ERS Group Blog ACS Estimates and Affirmative Action - Past, Present, Future

The U.S. Census Bureau announced program changes in its 2016 budget that include discontinuation of the 3-year American Community Survey (ACS). Federal contractors might wonder whether this discontinuation, if approved, will affect affirmative action plan preparation.The EEO Tabulations, which serve as the source of benchmark data for OFCCP-mandated affirmative action plans, are based upon ACS tabulations. Likely, the loss of the 3-year estimates will not substantially affect federal contractors. The Census is not discontinuing the ACS entirely, but simply terminating the release of one of the sets of estimates. The current (2010) EEO Tabulations are based on the ACS 5-year estimates, not the 3-year estimates. It is helpful to note that the Census Bureau's rationale for permanent termination of the "3-Year Data Product" is that "[t]he Census Bureau intended to produce this data product for a few years when the ACS was a new survey. Now that the ACS has collected data for nearly a decade, this product can be discontinued without serious impacts on the estimates for these communities" (see CEN 105-CEN108).Nevertheless, there are still several open questions as to the application of the ACS-derived EEO Tabulations to affirmative action plans. EEO Tabulations - The BasicsPrior to delving into these questions, let's review a few key features of the EEO tabulations. The tabulations are sponsored by a consortium of government agencies, including the EEOC, the DOL (OFCCP), and the U.S. Department of Justice. The tabulations serve three major goals as they:Allow for measuring the effect of anti-discrimination laws and regulations in the workplace,Provide ethnicity, race, and sex civilian workforce data mapped to occupational data within a specified geography and job category, andMay be used to create external availability statistics for affirmative action planning. The 2000 EEO Tabulations (and prior year's tabulations) were built upon the results of the Decennial Census long form data. The 2010 Census did not include the long form, so the consortium turned to the ACS as an alternative.EEO Tabulations - The New ApproachIn 2012 a new set of tabulations was released based on the 5-year ACS survey data (over the period January 2006 through December 2010). The new tables contain data organized by geography, industry, age, educational attainment, median earnings, citizenship, and unemployment status. Pre-calculated margins of error are also included.The new EEO Tabulations have several advantages and disadvantages over prior iterations. The primary advantages are all associated with the annual nature of the ACS survey (instead of the Decennial Census). Given that the ACS data evolves annually, it is possible (at least in theory) to update the EEO Tabulations annually. Furthermore, it is feasible to update the EEO tabulations with additional variables or tables on a more frequent basis.The new EEO Tabulations have one preeminent disadvantage relative to previous versions - a relative lack of precision. Unlike the EEO tabulations based on decennial long form data, the ACS-derived tabulations contain substantial sampling error. This shortcoming is made clear by the Census Bureau's decision to include calculated margins of error for each point estimate in the EEO tabulations. A review of these data reveals that these margins of error can be quite large, especially for narrower geographies and occupations. For example, the point estimate for the African American representation among Sheet Metal Workers in Cincinnati, Ohio is 36.2%, but the margin of error associated with the estimate is +/- 25.2%. This implies that we can't reject the hypothesis that an observed African American representation of say (for example) 15% is distinguishable from the estimated benchmark of 36.2%.EEO Tabulations - Going ForwardWhile the aforementioned cancellation of the ACS 3-year estimates will not likely directly impact the EEO Tabulations, there remain at least two critical outstanding questions related to these tabulations.First, will the OFCCP acknowledge the substantial measurement error in the current EEO Tabulation estimates? The female and minority availability measures generated from the new EEO Tabulations are the result of a survey. As such, each reported availability measure is the mid-point of a confidence interval in which we can be 90% sure the true population value falls within. (The decision to present 90% margins of error is not conventional, 95% (associated with two standard deviations) is the norm). Second, how frequently will the EEO Tabulations be updated? Recall from the above discussion that many of the advantages associated with basing the EEO Tabulations upon the ACS is that this approach allows for more frequent updates, certainly more frequent than the decennial updates associated with prior versions of the EEO Tabulations. However, it is not clear that the sponsoring agencies have the desire or the budget to release an update anytime soon. It is unclear whether the DOJ, OFCCP, and other sponsoring agencies will wait until 2020 to release an update of the EEO Tabulations or provide an update on a timelier basis. Visit our Data Analysis and OFCCP Compliance pages for information on Census data and the use of ACS data for affirmative action programs.