Understanding and measuring food insecurity

1The four pillars of food security

Food security exists when all people, at all times, have physical and economic access to sufficient, safe and nutritious food to lead an active and healthy life (World Food Summit, 1996). It is made up of four pillars:

• Availability: food exists in the area (production, stocks, imports, aid).
• Access: households can obtain it (income, prices, markets, road security).
• Utilisation: food is properly prepared and absorbed (water, health, practices).
• Stability: the other three pillars hold over time, despite shocks.

In emergencies, the problem is often access rather than availability: markets are full but households have lost their income.

2Livelihoods

Livelihoods are the capabilities, assets and activities that make a living possible: land, livestock, tools, skills, employment, social networks, savings. A shock (drought, conflict, flood, price spike) destroys them or makes them inaccessible.

To cope, households adopt coping strategies: cutting meals, borrowing, selling assets. Some are reversible; others, such as selling the last breeding females or seeds, or taking children out of school, compromise the future. One aim of a response is also to prevent these crisis strategies.

3Household indicators

Several standard indicators, used notably by WFP, measure the situation:

• Food Consumption Score (FCS): frequency and diversity of foods consumed over 7 days, classified as poor, borderline or acceptable.
• Reduced Coping Strategies Index (rCSI): frequency of five food-related strategies (fewer meals, smaller portions, borrowing food…).
• Livelihood coping strategies: stress, crisis or emergency.
• Household Hunger Scale (HHS) and food expenditure share: the higher it is, the more vulnerable the household.

WFP combines them in the CARI approach to classify households into four food security groups.

4The IPC, the common language of food crises

The Integrated Food Security Phase Classification (IPC) classifies areas and populations into five phases: 1 minimal, 2 stressed, 3 crisis, 4 emergency, 5 catastrophe or famine. An area is classified as famine when at least 20% of households face extreme food gaps, global acute malnutrition reaches at least 30% and mortality exceeds 2 deaths per 10,000 people per day.

The IPC analysis is carried out by a national technical working group, using all available data, and includes projections. Populations in Phase 3 or above need urgent action. In West Africa, the Cadre Harmonisé plays the same role.

5Conducting a food security assessment: the complete method

A food security assessment is not a one-off survey: it is a sequence of steps running from the review of existing data to the response recommendation, each with its own timeline, lead actors and limitations. The Sphere standard on food security assessment (Food Security and Nutrition chapter, standard 1.1) requires that situations be assessed using recognised methods, that the analysis covers food security, livelihoods, markets and local capacities, and that findings are shared with other actors. The core technical reference remains WFP's Emergency Food Security Assessment (EFSA) Handbook, complemented by Food Security Cluster guidance and, in refugee operations, by UNHCR-WFP Joint Assessment Missions (JAM).

The standard sequence after a sudden-onset shock

In a protracted crisis the sequence is different: there is no longer a "day zero" but regular cycles, often aligned with the agricultural calendar and the IPC analysis (for example a lean-season survey and a post-harvest survey). The key issue then becomes comparability over time: same questionnaire, same season, same sampling method.

Step 1: the secondary data review

Before sending a single team to the field, the analyst gathers what already exists: previous IPC or Cadre Harmonisé analyses, HEA livelihood zone profiles, early warning bulletins (national systems, FEWS NET where it operates, FAO's GIEWS), WFP price series and national market information systems, SMART nutrition surveys, agricultural production data, IOM displacement reports (Displacement Tracking Matrix), Demographic and Health Surveys. For each pillar a "pre-shock" baseline is written, followed by a hypothesis on the effect of the shock. This review makes it possible to cut the questionnaire down to the information that is genuinely missing, saving time for both households and teams.

Step 2: defining the questions and the analysis plan

An assessment answers decision-making questions, not general curiosity. The five classic questions are: how many people are food insecure? Who are they (profile, area, livelihood group)? Why (immediate and underlying causes)? How will the situation evolve over the next three to six months? What response is appropriate (modality, duration, timing)? The analysis plan, drafted before data collection, links each question to the indicators, disaggregations and tables expected. An indicator that does not appear in any table of the analysis plan has no place in the questionnaire.

Step 3: choosing the methods

• Quantitative household survey: prevalence figures (FCS, rCSI, LCS, food expenditure share, HHS), comparable across areas and over time. It answers "how many" and "who".
• Key informant interviews (village chiefs, agricultural extension workers, teachers, health workers, traders): trends, timing, access, services. Quick but exposed to positional bias.
• Focus group discussions separated by sex, age and livelihood group: causes, coping strategies, modality preferences, protection risks. They answer "why" and "how".
• Trader surveys and price monitoring: how markets function and their capacity to respond to increased demand.
• Direct observation: condition of crops and pasture, visible stocks, queues at water points, livestock body condition.

Step 4: data collection, quality control and analysis

Enumerator training of three to five days, including one day of field testing; mixed teams (at least one female enumerator per pair to interview women); daily supervision with checks on the data received that same evening; preliminary analysis within the week following the end of data collection; validation workshop with the cluster and government technical services. An assessment report is only credible if it describes its limitations: periods and areas not covered, non-response bias, problems in translating an indicator.

Proportionality rule: the depth of the assessment must match the decision to be taken. To decide on a four-week emergency distribution, a credible rapid assessment is enough; to design an 18-month programme and set outcome targets, a representative household survey is essential. Delaying a life-saving response to obtain perfect data is a failure, just as launching a multi-year programme without a baseline is.

6Analytical frameworks: livelihoods, causes of malnutrition and seasonality

Household indicators describe a state; they do not explain why the household is in that situation or what will happen next. Three analytical frameworks make it possible to move from description to explanation: the sustainable livelihoods framework, the conceptual framework on the determinants of nutrition, and the seasonal calendar.

The sustainable livelihoods framework

Formalised in the late 1990s by the UK Department for International Development (DFID), this framework describes five categories of assets available to a household, the vulnerability context (shocks, trends, seasonality) and the policies, institutions and processes that determine access to them.

The practical value of this framework is to determine which assets the response must protect or rebuild. A pastoralist household whose herd has fallen from 30 to 8 goats does not simply have a consumption problem: it has lost most of its physical capital and its future income capacity. A response limited to a food ration will keep it alive but will not lift it out of the crisis.

The conceptual framework on the determinants of nutrition

The UNICEF framework (1990, updated in 2020 as a framework on the determinants of maternal and child nutrition) distinguishes immediate causes (inadequate diet, disease), underlying causes (household food insecurity, inadequate care practices, unhealthy environment and insufficient access to health services) and basic causes (resources, governance, social norms). It is a reminder that improving food access is not enough to reduce acute malnutrition if water, sanitation, health and infant and young child feeding practices remain deficient: this is the "utilisation" pillar.

Benchmarks to know when reading nutrition data within a food security analysis:

• Prevalence thresholds for global acute malnutrition among children aged 6-59 months (WHO and UNICEF, 2018): very low below 2.5%; low from 2.5% to less than 5%; medium from 5% to less than 10%; high from 10% to less than 15%; very high at 15% and above.
• Phases of the IPC Acute Malnutrition scale: Phase 1 below 5%, Phase 2 from 5 to 9.9%, Phase 3 from 10 to 14.9%, Phase 4 from 15 to 29.9%, Phase 5 at 30% and above.
• Mortality thresholds used by Sphere: a crude mortality rate above 1 death per 10,000 people per day, or an under-five mortality rate above 2 per 10,000 per day, indicates an emergency; a doubling of the baseline rate is also a warning sign.
• Mid-upper arm circumference in children aged 6 to 59 months: below 115 mm for severe acute malnutrition, from 115 to less than 125 mm for moderate acute malnutrition.

A sound analysis cross-references the two: a GAM rate of 14% in an area where household consumption is acceptable should prompt a look at disease (measles, diarrhoea, malaria), water and practices; severely deteriorated consumption with a still-low GAM rate often signals nutritional deterioration in the following weeks, because malnutrition is a lagging indicator.

The seasonal calendar

The seasonal calendar is the analyst's simplest and most underused tool. It is built in focus group discussions with men and women from each livelihood group, month by month, for a normal year. It shows when the lean season starts, when prices peak, when agricultural labour is in demand and when diseases peak, and therefore when each intervention is useful.

How to read it: a survey conducted in November, just after the harvest, will almost always give a more favourable picture than one conducted in August. Comparing these two surveys to conclude that there has been a "deterioration" or an "improvement" is a classic mistake: a season should be compared with the same season of the previous year.

7Worked example: the Haut-Kéla district assessment

Fictional context: Haut-Kéla district (186,000 inhabitants, around 31,000 households) is made up of three livelihood zones: a rain-fed cereal farming zone in the south, an agropastoral zone in the centre and a pastoral zone in the north. The rainy season was severely deficient (62% of average), then armed clashes in August displaced some 21,000 people from the pastoral zone to the district capital, Doumbaré. The NGO "Terres Solidaires", a cluster member, is responsible, together with two partners, for the in-depth assessment in October.

1. Secondary data review (week 1)

• Latest Cadre Harmonisé (March): district in Phase 2, projected Phase 2 for the lean season.
• HEA profile (2021): in the agropastoral zone, very poor and poor households represent 45% of the population and obtain 55% of their food from their own crops in a normal year.
• Millet price in Doumbaré in September: 410 currency units per kilo, compared with 260 a year earlier (+58%) and a five-year average of 245.
• Previous year's SMART survey (same season): GAM of 9.1%.
• Gaps: no recent data on displaced people, no data on livestock losses.

2. Sampling plan

Four strata are selected: farming zone, agropastoral zone, pastoral zone (residents who stayed, partial access) and displaced people in Doumbaré. For each, with an expected prevalence of 50%, a precision of ±10 percentage points, a design effect of 2 and 10% non-response: 96 × 2 = 192, i.e. around 211 households, rounded up to 22 clusters of 10 households. Planned total: 880 households. In the pastoral zone, 6 of the 22 clusters are inaccessible for security reasons; they are replaced by telephone interviews with 48 households and by key informant interviews, and this limitation is reported.

3. Main findings

4. Analysis by pillar

• Availability: local production sharply reduced, but markets in the district capital remain supplied by traders coming from the regional capital; the trader survey indicates a restocking time of 7 to 10 days. Availability is a problem in the pastoral zone, where two out of three weekly markets are closed.
• Access: this is the most deteriorated pillar. Millet price up 58%, loss of income from agricultural labour (fewer crops to harvest), collapse in terms of trade for pastoralists. Displaced people have lost both their herds and their social networks.
• Utilisation: GAM has risen from 9.1% to 12.8% for a comparable season. WASH data show that 41% of displaced people drink untreated surface water.
• Stability: the next harvest is 12 months away; the lean season will start early, from April in the agropastoral zone. The conflict remains active in the north.

5. Reading severity and projection

Displaced people and the pastoral zone show a convergence of indicators consistent with IPC Phase 3 and, for a significant share of households, with Phase 4: more than a quarter of households using emergency strategies, mean rCSI above 19, more than a third of households with poor consumption. The agropastoral zone is rather in Phase 3; the farming zone in Phase 2, with a significant minority of households in Phase 3. The team does not classify the district itself: it passes these findings, with their method and limitations, to the Cadre Harmonisé analysis cell in November. It projects a deterioration from April, when remaining stocks will be exhausted.

6. Problem statement and recommendations

Agreed statement: "Around 48,000 people (displaced people in Doumbaré and poor and very poor households in the accessible pastoral zone) cannot meet their minimum food needs without resorting to emergency coping strategies, mainly because of the loss of their livestock and income and the 58% rise in cereal prices. A further 25,000 people in the agropastoral zone will fall into the same situation from April if no assistance is provided."

Recommendations passed to the cluster: immediate food assistance for displaced people (cash transfers in Doumbaré, where the market is functioning, to be confirmed by a rapid market assessment); in-kind food for the pastoral zone, where markets are closed; targeted lean-season assistance in the agropastoral zone from March to August; support to protect the core breeding stock; short-cycle seeds in May; coordination with the Nutrition Cluster for screening and with the WASH Cluster for water at displacement sites; a new follow-up survey in June.

Note what the team did not do: it did not add up percentages from zones of different sizes without weighting; it did not conclude that there was "famine" on the basis of a few severe indicators; it did not recommend a single modality for the whole district; it did not hide the partial coverage of the pastoral zone.

8Reusable tools: analysis matrix, review grid and checklist

Four-pillar analysis matrix

This matrix is used both to prepare the questionnaire and to structure the report. For each pillar, it identifies the indicators, their sources and the level of analysis (area, market, household, individual).

The warning thresholds in the last column are working benchmarks, to be adapted to the context and local references; they do not replace the IPC reference tables.

Secondary data review grid

Checklist before launching a household survey

1. The analysis plan is written and every question in the questionnaire links to a table in it.
2. The questionnaire uses the standard modules (FCS, rCSI, LCS, expenditure, HHS) in the version recommended by WFP or the cluster, without changing the wording or the recall periods.
3. The LCS strategies have been adapted to the context with the cluster (validated local list of stress, crisis and emergency strategies).
4. The questionnaire has been translated, back-translated and pre-tested with at least 20 households.
5. The sample size per stratum is calculated and justified; the sampling frame is documented.
6. Teams are mixed, trained for at least three days, and sign a code of conduct that includes protection from sexual exploitation and abuse.
7. Informed consent is read out, participation is voluntary and no promise of assistance is made.
8. A referral protocol exists for serious cases encountered (malnourished child, survivor of violence), with service contact details.
9. Personal data are minimised, encrypted on the phones and deleted according to a set schedule.
10. The date for sharing findings with the cluster and communities has been set.

Template structure for an assessment report (10 to 20 pages)

1. Two-page executive summary: key figures, severity, recommendations.
2. Context and objectives; assessment questions.
3. Methodology: strata, sample, dates, tools, limitations.
4. Findings by pillar, then by outcome indicator, disaggregated.
5. Profiles of food-insecure households: who, where, how many, why.
6. Outlook and most likely scenario over three to six months.
7. Recommendations: response, targeting, timing, further information needs.
8. Annexes: questionnaire, detailed tables, list of localities surveyed.

9Adapting the assessment to the context, common mistakes and donor expectations

Variations by context

Common mistakes

• Confusing severity with magnitude: an area where 45% of households have poor consumption but with only 8,000 inhabitants carries less weight, in number of people, than an area at 20% with 200,000 inhabitants. The report must present both.
• Aggregating without weighting: the average of four strata of different sizes is not the district prevalence. Each stratum is weighted by its population.
• Interpreting an indicator in isolation: an acceptable FCS achieved by selling all the livestock is not good news. The analysis always cross-references the current status (consumption) with coping capacity (strategies, expenditure).
• Comparing different seasons: see the seasonal calendar.
• Modifying standard questions: adding a ninth food group to the FCS or changing the rCSI recall period makes the results non-comparable and unusable for the IPC.
• Ignoring absent households: households that have migrated, or are too poor to be present in the villages surveyed, are often the most affected.
• Raising expectations: a survey that suggests aid will follow distorts the answers in subsequent surveys and damages the relationship with the community.
• Publishing too late: an assessment whose findings arrive after the decision window (IPC workshop, response plan, funding appeal) is useless. A preliminary note is produced within ten days.

What donors and evaluators look for

• ECHO: in the proposal, the needs assessment section must cite recent, disaggregated data, draw on joint assessments and the IPC or Cadre Harmonisé, and justify the choice of areas and groups. ECHO's policy on humanitarian food assistance emphasises analysis of causes and protection of livelihoods, not just consumption.
• BHA (United States): the Emergency Application Guidelines require the food security situation to be substantiated by recognised sources, a market analysis to justify the choice of modality, and standard indicators (FCS, rCSI) with a baseline.
• FCDO and UN agencies: consistency with the Humanitarian Response Plan and cluster priorities; quality of the methodology; data sharing with the common system.
• Evaluators: they check whether programming decisions actually flow from the assessment, whether data have been disaggregated by sex, age and disability, and whether communities were consulted on their priorities. The OECD DAC relevance criterion is judged largely on the quality of the initial analysis.

The question every auditor or evaluator asks: "Show me the chain that runs from the data to the decision." If you cannot link your areas, targeting criteria, transfer values and timing to documented assessment findings, the programme will be judged of limited relevance, even if it was well implemented.

10Interview questions and key references

Interview questions

1. "The markets are full, but households are not eating enough. How do you explain this and what do you recommend?"

A good answer identifies an access problem rather than an availability problem, cites possible causes (loss of income, price rises, debt, insecurity on the road to the market), proposes checking traders' capacity to respond to increased demand and physical access, and concludes that a cash-based response is probably appropriate, subject to a market analysis and preferences, with price monitoring.

2. "You have one week to give an order of magnitude of needs after a flood. How do you go about it?"

The answer describes a secondary data review (population, livelihood profiles, calendar, reference prices), a rapid assessment coordinated with the cluster through key informants and market observation, the use of impact maps to estimate affected populations, a range rather than a single figure, and a plan for a household survey at four to six weeks. It mentions coordination to avoid duplicate assessments.

3. "The FCS in your survey is good, but all the other indicators are bad. What do you conclude?"

The answer considers two hypotheses: a data collection problem (food groups misunderstood, foods eaten in very small quantities counted as consumption days, enumerator effect) or consumption maintained through destructive strategies. It explains how to decide between them: data checks by enumerator, comparison with the HDDS and expenditure, qualitative interviews. It concludes that nothing is published until the issue is understood.

4. "What is the difference between chronic and acute food insecurity, and why does it matter for programming?"

The answer explains that acute food insecurity is linked to a shock and a given period, and calls for an emergency response to save lives and protect livelihoods; chronic food insecurity is persistent, even without a shock, and calls for structural responses (social safety nets, agricultural development, employment). It cites the IPC Chronic Food Insecurity scale and shows that the same area can experience both, which requires combining emergency and longer-term programmes.

5. "How do you integrate gender and protection into a food security assessment?"

The answer cites mixed teams, separate focus group discussions, disaggregation by sex of the head of household, age and disability, questions on who decides how resources are used, intra-household food distribution, risks faced in accessing markets and fields, and a referral protocol. It avoids reducing gender to the single variable "female-headed household".

6. "A colleague suggests adding 40 questions to the standard questionnaire. What do you tell them?"

The answer refers back to the analysis plan: every question must inform a decision. It recalls the costs (interview length, respondent fatigue, quality that declines beyond 45 to 60 minutes) and proposes addressing some questions through qualitative methods.

Key references

• The Sphere Handbook (2018), Food Security and Nutrition chapter: assessment standards and standards on food security and livelihoods; Sphere Association.
• Emergency Food Security Assessment Handbook (second edition) and technical guidance notes on indicators: World Food Programme.
• Joint Assessment Mission Guidelines: UNHCR and WFP.
• IPC Technical Manual, version 3.1: IPC Global Partnership.
• Cadre Harmonisé Manual: CILSS.
• Sustainable Livelihoods Guidance Sheets: DFID (1999).
• Conceptual Framework on the Determinants of Maternal and Child Nutrition (2020): UNICEF.
• SMART methodology and guidance on interpreting acute malnutrition prevalence: SMART partners, WHO and UNICEF.
• Global Food Security Cluster guidance on assessments and monitoring: Global Food Security Cluster.
• Multi-Cluster/Sector Initial Rapid Assessment (MIRA) guidance: IASC.