Reduce Common Method Bias

Hi Folks,

I have a question regarding my research. I am looking into the behavior/personality of CEO on the firms general Behavior. My theory in short, a risk loving CEO will make the firm more innovative. In my dataset I have responses from both CEO and General Manager of each company.
Now, the behavior/personality response from the CEO is in the CEO source only. The question now is, would it be necessary to now take the datas representing the company variable (DV) from the general Manager only (to reduce common method bias)?
I would very much like to combine both responses on my DV, as I believe that the CEO has another insight/perspective on the company than the general manager, making the variable more complete if using them combined. This however would than mean that I could not use my multiple sources to control for common method bias. I read that one should use another source for the DV than for the IV if possible.. I however don't know what to do if you expect the sources to have different insights etc. as they belong to a different group/ different power.

I'd appreciate some help :)