Skip to content
Back to search results

Spatial Poverty Mapping

Introduced by
World Bank poverty-mapping programme; Bedi, Coudouel & Simler
Year
2007
Type
Spatial-statistical and GIS method for analysing poverty distribution
Family
Spatial poverty analysis
Font size

What does it do?

Spatial poverty mapping visualises and analyses the geographic distribution of poverty using geographic information systems and spatial statistics, turning poverty estimates into maps that reveal where the poor live at fine spatial scales.

It combines small-area poverty estimates with spatial covariates — remote-sensing data, night-time lights, accessibility, and terrain — examines spatial patterns and autocorrelation, and supports the geographic targeting of resources. Consolidated through the World Bank programme documented by Bedi, Coudouel, and Simler and energised by data such as the satellite night-lights series analysed by Henderson, Storeygard, and Weil, it has become a standard tool for evidence-based geographic targeting.

The intuition behind the method

Poverty is not spread evenly across a country; it clusters — in remote uplands, drylands, urban slums, and places cut off from markets and services. National poverty rates, and even provincial averages, hide this geography. Spatial poverty mapping makes it visible by estimating poverty for small areas and drawing it on a map, so that hotspots and the spatial correlates of deprivation — distance from roads, dim night-lights, poor soils, low rainfall — stand out. Once you can see where poverty concentrates, you can target scarce resources to those places rather than scattering them uniformly. The map turns an abstract statistic into a spatial decision tool, while also forcing analysts to ask whether nearby areas resemble one another (spatial autocorrelation) and whether the patterns are real or artefacts of the underlying estimation.

History

Spatial poverty mapping advanced in the late 1990s and 2000s when small-area estimation methods, notably those of Elbers, Lanjouw, and Lanjouw, made it possible to combine surveys and censuses into fine-grained poverty estimates that could be mapped. The World Bank and partners produced national poverty maps across many countries, and the 2007 volume edited by Tara Bedi, Aline Coudouel, and Kenneth Simler, More Than a Pretty Picture, drew together the methods and policy uses. The parallel growth of geospatial data — land cover, accessibility, and especially night-time lights, whose value as an economic proxy was demonstrated by Henderson, Storeygard, and Weil in 2012 — expanded the covariates and predictive possibilities, and machine-learning approaches later extended the field.

When to use it

Use spatial poverty mapping when you need to know where poverty concentrates within a country at finer scales than survey averages allow, and when programmes, transfers, or infrastructure are to be allocated geographically. It is well suited to designing targeted interventions, prioritising regions, and combining poverty estimates with spatial data on environment, accessibility, and economic activity. It is less appropriate when within-area heterogeneity is so high that geographic targeting would misallocate to many non-poor and miss poor households, when the underlying small-area estimates are too uncertain to map credibly, or when the map risks being read as more precise than the estimation supports.

Limitations and considerations

  • Maps inherit the uncertainty of the underlying small-area estimates, which is easily forgotten once drawn.
  • Geographic targeting misses poor households in better-off areas and includes non-poor in poor areas.
  • Choices of spatial scale and boundaries (the modifiable areal unit problem) can change the apparent pattern.
  • Requires linked census, survey, and geospatial data and technical capacity that many settings lack.

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

Strengths

  • Reveals the spatial concentration of poverty that national and regional averages obscure.
  • Enables evidence-based geographic targeting of programmes, transfers, and infrastructure.
  • Integrates poverty estimates with rich spatial covariates such as night-lights, accessibility, and land cover.
  • Communicates complex distributional information vividly to policymakers and the public through maps.

Applications

  • Geographic targeting of social transfers and poverty programmes to high-need areas.
  • Prioritising the location of infrastructure, schools, and health facilities.
  • Identifying poverty hotspots and their environmental and accessibility correlates.
  • Combining night-lights and remote sensing to estimate well-being where survey data are scarce.
  • Informing decentralised budget allocation and regional development planning.

How it works

How it works

1

The analysis begins by assembling geographically referenced poverty estimates, typically small-area estimates that combine detailed but small household surveys with comprehensive census data to produce poverty rates for administrative units far smaller than the survey alone could support. These estimates are joined to spatial boundaries to create the base poverty layer that everything else builds on.

2

Spatial covariates are overlaid on the poverty layer: remote-sensing products such as land cover and vegetation indices, night-time light intensity (which, as Henderson, Storeygard, and Weil showed, proxies for economic activity), market accessibility and travel time, elevation and terrain, rainfall, and infrastructure. These layers both help explain the geography of poverty and can serve as predictors where direct poverty estimates are sparse.

3

Spatial statistics characterise the patterns in the mapped data, testing whether poverty is spatially clustered rather than randomly distributed. Global measures such as Moran's I, defined as I = [n / (sum_i sum_j w_ij)] times [sum_i sum_j w_ij (x_i - x-bar)(x_j - x-bar)] / [sum_i (x_i - x-bar) squared] with spatial weights w_ij, quantify spatial autocorrelation, while local indicators identify specific hotspots and cold spots, guarding against treating spatially dependent data as if observations were independent.

Formula
I = \frac{n}{\sum_i\sum_j w_{ij}}\frac{\sum_i\sum_j w_{ij}(x_i-\bar{x})(x_j-\bar{x})}{\sum_i (x_i-\bar{x})^2}
4

Drawing the analysis together, the maps and pattern statistics inform geographic targeting — directing programmes, transfers, infrastructure, and services to the areas of greatest need. As Bedi, Coudouel, and Simler emphasise, a poverty map is more than a pretty picture: it becomes useful only when it feeds concrete decisions about where to allocate resources, balanced against the targeting errors and political-economy constraints that geographic allocation entails.

Common pitfalls

  • Presenting a smooth, confident map while hiding the wide confidence intervals of the small-area estimates beneath it.
  • Ignoring spatial autocorrelation and treating areal observations as independent in subsequent analysis.
  • Changing conclusions by switching spatial units without acknowledging the modifiable areal unit problem.
  • Treating the map as the decision rather than as one input weighed against targeting errors and local realities.

Frequently asked questions

How is spatial poverty mapping related to small-area estimation?

Small-area estimation is usually the engine that produces the poverty numbers a map displays. It combines a detailed household survey, which measures consumption or poverty well but only for a sample, with a full census, which covers everyone but lacks income detail, to predict poverty rates for small administrative units. Spatial poverty mapping then visualises and analyses those estimates geographically, adds spatial covariates, and tests for clustering. The mapping is only as reliable as the small-area estimates underneath it, so their uncertainty should travel with the map.

What is Moran's I and why does it matter here?

Moran's I is a global measure of spatial autocorrelation — the degree to which nearby areas have similar values. It ranges roughly from -1 (a checkerboard of dissimilar neighbours) through 0 (random) to +1 (strong clustering of similar values). Poverty is typically positively autocorrelated, with poor areas next to poor areas. Measuring this matters both substantively, because clustering justifies geographic targeting, and statistically, because ignoring spatial dependence violates the independence assumptions of ordinary regression and biases inference.

Why are night-time lights used in poverty mapping?

Henderson, Storeygard, and Weil showed that the intensity of night-time lights captured by satellites is a useful proxy for economic activity, correlating with income and growth. In poverty mapping, night-lights and other remote-sensing layers serve as low-cost, frequently updated, globally available covariates that help explain the geography of poverty and predict well-being in places where household survey data are sparse, outdated, or unavailable, complementing rather than replacing direct poverty estimates.

Related methods

Sources

  1. Henderson, J. V., Storeygard, A., & Weil, D. N. (2012). Measuring Economic Growth from Outer Space. American Economic Review, 102(2), 994-1028.

    Open DOI reference
    Copied.
  2. Bedi, T., Coudouel, A., & Simler, K. (Eds.). (2007). More Than a Pretty Picture: Using Poverty Maps to Design Better Policies and Interventions. Washington, DC: World Bank.

    ISBN: 9780821369319
    Copied.