Which control variables should I include?

In order to reduce any bias in observational studies or surveys, control variables are often included as independent variables in the evaluation in order to estimate the effect of the predictor(s) of interest more "cleanly". In practice, however, the selection of these variables is often quite arbitrary or demographic variables such as age, gender or education are used reflexively, which are relevant for the sample description but for which it is not clear how and why they are related to the central variables. However, this is not good practice. Instead, any control variables should be selected in a theoretically sound manner and justified accordingly. Together with the fact that surveys should generally be as short as possible (see also: How long should my questionnaire be? (FAQ)), the following applies to the inclusion of control variables in your survey: "When in doubt, leave them out". However, the first part of the sentence is also relevant: if you simply leave out an "undoubted" (i.e. obviously important and/or identified as relevant in many relevant studies) control variable in your survey, you also need a good reason for doing so.

Literature

Becker, Thomas E.; Atinc, Guclu; Breaugh, James A.; Carlson, Kevin D.; Edwards, Jeffrey R.; Spector, Paul E.,2016. Statistical control in correlational studies: 10 essential recommendations for organizational researchers. Journal of Organizational Behavior, 37: 157-167