AISP Toolkit Feb25 2025 - Flipbook - Page 37
population may choose to focus on highly disaggregated data categories to ensure that community
members are adequately represented.
Ask:
How do identities—self-identi昀椀ed, self-perceived, and/or externally perceived—interact
with the outcomes you’re studying? How have legacies of oppression and opportunity
in昀氀uenced these outcomes?
Are certain identities disproportionately harmed, or are some particularly visible (which
could mean over-surveillance) or invisible (erased), in patterns?
What kind of identity source (self-identi昀椀cation or observed, aka “street race”14) is
important here?
How can we validate and test answers to these questions with community leaders and
members?
What minimum population thresholds are needed to fully realize the framework’s purpose?
Frameworks with a focus on more complete disaggregation may have very small—even as
low as 1 or 0—minimum standards for disaggregated group size. Frameworks focusing on
broader population trends may need higher minimum sizes for aligned results.
Data Collection: Data landscape assessment for data reuse
Once the governance team has developed an initial sense of identity offerings for a RELD/SOGIE
framework, we encourage reviews of available data, including alignment (or lack thereof) among
datasets. The goal of this stage is to understand whether current data assets meet the purpose
described in the previous stage, to identify whether and what kind of supplemental collection or
detail could adapt assets to the purpose, or whether the project needs newly collected data.
Ask:
What categories of identity are available across datasets? How well do they align with the
minimum set described in the previous stage?
How well do the categories align among themselves? How do the different datasets handle
multiple responses, write-ins, “other,” unknown, blank, and “decline to respond”? How do
they handle changing responses over time?
CENTERING RACIAL EQUITY THROUGHOUT THE DATA LIFE CYCLE
How do multiple identities in a category, and intersections across categories, magnify or
mitigate disparate effects?
How do the data differ by report type: Are they self-reported, completed through proxy,
observed, and/or imputed? Is this consistent over time and across sets? Do the individual
sets change by source type over time?
What is the completeness, quality, and consistency of the data? How do responses for the
same person compare across sets?
14 López, N. (2024). What is street race? Institute for the Study of Race and Social Justice.
33
