Dataset

Household food security survey — lean season 2024

2,112 synthetic household interviews carrying the raw components of the Food Consumption Score, Household Hunger Scale and reduced Coping Strategy Index, so the composite indicators have to be computed rather than read off.

syntheticFood SecurityKoboToolboxFood Consumption Score (FCS)Household Hunger Scale (HHS)Coping Strategy Index (CSI)Cash assistanceIPC classification

Rows: 4,233
Variables: 35
Period: Lean season, 2024-06 to 2024-07
Licence: CC BY 4.0
Completeness: 98%

Standards and methodologies

Integrated Food Security Phase Classification (IPC)Sphere StandardsCore Humanitarian Standard (CHS)

Files

food-security-survey-2024.v1.csvCSV · 2,112 rows · 207 KB
livelihood-coping-2024.v1.csvCSV · 2,121 rows · 130 KB

Files are versioned by filename. A corrected release ships as .v2.csv rather than replacing the file in place, so an analysis pinned to v1 keeps reproducing.

Data dictionary

Variable	Type	Description	Allowed values
`household_id`	string	Pseudonymous household identifier, unique except where a household was enumerated twice.	—
`district`	categorical	District the household was surveyed in.	Nord-Ouest, Artibonite, Centre, Sud
`survey_date`	date	Date of the household interview, during the lean season.	—
`household_size`	integer	Number of people usually eating from the same pot.	—
`sex_head_of_household`	categorical	Sex of the head of household. One team used a different code.	f, m
`displacement_status`	categorical	Displacement situation of the household.	resident, idp, returnee, host
`main_livelihood`	categorical	Primary livelihood or income source reported by the household.	subsistence-farming, casual-labour, petty-trade, livestock, salaried-employment, remittances, fishing, no-stable-income
`fcs_cereals_tubers`days	integer	Days in the last seven that cereals, grains, roots or tubers were eaten. FCS weight 2.	—
`fcs_pulses`days	integer	Days in the last seven that pulses, nuts or seeds were eaten. FCS weight 3.	—
`fcs_vegetables`days	integer	Days in the last seven that vegetables or leaves were eaten. FCS weight 1.	—
`fcs_fruit`days	integer	Days in the last seven that fruit was eaten. FCS weight 1.	—
`fcs_meat_fish_eggs`days	integer	Days in the last seven that meat, fish or eggs were eaten. FCS weight 4.	—
`fcs_dairy`days	integer	Days in the last seven that milk or dairy was eaten. FCS weight 4.	—
`fcs_oils_fats`days	integer	Days in the last seven that oil or fat was consumed. FCS weight 0.5.	—
`fcs_sugar`days	integer	Days in the last seven that sugar or sweets were consumed. FCS weight 0.5.	—
`hhs_no_food_in_house`	integer	HHS question 1, no food of any kind in the house. Scored 0 never, 1 rarely or sometimes, 2 often.	0, 1, 2
`hhs_sleep_hungry`	integer	HHS question 2, went to sleep hungry. Scored 0 never, 1 rarely or sometimes, 2 often.	0, 1, 2
`hhs_day_and_night_without_eating`	integer	HHS question 3, a whole day and night without eating. Scored 0 never, 1 rarely or sometimes, 2 often.	0, 1, 2
`rcsi_less_preferred_food`days	integer	Days in the last seven relying on less preferred or less expensive food. rCSI weight 1.	—
`rcsi_borrowed_food`days	integer	Days in the last seven borrowing food or relying on help from others. rCSI weight 2.	—
`rcsi_limit_portion_size`days	integer	Days in the last seven limiting portion size at mealtimes. rCSI weight 1.	—
`rcsi_restrict_adult_consumption`days	integer	Days in the last seven restricting adult consumption so children could eat. rCSI weight 3.	—
`rcsi_reduce_meal_numbers`days	integer	Days in the last seven reducing the number of meals per day. rCSI weight 1.	—
`received_food_assistance_30d`	boolean	Whether the household received in-kind food assistance in the last 30 days.	true, false
`received_cash_assistance_30d`	boolean	Whether the household received cash or voucher assistance in the last 30 days.	true, false
`sold_household_assets`	categorical	LCS module, stress strategy: sold household assets such as radio or furniture in the last 30 days. In livelihood-coping-2024.v1.csv.	yes, no, not-applicable
`spent_savings`	categorical	LCS module, stress strategy: spent savings.	yes, no, not-applicable
`borrowed_money`	categorical	LCS module, stress strategy: borrowed money or bought food on credit.	yes, no, not-applicable
`sold_more_animals_than_usual`	categorical	LCS module, stress strategy: sold more animals than usual. Not applicable to a household with no animals.	yes, no, not-applicable
`sold_productive_assets`	categorical	LCS module, crisis strategy: sold productive assets or means of transport.	yes, no, not-applicable
`withdrew_children_from_school`	categorical	LCS module, crisis strategy: withdrew children from school. Not applicable to a household with no school-age children.	yes, no, not-applicable
`reduced_health_expenditure`	categorical	LCS module, crisis strategy: reduced essential health or education expenditure.	yes, no, not-applicable
`sold_house_or_land`	categorical	LCS module, emergency strategy: sold house or land.	yes, no, not-applicable
`begged`	categorical	LCS module, emergency strategy: begged.	yes, no, not-applicable
`sold_last_female_animals`	categorical	LCS module, emergency strategy: sold the last female breeding animals. Not applicable to a household with no animals, and the distinction is load-bearing.	yes, no, not-applicable

Provenance

Source: Synthetic, generated by scripts/generate/food_security_survey_2024.py
Collection method: Simulated household food security questionnaire, seven-day recall
Geography: Four anonymised districts
Period: Lean season, 2024-06 to 2024-07

Data quality

Fully generated. No real household is described.
FCS is the weighted sum of the eight consumption groups: 2 cereals + 3 pulses + 1 vegetables + 1 fruit + 4 meat/fish/eggs + 4 dairy + 0.5 oils + 0.5 sugar.
Which FCS thresholds you use changes the headline. On the standard 21/35 set about 1% of households are poor and 23% borderline. On the 28/42 set used where oil and sugar are consumed near-universally, 7% are poor and 39% borderline. Neither is wrong; failing to state which you used is.
HHS is valid only when all three questions are answered. Households with a partial response must be excluded, not zero-filled — zero-filling scores a hungry household as food secure.
FCS and rCSI correlate at about -0.45. They measure related but distinct things, and a household can eat monotonously without yet resorting to coping strategies, so reporting one as a proxy for the other loses real information.
The Livelihood Coping Strategies module ships as a second file keyed on household_id. A household is classified by the most severe strategy it used, not by a count and not by a weighted sum: 16.0% used none, 35.5% stress, 30.3% crisis and 18.2% emergency.
`not-applicable` is not `no`. A household with no animals cannot sell its last female animal, and per-strategy prevalence must be computed on the households the strategy is available to. Selling more animals than usual is 37.4% among households with animals and 27.5% if every household is counted.
This dataset supports the food consumption evidence used in an IPC analysis. It does not produce an IPC phase, which is assigned by a technical working group convening several outcome indicators against contributing factors.

Known issues

Twenty-three records hold a consumption value above seven days, which is impossible against a seven-day recall. A score computed without range-checking inherits the impossible value and inflates that household.
About 127 cells across dairy, fruit and meat are blank. Treating a blank as zero days scores the household as eating less than it did and can push a borderline household into poor — the most consequential silent failure in this dataset.
Thirty-seven households have a partially administered Household Hunger Scale, with one of the three questions unanswered.
One team in Sud recorded the head of household's sex as Female and Male rather than f and m, which fragments any sex-disaggregated table into four categories.
Twelve households were enumerated twice under different identifiers.
One district's enumerators recorded `not-applicable` as `no` throughout the coping module, so Centre has no not-applicable cells at all. The household-level classification survives this; per-strategy prevalence does not. Selling the last female animals reads 12.3% in Nord-Ouest and 6.5% in Centre on the correct denominator, and 6.8% against 6.4% on the wrong one — the gap disappears.
Fifty-eight households have the coping module partly administered, with the three emergency questions blank. A maximum-severity rule computed over what is present classifies them as crisis at worst, so the 18.2% emergency share is a lower bound.
Nine households appear in the coping module and not in the survey, because the module was administered from a separate listing. An inner join drops them silently.

Worked examples

Python

Computing FCS, HHS and rCSI from raw components

Range-checks the components, computes all three composite indicators, applies both FCS threshold sets side by side, and reports how many households the choice moves.

Run in Colab Notebook (.ipynb)Quarto source (.qmd)

Food insecurity by displacement status and livelihood

Disaggregates food consumption groups by displacement status, livelihood and sex of household head, and shows which cuts are large enough to act on.

Run in Colab Notebook (.ipynb)Quarto source (.qmd)

R

Computing FCS, HHS and rCSI from raw components

The same three indicators built with dplyr, including the exclusion rule for incomplete Household Hunger Scale responses.

Notebook (.ipynb)Quarto source (.qmd)