Skip to content
Back to search results

Spatial Exposure Index

Introduced by
Wendell Bell (P* indices); Douglas Massey & Nancy Denton (segregation dimensions)
Year
1954
Type
Segregation measure of the potential contact or isolation between population groups
Family
Residential segregation measurement
Font size

What does it do?

The exposure and isolation indices, written P*, measure residential segregation as the degree of potential contact between population groups across the neighbourhoods of a region.

Developed by Wendell Bell in 1954 and later codified by Massey and Denton in 1988 as the 'exposure' dimension of segregation, they answer a different question from evenness measures like the dissimilarity index: not how unevenly groups are distributed, but how much members of one group actually share neighbourhoods with members of another or only with their own. The interaction index gauges cross-group exposure while the isolation index gauges within-group concentration, each interpretable as a probability.

The intuition behind the method

Imagine a member of group X picked at random and ask: of the neighbours around this person, what fraction belongs to group Y? Average that over everyone in group X, weighting by where they live, and you get the exposure of X to Y — the typical chance an X person rubs shoulders with a Y person. Replace Y with X and you get isolation — how surrounded by their own group X members typically are. Unlike measures of even spreading, these indices care about the actual mix experienced on the ground, so they depend on group sizes: a small minority can be highly isolated only if it concentrates, whereas exposure to a large majority is high almost everywhere.

History

Wendell Bell introduced a probability model for ecological segregation in Social Forces in 1954, formalizing the P* exposure and isolation indices as the chance of intergroup or intragroup contact. The framework gained prominence through Stanley Lieberson's work in the 1980s and was systematized by Douglas Massey and Nancy Denton's influential 1988 Social Forces paper 'The dimensions of residential segregation', which placed exposure alongside evenness, concentration, centralization, and clustering as one of five distinct dimensions. Subsequent research, including spatial reformulations by Reardon, O'Sullivan, and Wong, addressed the checkerboard problem by embedding adjacency and distance-decay, yielding the spatially explicit exposure measures used today.

When to use it

Use exposure and isolation indices when the question is about potential intergroup contact or the experience of living among one's own group, rather than the evenness of a group's spread. They are the appropriate tool for studying the lived experience of segregation, for relating segregation to outcomes that plausibly operate through contact (such as language acquisition, intergroup attitudes, or access to majority-group resources), and for tracking changes in isolation as a minority grows or disperses. They are less suitable as a pure measure of distributional inequality — for which dissimilarity is preferred — and the unadjusted versions should be avoided where the checkerboard problem matters, in favour of spatially adjusted variants.

Limitations and considerations

  • Depend strongly on the relative sizes of the groups, complicating comparisons across cities or over time.
  • The basic indices ignore the spatial arrangement of units (the checkerboard problem) unless adjusted.
  • Like all area-based measures, they are sensitive to the modifiable areal unit problem.
  • Measure exposure and isolation but not the evenness dimension, so they give only a partial picture of segregation.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Directly capture potential intergroup contact, the dimension of segregation most tied to lived experience.
  • Each value has a clear probabilistic interpretation between 0 and 1.
  • Distinguish exposure to other groups from isolation within one's own group, two complementary perspectives.
  • Spatially adjusted versions can incorporate cross-boundary contact and the geography of neighbourhoods.

Applications

  • Measuring the typical exposure of an ethnic minority to the majority population across a city's neighbourhoods.
  • Tracking changes in the residential isolation of a group as it grows or disperses over decades.
  • Relating segregation to outcomes such as language acquisition, schooling, or intergroup attitudes.
  • Comparing the contact dimension of segregation across metropolitan areas alongside evenness measures.
  • Assessing school or workplace segregation in terms of potential contact between groups.

How it works

How it works

1

The interaction (exposure) index P*_xy is the population-weighted average, over all areal units i, of the proportion of unit i that belongs to group Y. Each unit's contribution is weighted by the share of all group-X members who live there (x_i over total X) and multiplied by group Y's local share (y_i over the unit total t_i). The result is the probability that a randomly chosen member of group X shares their unit with a member of group Y.

Formula
P^*_{xy} = \sum_i \frac{x_i}{X}\cdot\frac{y_i}{t_i}
2

The isolation index P*_xx uses the same construction but multiplies group X's residential weight by group X's own local share x_i over t_i. It is therefore the probability that a randomly chosen member of group X shares their unit with another member of group X — a direct measure of how much the group is exposed only to itself. Exposure to all groups including one's own sums to one, so isolation and the exposure indices to other groups are complementary.

Formula
P^*_{xx} = \sum_i \frac{x_i}{X}\cdot\frac{x_i}{t_i}
3

The basic P* indices treat each areal unit in isolation and so suffer the 'checkerboard problem': a perfectly alternating pattern of group-X and group-Y units scores the same whether the units are intermingled or segregated into two halves, because only within-unit composition counts. Spatial versions correct this by incorporating neighbouring units through distance-decay or adjacency weights, so that contact across unit boundaries contributes to measured exposure and the index reflects the true geography of potential interaction.

4

Each P* value is a probability between 0 and 1: the chance that a group member's randomly chosen co-resident in their unit belongs to the reference group. High isolation (near 1) means a group lives largely among itself; high cross-group exposure means frequent potential contact. Because the indices depend on the relative sizes of the groups, they should be interpreted alongside group proportions and, ideally, alongside an evenness measure such as the dissimilarity index to separate exposure from uneven distribution.

Common pitfalls

  • Comparing isolation across places with very different minority shares as if the index were size-neutral.
  • Using the unadjusted index where adjacency matters, so segregated and intermingled checkerboards score alike.
  • Confusing exposure (P*) with evenness measures such as the dissimilarity index, which answer a different question.
  • Ignoring how the choice of areal units inflates or deflates measured isolation.

Frequently asked questions

What is the difference between the exposure index and the dissimilarity index?

They measure different dimensions of segregation. The dissimilarity index gauges evenness — how unequally a group is distributed across units, independent of group size — and answers what share of one group would have to relocate for an even distribution. The exposure index P* gauges potential contact — the probability that a group member shares a unit with the same or another group — and depends heavily on the relative sizes of the groups. A city can be highly uneven yet, if the minority is tiny, still leave the majority with little exposure to it.

What is the checkerboard problem and how is it addressed?

The checkerboard problem is that the standard P* indices use only the composition within each unit and ignore which units are adjacent, so a perfectly alternating arrangement of group-X and group-Y units scores identically whether those units are intermingled or split into two homogeneous halves. Spatial exposure measures fix this by weighting the contribution of nearby units through adjacency or distance-decay functions, so that contact across unit boundaries is counted and the index reflects the actual spatial pattern.

Why do isolation and exposure depend on group size?

Because they are probabilities of contact, they are mechanically tied to how common each group is. A member of a large group will, on average, encounter their own group frequently almost regardless of spatial arrangement, so isolation tends to be high simply from size; conversely a small minority can only be highly isolated if it strongly concentrates. This size dependence is why analysts report exposure indices together with group proportions and often alongside a size-independent evenness measure like dissimilarity.

Related methods

Sources

  1. Bell, W. (1954). A probability model for the measurement of ecological segregation. Social Forces, 32(4), 357–364.

    Open DOI reference
    Copied.
  2. Massey, D. S., & Denton, N. A. (1988). The dimensions of residential segregation. Social Forces, 67(2), 281–315.

    Open DOI reference
    Copied.