Statistical Analysis of Adverse Impact: A Practitioner’S Guide

Statistical Analysis of

Adverse Impact:

A Practitioner’s Guide

STEPHANIE R. THOMAS, Ph.D.

Founder, Thomas Econometrics

AuthorHouse™

1663 Liberty Drive

Bloomington, IN 47403

www.authorhouse.com

Phone: 1-800-839-8640

© 2011 Stephanie R. Thomas, Ph.D. All rights reserved.

No part of this book may be reproduced, stored in a retrieval system, or transmitted by any means without the written permission of the author.

First published by AuthorHouse 4/25/2011

ISBN: 978-1-4567-6621-4 (sc)

ISBN 978-1-4567-6622-1 (hc)

ISBN 978-1-4567-6620-7 (e)

Library of Congress Control Number: 2011906431

Printed in the United States of America

Any people depicted in stock imagery provided by Thinkstock are models, and such images are being used for illustrative purposes only. Certain stock imagery © Thinkstock.

Because of the dynamic nature of the Internet, any web addresses or links contained in this book may have changed since publication and may no longer be valid. The views expressed in this work are solely those of the author and do not necessarily reflect the views of the publisher, and the publisher hereby disclaims any responsibility for them.

About the Author

A former New York University faculty member, Stephanie R. Thomas specializes in statistics and mathematical economics. Regarded as one of the leading experts on the analysis of equal employment opportunity issues, Dr. Thomas is a noted authority on compensation gender equity, the quantitative analysis of discrimination, and the mathematical examination of employment practices in the workplace.

For more than a decade, Dr. Thomas has provided consulting services to Fortune 500 companies, major law firms, and federal and state governments and agencies such as the Department of Justice and the FBI. She has testified as an economic and statistical expert in mediations, arbitrations, and federal and state courts throughout the United States.

Dr. Thomas frequently speaks to legal groups, industry organizations, and the media on equal employment opportunity and affirmative action compliance issues, compensation equity, and employment discrimination litigation avoidance.

Dr. Thomas is the founder of Thomas Econometrics, a full service quantitative consulting firm, and is the host of The Proactive Employer, the leading weekly broadcast dedicated to EEO compliance and employment litigation avoidance.

Contents

1. Introduction

2. The Disparate Impact Framework

3. The Basics of Statistical Inference

4. The Four-Fifths Rule

5. Statistical Testing of Disparate Impact

6. The Study Population

7. Hiring

8. Utilization Analysis

9. Promotions Analysis

10. Reduction in Force Analysis

11. Recruiting and Social Media

12. Current Employment Status as a Screening Tool

13. Employment Litigation Risk Management

1. Introduction

The idea of disparate impact in the employment context was first mentioned in 1966, shortly after the passage of the U.S. Civil Rights Act. Despite the fact that the concept of disparate impact has been highly refined and is now widely understood, employers are still faced with hundreds of disparate impact lawsuits every year. The reason for this is, I believe, twofold.

First, employers have traditionally approached discrimination prevention through training. Recent estimates indicate that companies in the United States spend nearly $300 million annually on diversity training programs. Employees, managers and supervisors spend countless hours in seminars discussing conflict resolution and communications skills, the importance of having a diverse and inclusive workplace, and so forth.

Disparate impact is not a problem that can be solved completely through diversity training or anti discrimination seminars. Disparate impact is not intentional discrimination and does not arise out of personal interactions. Because disparate impact resides in the policies and practices of the organization, conventional training for supervisors and managers will not effectively prevent disparate impact.

This is not to say that diversity and anti discrimination training is not necessary-it is very necessary, and can prevent occurrences of intentional discrimination. But anti-discrimination training on its own is not an effective prophylactic for disparate impact.

Second, employers may be unaware that one of their policies or practices is creating disparate impact until they are confronted with a disparate impact lawsuit. The presence of disparate impact is less obvious than other forms of discrimination. Disparate impact doesn’t manifest in the obvious ways that disparate treatment manifests. If employers are not examining their policies and practices with an eye toward proactive identification of potential problems, they may not learn of the disparate impact until it is too late.

Managing the risk of disparate impact litigation cannot effectively be done through traditional means. It requires a new approach built on proactive quantitative analysis.

This book was written with human resources professionals, in-house counsel, and employment lawyers in mind. The goal is to introduce the reader to statistical analysis of disparate impact through the use of concrete examples and simple calculations. In most cases, members of the intended audience will not be performing the statistical analysis of disparate impact themselves; rather, the calculations will be performed by a statistical consultant. Having a basic understanding of the statistical tools commonly used in disparate impact analysis will give human resources professionals and legal counsel a better understanding of the information their statistical consultants are providing and how to correct-and prevent-problem areas that may exist within the organization’s employment policies and practices.

The book is organized into three sections. Chapters Two through Five provide a discussion of the disparate impact framework and the various tools that are used to examine an employment policy or practice for disparate impact. Chapters Six through Ten present how those tools can be applied to various questions arising within the hiring process, workforce analysis, promotion process, and reduction in force planning. Chapters Eleven through Thirteen present a discussion of some current issues regarding disparate impact, and how to minimize the risk of employment-related litigation with proactive statistical analyses.

Chapter Two provides an overview of the disparate impact and adverse impact framework and explains how disparate impact is defined under Title VII of the Civil Rights Act and the Uniform Guidelines for Employee Selection Procedures. The seminal disparate impact court case, Griggs v. Duke Power Company, is discussed and used to illustrate the various stages of a disparate impact claim within the litigation context. Chapter Two concludes with an introduction to the process of statistical inference and statistical modeling.

In Chapter Three, the basic concepts of statistical inference are explored. In this non-technical discussion, readers are introduced to probability theory, standard deviation, and statistical significance through the use of a coin-flip example. The non-technical presentation will be of particular interest to those readers with little or no formal statistical background.

Chapter Four focuses on the most commonly used tool for examining disparate impact: the Four-Fifths Rule (also known as the 80% Rule). The origin of the Four-Fifths Rule is presented, along with simple example calculations. Advantages and disadvantages of using the Four-Fifths Rule to study questions of disparate impact are also discussed.

Chapter Five focuses on formal statistical tools commonly used in disparate impact analyses. The chapter begins with a discussion of shortfall and surplus measures in positive and negative selection processes, and a non-technical presentation of hypothesis testing. The remainder of the chapter provides a technical discussion of various test statistics drawn from the binomial, chi square and hypergeometric distributions. The chapter concludes with a non-technical discussion of the limitations of formal statistical testing. If the reader is not interested in the technical presentation of the various test statistics, the middle portion of this chapter can be omitted.

Chapter Six highlights the importance of a properly defined study population. Construction of a study population based on applicant flow, bidding data, incumbency, and the external labor market are discussed.

Chapter Seven provides a discussion of the types of disparate impact claims made surrounding the hiring process. Pre-employment testing is discussed, and examples of aggregated statistical analyses, multiple pools analyses, event analyses, statistical combination and stage analyses are presented.

In Chapter Eight, an overview of utilization analysis is presented. While a utilization analysis does not address a specific employment policy or practice in isolation, many of the techniques involved in conducting a utilization analysis are very similar to the techniques used in examining questions of disparate impact. For this reason, a brief discussion of utilization analysis is presented.

Chapter Nine presents a discussion of promotions and the ways in which the study population can be constructed for examining disparate impact in the promotions context. Examples of aggregated statistical analyses, multiple pools analyses, event analyses, and statistical combination are presented.

Chapter Ten focuses on reductions in force. The legal concerns associated with a reduction in force are briefly discussed, and an overview of the Age Discrimination in Employment Act is provided. Examples of aggregated statistical analyses, multiple pools analyses, and statistical combination are presented. A discussion of the various definitions of age-protected, and how those definitions can influence a statistical analysis, is also provided. The chapter concludes with a discussion of the importance of not only performing a formal statistical analysis but also examining the raw numbers of the reduction in force.

No discussion of disparate impact in employment decisions would be complete without mentioning recruiting and social media. Chapter Eleven outlines the advantages of using social media in the recruiting process, as well as how demographic differences between labor markets and typical social media users may create the potential for disparate impact.

Chapter Twelve discusses what some believe to be a new trend in recruiting and hiring-the use of current employment status as a screening tool. While there is some controversy over how widespread this practice is, regulatory agencies have taken notice of this issue. Chapter Twelve provides the highlights of a public meeting held by the Equal Employment Opportunity Commission to examine the exclusion of unemployed candidates from applicant pools and how this practice may have a disparate impact on women, racial and ethnic minorities, and individuals with disabilities.

The final chapter provides a discussion of how the various tools for examining questions of disparate impact can be used in litigation avoidance. Chapter Thirteen highlights the importance of managing the risk of employment litigation, and provides suggestions on how proactive statistical analyses can be used as part of an organization’s overall risk management strategy.

Note that the author is an economic and statistical consultant, not an attorney; the information presented herein should not be construed as legal advice.

2. The Disparate Impact Framework

There are two main legal theories of discrimination: disparate treatment and disparate impact. Disparate treatment is what we typically think of when we think about discrimination. Disparate treatment occurs when a decision-maker intentionally treats members of one group differently from members of another group. Examples of disparate treatment include intentionally denying employees over the age of forty promotional opportunities because of their age, refusing to consider applications of job candidates because of those candidates‘ religious beliefs, or paying women less than men because of their gender.

Theory of Disparate Impact

The second theory, disparate impact, focuses on the criteria used by decision-makers. Unlike disparate treatment, the intent of the decision-maker is irrelevant. In some respects,