Dummy variable or multilevel analysis?

[Gustav Egede Hansen]

Hi everybody
I am estimating the impact of outsourcing on employees’ job engagement moderated by occupational group. Here is my current model:

reg engagement i.outsourced##i.occupation controlvariables, vce(robust)
(outsourced: 0 = public and 1 = outsourced, occupation: 0 = occupation_A and 1 occupation_B)

This seems intuitive to me, but I wonder whether I need to do a multilevel analysis or in any other way account for the data being “organized” at occupational level. However, from my understanding, I cannot conduct a multilevel analysis as I only have two occupations (rules of thump state minimum that at least 15 groups at level-2 are required). Likewise, I cannot cluster occupation, again, due to the low number of clusters. But I am also curious if I even need a multilevel model with, e.g., random intercepts. From my understanding, the dummy for occupation provides two different estimates (and thus, in a way, two different intercepts). So, my question is: is the specified model correctly or can I account for the variation shared on level-2 (through multilevel analysis, clustering, or a third technique)?

Best
Gustav

[Clyde Schechter]

The specified model is appropriate for your situation. A random-effects model with only two value of the variable defining the higher level is inadvisable.

[Gustav Egede Hansen]

Thanks Clyde for the quick and precise answer!

[Gustav Egede Hansen]

Clyde Schechter, I have a short follow-up question if you don’t mind. Would it be correct to call the specified model a least squares dummy variable model (as it includes dummies for outsourcing and occupational groups) or is it only when conducting panel data analysis that this term applies?

[Clyde Schechter]

That would be accurate. Personally, I do not like the term "dummy variable"; I prefer to call them "indicators." But given that the term "dummy variable" is widely understood, and the term "indicator" can also be used to mean something different in the SEM context, I wouldn't press that upon you.

[Gustav Egede Hansen]

Perfect! Once again thank you.