Our August 15, 2023, Urban Wire post and November 17, 2023, report and fact sheet on LGBTQ+ homeownership rates contained inaccurate results. We revised the publications with corrected data.

What Happened

Our analysis relied on US Census Bureau Household Pulse Survey data for weeks 36 through 61, which were collected from August 18, 2021, through September 4, 2023.

The Household Pulse Survey uses two questions to identify respondents’ sex at birth and gender identity:

D6 What sex were you assigned at birth, on your original birth certificate?

Male (1)

Female (2)

D7 Do you currently describe yourself as male, female, or transgender?

Male (1)

Female (2)

Transgender (3)

None of these (4)

We identified respondents as transgender if they described themselves as transgender on the gender-identity question (D7) or if they answered the sex-at-birth (D6) and current-gender-identity (D7) questions differently. Using this method, we counted 10,493 transgender respondents.

But we had not accounted for the Census Bureau’s practice of filling in blank responses on the gender-assigned-at-birth (D6) question. When respondents do not answer the sex-at-birth question on the survey, the Census Bureau randomly assigns each respondent male or female, regardless of their answer to the gender-identity question (D7). As a result, our method for identifying transgender respondents by tracking respondents who had different responses to questions D6 and D7 incorrectly included some respondents who had their gender assigned at birth randomly assigned.

About 0.3 percent of all respondents in our sample, and 19.7 percent of the respondents we had previously identified as transgender, had a randomly imputed sex at birth.

How We Fixed It

When the Census Bureau randomly assigns a respondent a sex for answer D6, it adds a flag for that respondent called AGENID. We used this flag to remove all respondents with a randomly imputed sex at birth from our sample and computed results for the blog post and report again. We now identify 8,427 transgender respondents in our survey sample.

The trends and most of the numbers we reported in both publications were unaffected. After using the corrected data, we see a slightly larger gap in homeownership rates by gender identity in the second regression in our report. In addition, the transgender or nonbinary homeownership rate is 1 percentage point lower than we previously thought.

We would like to thank Bill M. Jesdale of the University of Massachusetts for writing a useful description of this issue and the solution.