Click here for search results

How are Poverty Maps Built?

The examples provided on this page, and the documents which support them, are only meant to be illustrative and do not indicate World Bank endorsement of any data, methodologies, or geographic boundaries.

This section briefly presents building blocks of poverty maps:

  • Information sources: Poverty maps can be built using censuses, surveys, administrative data and other sources of information.
  • Geographic Information Systems (GISs): GISs are software programs which allow to display information on the basis of their geographic coordinates. They allow to combine information from heterogeneous sources.
  • Small area estimation: A technique developed by the World Bank to combine information from surveys (which contain comprehensive information) and censuses (which allow fine disaggregation). This allows to present detailed information on poverty that is sufficiently disaggregated to capture heterogeneity.

Information sources

Surveys and Censuses

The main sources of information on socioeconomic indicators are typically censuses and surveys (see Types of Data for a guide to the different types of data available and Accessing Surveys for information on how to access some of these sources). Surveys typically provide very comprehensive information on a broad range of dimensions of living standards and their determinants or correlates. For instance, multi-topic household surveys typically report information on income and/or consumption, while Demographic and Health Surveys present information on anthropometric measures and other indicators of health status. On the down side, surveys only cover a relatively small subset of households or individuals. Typically, estimates are only representative at relatively aggregated levels – such as regions. 

Censuses provide information on all individuals and households in a country, thereby allowing for the finest geographic disaggregation. Census data can be compiled for small administrative areas, for communities, villages, and towns. On the downside, censuses are typically not carried out very frequently (usually once a decade) and they only collect information on a limited set of indicators (in particular, income or consumption are typically not available).

These two sources can also sometimes be combined to build on their respective strengths and obtain detailed information that is representative at a very low level of disaggregation (see Small Area Estimation Maps).

Administrative data

Another critical source of information to analyze the link between welfare indicators, their determinants and interventions is administrative data (e.g. information on schools, health facilities, markets, roads, etc.). For instance, information on the transport network and its quality can be used to estimate the distance or travel time that communities face to reach essential services or to access inputs or outputs markets. The map below for example shows the number of women in reproductive age groups by travel time from health centers that offer reproductive services in a region of Madagascar. Although other factors determine actual use of facilities (e.g. quality and cost), the resulting indicators of equity in access to services are useful for efforts to improve public infrastructure.

Other sources

Other data sources can also be central to the measurement of poverty, its understanding and policy design. These include information on rainfall and agro-climatic conditions which can be used to indicate communities’ susceptibility to food shortages. Several major initiatives have developed monitoring systems to assess food security and coordinate drought relief operations. Two examples are USAIDS's Famine Early Warning System (FEWS) and the Food Insecurity and Vulnerability Information and Mapping System (FIVIMS ) coordinated by FAO.

Geographic information systems

In order to present the disaggregated information on maps, one needs to have some kind of geographic location coordinate for each observation. Geographic information systems (GISs) are computer software programs designed to handle geographically referenced data. They are essentially database management systems that use geographic location as a reference for each database record. These systems are used to integrate information from very different sources (e.g. surveys, census, administrative data, satellite images, etc.) into a single platform, where each observation is matched with the identifier of the area it covers. They also permit the analysis of spatial association between different dimensions. In particular, they permit the simultaneous analysis of variables which are observed at different levels. For instance, poverty status might be observed at the district level while climate is recorded at the level of agro-climatic zones. Or some infrastructure might serve broad areas (hospitals, major roads) while others serve smaller zones (primary school or health post). The GISs allow the simultaneous analysis of information from heterogeneous sources, as long as they have geographic location coordinates. For instance, for each village, a GIS can generate the distance to the nearest market town, the average rainfall within a 20 kilometer radius, demographic indicators, and village-level estimates of income poverty. (See Documents and Links for selected readings on GIS).

Small area estimation maps

The key to the use of poverty mapping is to present information that is sufficiently disaggregated to capture heterogeneity (Lanjouw and Özler 2002). Small area estimation is a statistical inference technique that allows estimation for very small areas, by combining information from censuses and household surveys.

A team of researchers in the Development Economics Research Group, Poverty Cluster (DECRG-PO) at the World Bank, provides technical assistance, capacity building, and various free software tools to statistical institutes in developing countries upon demand. The team is also developing a “how-to manual” for poverty mapping, platform-independent software tools, and training courses. See Poverty Research: Small Area Estimation Poverty Maps for more information.

Small area estimation maps combine the depth of information in a survey (information on consumption and/or income) with the complete spatial coverage available from a census (without detailed information on welfare). Many countries have used small area estimation to create local welfare estimates and poverty maps (see Documents and Links for readings on policy applications in various regions).

This map illustrates head count poverty index in Ecuador, aggregated by region, province, and municipality. Source: National Statistical Office of Ecuador, Hentschel et al.

The methodology of small area estimation involves imputing into population census data—which does not have consumption data—a measure of per capita consumption from household survey data—which has a sample too small for small area disaggregation. The first step consists in the estimation of regressions that model expenditure or consumption using a set of explanatory variables that are common to both the household survey and the census (e.g. household size, education, housing and infrastructure characteristics and demographic variables). These regressions are estimated at the lowest geographical level for which the survey data is representative. The second step consists in using the estimated coefficients from these regressions (including the estimated error terms associated with those coefficients) to predict expenditure or consumption for every household in the census. Basically, the coefficients are used to “predict” the expenditure or consumption level of each household on the basis of the explanatory variables that are common to the census and the survey. These household-unit data are then used to compute poverty estimates for small areas. (See Demombynes' A Manual for the Poverty and Inequality Mapper Module and Zhao's User Manual for PovMap).

Given that indirect estimation of poverty indicators implies some degree of uncertainty, small area estimation data should be compared with information that describes more general characteristics of the communities. Such information is available from the censuses themselves and from other data sources. Besides cross-checking the sensibleness of the estimated data, such comparisons bring to light correlations between welfare and characteristics such as climate, the number of livestock per capita, the distance to the nearest health facility, and the number of water pumps per community—characteristics not necessarily included in censuses and surveys.

Back to Poverty Mapping

Permanent URL for this page: