Poverty Probability Index
- Introduced by
- Mark Schreiner; Grameen Foundation (now Innovations for Poverty Action)
- Year
- 2005
- Type
- Poverty-likelihood scoring instrument
- Family
- Poverty targeting and scoring
What does it do?
The Poverty Probability Index (PPI), formerly the Progress out of Poverty Index, is a simple, country-specific scorecard that estimates the likelihood that a household is living below a given poverty line.
Developed by Mark Schreiner and disseminated first by the Grameen Foundation and later by Innovations for Poverty Action, it reduces poverty measurement to ten easy-to-answer, verifiable questions about household characteristics. The answers produce a score from 0 to 100, which a calibration table converts into the probability that the household falls below national or international poverty lines — a low-cost alternative to a full consumption survey for organizations that need to track the poverty profile of the people they serve.
The intuition behind the method
A proper poverty assessment means a long expenditure survey that few field organizations can run on every client. The PPI accepts a trade-off: instead of measuring consumption directly, it picks a handful of household characteristics that, in a national survey, best predict whether a household is poor — things like family size, children's schooling, the materials of the dwelling, and ownership of a few key assets. Each answer carries pre-computed points; a field worker adds them to a score out of 100. The clever step is the calibration table: built from the same national survey, it says, for each score, what fraction of households with that score were actually below the poverty line. So a score is not a measure of poverty itself but an address that points to a poverty probability.
History
The PPI grew out of Mark Schreiner's work on credit scoring for microfinance, set out in the 2002 CGAP Occasional Paper 'Scoring: The Next Breakthrough in Microcredit?', which argued that the statistical scoring used in consumer credit could be adapted to predict poverty. The Grameen Foundation launched the Progress out of Poverty Index in the mid-2000s, building certified scorecards country by country from national household surveys. In 2016 stewardship passed to Innovations for Poverty Action and the tool was renamed the Poverty Probability Index to emphasize that it estimates a likelihood rather than a definitive status. Hundreds of organizations across dozens of countries have used it for social-performance management.
When to use it
Use the PPI when an organization — a microfinance institution, social enterprise, or NGO — needs a cheap, fast, repeatable estimate of the poverty profile of its clients or beneficiaries and cannot run a full consumption survey. It is well suited to targeting, segmenting, and tracking poverty outreach over time at the group level, and to social-performance reporting. It must use the certified, up-to-date scorecard for the specific country and poverty line, since the calibration is country- and survey-specific. It is not designed to classify an individual household as definitively poor or non-poor (it gives a probability, not a verdict), nor to replace official poverty statistics or measure non-monetary deprivation.
Limitations and considerations
- It estimates a probability for groups, not a definitive poor/non-poor classification for an individual household, so it is weak for hard eligibility decisions.
- Each scorecard is country- and line-specific and must be recalibrated as new national surveys appear, or it drifts out of date.
- Predictive accuracy depends on the strength of the indicator-poverty relationship in the original survey, which varies by country and poverty line.
- It measures monetary poverty likelihood only and captures no multidimensional or non-income deprivation.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Extremely low cost and fast: ten simple questions and a sum let non-expert staff estimate poverty without a full consumption survey.
- Statistically grounded — indicators and points are derived from a national survey calibrated against an actual poverty line, not chosen by intuition.
- Outputs an interpretable poverty probability and group poverty rate that can be tracked over time and compared to national and international lines.
- Standardized, certified, country-specific scorecards make results comparable across organizations using the same instrument.
Applications
- Microfinance institutions estimating and tracking the poverty profile of their borrowers for social-performance reporting.
- Social enterprises and NGOs targeting and segmenting clients by poverty likelihood without running consumption surveys.
- Monitoring movement of a client base across a poverty line over time as a program-impact signal.
- Donors and investors verifying that an organization is reaching its intended poor population.
- Designing simple proxy-means-test-style targeting where a full means test is impractical.
How it works
How it works
Construction starts from a recent national income or expenditure survey in which true poverty status (below a chosen line) is known. From dozens of candidate household characteristics, a small set — typically ten — is chosen that jointly predict poverty well while being simple to ask, easy to verify, and stable over time. The selection balances statistical power against practicality, favoring non-sensitive, objective indicators (household size, schooling, housing, durable assets) that a non-expert can collect in minutes.
The relationship between the candidate indicators and poverty status is estimated statistically — historically via logistic or related regression of being poor on the indicators — to learn how much each answer signals poverty. The fitted relationship is then transformed into a simple, integer point system: each possible answer to each question is assigned non-negative points so that the points sum to a score on a 0-to-100 scale, with higher scores indicating lower likelihood of poverty. This conversion trades a little statistical precision for great ease of field use.
Formula
\Pr(\text{poor}_i = 1 \mid x_i) = \frac{1}{1 + e^{-(\beta_0 + \sum_j \beta_j x_{ij})}}In the field, a household is scored by asking the ten questions and summing the points for its answers, S_i = sum of w_j times x_ij, giving an integer score between 0 and 100. No computation beyond addition is required, so the instrument can be administered on paper or a phone by enumerators with minimal training, which is the central practical advantage of the design.
Formula
S_i = \sum_{j=1}^{J} w_j\, x_{ij}, \qquad S_i \in [0, 100]The score is converted to a poverty likelihood using a lookup table calibrated on the original survey: for each score (or score band) the table reports the share of households at that score that were below the poverty line. Averaging these likelihoods over a sample of clients estimates the group's poverty rate, and re-measuring over time estimates movement across the line. The table is the only thing that carries poverty information; the score by itself is just an index pointing into it.
Common pitfalls
- Using an outdated or wrong-country scorecard, so the score-to-likelihood calibration no longer reflects the population being measured.
- Interpreting an individual household's score as a definitive poverty classification rather than as a probability to be averaged over a sample.
- Adapting or dropping questions from a certified scorecard, which invalidates the calibrated point system and the lookup table.
- Comparing PPI poverty rates across countries or across different poverty lines as if a score meant the same thing everywhere.
Frequently asked questions
Does a PPI score tell me whether a specific household is poor?
No. The PPI gives a probability, not a verdict. A score points into a calibration table that says what fraction of households with that score were below the poverty line in the reference survey. For a single household the right reading is 'this household has, say, a 60 percent likelihood of being below the line', and the tool is designed to be averaged over a sample to estimate a group poverty rate. Using it to make a yes/no eligibility decision for one household goes beyond what the instrument was built to do.
Why must I use the official scorecard for my country?
Both the choice of indicators and the points attached to each answer, as well as the score-to-likelihood lookup table, are estimated from a specific national household survey and a specific poverty line. Those relationships differ across countries and over time, so a scorecard built for one country or an old survey will mis-predict poverty elsewhere. Certified, country-specific scorecards are published and periodically refreshed as new national surveys become available, and using the current certified version is what makes the estimates valid and comparable.
How does the PPI relate to a proxy means test?
Both predict an unobserved welfare measure from easy-to-observe household characteristics. A proxy means test is usually used to make individual eligibility decisions for a program and predicts a continuous welfare level, whereas the PPI is calibrated to output the probability of being below a poverty line and is intended for estimating and tracking poverty rates at the group level. They share the same statistical logic but differ in purpose: classification versus poverty-rate measurement.
Related methods
Sources
Schreiner, M. (2016). The Poverty Probability Index (PPI): A Brief on Calculating Annual Poverty Rates and Movement Across a Poverty Line. Innovations for Poverty Action / PovertyIndex.org.
Open publication referenceCopied.Schreiner, M. (2002). Scoring: The Next Breakthrough in Microcredit? CGAP Occasional Paper 7. Consultative Group to Assist the Poor, Washington, DC.
Open publication referenceCopied.