A number of different surveys and other data sources can be used for analyzing income poverty and its correlates (see Data and Tools). In this page, we try and present the various types of analysis that can be done with different sources.
The discussion below refers to the attached Table 1.24: Income Poverty: Data Availability and Analyses Tools. The Table distinguishes cases of severe data limitation (#1) to a good data situation (#9). The data sources discussed and ranked include the population census, rapid monitoring surveys, income and expenditure surveys, demographic and health surveys, and multi-topic surveys.
Based on data availability, the Table identifies which tools can be used for poverty analysis. Income/expenditure/consumption poverty measurement is possible only if at least one multi-topic or income and expenditure survey exists. Other data sources–-such as a population census, demographic and health surveys, and rapid monitoring surveys–-do not lend themselves to poverty measurement. Even in cases where income and consumption poverty measurement is not possible, as the Table illustrates, several analysis tools can be applied that are important for policy making. For example, spatial poverty maps can in most cases be developed using proxies for income or consumption. Rapid monitoring surveys and demographic and health surveys also lend themselves to developing a basic profile of the poor.
Still, although many different surveys can be and are used for poverty and welfare analysis, it should be emphasized that a multi-topic survey is a key tool for measuring and understanding a wide range of issues related to poverty. In the short run, demographic and health surveys or more specialized surveys can supply important information but in the long run, the availability of a multi-topic survey is essential.
Apart from the type of survey available, it matters whether analysts have access to only one single cross-section of data, several cross-sections, or panel data. In principle, insights into the dynamics of poverty require the availability of several multi-topic household data sets collected at different times. Such information allows for measuring changes in poverty as well as the underlying characteristics causing these changes (cases #8 and #9).
In countries where only one cross-sectional survey is available (cases #5 and #7),quasi-panel data can sometimes be derived if income and consumption are recorded at different points in time. Surveys sometimes record information on demographics, activities, and income in a first visit, and repeat the income module quarterly thereafter, for a year. Also, some surveys ask households to recollect their income or consumption for previous time periods. Even when no quasi-panel components are available, it may be possible to build measures of household vulnerability that rely on the variation within communities or other subgroups, or on external information on the seasonality of prices and production.
More can be done when two or more cross-section surveys are available (cases #6 and #8), because changes in the levels and patterns of poverty over time can be analyzed. As mentioned earlier, poverty comparisons over time require careful analysis, but they give insights into the dynamics of poverty and its determinants, and they can be used for evaluation. While repeated cross-sections reveal trends for population groups, they but do not allow the tracking of individuals or households over time. They reveal aggregate changes, but they do not capture individual movements into or out of poverty.
Panel data (#9) follow the same individuals or households over time, so that one can relate their patterns of consumption and income to changes in other characteristics, such as demographics, migration, labor market situation, durable goods ownership, access to services, and health and education status. Panel data have advantages over repeated cross-sectional surveys. They permit the analysis of the factors that underlie mobility. They also record information on past events more precisely than the retrospective questions sometimes included in cross-sectional surveys, and they help in assessing the impact of public programs and services on poverty outcomes. Only panel data allow analysis of the determinants of poverty, while cross-sectional data are limited to revealing correlates of poverty. Correlates are characteristics that are found to be closely linked to poverty-- for example, family size might be linked to poverty--but no causality pattern can be inferred from their analysis. For example, it is impossible to say whether a family is poor because it is large or whether a family is large because it is poor. On the contrary, determinants of poverty provide information on the causes of poverty and can be analyzed by looking at households over time and analyzing their welfare changes in light of their characteristics.
Some limitations of panel data are that households can change over time, disappear entirely from the sample (because of death or migration), or split or regroup because children grow up or household members are married or divorced. If the disappearance from the panel (attrition) is linked to certain characteristics—for example households with good education move away from poor neighborhoods—then the estimation results of panel regressions have to be treated with care. Also, as time passes by, panel surveys can become less representative if they fail to include new members of the population--new births or immigrants. As other surveys, panel data can also suffer from measurement errors, especially those related to household income and consumption, which can affect the quality of mobility statistics.
- See Data and Tools for links to actual poverty, links to household surveys that can be used to study poverty and an explanation of different types of data and data sources.
Back to Measuring Poverty