Agreeing on a definition of poverty is central to producing valid and reliable measures. Defining poverty, and a normative adequate standard of living, also carries practical significance for targeting assistance and tracking outcomes. Common measures like Gross Domestic Product (GDP) and the Gini coefficient are aggregates and cannot differentiate the global poor from the non-poor, nor were they intended to. Poverty thresholds like the population below $1.90-per-day use a monetized consumption or income threshold and focus on low-income individuals and households, but are criticized for excluding poverty along other measures, such as health and educational deprivations. Additionally, some argue that these thresholds represent an extreme level of deprivation and thus generally do not accurately represent poverty levels or lead to programs that meaningfully decrease poverty. In light of the ongoing debate over the definition of poverty and the seemingly many competing measures, EPAR reviewed 36 poverty indicators to analyze some of the trade-offs.
At the lower $1.90/day consumption threshold (red line), poverty has declined between 1990 and 2018, from 55% to 39% of the population, though the overall headcount has increased by almost 50%, rising from 280 million to over 420 million. Some researchers argue that the higher $5.50 and $7.40 thresholds better reflect levels of poverty. These higher threshold ratios (blue and purple lines) have only declined slightly while the number of people experiencing poverty at these thresholds has more than doubled. Data come from the World Bank’s PovCalNet.
Landscape Assessment of Indicators
We searched for poverty indicators and identified 139 candidates; of these, 36 indicators met the following study inclusion criteria:
-
Must be current,
-
Must be a calculated measure that is not merely a proposal, as evidenced by discussion or adoption by organizations, and
-
Must either have a monetary component or measure a proxy for poverty such as food security, assets, or economic opportunity.
The included indicators utilize at least 26 different sources of household and administrative data, though household-level survey data are most commonly used for indicator construction. Indicators are typically based on quantitative estimates of income or consumption; however, an increasing number of measurements classify households according to deprivation of assets, food, or access to services and basic infrastructure. Most indicators in our review are multidimensional indices with the poverty line and poverty gap measures being key exceptions that rely exclusively on income and consumption to identify poverty.
Indicators were evaluated on the bases of calculation simplicity, simplicity of interpretation, comparability over time and geography, organizational uptake, number of deprivations, and whether a given measure reflects poverty rates (#/population), severity or depth (distance from a poverty line), or both. Not surprisingly, no single indicator dominates on all criteria. Rather, indicator selection requires considering tradeoffs for a given use case. Indicators that rank highest for organizational uptake tend to have high levels of comparability across time and geography, despite data and context differences. Simplicity of direct interpretation, rather than calculation, tends to be strongly associated with organizational uptake, though indicators meeting these criteria are also some of the oldest indicators, which influence utilization. Some emerging indicators based on complex methods with lower current utilization such as the Facebook Relative Wealth Index and machine-learning guided classification of satellite imagery may gain organizational traction in the future.
Data Collection Opportunities
Given the heavy reliance on household-level survey data for the construction of many widely used poverty indicators, opportunities exist to enhance poverty indicator utility via increased survey data collection efforts. Household survey data are particularly useful for indicator construction since they tend to be relatively accessible, easy to work with, and easily disaggregated by household demographic traits like gender and age. However, these data are time- and resource-intensive to collect, which limits their geographic coverage, especially in countries where poverty rates may be highest, and reduces survey frequency and their timeliness for informing the policy-making process. The reliability, multidimensionality, and comparability of poverty indicators could thus be strengthened by more frequent and widespread household survey implementation cycles.
Access the paper and supporting spreadsheet for EPAR Project #424 here.
By Lucero Marquez and Helen Ippolito
Summarizing original EPAR research by Kelsey Figone, Ushanjani Gollapudi, Helen Ippolito, Lucero Marquez, Aline Meysonnat, Andrew Tomes, Sebastian Wood Vilaseca, Jacob Wall, and C. Leigh Anderson.