During a recent release of the revised Section 503 and VEVRAA FAQs on the OFCCP website, the agency outlined their interpretation of how non-responses to the invitation to self-identify as an individual with a disability should be treated in an analysis. As most contractors are aware, under the Section 503 regulations, each covered federal contractor is required to “invite each of its employees to voluntarily inform the contractor whether the employee believes that he or she is an individual with a disability.” In March 2014, DCI staff members wrote a blog outlining the concerns with non-responses to the disability self-identification form. DCI is adamant that it would be improper to include those who do not complete the survey as a part of the utilization analysis.
In the recent FAQ release, OFCCP indicated that non-responses should essentially be treated as “no” responses, and should be reflected in the disability utilization analysis totals. The full FAQ is outlined below:
How should non-responses to the invitation to self-identify as an individual with a disability be treated when conducting the utilization analysis?
The regulations require contractors to conduct an annual utilization analysis to determine the representation of people with disabilities in each job group, or if it has 100 or fewer employees, in its workforce as a whole. To calculate the percentage of a job group (or workforce) that is comprised of people with disabilities contractors should use the same methodology used to calculate the percentage of a job group (or workforce) that is comprised of any other specific demographic group. Specifically, contractors should compare the number of individuals identified as having a disability to the total number of employees in the job group. Non-responses should be counted solely in the job group (or workforce) total, unless the contractor has actual knowledge that a particular non-responsive individual(s) has a disability. The contractor may count as an individual with a disability any individual who it actually knows to have a disability, whether or not the individual chose to self-identify.
If OFCCP’s expectation is that non-responses should be treated as “no disability” responses when the actual disability status is unknown, it would be logical to assume that this approach is consistent with how other federal agencies approach non-response information. DCI investigated how the United States Census Bureau approached non-responses to disability status questions during their assimilation of disability data for the American Community Survey (ACS). The ACS is a mandatory, ongoing statistical survey reaching over 3.5 million households per year. It collects information on demographic, economic, social, housing, and financial characteristics. It is also the source of OFCCP’s mandated disability utilization goal of 7%.
According to a report outlining measurement of disability in the ACS, the Census Bureau handles invalid and missing disability data in the ACS by filling-in those data points with the data points they would expect, based on the general pattern of responses in the survey. More specifically, respondents who are missing disability data are allocated best guess responses based upon comparisons to similar respondents in terms of age, sex, employment status, and school enrollment status. Imputation techniques such as this are very common in survey research and serve to address some of the consequences of encountering missing data that we wrote about previously, such as “censoring” where the pattern of “missingness” is closely related to the missing variable—a phenomenon that cannot be ignored, especially when collecting disability-related information in the employment context. When imputation is not feasible or when a variable has too many missing values, researchers may instead remove missing cases or entire variables from the data. Alternatively, a researcher may consider a weighting scheme for variables according to the rate of missing data. It is important to note, however, that OFCCP does not endorse any of these approaches and will expect contractors to essentially treat disability status as a binary outcome (i.e., “yes” or “no”). We contend that this approach is flawed and may impede the ability of contractors to reasonably assess whether and where impediments exist in their workforce.
Limitations with comparing ACS Survey to employment context
It is clear that the ACS did not simply count non-responses as “not disabled” in their estimates of disabled individuals in the population. It is also important to note that the ACS, due to its association with the Census Bureau, would likely receive a higher percentage of individuals who would choose to self-identify their disability status than an employer’s voluntary self-identification form. American citizens are legally obligated to answer all of the questions as a part of the ACS, as accurately as possible, per the relevant laws at Title 18 U.S.C Section 3571 and Section 3559, which amends Title 13 U.S.C. Section 221. Conversely, the disability self-identification form offered to applicants and employees is considered voluntary and completion of it is not legally mandated. In light of this difference, it is logical to assume a higher percentage of individuals would opt not to complete the form in a voluntary employment context than individuals that chose not to identify disability status in the ACS. Depending on which metric is used, there were between 2.6%-5.4% ACS participants who did not identify disability status. Contractors would likely expect to see a higher percentage not identifying disability status in their Affirmative Action data.
By Joanna Colosimo, Senior Consultant and David Morgan, Senior Consultant, DCI Consulting Group