What are the USAID Poverty Assessment Tools (PAT)?

The USAID Poverty Assessment Tools include short, country-specific surveys that gather household data on indicators that have been identified as the best predictors of whether a given set of households is very poor, according to the legislative definition of extreme poverty applicable to the country in question. The data gathered through these surveys is then entered into a template, at which point a software package (Epi Info) estimates the share of households living below the applicable poverty line. The tools greatly simplify the lengthy and time-consuming household budget surveys or composite surveys, which request information on hundreds of potential indicators of income and consumption, focusing instead on a small number of indicators that match the results obtained by these longer surveys. Each tool is meant to be administered in twenty minutes or less, and produce data which can be easily used by partner organizations to determine generally what percentage of clients fall into the definition of very poor according to the legislation.

What is the mandate for the Poverty Assessment Tool development project?

Recent U.S. Congressional law requires USAID to develop and certify at least two tools for assessing the poverty level of its microenterprise beneficiaries.

USAID/EGAT/MD contracted IRIS to develop, test and disseminate poverty assessment tools that meet Congressional requirements for accuracy and practicality. The Developing Poverty Assessment Tools project began September 16, 2003, and the first phase (PAT I) ended October 31, 2006.

Evidence from the first phase of the project indicated that country-specific tools tend to be much more accurate than international tools; attempting to use the same tool in multiple countries yields very inaccurate poverty estimates even if those countries have comparable income levels. On the basis of that evidence, USAID accepted IRIS’recommendation that a specific poverty assessment tool be developed, tested, and regularly updated for each country with substantial microenterprise activity funded by USAID.

The project is now developing additional country tools as a second phase of the Developing Poverty Assessment Tools project (known as PAT II) that will continue until 2010.

How were the practicality and accuracy of the USAID PATs tested?

The testing methodology developed for the poverty assessment tools project involved two rounds of testing: tests of accuracy and field tests of practicality. The accuracy tests involve gathering data on a large number of potential poverty indicators and also a benchmark value for household consumption expenditures or household income on the same households. Statistical methods are then used to test how well various indicators, and various estimation methods, perform in estimating poverty levels. The accuracy of these predictions were gauged by both existing measures and innovative methods developed by the IRIS team over the course of the project.

The tests of practicality ran from late 2004 to mid 2006 and involved microenterprise practitioners implementing multiple prototype tools in order to judge the practicality of gathering household data on the best-performing indicators, as well as the practicality of implementing the process as a whole.

What data are used to develop each country tool?

A country-specific poverty assessment tool can be developed from either existing survey data or an IRIS-managed survey specifically designed for that purpose. Of the twelve tools completed in the original PAT project, eight were from existing Living Standard Measurement Surveys (LSMS) coordinated by the World Bank, while the remaining four country tools were developed by IRIS using original data collected in field surveys. In either case, the data must include a household consumption expenditure or income benchmark by which to classify households as truly very poor or not very-poor, along with indicators about the household (on demographics, housing characteristics, consumer durables or assets owned, and other categories). From this initially large pool of indicators, statistical methods are used to identify which 10 or 15 provide the strongest clues as to the poverty status of each household. The resulting short list of indicators provides the basis for the poverty assessment tool for that country.

How were indicators selected for each USAID PAT?

Indicators for each country-specific USAID Poverty Assessment Tool (PAT) are selected from large sets of indicators found in national poverty assessment surveys. These surveys are administered on a nationally representative sample of a country’s population, and the data collected from them reflects how the majority of a country’s population would answer specific questions. Analysis could be conducted on how the majority of the population of a specific region of a country would answer a question, but the PATs were constructed using only nationally representative data. The purpose of using national data instead of regional data is to achieve the goal of creating one tool per country which assesses poverty in a standard way throughout the entire country.

Indicators were selected by running a statistical program that selected the top fifteen indicators out of the entire pool of indicators from a national poverty assessment survey that correlated with a poverty level that defines a household as “very poor.” The key objective of each tool is to differentiate the “very poor” from the “not very poor.”

How does the PAT development team decide which country tools to develop next?

Prioritization of countries for tool development is conducted through discussions between the IRIS project team and USAID's Office of Microenterprise Development. Countries with higher levels of USAID microenterprise funding are assigned higher priority to receive an assessment tool. Among these countries, the project team searches for suitable, existing integrated household expenditure or income surveys. If none can be found for a high-priority country, the IRIS Center arranges its own household field survey to collect the necessary data for developing an assessment tool.

Because creating a tool from existing data is less time consuming, to date these countries have taken priority in order to get as many tools ready as quickly as possible. However, especially as suitable existing data sources are exhausted, the development team will also pursue original survey data collection in the future.

What do the USAID PATs measure?

The USAID Poverty Assessment Tools (PATs) measure poverty by estimating the percentage of a client population who are “"very poor" that is, the percentage who fall below the extreme poverty line of either a country's national poverty line or the International Poverty Line (PPP$1/day), depending on the particular country. Using simple questions and accurate models, the USAID PATs categorize client households as “"very poor”" or "not very poor"”—in other words, as being above and below the extreme poverty line—based on a short client surveys and an accompanying data entry program.

How do the PATs measure poverty? How do the PATs define "poverty"?

The USAID Poverty Assessment Tools (PATs) define poverty either by the median poverty line (with "very poor" defined as the bottom 50% below the national poverty line) or by the $1/day measure, also known as the International Poverty Line. The USAID PATs take the higher threshold for each particular country tool.

What is the median poverty line? How is it determined?

Poverty lines are generally set by national governments, and used together with household survey data to measure the prevalence of poverty. Median extreme poverty lines are set by the USAID Poverty Assessment Tools team using national survey data. These lines correspond to the amount of per capita income or consumption expenditures that divides the poorest half of those living below the national poverty line from the less-poor half of the officially poor.

What is the $1/day or International Poverty Line? How is it determined?

The international poverty line was first developed by World Bank researchers working on the 1990 World Development Report (WDR), which focused on poverty. In order to estimate the share of the world's population living in extreme poverty, the WDR team examined the national poverty lines that low-income countries were themselves using to track poverty among their own citizens. To compare these national poverty lines, they first had to be converted to a common currency; for this conversion, the WDR team used Purchasing Power Parity (PPP) exchange rates, which are adjusted for differences in the purchasing power of currencies in domestic markets (more on PPP exchange rates below). Once this adjustment was made, it turned out that the poverty lines of the poorest countries spanned a range between $275 and $370 per year in PPP terms at 1985 prices. These two values were used as the basis for constructing the first solid estimates of the global prevalence of extreme poverty in reference to a common poverty line. Because the higher of these two values was so close to $1 per day, it soon became known as the “dollar a day line.”

In 2000, the World Bank significantly revised the international poverty line, in order to take advantage of new PPP data covering a much larger number of developing countries. The researchers set the new line at the median among the national poverty lines of the 10 poorest countries. The resulting line was $1.08 per day at 1993 PPP- that is, at purchasing power parity based on 1993 prices. Although the new line differs from the original version, it continues to be called the “dollar a day line,” and serves as the common international standard of extreme poverty.

It is essential to notice that PPP exchange rates tend to be very different from standard market exchange rates, especially for countries at very different levels of real income. For example, at mid-2005 prices, the market exchange rate to the US dollar was 1.4 times the PPP exchange rate in Albania, 2.5 times in Bolivia, 3.7 times in Mozambique and Uganda, and 5.5 times in Vietnam. That means that using market exchange rates to compute the local currency equivalent of the "$1 a day line" produces values completely different than those resulting from using PPP exchange rates as required in the legislation. The main source of the difference is that many services tend to be relatively much cheaper in poor countries than in rich countries; because those services are not traded internationally, the difference in their relative cost affects PPP exchange rates but not market exchange rates.

The key to computing these values is the 1993 consumption PPPs published by the World Bank at http://iresearch.worldbank.org/PovcalNet/jsp/index.jsp. To convert these values into the local-currency equivalent of the international poverty line at current prices, multiply the published PPP by 1.08, and then by the ratio of the current Consumer Price Index to its average value in 1993.

For the convenience of practitioners, a detailed explanation of this calculation, along with recent values for a wide range low- and middle-income countries, is contained in the Annex of National and International Poverty Lines: An Overview [PDF: 106KB] by Don Sillers. A downloadable file [Excel: 90KB] shows the data and calculations used to produce that Annex.

Do the PATs measure individual client poverty?

Though they could be reconfigured to predict the poverty level of an individual household, the tools are designed to provide a statistically accurate assessment of poverty among a pool of clients, rather than of individual households. This reflects the Congressional requirement that the tools measure the total percentage of very poor households among each organization's microenterprise clients.

All tools designed to assess human and social phenomena- poverty being only one of them- are subject to measurement errors. In the case of poverty assessment, these errors are of two types: misclassifying a very poor household as not very-poor; and the opposite, misclassifying a not very-poor household as very poor. Finding ways to reduce the size of these errors in the poverty assessment tools has been a key task for IRIS. One of the accomplishments of the project is the development of statistical techniques that identify poverty indicators that ensure that the number of very poor households misclassified as not very-poor is the same as the number households misclassified in the other direction. This means that, over a large enough sample, individual misclassifications do not matter- since the two types of errors would cancel each other out, regardless of how large they are. However, these misclassifications still matter at the individual level. This explains why the same tool can be very accurate when measuring poverty at a collective level, but very inaccurate if used on an individual basis.

Consider this example: if a tool is used to report on the number of very poor clients in a group, it does not matter that two very poor clients are incorrectly classified as not very-poor, as long as two not very-poor clients are misclassified as very poor; the errors cancel out and do not affect the aggregate result. If the tool, however, is used for individual assessment (or targeting), the errors add up: four people have been incorrectly identified, with the result that two very poor clients will not receive the services designed for them, while two not very-poor clients will receive services they may not need.

How do the PATs measure poverty accurately? How do the PATs define accuracy?

The accuracy of a particular set of indicators—of a USAID Poverty Assessment Tool——is assessed by comparing the predicted poverty status with the“"true” poverty" status as established by the benchmark national survey data. This technical note [PDF:55KB] explains the project's measure of accuracy, and the different econometric techniques that the team has developed to increase the accuracy of the indicators—of each USAID country Poverty Assessment Tool.

Do the PATs measure impact?

The PATs measure the percentage of extreme poverty of a group of people at one point in time. A tool can be used to measure the poverty level of a specific group of people at one point in time, then again on the same group of people at a second point in time, but an organization must remember that even if the two measurements indicate a change in poverty level of this specific group of people between these two points in time, the results do not establish causation for the change in poverty level. The PATs only measure poverty levels; they do not measure why there has been a change in poverty level. USAID and the IRIS Center encourage organizations to use the PATs for other uses outside of reporting to MRR, but only if they have a full understanding of the limitations of the tool.

If an organization uses the PAT as one component of an impact study, it should only be used under the following conditions: a) impact is assessed on the same, sufficiently large, group of clients, and not for individual clients; b) the tools are calibrated for the country or region in question; c) the poverty line across which the movement is measured is the line used to calibrate the tools; and d) tools are kept up-to-date between the first and the subsequent measurements (so that the poverty measurement continues to capture changing economic conditions).

Even if these conditions are met, it is unclear that the tools will be able to identify real changes in poverty over time due to their inherent measurement errors. Unless the changes in poverty rate are exceptionally large and the tools exceptionally accurate, the changes identified are likely to be contained within the margin of error.

What do the USAID PATs look like?

Existing tools are drafted in an MS Excel spreadsheet, and the data entry templates are created in the Epi Info software. Both tools and templates are posted on the povertytools.org website here. As additional tools are developed, they will be posted to the same web page.