Exposure & Isolation Index Calculator

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.

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.