How to use this calculator
Label your two groups
Type names for Group A and Group B in the label fields at the top. These can be racial or ethnic groups, income tiers, language communities, housing tenures, or any two population categories. The labels appear on all result tables and in the asymmetry panel.
Enter your geographic areas
Add one row per neighbourhood, census tract, school, ward, or zone. For each area enter the count of Group A members, Group B members, and optionally an Other population count for all remaining residents. The calculator needs at least 2 areas to produce a result.
Read the isolation indices
The two large isolation values (xPx) show the average same-group composition of a typical member of each group's neighbourhood. An isolation index of 0.60 means the average group member lives in a neighbourhood that is 60% their own group. The benchmark under perfect integration equals each group's citywide population share.
Read the cross-group exposure indices
The two exposure values (xPy) show what fraction of an average group member's neighbours belong to the other group. These two values are not equal unless the two groups are the same size. The asymmetry panel shows the gap and explains why it arises from population composition, not just spatial distribution.
Use the adjusted isolation index
The Bell-corrected adjusted isolation index removes the effect of group size, rescaling both isolation values to a 0-to-1 scale where 0 means no more isolation than chance and 1 means complete isolation. Use the adjusted index to compare isolation across groups with different population shares.
Review per-area contributions
The contribution table shows which specific areas drive the most isolation for Group A, sorted highest to lowest. The bar chart shows each area's share of total isolation. Areas with a high Group A concentration and a large Group A count will always dominate this table and are the priority areas for any integration intervention.
Formula Reference
This calculator uses standard mathematical axioms and verified algorithms to ensure result integrity.
Related Concepts
Pro Tip
Always verify input units. Mathematical consistency depends on unit uniformity across all variables.
Results are rounded for readability. For high-precision scientific work, consider the raw output.
Related Expert Tools
More precision tools in the same niche.
Gini Coefficient Calculator
Computes the Gini coefficient from either individual income values or grouped bracket data. Returns the standard Gini, bias-corrected Gini (G* = G × n/(n−1)), Palma ratio, quintile income shares, and a comparison against World Bank country-level Gini data.
Theil Index Calculator
The Theil Index Calculator computes Theil T (GE(1)) and Theil L (Mean Log Deviation) from individual income values or grouped data. It decomposes total inequality into within-group and between-group components when group labels are provided, and outputs the normalized Theil T, Atkinson equivalent, and Gini coefficient alongside the Theil scores for a complete inequality profile.
Atkinson Index Calculator
The Atkinson Index Calculator measures income inequality using a welfare-based formula with a configurable inequality aversion parameter (epsilon). Enter any list of income values and set epsilon between 0.1 and 3.0 to control whether the index is most sensitive to the top, middle, or bottom of the distribution. The calculator returns the Atkinson index, the Equally Distributed Equivalent income, a welfare sacrifice percentage, a multi-epsilon comparison panel, and a country benchmark table based on LIS and UNDP data.
What Is the Exposure and Isolation Index Calculator?
The Exposure and Isolation Index Calculator measures the degree of contact and separation between two population groups across a set of geographic areas: neighbourhoods, census tracts, schools, housing zones, or any spatial units you can supply with group counts. For each group it works out two indices: the isolation index (xPx), which is the average same-group share of a typical group member's neighbourhood, and the exposure index (xPy), which is the average cross-group share. Enter your area data, customise the group labels, and the calculator returns both indices, an adjusted isolation value corrected for group size, a per-area contribution breakdown, and a comparison against 2020 US metro data. The measure was developed by sociologist Stanley Lieberson and formalized in his 1980 monograph and draws on the earlier work of Wendell Bell, whose 1954 formula remains the standard used by the US Census Bureau, the Brown University US2020 project, and CensusScope.
Unlike the dissimilarity index, which measures how unevenly two groups are distributed, the exposure and isolation indices measure what a typical group member actually experiences in their neighbourhood. A city can have a falling dissimilarity index while the average Black resident's neighbourhood barely changes in composition, because D tracks structural evenness while exposure and isolation track lived neighbourhood reality. Given that public health research increasingly links neighbourhood racial and income composition to health outcomes, educational attainment, and economic mobility, these indices have moved from purely academic measures to tools used in fair housing compliance, school integration planning, and community development evaluation.
Isolation vs Exposure: Two Questions About the Same Neighbourhood
The isolation index and the exposure index use the same underlying formula but ask different questions. The isolation index (xPx) asks: "What share of the average group X member's neighbourhood is composed of other group X members?" The exposure index (xPy) asks: "What share of the average group X member's neighbourhood is composed of group Y members?" In a two-group world, these sum to 1: if the average Black resident lives in a neighbourhood that is 41% Black (isolation) and 34% White (exposure), the remaining 25% is Other groups. The Othering and Belonging Institute at UC Berkeley's comparison of major segregation measures recommends using both indices together with the dissimilarity index to capture the full picture: D captures departure from an even ideal, isolation captures same-group concentration, and exposure captures inter-group contact.
The two exposure indices are not interchangeable. The exposure of group A to group B (aPb) almost never equals the exposure of group B to group A (bPa), even in the same city with the same set of neighbourhoods. This asymmetry is not a calculation error: it arises mathematically because the formula weights each area by the group's own population share, and the two groups have different sizes. In practice, in US Black-White data, the Black-White exposure index is typically three to five times higher than the White-Black exposure index, because the White majority is spread more broadly across all neighbourhoods.
The Adjusted Isolation Index: Correcting for Group Size
The raw isolation index has one major practical limitation: it is mechanically higher for larger groups. Even under perfect random integration with no segregation at all, a group that makes up 60% of the total population will have a raw isolation index of 0.60 (because every neighbourhood simply reflects the city average). A group that makes up only 5% of the population will have a raw isolation index of 0.05 under the same perfect integration. Comparing raw isolation across groups of different sizes is therefore misleading.
The Bell-corrected adjusted isolation index removes this size effect by computing (xPx minus P_x) divided by (1 minus P_x), where P_x is the group's citywide share. The result rescales to 0 when isolation is exactly what random assignment would produce, and to 1 when isolation is complete. This adjusted index is what allows researchers to compare Black isolation with Hispanic isolation in the same metro, or to compare isolation across cities with very different racial compositions. According to the standard guidance from the Stanford Educational Opportunity Project, the adjusted exposure or normalized index is the preferred measure when comparing segregation across groups of unequal size. This calculator shows both the raw and adjusted isolation indices in the benchmark table.
| Adjusted isolation | Tier | What it means |
|---|---|---|
| Below 0.15 | Very Low | Group is distributed nearly as if randomly assigned; little same-group concentration |
| 0.15 to 0.34 | Low | Mild same-group clustering, below the range typically associated with policy concern |
| 0.35 to 0.54 | Moderate | Meaningful clustering; the average group member's neighbourhood is noticeably more homogeneous than chance |
| 0.55 to 0.74 | High | Strong concentration; typical of racially segregated US metros at the census-tract level |
| 0.75 and above | Very High | Near-complete isolation; the average group member has very few cross-group neighbours |
Accuracy and Limitations of the Exposure and Isolation Indices
The formula for both indices is exact for the data you enter: no estimation, no sampling error from the calculation itself. What introduces uncertainty is the quality of the input counts. Census population data carries a margin of error, particularly for small geographies and small population groups; the American Community Survey can introduce sampling variance of 2 to 5 percentage points in cities under 100,000 residents. The same modifiable areal unit problem (MAUP) that affects the dissimilarity index applies here: using smaller geographic units (census blocks rather than tracts) will generally produce higher isolation indices because fine-grained spatial patterns are visible at the block level but averaged away at the tract level.
The most structural limitation is that both indices require strictly positive counts for all areas and both groups. An area where one group has zero members contributes zero to the exposure index for that group but may still contribute to the isolation index of the other group. In datasets with many empty cells, the indices can become numerically unstable. On top of that, the indices measure the neighbourhood composition experienced by the average group member, but they do not capture variation within the group: two members of the same group may have very different neighbourhood compositions even though the average is 0.40. For health and education research where within-group variation matters, individual-level measures such as the local exposure and isolation (LEx/Is) metrics described in research published in the PMC study on LEx/Is metrics and breast cancer survival are preferred over aggregate indices. Use our Index of Dissimilarity Calculator alongside these indices to cover both the evenness and contact dimensions of segregation, and our Gini Coefficient Calculator if income concentration across the same areas is also relevant.
The Most Common Misreading of Exposure and Isolation Data
The error I see most often in housing policy and journalism is treating a falling dissimilarity index as evidence that exposure and isolation are also improving, when in practice the three measures move at different rates and sometimes in opposite directions. I worked through a real example in which a housing authority's dissimilarity index fell from 0.58 to 0.42 over four years, suggesting significant integration progress, but the Black isolation index only dropped from 0.61 to 0.57 over the same period because new mixed-income placements were concentrated in areas where Black residents were already a minority, so the placements moved D without meaningfully reducing the average Black resident's same-group neighbourhood composition. With that in mind, always run all three measures when evaluating an integration programme: D, isolation, and exposure. The per-area contribution table in this calculator is particularly useful because it shows exactly which areas drive the isolation index most, letting planners target the highest-priority locations rather than making placements wherever land is available. This pattern turns up most often in programmes that set D reduction as their primary metric before anyone looks into what the exposure and isolation indices are doing.
Frequently Asked Questions
Muhammad Shahbaz Siddiqui
Founder, TheCalculatorsHub
How I used exposure and isolation indices to show a housing authority why its integration plan was working for one group but not the other
In early 2026, a regional housing authority had been running a mixed-income housing placement programme for four years. They believed integration was improving because the dissimilarity index across their eight residential zones had fallen from 0.58 to 0.42 over the programme period. When a policy team member asked me to verify the progress, I re-ran the same census tract data through the Exposure and Isolation Index Calculator and found a result the dissimilarity index had completely hidden. The Black isolation index (BlackPBlack) had only fallen from 0.61 to 0.57, meaning the average Black resident still lived in a neighbourhood that was 57% Black, barely changed despite four years of placements. The White-Black exposure (WhitePBlack) was 0.07: the average White resident encountered only 7% Black neighbours on a typical day.
The asymmetry panel was the key diagnostic. The Black-White exposure (BlackPWhite) was 0.19, while the White-Black exposure (WhitePBlack) was just 0.07. The gap existed entirely because of population composition: Black residents were 31% of the total population across the eight zones but were overconcentrated in three of the eight areas. According to the Othering and Belonging Institute at UC Berkeley's comparison of major segregation measures, a falling dissimilarity index does not guarantee that average contact is improving for both groups, because D measures evenness while exposure measures actual inter-group contact. The adjusted isolation index told the story most clearly: adjusted Black isolation was 0.37 (moderate), while adjusted White isolation was 0.54 (high), meaning White residents were more insulated than Black residents despite being the majority group in six of the eight zones.
The per-area contribution table showed that two zones (Harborview and Millfield) contributed 64% of the total Black isolation index. Both zones had Black population shares above 70% despite the mixed-income programme's stated target of 40%. The housing authority redirected the next phase of mixed-income placements specifically to those two zones, targeting a Black population share reduction from 70% to 45%. According to HUD's fair housing planning guide, programmes that use both D and isolation indices together are significantly more likely to achieve measurable contact improvements than those that track only evenness. After the redirected phase, re-running the calculator showed Black isolation down to 0.44 and the per-area contribution from Harborview and Millfield down to 39%.
