Abstract
Keywords
Illustrative legal case
The following narrative is abridged from an EEOC press release (EEOC, 2010):
In 2005, a Southern waste removal trucking company was charged by a former truck driver with workplace discrimination under the Americans with Disabilities Act (ADA). Specifically, he alleged that the company fired him because he was dyslexic, even though he was able to perform the essential functions of the job. Four hours after telling his supervisor that he had this impairment, he was fired because the company did not want him to see things “swirly” and have an accident. For five years, the company contended that the employee had no disability and that he was not fired for reasons of disability. An ADA was filed in 2009 and finally, in 2010 shortly before the scheduled trial date, the company admitted that:
the employee does have a ADA protected disability,
the employee was at all relevant times qualified to do his job,
the employer did dismiss him because of his disability and in violation of federal law, and
the supervisor who fired him had failed to engage in the interactive process regarding reasonable accommodation.
Under the court-ordered consent decree settling the lawsuit suit, the company was required to:
provide annual ADA training to all human resources personnel and managers,
report to the EEOC for two and a half years on all complaints of disability discrimination and all requests for reasonable accommodations,
permit the EEOC to inspect the company’s facilities to ensure compliance with federal law, and pay $95,000 to settle the suit.
The EEOC’s senior trial attorney involved in this case commented, “This is a classic case of an employer firing a worker with a disability because of its own misconceptions. Employees with disabilities such as dyslexia are every bit as protected under the ADA as those with more obvious, visible impairments such as blindness or being in a wheelchair.”
The National EEOC ADA Research Project (NEARP)
The National EEOC ADA Research Project (NEARP) is an exhaustive data-mining effort which relies upon the master database used by the Equal Employment Opportunity Commission (EEOC) to track investigations of workplace discrimination. NEARP investigators seek to develop disability or industry-specific profiles of employment discrimination, explore the contentious issues involved in workplace discrimination, document the interface of disability with other demographics, evaluate extant theories of stigma, and predict EEOC investigatory outcomes. A recent article by McMahon and McMahon (2016) describes the salient features of the NEARP database (N = 547,866 closed investigations) which themselves reveal new insights into the nature and scope of disability discrimination. The first comprehensive review of workplace discrimination concerning individuals with learning disabilities (LD) was conducted by Conway (2009). Based in part upon Conway’s seminal work, recent NEARP studies specific to LD describe characteristics of the charging parties and employers (McMahon, McMahon, West, & Conway, 2016) and the nature of LD allegations (McMahon, McMahon, West, Conway, Lemieux, 2017). The present article is focused squarely upon the matter of documenting and predicting the outcome of all reported EEOC allegations from 1992 (effective date of ADA employment provisions) through 2011.
Research questions and findings
Parametric and descriptive findings regarding merit vs. non-merit resolutions: LD vs. general disability
Every allegation of workplace discrimination investigated by EEOC results in an outcome which is either merit (favors the charging party) or without merit (favors the employer). Research questions and hypotheses stated in the null form follow directly: Research Question 1. Do differences exist in merit resolutions between allegations derived from LD charging parties in contrast to charging parties from a general disability population? Null Hypothesis 1. Compared to allegations derived from charging parties from a general disability population, LD allegations display no differences in the degree or nature of merit resolutions.
NEARP views an allegation with a non-merit resolution as perceived discrimination, and an allegation “resolved with merit” as actual discrimination as judged by EEOC, the primary ADA enforcement agency. Table 1 summarizes a comparison of Title I allegations from 1992 to 2011 derived from charging parties with LD (N = 9,480) with those of a comparison group of 313,480 allegations derived from charging parties with known impairments from a general disability population (GENDIS). Nonspecific impairments or those involving alternative prongs of the ADA definition of disability were not included in the GENDIS extraction. Nonparametric tests of proportion were performed (not reliant upon assumptions of normal distributions) and alpha levels were established at <.01.
Of the 9,480 allegations of workplace discrimination filed by charging parties with LD, 2,160 (22.8%) were resolved with merit. This rate is only slightly elevated from the 22.3% rate for the general disabilities comparison group, indicating that ADA appears to offer a consistent level of relief for persons with disabilities when it comes to EEOC investigations. When examining the subcategories of merit outcomes, charging parties with LD received settlements with benefits (e.g., as in the case study introducing this article) at a much higher rate than the comparison group (11.2% vs. 9.8%), a relative value difference of +14.3%. Charging parties with LD also experienced conciliation failure outcomes at a much lower rate than the comparison group (3.7% vs. 4.2%) a relative value difference of – 13.5%. An investigation outcome of conciliation failure indicates that the EEOC found reasonable cause for the merit resolution although the employer did not accept the recommended remedy. Nonetheless, the authors code such a finding as meritorious. The case study above is an example of a successful conciliation in which the remedy was accepted by the employer and a trial was obviated.
Conversely, the non-merit resolution category also provides some interesting insights. A total of 7,320 allegations of workplace discrimination involving charging parties with LD were closed as non-merit resolutions (77.2%). This is entirely consistent with the 77.7% rate for the GENDIS comparison group, indicating that the ADA appears to offer a consistent level of relief for employers when it comes to LD workplace discrimination investigations. When examining the subcategories of non-merit resolutions, we find that the overwhelming majority (88.2%) involve an EEOC finding of “no cause,” an outright employer victory in that evidence of discrimination was insufficient. The remaining 12.0% of non-merit closures involve a host of administrative technicalities which included only two significant differences: For allegations derived from charging parties with LD there occurred somewhat lower levels of processing problems and lower levels of allegations for which the EEOC lacks jurisdiction.
On balance, the strong similarities in investigatory outcomes are so striking that one may accurately conclude that learning disability has a somewhat typical profile when compared to all other known impairments in the EEOC database. Any history of LD as an “atypical” condition that is poorly understood or of questionable legitimacy is not confirmed by the “behavior” of the ADA implementation process.
Drivers of merit vs. non-merit resolutions: CHAID analysis
Research Question 2. What, if any, are the drivers of actual discrimination (merit resolutions) in allegations of hiring discrimination brought by Americans with disabilities? Null Hypothesis 2. There are no variables in the NEARP basis that differentiate merit vs. non-merit outcomes within group of allegations derived from charging parties with LD.
A data mining approach was used to further analyze the data in this study. Nong (2003) defines data mining as an array of techniques that are used to extract hidden predictive information from large databases. Specifically, data mining is concerned with inductive model building by the ex post facto explanation of a basic set of interrelated propositions. The explanation forms a middle range theory that becomes the basis for future hypothesis testing. Data mining has been used extensively in business for evaluating credit, predicting wages, segmenting the marketplace, and solving pattern recognition problems in healthcare (Forthover & Bryant, 2000; Ma, 2000; Melchoir et al., 2001; Nong, 2003; SPSS, 1998).
Chi Square Automatic Interaction Detection (CHAID) analysis has been used in NEARP to differentiate merit outcomes in physical vs. mental impairments (Chan, McMahon, Cheing, Rosenthal, & Bezyak, 2005), and predicting merit outcomes in such issues as sexual harassment (McMahon et al., 2006), hiring (McMahon et al., 2008a; McMahon et al., 2008b) and firing (Hurley, 2010). In this study, CHAID uses the same systematic algorithm that is utilized to search for predictors of merit and to identify and prioritize those predictors which show the most differentiation in the criterion variable: LD merit vs. LD non-merit resolution. The primary goal of CHAID is to obtain the most accurate prediction possible. However, CHAID is more flexible and does not require the distributional assumptions of traditional data mining approaches, such as discriminant or cluster analysis.
In CHAID, the degree of differentiation of predictor variables is depicted sequentially in a decision tree format to show the optimally split predictors. Thus, homogeneous groups of allegations (known as end groups) are identified in terms of their observed levels on the outcome variable. The alpha level for all statistical tests is 0.01, which corrects for the number of statistical tests within each predictor using a Bonferroni correction. Figure 1 provides the resulting CHAID decision tree for what contributes to merit outcomes for charging parties with learning disabilities.
In this analysis, many variables were available to evaluate as potential predictors of merit. These include: Charging party characteristics such as age, gender, and ethnicity; Employer characteristics such as number of employees, industry designation, and regional geographic location; Nature of the allegation such as hiring, firing disability harassment, wages, etc. (23 variations); and Timeline trends such as year of investigation closure.
This particular CHAID decision tree was the first of its kind for NEARP. As we have noted in the findings for Research Question 1, there is no statistical difference between the overall merit rate for LD and that of a general disability. It is not surprising, therefore, that only one predictor value was associated with a significant differentiation of the merit rate within LD: the age of the charging party.
Discussion and conclusion
The tree can be interpreted as follows. For merit determination in EEOC investigations, charging party age is the only level of differentiation. There are three distinct end groups. Node 1 and Node 2 are different in their age group composition, neither being continuous nor specific to older or younger age charging parties. They are similar, however, in that they are very close to the overall merit rate both for LD (22.8%) and to the overall merit rate for the aforementioned general disability population (22.3%). Node 1 and 2 merit rates serve to support the LD merit rate and render it similar to most impairment conditions.
Strikingly, the age group 18–21 shows a merit rate of 31.3%, which translates to a relative value of 1.37 times or +37% larger than the base LD merit rate. It is difficult to ignore that this same age group represents the school-to-work “transition age” population of LD Americans. It suggests that they are keenly aware of their legitimate ADA status, they are willing to file and persist in the processing of ADA allegations, and they are deemed by EEOC in the end to experience more “actual” vs. “alleged” discriminatory events. It is also possible that the other LD age groups have more difficulty establishing their ADA legitimacy as Americans with disabilities, and the older among them (60 and older) may have difficulties in articulating their “impairment x employment problems” or persisting in the investigatory process, particularly before the 2009 ADA Amendments expanded the breadth of impairments protected.
By examining the allegations filed under Title I of the EEOC using the NEARP database, this study points to several key findings. First, charging parties with LD show comparable rates of merit resolution compared to the GENDIS group. There was no statistical difference in merit outcomes between the LD and GENDIS groups, and charging parties with LD even experienced lower rates of conciliation failure and greater success in obtaining settlements with benefits. Employers receive approximately equal relief from allegations filed by charging parties with LD compared to those filed by the GENDIS group, which is demonstrated by their similar profiles of non-merit resolution rates. Finally, with regard to the second research question, age was the only factor associated with a higher probability of merit resolution (or actual discrimination) in cases filed by charging parties with LD. These findings provide an informative profile of merit vs. non-merit resolutions regarding charging parties with LD and can better inform the expectations of both charging parties and employers when it comes to avoiding and resolving such issues.
Conflict of interest
The authors have no conflict of interest to report.