Since you are doing a quantitative analysis, you can create a variable to capture the stakeholder membership for each respondent, and then examine the effects of belonging to different groups.

The question is whether you have something to gain by combining the different groups. (You can compare results without combining the datasets.) You could, say, create a regression with dummy variables attached to each independent variable, but the result would be the same as if you had run each regression separately. For this to be useful, there would need to be some common factors that have the same relationships across multiple stakeholders and thus that would not require dummy variables, so that you are gaining some explanatory power by combining the data together. If you combine the data, you need to be very wary of how you interpret the results (e.g., what population you are generalizing to), and of how the different populations and samples compare in size; otherwise, the largest group might tend to dominate the results (depending on how the model is specified).

Running a regression with dummy variables (which is equivalent to what I suggested above) is not identical to running separate regressions in each subset, because that requires interactions effects. In addition, the dummy variables would provide empirical tests of the differences between the subgroups.

If every independent variable in the regression has a dummy variable applied interactively to identify the stakeholder group (as I was discussing), then you should get the same results through one combined regression or multiple separate regressions. If only a single variable is added as a new independent variable (which is what I thought David Morgan was suggesting), then you will get only an overall average measure of how the stakeholder groups differ, but you will then implicitly be assuming that the models are the same for all stakeholder groups except that they might have different starting points. That assumption seems highly likely to be untrue: e.g., because different stakeholder groups may have different priorities. At the least, I think you would need to test whether the stakeholder groups all have similar models before applying such an approach.

You would also want to think about how to make comparisons among the groups. That depends on the software you use and how you set it up. For example, dummy variables might make it easy to compare group B with group A, and group C with group A, but not group B with group C.

With any one set of dummy variables, each dummy will be compared to the reference (omitted) category, but you can always choose a different reference category to set up any comparison that you choose.