Disability Utilization Goals Gone Wild

Beginning March 24, 2014, federal contractors will be required to collect disability information from their workforce on an annual basis and use those data to conduct disability utilization analyses. Under the new 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.” As the regulations specifically use the word “invite,” contractors are required only to extend an invitation to an employee, and employees cannot be forced to complete the self-identification form. Therefore, it is highly likely that contractors will have disability status information on only a subset of the workforce. Unlike utilization analysis conducted under EO 11246 in which a contractor has the race/ethnicity and sex of all of its employees (due to a requirement to visually identify race/ethnicity and sex of those who do not disclose such information), utilization analysis under Section 503 will be conducted on partial workforce data.

A further wrinkle pertains to the response options that may be chosen if an individual chooses to complete the self-identification. Instead of having only two options (yes/no), those choosing to complete the form have three response options:

• YES, I HAVE A DISABILITY (or previously had a disability)
• NO, I DON’T HAVE A DISABILITY
• I DON’T WISH TO ANSWER

Below are some basic questions to consider before reading the rest of this blog to get you thinking about some self-identification challenges with the new disability form and process.

1. What percent of the workforce will accept the invitation to self-identify? (We are estimating that, depending upon the industry, it will range from 10% - 40%.)
2. Are the employees who complete the self-identification form more likely to be disabled or non-disabled? (We project that they are more likely to be disabled.)
3. How should employees who do not self-identify be treated in the utilization analysis? (We and OFCCP remove them from the applicant adverse impact analysis.)
4. For analysis purposes, should employees choosing “I don’t wish to answer” be treated differently from those who do not complete the form?

Now that your mind is sufficiently exercised, let’s discuss why data collected using OFCCP’s prescribed Voluntary Self-Identification of Disability form may pose considerable analytic difficulty for federal contractors trying to conduct disability utilization analyses.

OFCCP has set a 7% goal against which federal contractors will be expected to compare the percentage of their workforces that are disabled. As noted earlier, there are two issues with conducting utilization analyses based on the disability data to be collected that may limit the reliability of the percentages calculated by federal contractors:

1. A portion of the workforce will not complete the form (i.e., survey non-response)
2. A portion of those responding will select ‘I DON’T WISH TO ANSWER’ (referred to as IDWTA)

Assumptions made about the types of employees choosing not to complete the form and the types of employees selecting the IDWTA will influence the utilization statistics observed and the conclusions about whether goals are met. In the following illustration, we simulate different types of response patterns to highlight the different conclusions at which one may arrive depending on what one assumes about missing and IDWTA data and how those data are treated in the calculation of utilization percentages.

Baseline Simulation Data

As a basis for our illustration, we present a hypothetical contractor establishment with 1,000 employees in the Professionals job group. Of the 1,000 employees in this job group, 926 do not have a disability and 74 do have a disability.

Ideal Utilization Analysis

In an ideal world (presumably, the one which OFCCP envisions), 100% of the workforce responds to the invitation to voluntarily self-identify in either the affirmative [YES, I HAVE A DISABILITY (or previously had a disability)] or the negative [NO, I DON’T HAVE A DISABILITY]. In such a case, 926 of the employees in the job group would indicate they do not have a disability and 74 would indicate that they do have a disability (or have had one in the past). Thus, our utilization analysis would show that 7.4% of the total job group has identified as having or having had a disability. The 7.4% figure can be meaningfully compared to the 7% utilization goal to conclude that the contractor is exceeding that goal.

What a straightforward, potentially useful analysis! Let’s now consider some more realistic scenarios.

Utilization Analysis with “I DON’T WISH TO ANSWER”

It is unlikely that all individuals providing a response will wish to self-identify as either disabled or not disabled. In fact, that is the precise reason for the IDWTA option provided on the self-identification form. The fact that individuals may not want to self-identify is consequential to the calculation of utilization, as how such individuals are treated in the analysis will influence the utilization percentage.

Imagine a scenario in which 20% of those who are disabled and 20% of those who are not disabled choose the IDWTA option. Utilization percentages look very different depending on whether employees not wishing to identify are included in the denominator of the utilization calculation or excluded from the analysis. We present the two different utilization scenarios below.

As is clear, if those choosing the IDWTA option are assumed to not be disabled and are included in the denominator of the calculation, the utilization percentage falls well short of the 7% goal (which we know is also well below the actual utilization in the hypothetical job group).

The issue of how to treat those who do not wish to identify becomes even more complicated if we do not assume that equal percentages of disabled employees and non-disabled employees choose the IDWTA option. For example, it is possible that a higher percentage of individuals with disabilities choose the IDWTA option compared to those without disabilities.

Imagine a scenario in which 20% of those who are disabled choose the IDWTA option, but only 10% of those who are not disabled choose the IDWTA option. As shown below, if such a case were to occur using our example data, the utilization percentage would fall short of OFCCP’s goal regardless of whether the denominator included the individuals not wishing to identify.

Utilization Analysis with Survey Non-Response

To make the scenario even more realistic, and substantially more complex, consider the fact that a portion of the employees in the job group will choose not to complete the form. Because this is a voluntary invitation, it is unlikely that all employees will provide a response to the survey. This means that our utilization analysis will not only be affected by the number of individuals choosing the IDWTA option, but also by those not providing any response. Again, the treatment of such individuals in the analysis will influence the utilization percentage.

To illustrate, assume that the survey yields a 40% response rate. In other words, 600 individuals do not complete the survey, and there are 400 individuals falling into one of the three response categories. Also assume that the ‘non-responders’ are found equally in the individuals with disabilities and non-disabled groups. Last, assume that of those deciding to provide a response, 20% of both individuals with disabilities and those who are non-disabled choose the IDWTA option (as in Scenario 2 and Scenario 3). As shown in the table below, one would arrive at three very different conclusions, depending on who was included in the denominator of the utilization analysis. Scenario 8 is the only case in which one would correctly conclude that the contractor meets the 7% utilization goal set by OFCCP. Calculations under Scenario 6, in which those who didn’t respond and those who chose the IDWTA option are assumed to be non-disabled and are included in the denominator, erroneously suggest that the contractor has a long way to go to meet the 7% goal.

As with the case of those who choose the IDWTA, it is unlikely that there will be equal percentages of disabled and non-disabled individuals choosing not to respond. More likely is the scenario in which a certain percentage of disabled individuals choose not to respond, a certain percentage of non-disabled individuals choose not to respond, and those percentages are not equal to one another. Such a case coupled with the unlikelihood of equal percentages of disabled and non-disabled individuals choosing the IDWTA option make the analyses all the more complicated.

As a final illustration, assume the survey yields a 40% response rate overall, but that the 40% response rate is produced by 39% of non-disabled responding (total of 361 responding) and 53% of disabled responding (total of 39 responding). Further assume, as in Scenarios 4 and 5, 20% of the disabled responders choose IDWTA (8 choose IDWTA), whereas only 10% of the non-disabled responders choose IDWTA (36 choose IDWTA). The table outlining Scenario 9 through 11 highlight the VERY different utilization percentages calculated, depending on who is included in the denominator of the calculation. Only Scenario 10 shows a realistic approximation of the true job group utilization outlined in Scenario 1. Scenario 9 shows substantial under-utilization and Scenario 11 shows substantial over-utilization. Thus, it should be clear that realities surrounding non-response and IDWTA choices will have material influence on the utilization percentages.

To our knowledge, OFCCP has not yet published guidance on how they will conduct the analysis. DCI is adamant that it would be improper to include those who do not wish to identify in the denominator without accounting for the fact that many of them are likely disabled. Similarly, DCI is also adamant that it would be improper to include those who do not complete the survey in the denominator without accounting for the fact that many of them are likely disabled. A more reasonable option may be assuming that the percentages of disabled/non-disabled in the IDWTA and non-response categories match the percentages observed based on the affirmative and negative responses. Alternatively, it may be appropriate to (a) assume that the IDWTA and non-response samples comprise a percentage disabled equal to OFCCP’s goal of 7% and (b) calculate overall job group percentages based on that estimate.

In closing, until OFCCP shares guidance on appropriately conducting a utilization analysis based on the Voluntary Self-Identification of Disability form, it is recommended that contractors conduct such an analysis based solely on the data collected that indicate the affirmative [YES, I HAVE A DISABILITY (or previously had a disability)] and negative [NO, I DON’T HAVE A DISABILITY]. Stay tuned for formal DCI recommendation on how to best address the issues presented here.

by Jana Garman, M.A., Associate Consultant; Kayo Sady, Ph.D., Senior Consultant; and David Cohen, President, DCI Consulting Group