Monday, Aug 4: 2:00 PM - 3:50 PM
0768
Topic-Contributed Paper Session
Music City Center
Room: CC-106A
salary equity studies
Applied
Yes
Main Sponsor
Section on Teaching of Statistics in the Health Sciences
Co Sponsors
Caucus for Women in Statistics
History of Statistics Interest Group
Section on Statistics and Data Science Education
Presentations
In the adversarial justice system of the United States, it is often left to the courts to resolve disputes, in particular to determine whether the results of statistical models support or fail to support the legality of the practices that led to the models. Does a model demonstrate discrimination, legitimate differences, a random process? Can this be resolved by competing expert testimony or by reference to previous legal decisions? The heart of statistics is probability, certainty that of the law.
Keywords
judicial remedies
legal impact of models
Many U.S. colleges and universities conduct faculty salary-equity studies, using institutional data to identify and remediate pay disparities for women and other disadvantage groups. This session will discuss the unique challenges of managing data for such studies, including access to identity variables, coverage for variables of interest, and the types of systems used to manage faculty information in colleges and universities. Finally, the presenter will discuss how the American Association of University Professors can support college and university administrators, faculty members, and collective bargaining units engaged in salary-equity studies.
Speaker
Glenn Colby, American Association of University Professors
Faculty salary equity studies are routinely conducted to examine whether there are significant differences (statistical and/or practical) associated with gender and race/ethnicity at the institutional level or at finer scales (divisions, departments or other units/categories). There are a number of statistical issues that we highlight in planning such studies from data availability and quality, choice of relevant populations to aspects regarding the choice of models such as which variables to include, transformations of variables, interactions, and treatment of outliers, among others.
Modern approaches to salary equity studies, such as proposed by Billard (2017), frequently use statistical ANCOVA-like models to quantify salary differences between and within subgroups of employees of interest. In this talk, we look at some of the seminal research by Scott (1977) and promoted extensively by Haignere (2002) under the name of ``white male only model.'' We show that it too fits within the ANCOVA framework, thereby facilitating understanding of the statistical properties of the ``white male only model'' approach and enabling comparisons with other modeling approaches. In particular, we show that inference for the WMO model fits within the ANCOVA linear model thereby rendering the calculation of standard errors, confidence intervals, and hypothesis tests immediate. We will also discuss statistical power for detecting group differences. Through simulations and examples, we provide guidance on which modeling techniques to use under different scenarios.
When inequities are found in salary structures, it behooves institutions to implement corrective actions. It is the law. How to make the necessary corrections however is not as easy as it might at first appear. There are several approaches that could be adopted, some less satisfactory than others. This talk considers options and their implementation, taking into consideration what the courts dictate along with scientific dictates.