I am implementing procedures that require applicants to fill out an EEO self-identify form regarding their gender and race/ethnicity.  I attended a conference last year in which we were advised that if an applicant declines to self-ID, it is okay to "guess" based on observation.  This made sense to me as a solution to the dilemma of being required to report this data in an affirmative action plan but not offered a category called "other," "non-response," etc.

Some staff are having a hard time understanding why this kind of guessing is permitted, so I'm checking to see if others do visual observation or if there are better solutions to this dilemma.  And I wonder, what do folks do who have an online system for submitting resumes and applications to positions?

 Most people look at me and say Hispanic or White. I self identify as African-American. Visual observation is tricky. I have so many friends who are middle eastern and people think they are African American, Hispanic ect...ect.

Much of race discrimination is actually based on perceived race, not claimed race or actual race.  Therefore, the EEOC is interested in ER's perspective on race if EE refuses to provide it.

Actually any self-identification should have the option of "non-response" or unknown which allows applicants to opt out for whatever reason. If they opt out and there has been no visual identification, they should not be included in any adverse impact analysis that you analyze for your applicants/hires.

Here's what we tell our clients: Provide all applicants the option to self-identify either electronically or by paper. If they opt out electronically, but show up for an in-person interview, provide them another opportunity. If they opt out again, then visually identify but do not make questionable guesses. It is really hard in some cases to identify the race/ethniciy but gender you can at least identify. It is possible you may have only gender, but the ethnicity is unknown for analysis purposes. Finally, anyone that is hired is given the opportunity to identify again and at this point there should be little reason to opt out however, you can't force them to identify so you may be left with the visual.

The visual identification has very little to do with how we "see" people and discrimination but has more to do with having an accurate data set for performing adverse impact analysis and making sure your recruiting efforts are attracting a representitive sample of the population and your hiring decisions are fair and unbiased. Without any kind of visual identification or collection of the data, there is no way to effectively measure this.

Speediware:

The visual identification has very little to do with how we "see" people and discrimination but has more to do with having an accurate data set for performing adverse impact analysis and making sure your recruiting efforts are attracting a representitive sample of the population and your hiring decisions are fair and unbiased. Without any kind of visual identification or collection of the data, there is no way to effectively measure this.

That's an interesting claim.  The visual identification is based entirely on how the rater "sees" the applicant/employee and nothing to do with accuracy except an accurate record of how the rater perceives the applicant/employee's race.  That means that any EEO database is full of self-report data that is intermingled with rater data that may often be very different from self reported data.  I can recall numerous instances of self-reported classifications that did not have much to do with the person's easily visible characteristics (one of the most common I've seen in the Southwest is Native American self-report of a person who looks to be White, Non-Hispanic).

EEO data is socially constructed whether it's self-reported or based on the "findings" of a "rater" and, therefore, any analysis can only be as accurate as such ratings.  You cannot collect data using self reports from entirely different reporting sources and then talk about the data and analysis of the data as if the data were composed of genetic mappings of individuals.  Arguably, you could MORE effectively measure the things you discuss by having a non-report category or attribution flag so that you can analyze your data with and without the information based on rater input to see if there is any substantive difference.  Can you imagine a sociologist doing a study on education and, for each person who opted not to answer the education question on the survey instrument, the sociologist would simply take a stab at how educated the person was based on a conversation with them?  EEO and DOL race data have lofty goals that simply cannot be met by their data collection methods.  Both agencies have lost some cases when their strict reliance on overly simplistic statistics and unduly high regard for the quality of data collected in this fashion were called into question by people with serious research backgrounds.

Interesting comments from you all, and shows why the matter of visual observation was driving my staff nuts.

 I consulted with my affirmative action plan consultant, and he sent me a memo from the OFCCP that says visual observation is not required, or even recommended, for collecting applicant data for the purposes of the affirmative action plan.  So therefore we can, and will, report those who opt-out as just that.   However, he said that for the purposes of reporting on the EEO-1 form one can and must use visual observation, as there is no option to report "other", "did not respond", etc.  Still faulty for all the reasons you all state, but I don't see that we authorized to just omit that group on the EEO-1 form.

CarolL1999:

Interesting comments from you all, and shows why the matter of visual observation was driving my staff nuts.

 I consulted with my affirmative action plan consultant, and he sent me a memo from the OFCCP that says visual observation is not required, or even recommended, for collecting applicant data for the purposes of the affirmative action plan.  So therefore we can, and will, report those who opt-out as just that.   However, he said that for the purposes of reporting on the EEO-1 form one can and must use visual observation, as there is no option to report "other", "did not respond", etc.  Still faulty for all the reasons you all state, but I don't see that we authorized to just omit that group on the EEO-1 form.

It's been 7 years since I've done an AAP but I've done an EEO-1 nearly every year for the last decade.  Whether on a spreadsheet or with a HRIS, I've always flagged the ones without a self-assigned race category.  I don't recall how I handled that on the last AAP I did, although that was with access to corporate counsel so I probably just asked them.

BTW: what DOL says you shouldn't do isn't always the same as what you can't do and occassionally it's what you can and should do, depending on the legal context of whatever it is that you are doing.  Grain of salt and all that.