It's worth reading this short paper: Jamieson, S., 2004. Likert scales: how to (ab)use them. Medical Education, 38, pp.1217-1218. It is argued by some people that calculating means and standard deviations for Likert scale data is invalid, because the intervals between the points on the scale cannot be assumed to be equal. However, others take the opposite view!

I think the paper Brian shared is a good one to read. I could perhaps offer some advice, but it depends on what your data looks like and what you are trying to do with it (I'm not 100% sure I follow).

Are you definitely going to use Likert-scale questions? If so, I think you have to look at your data before making decisions about interpretation, but you can mostly just report the results and interpret them subjectively can't you? For example, if there you are using a Likert-scale question 1 (very dissatisfied) ~ 5 (very satisfied) and 80% of your sample selected "5" it would make the most sense to report this result in this way and interpret it to mean that most people said they were satisfied with their job.

If you are asking about how to compare responses amongst sub-groups, then there are various options, but again, it will depend on your data. In the example above, if 80% selected "5" and 18% selected "1" or "2" in one subgroup and 80% selected "1" and 18% selected "5" in another subgroup, I think it would make more sense to convert this into a 2-point scale: generally satisfied and generally dissatisfied.

If you are asking about using multiple Likert-scale questions that ask about job satisfaction and then asking how to combine them all into one measurement of "general job satisfaction" then that is a completely separate issue. In that case, you should probably check with the creator of that particular survey/metric and see how they weight the questions. However, you could also consider converting the responses into overall Z-scores, which would probably allow you to add/subtract them into a single new score much more easily.

In part, I think that the answer to your question depends on your research questions. Depending on your question(s) one or the other analysis method may answer the question better.

I general terms, I like your choice A because it provides more categories and thus more nuances to your data. But I also like looking at the standard deviation as an indication of how widely the answers ranged above and below average, and therefore how much agreement there was in the answers.

I agree it is your study’s objectives that drive the analysis and interpretation of the data of your instruments. That being said, the scales you have pointed out in your query (i.e., Generic Job Satisfaction Scale, Work Identity Scale, and Passion Scale) were originally developed and validated using 5-point Likert or Likert-type scales, employing different response options.

Since the response options for the mentioned scales are not the same—not all of them agreement-oriented—interpreting each scale relies on how you and your respondents construe the midpoint. For informative insights, you could refer to Chyung et al. (2017) and Nadler et al. (2015). Both articles are referenced below.

Chyung, S. Y. Y., Roberts, K., Swanson, I., & Hankinson, A. (2017). Evidence-based survey design: The use of a midpoint on the Likert scale. Performance Improvement, 56(10), 15–23. https://doi.org/10.1002/pfi.21727

Nadler, J. T., Weston, R., & Voyles, E. C. (2015). Stuck in the middle: The use and interpretation of mid-points in items on questionnaires. The Journal of General Psychology, 142(2), 71–89. https://doi.org/10.1080/00221309.2014.994590

Good luck,

Thank you for all your responses! I appreciate that you took the time to answer my question.