Seemingly simple questions: determining the equity of pay using advanced modeling
Publisher
University of Tennessee at Chattanooga
Place of Publication
Chattanooga (Tenn.)
Abstract
Determining whether employees are paid equitably ought to be straightforward. We would like to simply use a t-test to determine, for example, if women are paid less than men. In practice however, when large organizations employ many individuals whom perform a diverse array of jobs, with each individual bringing genuinely unique qualifications to their job, it can be difficult to discern whether employee pay is determined by demographic traits (e.g., ethnicity), or can instead be explained by other factors, such as experience. To assess which factors account for variability in pay, we have leveraged a regression-based model and because employees are naturally nested within an organizational hierarchy, we used hierarchical-modeling techniques. This allowed us to better capture the context surrounding employees’ pay. Finally, we have abandoned null-hypothesis-significance testing in favor of a Bayesian approach to thinking about information. This final step introduced several key benefits, including allowing us to ask probabilistic questions of the model (e.g., how likely is it that women in job X are underpaid?), providing us a method for supporting a lack of equity issues where they legitimately do not exist (i.e., null effects), and forcing us to think critically and transparently about our a-priori beliefs.
Date
October 2019
Subject
Industrial and organizational psychology
Document Type
presentations
Language
English
Rights
http://rightsstatements.org/vocab/InC/1.0/
License
http://creativecommons.org/licenses/by/4.0/
Seemingly simple questions: determining the equity of pay using advanced modeling
Determining whether employees are paid equitably ought to be straightforward. We would like to simply use a t-test to determine, for example, if women are paid less than men. In practice however, when large organizations employ many individuals whom perform a diverse array of jobs, with each individual bringing genuinely unique qualifications to their job, it can be difficult to discern whether employee pay is determined by demographic traits (e.g., ethnicity), or can instead be explained by other factors, such as experience. To assess which factors account for variability in pay, we have leveraged a regression-based model and because employees are naturally nested within an organizational hierarchy, we used hierarchical-modeling techniques. This allowed us to better capture the context surrounding employees’ pay. Finally, we have abandoned null-hypothesis-significance testing in favor of a Bayesian approach to thinking about information. This final step introduced several key benefits, including allowing us to ask probabilistic questions of the model (e.g., how likely is it that women in job X are underpaid?), providing us a method for supporting a lack of equity issues where they legitimately do not exist (i.e., null effects), and forcing us to think critically and transparently about our a-priori beliefs.
Department
University of Tennessee at Chattanooga. Dept. of Psychology