Disaggregation

When Should You Disaggregate Data?

Disaggregation should be built into every monitoring system where equity matters - which is virtually all development and humanitarian programs. Plan the relevant disaggregation when:

• The applicable agreement or reporting standard explicitly specifies subgroup outputs; identify the source, scope and version
• The program targets specific sub-groups (women, children, people with disabilities, ethnic minorities)
• The theory of change explicitly aims to reduce inequities
• Previous monitoring has shown that aggregate improvements masked disparities

Disaggregation is not just a technical practice - it is an equity commitment operationalised in data systems.

How It Works

Step 1: Identify required and meaningful disaggregation variables

Not every indicator needs every disaggregation variable. Decide which variables are mandatory (donor requirements), which are program-relevant (age for a youth program), and which are analytically meaningful (location for geographically targeted interventions). Standard disaggregation variables include:

• Sex, where relevant and collected appropriately: male, female, other/prefer not to say
• Age group: define nonoverlapping categories suited to the question, such as 0-14, 15-24, 25-59 and 60+; these are illustrative analytic bands, not universal legal age definitions
• Geographic location: district, urban/rural, program zone
• Wealth/vulnerability status: wealth quintile, food security status, displacement status
• Disability status: an appropriate named, versioned functioning instrument with its own coding rules

Step 2: Design data collection to capture sub-group data

Disaggregation requires that data collection instruments capture the sub-group variables for every respondent or unit of analysis. This means adding demographic questions to surveys, ensuring rosters capture sex and age, and training enumerators on consistent categorisation.

Step 3: Build disaggregation into the analysis plan

Specify which indicators will be analyzed by which variables. Document this in the M&E plan and ensure the data management system can produce disaggregated tables.

Step 4: Report and act on disaggregated findings

Disaggregated data has no value if it stays in spreadsheets. Include disaggregated tables in program reports and flag significant disparities. When disaggregation reveals that women, children, or a particular geographic group is underperforming relative to the aggregate, treat this as a program management signal requiring response.

Key Components

• Disaggregation plan: specifying which variables will be collected and which indicators will be disaggregated
• Data collection instrument design: ensuring demographic variables are collected for all respondents
• Reporting templates: standard tables showing aggregate and disaggregated results side by side
• Minimum threshold guidance: design-aware uncertainty, effective sample and disclosure safeguards for subgroup reporting; there is no universal n=30 rule
• Equity analysis: comparison of outcomes across sub-groups to identify and address disparities

Best Practices

Disaggregate by sex as a minimum. Sex disaggregation is the most universal requirement and the most commonly missing. If you can only do one disaggregation, start here.

Collect data at the right level for disaggregation. Aggregate household data cannot be disaggregated to individual women's outcomes. Design data collection at the individual level if individual disaggregation is required.

Be consistent across time points. Disaggregation categories must be identical at baseline, midline, and endline to enable comparison.

Sample with sub-groups in mind. If a sub-group constitutes only 5% of the population but you need statistically meaningful results for them, you need to oversample that group. Random sampling of the full population will not produce adequate sub-group sample sizes.

Report disparity, not just disaggregated numbers. Saying "60% of women and 75% of men achieved the outcome" is more useful than listing both numbers in separate columns without comment.

Common Mistakes

Collecting disaggregation variables but not analyzing them. Many programs dutifully record sex and age in data collection but produce only aggregate numbers in analysis. Build disaggregated analysis into the reporting template so it cannot be skipped.

Insufficient sub-group sample sizes. When a sub-group has fewer than 30 respondents, statistical conclusions are unreliable. Plan sample sizes to enable meaningful sub-group analysis.

Too many disaggregation variables. Disaggregating every indicator by sex, age, location, and wealth simultaneously is analytically valuable but practically overwhelming. Prioritize which indicators need which disaggregation based on the program's equity objectives.

Examples

The following scenarios, settings, and numbers are illustrative; they are not documented program findings or standard design requirements.

Health program, Sub-Saharan Africa. A donor-funded HIV prevention program reported aggregate HIV testing rates in Year 1 that appeared strong (68% of target population tested). Disaggregation by age group revealed that testing rates among 15-24 year olds were only 41% - well below the program target - while rates among adults 25-49 were 82%. This finding prompted targeted youth engagement activities; recorded youth testing rates reached 67% by Year 2. The monitoring trend alone does not identify the activities' causal effect.

Education program, South Asia. A donor-funded girls' education program in Pakistan disaggregated school attendance data by wealth quintile in addition to sex. The data revealed that girls from the poorest quintile had attendance rates 30 percentage points lower than girls from the middle quintile, despite the program providing stipends to all girls. Investigation revealed that stipend payment delays were disproportionately affecting the most remote villages - a logistical issue that was corrected in the subsequent term.

Proposal Context

Justify each proposed disaggregation by equity and decision needs, sample design, precision, missingness and disclosure risk. Check any required categories in the applicable reporting agreement. Keep definitions comparable across rounds and identify subgroups whose evidence cannot support a safe or precise estimate. A nominal sample size alone does not establish subgroup precision. Plan how findings will inform decisions. Pair with smart-indicators and sampling-methods.