Mastering Control Variables

mastering control variables

In fields that use statistical analysis, such as business, education, health science and social science, it is common practice to include control variables. Generally, extra variables that are not the target of the investigation are added to analyses in the hope that they will control for extraneous factors and yield more accurate results. Much has been written over the past century about limitations in how control variables are used, and that blindly including extra variables does not provide more accurate results. There is a better way as I discuss in my most recent paper on mastering control variables, published in Journal of Business and Psychology, I describe a 7-step procedure for making the most of control variables.

Limitations of Control Variables

Entering control variables into a statistical analysis is a mathematical means of removing the effects of extraneous factors from results. It is based on an underlying assumption about the role that control variables have in the phenomenon being studied. Suppose you are interested in whether two variables are related, say stressful job conditions and job dissatisfaction. You might include some control variables, but this only makes sense if those control variables influence both job conditions and feelings. If they do, then controlling for them allows you to see if there is any relationship between the target variables that is not caused by the controls. If on the other hand, the control variables do not influence the target variables, but are related to them for a different reason, such as the target variables influence the controls, then including control variables will lead to an erroneous conclusion. In most cases where control variables are used, one does not know if the assumption of the control variables influencing the target variables is correct.

Mastering Control Variables

 The main problem with the use of control variables is that they are used blindly. That is, researchers add several extra variables with little concern for the underlying assumption. Often choice of control variables is based on common use in a research field, or on the fact that the controls are related to the target variables. However, there is a better way to use control variables to rule in or rule out alternative explanations for research results.

Control variable choice should never be haphazard but should be based on a detailed analysis of why target variables might be related. Rather than including controls merely to eliminate the influence of extraneous variables on results, control variables can be used to test alternative explanations. For example, we might wish to understand why we find with our employee survey that stressful job conditions are related to job dissatisfaction. One explanation is that stress leads to dissatisfaction, but there are other alternatives that we might want to test. One is that job engagement drives both perceptions of stress and satisfaction. Employees who are disengaged will find the work stressful and be dissatisfied. If engagement is added to a statistical analysis, one can test if engagement might drive both target variables.

A Systematic Approach

Control variables are best used in a programmatic manner in which a researcher first establishes that target variables are related, and then generates a series of potential explanations that involve other variables. Those other variables can be assessed and entered in analyses as control variables in order to rule them in or out as potential explanations. Those that cannot be ruled out would serve as targets for additional research to further explore their role.

Too often researchers use control variables blindly, hoping that it will produce a more rigorous and trustworthy results. Such use might be considered scientific rigor theater in that it gives the false illusion that results are more conclusive than they are. Control variables can serve a more useful role if they are carefully chosen to test pre-planned explanations for results concerning target variables.

Photo by Lukas from Pexels

Subscribe to my blog. Enter your e-mail and click SUBSCRIBE

Join 1,173 other subscribers

2 Replies to “Mastering Control Variables”

Leave a Reply

Your email address will not be published. Required fields are marked *

The reCAPTCHA verification period has expired. Please reload the page.