Most public discussion of inequality relies on one number per group. That number is usually an average, and averages discard exactly the information that determines who needs help.
The mean and the median tell different stories
A mean is pulled by extreme values at either end of a distribution. Where a small number of very high earners exist, the mean sits well above what a typical member of the group receives.
A median describes the middle position and ignores how far the extremes stretch. Reporting both together reveals skew, which is why serious statistical series usually publish the pair.
Where only one is cited, the choice can shift the apparent size of a gap substantially. Two commentators using different measures on identical data can reach opposite conclusions honestly.
Broad categories bundle unlike populations
Large official categories often combine communities with very different histories, migration patterns and regional concentrations. The category average then describes nobody in particular.
Within a single broad grouping, outcomes on income, schooling and health can vary as much as they do between groupings. Aggregation flattens that variation into a single unhelpful figure.
Disaggregated reporting reveals the internal spread, which is why advocacy for finer categories is a recurring theme in statistical policy debates across many countries.
Aggregation can reverse the direction of a finding
A well-known statistical result holds that a pattern present in every subgroup can reverse when the subgroups are combined. Composition, not the underlying relationship, produces the reversal.
This matters for institutions comparing outcomes across departments or regions. A body can appear to favour one group overall while every individual unit shows the opposite pattern.
The practical response is to examine data at the level where decisions are actually made. Aggregate figures are useful for describing scale but poor for locating a cause.
Programmes designed around averages misdirect resources
Eligibility thresholds set using group averages assume the group is clustered near that value. Where the spread is wide, a threshold either excludes many in need or admits many who are not.
Geographic targeting has the same weakness. An area average conceals affluent households in deprived districts and struggling households in prosperous ones, and both are then treated incorrectly.
Programmes that combine a group signal with an individual assessment tend to reach the intended population more reliably, though the added administration carries real cost.
Disaggregation has genuine limits
Splitting data into finer categories reduces the number of observations in each, and small samples produce unstable estimates that swing between collection rounds.
Fine categories also raise privacy questions, since a small cell in a published table can identify individuals. Statistical agencies suppress or blur such cells for that reason.
The workable position is usually a compromise: publish detail where sample sizes allow, state the uncertainty plainly, and avoid treating any single figure as a complete description.