composite justice!

Still not quite got my PhD by the way! Hopefully 2018.

All the best,

Sam

Hello

you can see these good contributions

Measuring Justice and Fairness

Multidimensional constructs in OB reseach

best

I must agree with Stephen Joy! Although, there is some theoretical reason as well...In fact a lot of things also depend upon the level of abstraction of a construct itself. If we apply the fractal network theory in understanding the psychological mechanisms, one can easily claim that big structure have underlying small but similar shape structures. So in the same manner, it can be stated that the constructs also have underlying sub structures. Now there is also a need to understand the need for underlying structure itself. They basically hold the whole structure by providing it some internal strength so it does not collapse. Thus, if you will collapse a structure back to a single structure, one idea is that it can restrict the research advancements. I must assimilate it like a nuclear explosion which can not be stopped. This could also destroy the overall theoretical existence, structure and need for a construct.

However, there may be another point of view as well. There should be no harm in it if done in a systematic manner, but that should be consensual. Particularly, there is a need to carefully examine if collapse results in something different from the original construct, as stated by Stephen as original construct developed by original author. Why? Because, at times a collapse me lead to a different construct than developed by an original one. I must assimilate it with volcano which when releases its volcanic power from inside, results in a completely different worldview.

A third perspective, these days a lot of short versions of scale are used by different researchers. The researchers combine some items of different dimensions of a construct to measure it for the ease of respondents. If by collapse, you mean same, there is somehow logic in it.

Then, the question that arises is that why not to rely on big structures and Why to collapse? My personal opinion is that when a sufficient theoretical and empirical explanation and evidence becomes available, the construct can be collapsed into a single construct but carefully by retaining its original characteristics.

Regrads