Measuring Poverty in the East African Community

Poverty is a complex, multifaceted condition. Over the years scholars, institutions and policymakers have built different frameworks to define and measure it, from narrow income-based views to broad human-development perspectives. Two frameworks dominate practice.

Examples include the global Multidimensional Poverty Index (MPI) from OPHI and UNDP, and UNICEF's Multiple Overlapping Deprivation Analysis (MODA). There is no single agreed definition of poverty — from narrow to broad, objective to subjective. Together these frameworks give a fuller picture and help ensure no one is left behind.

Measuring poverty is crucial for understanding the scale of the problem, informing effective policy and assessing impact. It tells us who is poor and where, so interventions can be targeted, progress tracked, and the economy's performance judged.

The East African Community Partner States

Monetary poverty can be measured in absolute or relative terms — a distinction that changes who counts as poor.

Absolute poverty is a condition where a person cannot meet specific minimum requirements, defined by a fixed poverty line — for example, living on less than $3.00 per person per day (2021 PPP, the World Bank line). It is anchored to basic human needs such as food and shelter, and is most relevant in low- and middle-income countries.

Relative poverty is a condition where a person's resources are significantly below the average standard of living in their society — for example, earning less than 60% of national median income. It moves with society's average.

Poverty can be judged by an external standard or by people's own perceptions.

Objective measures come in three common flavours: consumption-based (how much a household spends), income-based (income vs a threshold), and deprivation-based (inability to afford basic needs such as clean water, sanitation or education).

Two widely used tools are the Minimum Income Question (MIQ) and the Ladder of Life:

Chronic and transient poverty describe how long and how often people experience poverty.

Vulnerability is forward-looking: it is not about who is poor now, but who might become poor.

Poverty and inequality are related but distinct — and confusing them leads to poor policy.

Measuring monetary poverty always follows the same three steps. This module is step 1; Modules 3 and 4 complete the journey.

The first step is to choose an indicator of welfare. Two candidates compete: income and consumption.

When consumption is the better choice

Consumption is recommended in low- and middle-income countries where informal employment is common, households grow or barter food, income is seasonal or irregular, and respondents may under-report income due to recall problems or mistrust.

When income works

Income suits high-income, formal economies where wage labour dominates and is reported through tax systems, employment is stable, and surveying income is easier than collecting detailed expenditure.

Surveys such as the LSMS or Household Budget Survey collect detailed data so we can build the total value of everything a household consumes over a reference period. It decomposes into four parts.

Accurate price data does two essential jobs in building the aggregate.

In rural and subsistence economies a large share of consumption never passes through a market. These non-cash food sources must be valued and included, or rural welfare is badly understated.

Homeowners and rent-free households consume housing services too. If we ignore that, owners look poorer than renters simply because they report no rent. So we impute a rental value.

Two ways to impute

Self-reported rent: ask owners "If you rented this home, what would it cost?" Simple, but prone to guesswork, sentimental value, over/under-estimation and social-desirability bias.

Hedonic regression (market-rent approach): use data from households that do pay rent to learn how housing features (rooms, area, urban/rural, toilet) drive rent, then predict rent for owner-occupiers.

Durables — fridges, TVs, bicycles, vehicles — are long-lived: they deliver benefits over years, not all at once. Counting them properly avoids both over- and under-stating welfare.

Some spending is genuine but does not represent welfare-enhancing consumption in the period, so it is left out.

The raw aggregate is nominal. To compare welfare fairly it must become real.

Data is collected at the household level, but we want to compare individuals. Households differ in size and composition, so total household consumption can't be compared directly.

Two adjustments are common. Per capita divides by the number of people — simple, but treats children and adults the same and ignores that larger households share resources (economies of scale). Per adult equivalent weights members to reflect children's lower needs and those economies of scale.

Households in different places — or surveyed at different times — pay different prices for the same goods. We deflate nominal consumption with a price index so we compare real quantities.

Spatial adjustment (the Paasche index)

For poverty analysis the household-level Paasche index is recommended. It uses each household's own quantities and answers: how much more or less does this household pay for what it actually consumed, versus a base region?

Temporal adjustment

Because surveys run over ~12 months, prices drift with inflation and seasonality. Temporal adjustment keeps purchasing power constant across the fieldwork period so comparisons aren't distorted by when a household was visited.

A poverty line turns a welfare measure into a verdict: above the line, not poor; below it, poor.

An absolute poverty line guarantees that comparisons are consistent — two people with the same level of welfare are treated the same way.

Three approaches

An absolute poverty line is a fixed threshold based on the minimum income or consumption needed to meet basic human needs — food, shelter, clothing, healthcare.

National poverty lines are country-specific, but EAC Partner States build them the same broad way — the Cost of Basic Needs approach, aligned with international best practice.

CBN works in two moves: estimate the cost of a food basket that meets minimum calorie needs (the food poverty line), then add an allowance for essential non-food spending (the non-food poverty line).

The food poverty line is the cost of meeting minimum nutritional needs. It is built in steps.

People need more than food. But there's no calorie-style anchor for non-food, so it is estimated indirectly from how households behave.

The idea: look at households whose food spending is about equal to the food poverty line. They are just meeting their food needs, so whatever else they spend reveals essential non-food spending. To smooth out noise, the calculation averages across widening bands (±0%, ±1%, ±2% … up to ±10% of the food line) and then averages those averages.

Add the two pieces and you have the line that defines absolute poverty.

To count the world's poor you can't just add up each country's national rate — that would use a different yardstick in every country. The International Poverty Line (IPL) is one standard for all.

Both measure poverty — but they answer different questions and shouldn't be confused.

Once you know each household's welfare and the poverty line, you know who is poor. But policymakers need this aggregated into a few headline statistics. The standard set is the Foster–Greer–Thorbecke (FGT) family.

where z is the poverty line, yᵢ the consumption of poor person i, and n the total population. The single parameter α switches between three indicators:

The most-used measure: the share of the population living below the poverty line — the incidence of poverty.

P₁ measures the depth of poverty — how far, on average, the poor fall below the line, expressed as a proportion of the line.

Also called the poverty severity index, P₂ averages the squares of the poverty gaps. Squaring gives extra weight to those furthest below the line — so it captures inequality among the poor.

Let's compute all three indicators for ten individuals against a $100 line.

Each indicator points to a different policy lever.

Putting them to work

Disaggregated indicators enable precision targeting (e.g. high P₀ in rural northern regions justifies investment in agriculture, roads and school feeding). Countries use P₀ to set and track national goals ("cut poverty from 35% to 25% by 2030"), P₁ to budget social protection, poverty mapping and small-area estimation to prioritise geographically, and before/after analysis to evaluate programmes such as conditional cash transfers.

Reading the indicators together across time reveals what is really happening — a falling headcount alone can mislead.

Poverty indicators describe the bottom of the distribution. The Gini index describes the whole of it.

Multidimensional poverty recognises that people can suffer several deprivations at the same time. Instead of asking "how much income or consumption?", it asks "what essential aspects of well-being are people lacking?"

Multidimensional measures don't replace monetary ones — they work alongside them.

Why measure multidimensional poverty?

Four frameworks underpin multidimensional measurement.

The MPI uses the Alkire–Foster method — counting weighted deprivations. It is built in clear steps.

An MPI reports three statistics that, together, capture both how many are poor and how deeply.

There is no single 'correct' set of dimensions — choices reflect what matters locally.

Dimensions and indicators are selected through consultation with stakeholders and citizens, alignment with national and regional development goals and the SDGs, and statistical analysis of national survey data.

The same Alkire–Foster method powers MPIs at three levels.

Each statistic guides a different policy choice.

Measuring poverty needs reliable, comprehensive data — most often from household surveys. They are the backbone of national and international poverty monitoring, supplying the data for both monetary and non-monetary indicators.

Different surveys serve different needs, from tracking income and consumption to assessing access to services and living conditions. The right choice depends on the poverty dimension being studied, the frequency of data needed, and the resources available.

Several survey types contribute to poverty measurement, each with a different focus. Examples from EAC national statistics offices are shown.

How often?

The recommended frequency for poverty surveys is usually 3 to 5 years — enough to track progress, inform policy and update SDG/national reporting. Annual data helps rapid monitoring but is less feasible for full multidimensional assessments.

Good design starts from how the data will be used. For poverty surveys, three foundations come first.

Sampling selects a subset of households to represent the whole population, with a known level of precision.

The questionnaire is a structured set of questions that collects information consistently and reliably. A good poverty questionnaire is clear, comprehensive, context-specific and logically ordered, covering both monetary and non-monetary dimensions while minimising respondent burden.

Food consumption can be captured two ways — each with trade-offs.

Recall periods by module

The recall period — how far back respondents must remember — affects accuracy, comparability and respondent burden. Shorter periods suit frequent events; longer periods suit rare ones.

Data quality is decided largely by the field team and by testing the instrument before launch.

Recruiting & training enumerators

Recruit for at least secondary education, basic competence with tablets/digital tools, strong communication, and fluency in relevant local languages — prior experience helps but adaptability is key. Training builds understanding of the survey's purpose, mastery of the questionnaire (sections, skip patterns, coding), interviewing and ethics, with practical sessions, mock interviews, a pilot test and a final evaluation.

Pre-tests vs pilots

Fieldwork quality determines how much cleaning is needed afterwards — so quality controls are built into the interview itself.

Mode of collection

The two main modes are PAPI (paper-and-pencil) and CAPI (computer-assisted personal interviewing). CATI (telephone) is generally not recommended for poverty measurement. CAPI is preferred because it makes data available faster, builds real-time validity checks into the interview, and programs skip patterns to minimise enumerator error.

Supervision & checks

Quality is protected through trained supervisors, real-time dashboards (with CAPI), checks on interview duration and unusual patterns, re-interview and back-checks, spot checks (supervisors observing interviews), quality tables to detect issues like age heaping and under-reporting, and daily feedback to enumerators.

Raw survey data is never ready to use. Processing turns it into a clean, analysable dataset; quality control makes the results credible.