Dr. John Antonakis is Professor of Organizational Behavior in the Faculty of Business and Economics of the University of Lausanne, Switzerland. His research is currently focused on the psychology of leadership including: Charisma, predictors of leadership (e.g., personality, intelligence), leader development, as well as on research methods. He has published articles in many prestigious journals including Science, Nature Human Behavior, Psychological Science, Academy of Management Journal, Journal of Management, Journal of Applied Psychology, Journal of Operations Management, Management Science, Entrepreneurship Theory and Practice, and Harvard Business Review, among many others; he has also published four books, and dozens of book chapters and conferences proceedings. He has received more than US$2.45 million in funding for his research. He teaches organizational behavior (Bachelor level), leadership (Master level), and causal analysis and statistics (Ph.D. level). Prof. Antonakis is former Editor in Chief of The Leadership Quarterly. He was previously associate editor for The Leadership Quarterly and Organizational Research Methods, and has served too on the editorial boards of several top journals in management and applied psychology. In 2019 he was named a highly cited scientist by Clarivate-Web of Science group. In 2020, 2021, and 2022 he was listed by PLOS Biology in the world’s top 2% of scientists for career-long impact.
In leadership, or other areas of management and applied psychology, researchers often obtain ratings behaviors of target individuals using questionnaires. I will show why variation in questionnaire ratings of leadership (“x”)—or of any other unit of study for that matter—should not be used as a gauge of variation of actual behavior. What x measures is not just leader behavior but much more. It is an evaluative judgment, which is influenced by many omitted causes at the target, the ratee, and the context in which the rating occurs. For example, suppose we wish to measure leader charisma using questionnaires. At the target level, it is possible that smarter, more extraverted, and more attractive individuals (let’s call these variables S) rated as having more charisma. At the rater level, mood, rater personality, or other idiosyncratic factors (let’s call these variables Z) may influence how the target is rated. Raters can also be biased by performance signals; their knowledge of how effective the leader is or how well the organization is going (let’s call the P) may bias how charismatic they see the target to be. The problem is this: Are S, Z, and P measured and put alongside x in the regression model? What if S, Z, and P also cause y? If these omitted variables are not measured and included in the model, then we may find a significant correlation between x and y; however, this correlation may be driven by S, Z, or P. Failure to account for omitted causes misleads theory and practice. It may be that x causes nothing and that it is S, Z, or P that are driving both x and y. We will discuss the implications of this problem and how it affects other areas of research in management and applied psychology.