Sept 19 -- The Census Bureau invites comments from the public by November 20, 2023 regarding the proposed American Community Survey Methods Panel: 2024 Sexual Orientation and Gender Identity Test.
The American Community Survey (ACS) is an ongoing monthly survey that collects detailed housing and socioeconomic data from about 3.5 million addresses in the United States and about 36,000 addresses in Puerto Rico each year. The ACS also collects detailed socioeconomic data from about 195,000 residents living in group quarters facilities in the United States and Puerto Rico each year. Data is collected via self-response modes (internet and paper) as well as interviewer-administered modes via telephone and in person. Resulting tabulations from this data collection are provided on a yearly basis. The ACS allows the Census Bureau to provide timely and relevant housing and socioeconomic statistics, even for low levels of geography.
The Census Bureau plans to request Office of Management and Budget (OMB) approval to conduct a test of sexual orientation and gender identity questions on the ACS. The test is referred to as the 2024 ACS Sexual Orientation and Gender Identity (SOGI) Test. Federal agencies have expressed interest in and identified legal uses for this information, including civil rights and equal employment enforcement.
The Census Bureau proposes to test question wording, response categories, and placement of sexual orientation and gender identity questions on the questionnaire. Of specific interest is how the questions perform when completed by proxy respondents. In the ACS, one person at an address typically answers questions about everyone living there. When one person answers a survey about others, we call this “proxy reporting.” Questions on sexual orientation and gender identity are not currently asked on any federal surveys that use proxy reporting. The test will build on existing qualitative research conducted throughout the federal government and private sector, including work by the Federal Interagency Working Group on Measuring SOGI, the Measuring SOGI Research Group as part of the Federal Committee on Statistical Methodology, and the National Academies of Sciences, Engineering, and Medicine (NASEM).
We are proposing to test up to two versions of the questions using a nationally representative, split-panel test. A sample of housing units will be selected for this test; housing units in sample for the ACS will not be eligible. We are also proposing to test a variation in display of the questions in the internet mode. A follow-up reinterview will also be conducted to assess the reliability and quality of responses.
Building on recommendations from NASEM and OMB Best Practices, the Census Bureau proposes to test a two-step gender identity question: first asking about sex assigned at birth and then asking about current gender. These questions will replace the existing question on sex.
The proposed ‘sex assigned at birth’ question would ask: What sex was <Name> assigned at birth? Mark (X) ONE box. The Census Bureau is proposing to omit the text, “on your original birth certificate” from the sex assigned at birth question as proposed by NASEM and OMB. Asking the question without this phrase has been used by some federal surveys and cognitively tested without issue (Asking About SOGI in the CPS: Cognitive Interview Results (census.gov)). Removing the reference to the birth certificate is more culturally sensitive to non-English speakers, some of whom may not have a birth certificate or have seen it. In recent years, many people who have a designation of X on their birth certificate were assigned male or female at birth before their parents chose to use X instead. Removing the birth certificate reference may help parents of those children answer this question. The Census Bureau also proposes to keep the existing order of the male and female categories to address a serious concern about how an inconsistency in the order of male/female categories across Census Bureau surveys could lead to interviewers accidentally selecting the wrong response category. Keeping the male/female order also minimizes the number of changes being tested at once.
The ‘current gender’ question will be asked only of people who are 15 and older. The proposed question is: What is <Name's> current gender? The response categories will be Male, Female, Transgender, Nonbinary, and “This person uses a different term” (with a space to write in a response). The proposed question stem aligns with the NASEM report recommendations. It refers to the subject of the question in the stem so that respondents are not left to reason what the question is asking based solely on the response options. The wording is also more concise when administered in the proxy version of the question as shown above.
The Census Bureau proposes to add “nonbinary” as a response option. Estimates suggest that over one million adults in the U.S. use this term to describe themselves. In previous research (e.g., CPS pretesting), respondents have commented that this category should be added and that not every person who is nonbinary considers themselves transgender. Data on the nonbinary population was also part of requests from federal agencies.
The OMB best practices suggest using a ‘mark all that apply’ instruction for the gender question. However, most federal surveys do not use ‘mark all that apply’ for this question. Additionally, the California Health Interview Survey (CHIS) decided not to implement ‘mark all that apply’ in part due to concerns about “the potential for increases in gender minority reporting from those who do not primarily identify as non-cisgender” artificially inflating estimates. Given the lack of consensus in this area, the Census Bureau proposes to test two treatments. A treatment that allows only one response category to be marked will be compared to a treatment that allows multiple categories to be marked.
Consistent with recommendations, a verification question will also be asked for anyone whose answer to the ‘sex assigned at birth’ question and ‘current gender’ question does not match. In addition, for evaluation purposes, the Census Bureau is considering asking the verification question to a sample of respondents whose answers are the same in the two questions. The Census Bureau is also proposing to add an open-ended write-in question to gather additional information about a person's gender identity for research purposes.
The ‘sexual orientation’ question will be asked only of people who are 15 and older. The proposed question is: Which of the following best represents how <Name> thinks of themselves? With response categories of: Gay or lesbian, Straight—that is not gay or lesbian, Bisexual, and This person uses a different term (with space to write-in a response). This question is in alignment with current recommendations of how to ask about sexual orientation, however it omits an explicit “I don't know” response category. This approach follows the conventions of the ACS, which does not offer an explicit “don't know” response option for any topics in the survey to minimize item nonresponse and increase data quality. Respondents can skip this question on the internet and paper modes. Interviewers can mark “don't know” and “refusal” in the computer-assisted personal interview (CAPI) instrument.
To help address sensitivity in interviewer-led modes, especially if other household members are present, the Census Bureau is also proposing to use a flashcard for in-person interviews and use numbered response categories for both in person and telephone interviews so that respondents can indicate a response category number to select the appropriate category or categories.
Both the ‘current gender’ and ‘sexual orientation’ question allow a write-in response. In the internet mode we will test two versions of the write in. In the first version the respondent will see the question, response categories, and the write-in field when they get to the screen with the question. In the second version, the respondent will only see the question and response categories when they get to the screen with the question. If the respondent selects the “This person uses a different term” category, then the write-in space will display, and the respondent can provide a write-in response. This display experiment will be embedded within the question wording experiment.
A follow-up content reinterview is also proposed for this test. A subset of the ACS questions will be re-asked in the reinterview to measure response reliability. For half of the reinterview sample, we will reinterview the sample respondent as the original interview, for the other half we will reinterview another adult member of the household. This will allow us to compare response reliability for proxy vs. self-responses. We will also compare the reliability of the SOGI questions to other questions in the ACS.
We estimate that 271,680 sampled housing units will respond to the test of the 480,000 housing units sampled for the test. Additionally, responding housing units are eligible for a content follow-up reinterview. We estimate that 65,280 sampled housing units will also complete the reinterview.
FRN:
https://www.federalregister.gov/d/2023-20256
Statistical Journal of the IAOS - Volume 39, issue 3: Beyond the binary: Sex and gender diversity in population projections
https://content.iospress.com/articles/statistical-journal-of-the-iaos/sji230044