Data Equity Principle 3
Disaggregate data on both outcomes and system conditions to analyze disparities, monitor progress, and guide action.
Description
Data users must acknowledge the diversity of experiences among priority communities to uncover disparities that can be hidden in aggregate data. Data analysis often starts by measuring outcomes for broad populations of individuals, but results can vary—sometimes significantly—across certain populations or groups with unique experiences and histories. Taking a passive stance in data analysis can lead data users to draw different conclusions. Without disaggregation, they may miss the opportunity to identify, address, and monitor disparities. The E-W Framework offers guidance on key disaggregates data systems should collect, including race and ethnicity, gender, income level, disability status, English proficiency, and LGBT status.
Though data systems must collect or link data on multiple relevant background characteristics, which factors are analyzed through disaggregation and how they are analyzed depend on the local context. Data analysis may require more than one level of disaggregation to capture the intersectional nature of individuals’ lived experiences. For example, a school district might explore whether high school graduation rates differ for students with disabilities by race. In contexts with smaller populations, disaggregating across multiple levels can be challenging as subgroup sizes grow smaller with each level of disaggregation, making it harder to reliably compare trends over time. However, data users must still consider the experiences of smaller groups, such as American Indians and Alaska Natives, and not simply default to grouping them under an “other” category that does not receive careful attention. Privacy-enhancing statistical techniques for small sample sizes can include data suppression (removing information for small subgroups), data masking or perturbation (introducing uncertainty into the dataset), blurring (reducing data precision), or Bayesian statistical methods (see Disaggregates In Depth for more).
Decisions about how to disaggregate data should also be guided by a theoretical understanding of a problem of practice and potential root causes to avoid perpetuating existing stereotypes and deficit narratives or framing that advertently or inadvertently blame particular groups rather than systems for disparate outcomes. In addition to disaggregating outcome data, data users should break out data on E-W and adjacent system conditions to reveal other underlying disparities. For example, system conditions such as access to school support staff may be relevant to the graduation rates of students with disabilities, and these indicators should also be disaggregated further by race. However, disaggregation alone is not enough to reveal causes or solutions for inequities, as described in Principle 4 on examining social and historical contexts to identify root causes of disparities and data-driven solutions.
When "standard" disaggregation is insufficient
Data users should consider whether standard categories commonly used to disaggregate data, such as broad racial categories, may not be appropriate for all groups and contexts. For example, an analysis of census data on four year postsecondary degree completion by race would show that more than half of Asian Americans have a bachelor's degree or higher, the highest rate among any racial group. However, this rate masks significant variation within different communities of Asian Americans: for instance, less than 15 percent of Laotian Americans obtain bachelor's degrees. Disaggregating data by both race and detailed ethnicity categories shows that certain groups of Asian Americans, including Laotian, Cambodian, Hmong and Vietnamese Americans, experience educational attainment on par with other minoritized groups. To put these differences in to context, users should also collect and disaggregate data on potential root causes that drive educational attainment for different ethnic groups, such as their reasons for immigration, generational status, neighborhood resources, or access to financial aid.
Applying this Principle
Work with community members to determine which characteristics to measure during data collection or to link into the data (if already available), and how to label these characteristics in data collection tools as well as eventual reporting (for example, Hispanic, Latino/a, Latinx).
Disaggregate both outcome and systems data at multiple levels to illuminate any disparities. Include qualitative research or input from the community so that readers can contextualize disaggregated data with individuals’ lived experiences and the root causes of any observed disparities.
When reporting disparities by subgroup, connect these to the system and root causes, not people. Use data visualization to clearly communicate disparities while avoiding perpetuating deficit narratives (see Principle 6).
Reflection Questions
- Who is or is not included within the categories representing the population of study?
- How can disaggregated data help us think about intersectional issues (for example, how outcomes might differ for Black boys versus Black girls)?
- Have we analyzed both outcome and structural disparities between subgroups and avoided placing blame or perpetuating stereotypes?
- When is it appropriate to compare data within versus between groups (for example, comparing outcomes for Latino high school graduates and Latino college graduates versus comparing outcomes for Latino and non-Latino college graduates)? Which comparisons would best answer your research questions and inform future action?
Be On The Lookout
Data users should tailor plans for disaggregation to each community and not simply report on mandated categories. For instance, defaulting to disaggregating data by just race and income would not provide much additional insight in a community comprised almost exclusively of Latino families with low incomes. Depending on the community’s local context and the problem of practice being considered, further disaggregation by factors such as English proficiency and newcomer status may reveal hidden disparities that systems should understand and address.
Additional Resources
- Understanding Minimum N-Size and Student Data Privacy: A Guide for Advocates. This guide from the Data Quality Campaign includes information on techniques for protecting student privacy when disaggregating data and dealing with small sample sizes. It introduces concepts including data suppression, perturbation, and blurring and describes advantages and disadvantages of each.
- Disaggregated Data: Not Just a Box Checking Exercise. This three-page brief by the Data Quality Campaign, Learning Heroes, and National Parent Teacher Association details what data disaggregation is, why it matters in K–12 education, which subgroups are required for disaggregation under the Every Student Succeeds Act, and how to communicate the value of disaggregated data to interested groups (including examples from multiple states).
- The Essentials of Disaggregated Data for Advancing Racial Equity. This Race Matters Institute blog post offers guidance on how far to go in data disaggregation, deciding which data to disaggregate, and presenting disaggregated data.
- By the Numbers: A Race for Results Case Study. This Annie E. Casey Foundation report shares two cases studies of how data users have disaggregated data to inform policies, practices, and decision making for their populations of focus.
- The Importance of Disaggregating Data. This short report by Safe Schools Healthy Students addresses the importance of disaggregating data (including examples), common disaggregates, and limitations of data disaggregation.
References
The framework's recommendations are based on syntheses of existing research. Please see the framework report for a list of works cited.