On Tuesday morning, the US Census Bureau announced that questions about sexual orientation and gender identity (SOGI) would be included in the 2020 Census—and then did an abrupt about-face that afternoon. The retraction is unfortunate, because these data are essential for crafting inclusive policy to support lesbian, gay, bisexual, and transgender (LGBT) Americans.

SOGI data help researchers understand the pay disparities LGBT workers face relative to their heterosexual counterparts. LGBT populations also face unique health care needs that cannot be fully understood without including SOGI questions on large-scale datasets like the Census or in electronic health records. Recent federal recognition of same-sex marriages also affects LGBT Americans’ access to a wide variety of federal benefits (and therefore, the federal budget).

Including SOGI questions on the Census is especially critical because LGBT individuals probably make up a low share of the US population, making them difficult to study with smaller-scale survey data. Of course, even this point is contestable due to the lack of data. Including SOGI questions on the Census would help solve this problem, because every American must fill out the decennial Census. The dataset is so large that even the scaled-down Census file made available to researchers would be sufficient to study LGBT and other small populations.

Although the 2020 Census would have been the first to include explicit SOGI questions, the Census Bureau has made prior attempts to estimate the number of same-sex couples in the United States.

Using data from the 2010 Census and subsequent American Community Surveys (ACS) on marital status and the reported sex of a respondent’s spouse, the Census Bureau has produced regular estimates of the number and characteristics of same-sex couples. This Census Bureau work built on path-breaking research by the Urban Institute’s Gary Gates and Jason Ost that used the 2000 Census to make similar inferences about same-sex couples.

These estimates are indirect and imperfect, and they only represent a subset of the broader LGBT population. However, researchers have always considered this prior work to be a statistical stepping stone to the future inclusion of SOGI questions on the Census and the ACS.

We don’t know what motivated the Census Bureau’s retraction. Nevertheless, this episode provides another opportunity to reiterate the importance of federal data collection. All Americans deserve policy that is evidence-based, and we need strong, comprehensive data to generate evidence.