Data dictionary

Every data file this site publishes, column by column: what each column means, its type and its units, where the figures come from and on what terms they may be reused. The columns' meanings are written once, with the data warehouse's tables, and this page is built from them, so it changes when the data do. Use your browser's find (Ctrl+F, or Cmd+F on a Mac) to look up a column.

How to read the figures

dhs2024_state_prevalence.parquet: Disability by state, DHS 2024

Curb Cut's survey estimates for every state from the Nigeria DHS 2024: DISABILITY3 and each of the six activities, for adults 18+ and children 5-17, both sexes and each sex. One row per state, measure, age group and sex. No extra breakdown (urban/rural and wealth are on the state pages). The data explorer (explorer.html) reads this file in the browser; the sensitivity page (sensitivity.html) uses its DISABILITY3 cells for both sexes, and population_2023, through data/sensitivity.json.

File: dhs2024_state_prevalence.parquet (Parquet, 46 KB, 1,554 rows). Made from the warehouse's mart_prevalence_state.

Sources and terms

The 18 columns of dhs2024_state_prevalence.parquet
NameTypeMeaningUnits
statetext (VARCHAR)Canonical state name (one of the 36 states or the Federal Capital Territory).not applicable
state_codetext (VARCHAR)ISO 3166-2 code of the state, e.g. NG-KN for Kano and NG-FC for the Federal Capital Territory.not applicable
zonetext (VARCHAR)Geopolitical zone of the state: North Central, North East, North West, South East, South South or South West.not applicable
measuretext (VARCHAR)What is measured. disability3: the Washington Group Short Set DISABILITY3 measure, "a lot of difficulty" or "cannot do at all" in at least one of six domains. <domain>_3plus: one domain alone at that level (seeing, hearing, walking, cognition = remembering or concentrating, selfcare, communication). <domain>_level_1 to _level_4 (national only): the share at each level (no difficulty, some, a lot, cannot do at all). MPI 2021 uses its own names (mpi_...), never DISABILITY3, because its questions are not the Washington Group set.not applicable
age_grouptext (VARCHAR)Age group of the people the cell covers, "18+" (the headline) or "5-17".years of age
sextext (VARCHAR)Sex of the people the cell covers, "all", "female" or "male".not applicable
estimatenumber (DOUBLE)Survey-weighted share of people with the measured difficulty. Null when the cell is suppressed or the state has no cell (has_data false).proportion (0 to 1)
senumber (DOUBLE)Design-based standard error of the estimate (Taylor linearisation, R survey), reflecting the survey's weights, strata and clusters. Null when suppressed.proportion (0 to 1)
ci_lownumber (DOUBLE)Lower end of the 95% confidence interval (Korn-Graubard). Null when suppressed.proportion (0 to 1)
ci_highnumber (DOUBLE)Upper end of the 95% confidence interval (Korn-Graubard). Null when suppressed.proportion (0 to 1)
n_unweightedwhole number (INTEGER)Unweighted number of respondents in the denominator (people in the cell with the item answered). Null when suppressed, since a count could be recovered by subtraction.respondents
n_effectivenumber (DOUBLE)Effective sample size, p(1-p)/se squared, capped at n_unweighted (NCHS). Null when suppressed.respondents
dfwhole number (INTEGER)Design degrees of freedom: survey clusters (PSUs) with a respondent in the cell minus their strata. Null when suppressed.degrees of freedom
precision_flagtext (VARCHAR)How far the estimate can be trusted, under the NCHS Data Presentation Standards for Proportions plus one stricter project rule: reliable; review (shown, but rests on 0 events, all events, fewer than 8 degrees of freedom, or 1 to 9 unweighted events); unreliable (n or effective n under 30, or an interval too wide in absolute or relative terms). "suppressed" when the disclosure check removed the numbers; "no_data" where a state has no cell for the source.not applicable
precision_reasontext (VARCHAR)The rule that set the flag, e.g. "low_events<10", "zero_events", "df<8", "n_effective<30", "relative_ci_width>130%"; for suppressed cells "n<30 (disclosure)" (fewer than 30 unweighted respondents) or "complementary suppression (disclosure)" (withheld so a neighbouring small cell cannot be worked out by subtraction). Null when the estimate is reliable.not applicable
suppressedtrue or false (BOOLEAN)True when the disclosure check removed the cell's numbers: fewer than 30 unweighted respondents, or withheld to protect such a cell. The numbers are then null; they are never estimated or filled.not applicable
has_datatrue or false (BOOLEAN)False where the source has no cell for this state, measure, age group and sex; the row exists so that all 37 states are always present.not applicable
population_2023whole number (BIGINT)2023 population projection for the state and the row's sex (all ages; both sexes when sex is "all"), NBS Demographic Bulletin 2023, Annex 3. Not the denominator of the estimate.persons

dhs2024_national_zone.parquet: Disability in Nigeria and its zones, DHS 2024

The same estimates for Nigeria as a whole and for each of the six geopolitical zones, including (nationally) the share at each of the four levels of difficulty. The data explorer (explorer.html) reads this file in the browser.

File: dhs2024_national_zone.parquet (Parquet, 15 KB, 342 rows). Made from the warehouse's mart_prevalence_national_zone.

Sources and terms

The 15 columns of dhs2024_national_zone.parquet
NameTypeMeaningUnits
geography_leveltext (VARCHAR)Level of the cell, "national" or "zone".not applicable
geographytext (VARCHAR)Where the cell is: "NG" (Nigeria) at national level, otherwise the zone's name (North Central, North East, North West, South East, South South, South West).not applicable
measuretext (VARCHAR)What is measured. disability3: the Washington Group Short Set DISABILITY3 measure, "a lot of difficulty" or "cannot do at all" in at least one of six domains. <domain>_3plus: one domain alone at that level (seeing, hearing, walking, cognition = remembering or concentrating, selfcare, communication). <domain>_level_1 to _level_4 (national only): the share at each level (no difficulty, some, a lot, cannot do at all). MPI 2021 uses its own names (mpi_...), never DISABILITY3, because its questions are not the Washington Group set.not applicable
age_grouptext (VARCHAR)Age group of the people the cell covers, "18+" (the headline) or "5-17".years of age
sextext (VARCHAR)Sex of the people the cell covers, "all", "female" or "male".not applicable
estimatenumber (DOUBLE)Survey-weighted share of people with the measured difficulty. Null when the cell is suppressed or the state has no cell (has_data false).proportion (0 to 1)
senumber (DOUBLE)Design-based standard error of the estimate (Taylor linearisation, R survey), reflecting the survey's weights, strata and clusters. Null when suppressed.proportion (0 to 1)
ci_lownumber (DOUBLE)Lower end of the 95% confidence interval (Korn-Graubard). Null when suppressed.proportion (0 to 1)
ci_highnumber (DOUBLE)Upper end of the 95% confidence interval (Korn-Graubard). Null when suppressed.proportion (0 to 1)
n_unweightedwhole number (INTEGER)Unweighted number of respondents in the denominator (people in the cell with the item answered). Null when suppressed, since a count could be recovered by subtraction.respondents
n_effectivenumber (DOUBLE)Effective sample size, p(1-p)/se squared, capped at n_unweighted (NCHS). Null when suppressed.respondents
dfwhole number (INTEGER)Design degrees of freedom: survey clusters (PSUs) with a respondent in the cell minus their strata. Null when suppressed.degrees of freedom
precision_flagtext (VARCHAR)How far the estimate can be trusted, under the NCHS Data Presentation Standards for Proportions plus one stricter project rule: reliable; review (shown, but rests on 0 events, all events, fewer than 8 degrees of freedom, or 1 to 9 unweighted events); unreliable (n or effective n under 30, or an interval too wide in absolute or relative terms). "suppressed" when the disclosure check removed the numbers; "no_data" where a state has no cell for the source.not applicable
precision_reasontext (VARCHAR)The rule that set the flag, e.g. "low_events<10", "zero_events", "df<8", "n_effective<30", "relative_ci_width>130%"; for suppressed cells "n<30 (disclosure)" (fewer than 30 unweighted respondents) or "complementary suppression (disclosure)" (withheld so a neighbouring small cell cannot be worked out by subtraction). Null when the estimate is reliable.not applicable
suppressedtrue or false (BOOLEAN)True when the disclosure check removed the cell's numbers: fewer than 30 unweighted respondents, or withheld to protect such a cell. The numbers are then null; they are never estimated or filled.not applicable

prevalence_comparison.parquet: Assumed and measured figures compared

The main page's comparison: JONAPWD's assumed 15%, WHO's global 16%, the NDHS 2024 report's figures for ages 15+, and Curb Cut's DHS 2024 estimate for adults 18+. Values are percentages, not proportions. The sensitivity page (sensitivity.html) quotes the JONAPWD and WHO rows, through data/sensitivity.json.

File: prevalence_comparison.parquet (Parquet, 4 KB, 5 rows). Made from the warehouse's mart_prevalence_comparison.

Sources and terms

The 11 columns of prevalence_comparison.parquet
NameTypeMeaningUnits
figure_idtext (VARCHAR)Stable identifier of the figure (e.g. jonapwd_2025, dhs2024_18plus_wg).not applicable
sort_orderwhole number (INTEGER)Order of the figure on the main page (1 first).not applicable
labeltext (VARCHAR)The figure's short name as the page shows it.not applicable
kindtext (VARCHAR)What sort of figure it is: "assumed" (an assumption, not a measurement), "global estimate" (a world figure) or "measured" (from a survey of Nigeria).not applicable
value_percentnumber (DOUBLE)The figure itself.percent (0 to 100)
ci_low_percentnumber (DOUBLE)Lower end of the 95% confidence interval; only for Curb Cut's own DHS 2024 estimate, null for the other figures.percent (0 to 100)
ci_high_percentnumber (DOUBLE)Upper end of the 95% confidence interval; only for Curb Cut's own DHS 2024 estimate, null for the other figures.percent (0 to 100)
populationtext (VARCHAR)Who the figure is about (e.g. "Household members aged 18+ (de facto)").not applicable
definitiontext (VARCHAR)What the figure counts, in words, as its source defines it.not applicable
sourcetext (VARCHAR)Citation of the figure's source.not applicable
source_urltext (VARCHAR)Web address of the figure's source.web address

spatial_state.parquet: Spatial results by state

Exploratory and descriptive spatial results for each state, from the DHS 2024 direct estimates: the local Moran's I for DISABILITY3 18+, the profile cluster and GRID3 health facilities.

File: spatial_state.parquet (Parquet, 4 KB, 37 rows). Made from the warehouse's mart_spatial_state.

Sources and terms

The 12 columns of spatial_state.parquet
NameTypeMeaningUnits
statetext (VARCHAR)Canonical state name (one of the 36 states or the Federal Capital Territory).not applicable
state_codetext (VARCHAR)ISO 3166-2 code of the state, e.g. NG-KN for Kano and NG-FC for the Federal Capital Territory.not applicable
zonetext (VARCHAR)Geopolitical zone of the state: North Central, North East, North West, South East, South South or South West.not applicable
disability3_18plusnumber (DOUBLE)The state's DHS 2024 DISABILITY3 estimate for adults 18+ (both sexes), the value the local analysis uses.proportion (0 to 1)
local_inumber (DOUBLE)Exploratory local Moran's I for DISABILITY3 18+ (Anselin, 1995): positive when the state is like its neighbours, negative when unlike them.index (no unit)
lisa_p_permnumber (DOUBLE)Two-sided conditional permutation p-value of local_i (9,999 permutations).probability (0 to 1)
lisa_q_bhnumber (DOUBLE)lisa_p_perm adjusted with Benjamini-Hochberg across the 37 states.probability (0 to 1)
lisa_quadranttext (VARCHAR)Quadrant of the Moran scatterplot whatever the significance (spdep, "mean"): High-High when the state and the average of its neighbours are both above the mean, Low-Low when both are below, High-Low or Low-High when they differ.not applicable
lisa_classtext (VARCHAR)lisa_quadrant when lisa_q_bh is below 0.05, otherwise "Not significant". Exploratory: with 37 states, local tests have low power.not applicable
profile_clusterwhole number (INTEGER)Descriptive cluster of the state's six domain prevalences for adults 18+ (Ward's method on their z-scores; the number of clusters chosen by mean silhouette). The number is a label, not an order.not applicable
facilitieswhole number (INTEGER)Health facilities in the state in GRID3 NGA Health Facilities v2.0 (all ownership).facilities
facilities_per_100knumber (DOUBLE)facilities per 100,000 people of all ages (NBS 2023 population projection). General health facilities, not disability services.facilities per 100,000 people

spatial_moran.parquet: Global Moran's I by measure

Whether states with similar estimates tend to be neighbours, one test per measure, with how many state estimates are imprecise.

File: spatial_moran.parquet (Parquet, 3 KB, 8 rows). Made from the warehouse's mart_spatial_moran.

Sources and terms

The 16 columns of spatial_moran.parquet
NameTypeMeaningUnits
measuretext (VARCHAR)The DHS 2024 measure tested: disability3 or one domain at "a lot of difficulty" or worse (seeing_3plus, hearing_3plus, walking_3plus, cognition_3plus, selfcare_3plus, communication_3plus).not applicable
age_grouptext (VARCHAR)Age group of the people the cell covers, "18+" (the headline) or "5-17".years of age
test_roletext (VARCHAR)The test's role: "primary" (DISABILITY3 18+), "domain" (a single domain, 18+) or "secondary" (DISABILITY3 5-17).not applicable
moran_inumber (DOUBLE)Global Moran's I of the 37 state estimates (queen contiguity, row-standardised weights): above expected_i means similar states tend to be neighbours.index (no unit)
expected_inumber (DOUBLE)Moran's I expected with no spatial pattern, -1/(n-1) for n = 37 states.index (no unit)
p_permnumber (DOUBLE)One-sided permutation p-value (positive clustering only), from nsim permutations.probability (0 to 1)
p_bhnumber (DOUBLE)p_perm adjusted with Benjamini-Hochberg across the 8 measures tested.probability (0 to 1)
nsimwhole number (INTEGER)Number of permutations behind p_perm.permutations
n_stateswhole number (INTEGER)Number of state estimates in the test; must be 37.states
n_reliablewhole number (INTEGER)States whose estimate for this measure is flagged reliable.states
n_reviewwhole number (INTEGER)States whose estimate is flagged review.states
n_unreliablewhole number (INTEGER)States whose estimate is flagged unreliable.states
n_zero_estimatewhole number (INTEGER)States whose estimate is exactly 0.states
n_imprecisewhole number (INTEGER)States whose estimate is not flagged reliable or is 0. Noisy estimates pull I towards 0, so a weak result with many imprecise states is not evidence of no pattern.states
measure_labeltext (VARCHAR)The measure in words, as the site shows it (e.g. "Difficulty walking or climbing steps").not applicable
sort_groupwhole number (INTEGER)Display order group (1 primary, 2 domains, 3 secondary).not applicable

jonapwd_state_by_type.parquet: JONAPWD's assumed figures by state and type

JONAPWD's 2025 figures, as published: an assumption (15% of a population projection, split by type), not a count. The site cites them; it does not offer them as a download.

Not offered for download: JONAPWD's tables by type are cited from an all-rights-reserved publication, and the site reads them only for its own tables and charts. Made from the warehouse's mart_jonapwd_state_by_type.

Sources and terms

The 10 columns of jonapwd_state_by_type.parquet
NameTypeMeaningUnits
statetext (VARCHAR)Canonical state name (one of the 36 states or the Federal Capital Territory).not applicable
state_codetext (VARCHAR)ISO 3166-2 code of the state, e.g. NG-KN for Kano and NG-FC for the Federal Capital Territory.not applicable
zonetext (VARCHAR)Geopolitical zone of the state: North Central, North East, North West, South East, South South or South West.not applicable
disability_typetext (VARCHAR)JONAPWD's disability category as a code: physical, hearing, visual, leprosy, albinism, spinal_cord_injury, intellectual, others, deaf_blindness or little_stature.not applicable
disability_type_labeltext (VARCHAR)The category's display label on the site (e.g. "Physical disability").not applicable
sort_orderwhole number (INTEGER)Display order of the category (1 to 10), as on the site.not applicable
personswhole number (BIGINT)JONAPWD's assumed number of persons with this type of disability in the state, as printed in the publication. An assumption (15% of a 2025 population projection, split by type), not a count or a survey estimate.persons (assumed)
state_totalwhole number (BIGINT)JONAPWD's assumed total for the state, the sum of its ten categories.persons (assumed)
share_of_statenumber (DOUBLE)This category's share of the state's JONAPWD total.proportion (0 to 1)
source_pagewhole number (INTEGER)Page of the JONAPWD PDF (counted from 1) the figure was read from.page number

jonapwd_national_by_type.parquet: JONAPWD's assumed figures by type, all states

JONAPWD's 2025 figures summed over the 37 states, by type. An assumption, not a count; cited, not offered as a download.

Not offered for download: JONAPWD's tables by type are cited from an all-rights-reserved publication, and the site reads them only for its own tables and charts. Made from the warehouse's mart_jonapwd_national_by_type.

Sources and terms

The 6 columns of jonapwd_national_by_type.parquet
NameTypeMeaningUnits
disability_typetext (VARCHAR)JONAPWD's disability category as a code: physical, hearing, visual, leprosy, albinism, spinal_cord_injury, intellectual, others, deaf_blindness or little_stature.not applicable
disability_type_labeltext (VARCHAR)The category's display label on the site (e.g. "Physical disability").not applicable
sort_orderwhole number (INTEGER)Display order of the category (1 to 10), as on the site.not applicable
personswhole number (BIGINT)JONAPWD's assumed number of persons with this type of disability, summed over the 37 states. An assumption, not a count.persons (assumed)
share_of_totalnumber (DOUBLE)This category's share of JONAPWD's all-state total.proportion (0 to 1)
stateswhole number (BIGINT)Number of states contributing; must be 37.states

states.json: The 37 states

One object per state, ordered by name: the state card's facts, JONAPWD's assumed total, and the DHS 2024 headline (DISABILITY3, adults 18+, both sexes).

File: states.json (JSON, 24 KB, 37 entries). Made from the warehouse's mart_prevalence_state and mart_state_profile.

Sources and terms

The 20 fields of states.json
NameTypeMeaningUnits
[]objectOne object per state (36 states and the Federal Capital Territory), by name.not applicable
[].statetextCanonical state name (one of the 36 states or the Federal Capital Territory).not applicable
[].state_codetextISO 3166-2 code of the state, e.g. NG-KN for Kano and NG-FC for the Federal Capital Territory.not applicable
[].zonetextGeopolitical zone of the state: North Central, North East, North West, South East, South South or South West.not applicable
[].capitaltextThe state capital.not applicable
[].capital_source_urltextWeb page the capital was taken from.web address
[].slogantext or nullThe state's slogan. Null where two sources did not agree; the card then leaves the line out.not applicable
[].slogan_source_urltext or nullWeb page the slogan was taken from. Present exactly when slogan is present.web address
[].population_2023whole number2023 population projection for the state, all ages, both sexes (NBS Demographic Bulletin 2023, Annex 3; source: National Population Commission).persons
[].jonapwd_totalwhole numberJONAPWD's assumed number of persons with disabilities in the state (its ten categories summed). An assumption (15% of a 2025 population projection), not a count.persons (assumed)
[].share_of_nationalnumberThe state's share of JONAPWD's total for all 37 states.proportion (0 to 1)
[].rank_of_37whole numberRank of the state's JONAPWD total among the 37 (1 = largest; tied totals share a rank).rank (1 to 37)
[].dhs2024objectThe state's DHS 2024 headline cell: DISABILITY3, adults 18+, both sexes. The whole object is null where the state has no published headline (suppressed, or no cell); when present, none of its numbers is null.not applicable
[].dhs2024.estimatenumberSurvey-weighted share of adults 18+ with a lot of difficulty, or unable, in at least one of six activities (DISABILITY3).proportion (0 to 1)
[].dhs2024.senumberDesign-based standard error of the estimate (Taylor linearisation, R survey).proportion (0 to 1)
[].dhs2024.ci_lownumberLower end of the 95% confidence interval (Korn-Graubard).proportion (0 to 1)
[].dhs2024.ci_highnumberUpper end of the 95% confidence interval (Korn-Graubard).proportion (0 to 1)
[].dhs2024.n_unweightedwhole numberUnweighted number of respondents in the denominator (at least 30).respondents
[].dhs2024.precision_flagtextHow far the estimate can be trusted (NCHS standards for proportions plus a fewer-than-10-events rule): "reliable", "review" or "unreliable". Suppressed cells never appear here.not applicable
[].dhs2024.precision_reasonnull (in this build)The rule that set the flag (e.g. "low_events<10"); null when the estimate is reliable.not applicable

prevalence.json: The national headline and the comparison

The main page's headline figure for Nigeria (DHS 2024 DISABILITY3, adults 18+, both sexes) and the comparison rows.

File: prevalence.json (JSON, 4 KB). Made from the warehouse's mart_prevalence_national_zone and mart_prevalence_comparison.

Sources and terms

The 24 fields of prevalence.json
NameTypeMeaningUnits
measuretextThe headline's measure, "disability3".not applicable
age_grouptextThe headline's age group, "18+".years of age
sourcetextCitation of the headline figure.not applicable
nationalobjectThe national headline cell (DHS 2024, DISABILITY3, adults 18+, both sexes); always published.not applicable
national.estimatenumberSurvey-weighted share of adults 18+ with a lot of difficulty, or unable, in at least one of six activities (DISABILITY3).proportion (0 to 1)
national.senumberDesign-based standard error of the estimate (Taylor linearisation, R survey).proportion (0 to 1)
national.ci_lownumberLower end of the 95% confidence interval (Korn-Graubard).proportion (0 to 1)
national.ci_highnumberUpper end of the 95% confidence interval (Korn-Graubard).proportion (0 to 1)
national.n_unweightedwhole numberUnweighted number of respondents in the denominator (at least 30).respondents
national.precision_flagtextHow far the estimate can be trusted (NCHS standards for proportions plus a fewer-than-10-events rule): "reliable", "review" or "unreliable". Suppressed cells never appear here.not applicable
national.precision_reasonnull (in this build)The rule that set the flag (e.g. "low_events<10"); null when the estimate is reliable.not applicable
comparisonlistThe comparison rows, in the order the main page shows them.not applicable
comparison[]objectOne figure in the comparison.not applicable
comparison[].figure_idtextStable identifier of the figure (e.g. jonapwd_2025, dhs2024_18plus_wg).not applicable
comparison[].sort_orderwhole numberOrder of the figure on the main page (1 first).not applicable
comparison[].labeltextThe figure's short name as the page shows it.not applicable
comparison[].kindtextWhat sort of figure it is: "assumed" (an assumption, not a measurement), "global estimate" (a world figure) or "measured" (from a survey of Nigeria).not applicable
comparison[].value_percentnumberThe figure itself.percent (0 to 100)
comparison[].ci_low_percentnumber or nullLower end of the 95% confidence interval; only for Curb Cut's own DHS 2024 estimate, null for the other figures.percent (0 to 100)
comparison[].ci_high_percentnumber or nullUpper end of the 95% confidence interval; only for Curb Cut's own DHS 2024 estimate, null for the other figures.percent (0 to 100)
comparison[].populationtextWho the figure is about (e.g. "Household members aged 18+ (de facto)").not applicable
comparison[].definitiontextWhat the figure counts, in words, as its source defines it.not applicable
comparison[].sourcetextCitation of the figure's source.not applicable
comparison[].source_urltextWeb address of the figure's source.web address

spatial.json: Spatial results in words

The main page's spatial section: each result with the plain-English sentence the page shows verbatim (from data/public/spatial/summary.json), cluster members named, and the Moran table as the page renders it.

File: spatial.json (JSON, 11 KB). Made from data/public/spatial/summary.json, with names and labels from the warehouse's dim_states and mart_spatial_moran.

Sources and terms

The 39 fields of spatial.json
NameTypeMeaningUnits
notetextWhy every result uses the direct state estimates and none uses JONAPWD.not applicable
moranlistGlobal Moran's I, one entry per measure tested.not applicable
moran[]objectOne measure's test.not applicable
moran[].measuretextThe DHS 2024 measure tested (disability3 or a domain at "a lot of difficulty" or worse).not applicable
moran[].age_grouptextAge group, "18+" or "5-17".years of age
moran[].roletextThe test's role, "primary" (DISABILITY3 18+), "domain" or "secondary" (DISABILITY3 5-17).not applicable
moran[].moran_inumberGlobal Moran's I of the 37 state estimates.index (no unit)
moran[].p_permnumberOne-sided permutation p-value (positive clustering only).probability (0 to 1)
moran[].p_bhnumberp_perm adjusted with Benjamini-Hochberg across the measures tested.probability (0 to 1)
moran[].n_imprecisewhole numberState estimates not flagged reliable, or equal to 0.states
moran[].labeltextThe measure in words, as the page shows it.not applicable
moran[].sentencetextThe result in one plain-English sentence, shown on the page as written.not applicable
moran_tableobjectThe Moran table as the page draws it (and as its download is written).not applicable
moran_table.captiontextThe table's caption.not applicable
moran_table.columnslistThe table's columns, in order.not applicable
moran_table.columns[]objectOne column.not applicable
moran_table.columns[].labeltextThe column heading.not applicable
moran_table.columns[].typetextHow the page formats the column, "text", "number" or "integer".not applicable
moran_table.columns[].decimalswhole numberDecimal places shown, for number columns.not applicable
moran_table.rowslistThe table's rows.not applicable
moran_table.rows[]list of text or numberOne row, one value per column (unrounded numbers).not applicable
lisaobjectThe exploratory local test (LISA) for DISABILITY3 18+.not applicable
lisa.classesobjectHow many states fall in each LISA class.not applicable
lisa.classes.{lisa_class}whole numberStates in the class (High-High, Low-Low, High-Low, Low-High or Not significant, at Benjamini-Hochberg q below 0.05).states
lisa.significant_stateslist (empty in this build)Names of the states with a significant local result (may be empty).not applicable
lisa.sentencetextThe result in one plain-English sentence.not applicable
clustersobjectThe descriptive profile clusters.not applicable
clusters.kwhole numberNumber of clusters, chosen by the highest mean silhouette of k = 2 to 6.clusters
clusters.groupslistOne entry per cluster.not applicable
clusters.groups[]objectOne cluster.not applicable
clusters.groups[].clusterwhole numberThe cluster's number (a label, not an order).not applicable
clusters.groups[].n_stateswhole numberStates in the cluster.states
clusters.groups[].stateslist of textThe states in the cluster, by name.not applicable
clusters.sentencetextThe result in plain English.not applicable
servicesobjectHealth facilities per 100,000 people against DISABILITY3 18+ prevalence, by state (ecological, descriptive).not applicable
services.spearman_rhonumberSpearman rank correlation across the 37 states.correlation (-1 to 1)
services.p_perm_two_sidednumberTwo-sided permutation p-value of the correlation.probability (0 to 1)
services.caveatslist of textLimits of the comparison, each in one sentence.not applicable
services.sentencetextThe result in plain English.not applicable

poverty.json: Disability and poverty (P5)

The pre-registered poverty analysis for the poverty page and the brief: every estimate by its id, every scalar result by analysis and id, and a few values derived from them. Each estimate is an adjusted association, not a causal effect.

File: poverty.json (JSON, 67 KB). Made from the warehouse's mart_p5_estimates and mart_p5_summary.

Sources and terms

The 50 fields of poverty.json
NameTypeMeaningUnits
notetextWhat the results are and are not (adjusted associations; MPI's own disability questions).not applicable
estimatesobjectEvery published estimate, keyed by its estimate_id (e.g. "primary").not applicable
estimates.{estimate_id}objectOne estimate.not applicable
estimates.{estimate_id}.analysistext"mpi" (the pre-registered poverty analysis) or "ghs" (the exploratory GHS-Panel comparison, E1).not applicable
estimates.{estimate_id}.estimate_grouptextThe estimate's group on the page: headline, secondary, dimension, robustness, leave_one_zone_out, subgroup, sensemakr (bounds under assumed unmeasured confounding), e1 (GHS) or r_check (R's check of an E1 estimate).not applicable
estimates.{estimate_id}.labeltextThe estimate in words, as the page shows it.not applicable
estimates.{estimate_id}.outcometextCode of the outcome (pre-registration section 3.3): poor (multidimensionally poor), score (deprivation score), dim_<dimension> (one MPI dimension), the robustness variants poor_fuel_alt and poor_shock_dk0; for GHS, dwork (change in working) and dlog_consumption (change in log consumption per person). label says it in words.not applicable
estimates.{estimate_id}.exposuretextCode of the exposure definition (pre-registration sections 1 and 3.2). primary: a household member aged 5+ has "a lot of difficulty" or "cannot do at all" in self-care, remembering or concentrating, communication, or (when MPI asked the severity) seeing, hearing or walking; people whose severity was not asked count as not severe. s2: NBS's broad definition, "some difficulty" or worse in any item, including lifting, using the hands and the skin item (albinism or vitiligo). s3: the household head has a severe difficulty. s5: as primary, but people whose seeing, hearing or walking severity was not asked count as severe. s6: routing-free, a severe difficulty in self-care, remembering or communication only. For GHS, onset (disability that began between the waves).not applicable
estimates.{estimate_id}.scaletextThe scale of estimate, se, ci_low, ci_high and mu0, in words (e.g. "risk difference (proportion; x100 = percentage points)").not applicable
estimates.{estimate_id}.learnerstext or nullMachine-learning models used for the outcome and the exposure (e.g. "outcome boosting, exposure logistic"). Null where none were used (the survey benchmark, the sensemakr bounds and GHS).not applicable
estimates.{estimate_id}.estimatenumberThe adjusted difference between exposed and unexposed households (or people for GHS), on the row's scale (a proportion, a 0-1 score, or log points for consumption). For the MPI rows estimated with DoubleML it is standardised to the whole population (population-weighted). Null when suppressed.as in scale; multiply by display_factor for the page
estimates.{estimate_id}.senumber or nullStandard error of the estimate. Null when suppressed.as in scale; multiply by display_factor for the page
estimates.{estimate_id}.ci_lownumberLower end of the 95% confidence interval. Null when suppressed.as in scale; multiply by display_factor for the page
estimates.{estimate_id}.ci_highnumberUpper end of the 95% confidence interval. Null when suppressed.as in scale; multiply by display_factor for the page
estimates.{estimate_id}.pvalnumber or nullTwo-sided p-value for a difference of 0 (DoubleML, standard errors clustered by enumeration area). Only the primary estimate's is a pre-registered test; on the secondary, dimension, robustness, leave-one-zone-out and subgroup rows it is descriptive. Null on the survey benchmark, sensemakr and GHS rows, and when suppressed.probability (0 to 1)
estimates.{estimate_id}.holm_pnumber or nullpval adjusted with Holm's method across S1, S2 and S3 (the pre-registered family for saying a secondary difference "is supported"); null on every other row.probability (0 to 1)
estimates.{estimate_id}.supportedtrue or false or nullTrue when holm_p is below 0.05; only for S1 to S3, otherwise null.not applicable
estimates.{estimate_id}.mu0number or nullThe outcome if no household had a member with a severe difficulty, standardised to the whole population (weighted doubly robust estimate): for poverty, the share of people who would be multidimensionally poor. mu0 plus the estimate is the same standardised share with such a member; neither is the observed rate of either group. Null where not computed (survey benchmark, sensemakr bounds, GHS).proportion (0 to 1)
estimates.{estimate_id}.nwhole numberUnits in the analysis (households for MPI, people for GHS); n_exposed plus n_unexposed.households (MPI) or people (GHS)
estimates.{estimate_id}.n_exposedwhole numberUnweighted exposed households (MPI) or people with onset (GHS). The estimate is suppressed if this or n_unexposed is under 30; both counts are then null.households (MPI) or people (GHS)
estimates.{estimate_id}.n_unexposedwhole numberUnweighted comparison households or people; null when suppressed.households (MPI) or people (GHS)
estimates.{estimate_id}.suppressedtrue or falseTrue when the exposed or the comparison group has fewer than 30 unweighted units. The estimate, se, interval, pval, holm_p, supported, mu0, n_exposed and n_unexposed are then null; n, display_factor and the descriptive columns stay.not applicable
estimates.{estimate_id}.display_factorwhole numberMultiplier for display: 100 turns proportions into percentage points and 0-1 scores into points out of 100; 1 for GHS log consumption (log points).not applicable
summaryobjectEvery scalar result, keyed by analysis ("mpi" or "ghs").not applicable
summary.{analysis}objectOne analysis's results, keyed by item_id.not applicable
summary.{analysis}.{item_id}objectOne result.not applicable
summary.{analysis}.{item_id}.sectiontextThe part of the analysis the result belongs to (e.g. table7, evalue, sensitivity, placebo, overlap, sample, registration).not applicable
summary.{analysis}.{item_id}.labeltextThe result in words, including its unit where it has one.not applicable
summary.{analysis}.{item_id}.kindtextHow to read value: count, share (a proportion, 0 to 1), number, flag (1 yes, 0 no) or text (in value_text).not applicable
summary.{analysis}.{item_id}.valuenumber or nullThe result as a number, read as kind says. Null only for text items and suppressed counts.depends on kind and label
summary.{analysis}.{item_id}.value_texttext or nullThe result as text, for kind "text" (e.g. the OSF address); otherwise null.not applicable
summary.{analysis}.{item_id}.suppressedtrue or falseTrue when the result is a count under 30 (above 0); value is then null.not applicable
derivedobjectValues computed from the estimates and results above; the page and the brief do no other arithmetic.not applicable
derived.verdicttextThe pre-registered wording rule (section 4) applied to the primary estimate: "no meaningful difference" only if the whole 95% interval lies within 2 percentage points of 0; otherwise "higher" or "lower" if the interval excludes 0, else "inconclusive".not applicable
derived.primary_impliednumberThe share of people who would be multidimensionally poor if every household had a member with a severe difficulty, standardised to the whole population (the primary estimate's mu0 plus the estimate); not the observed rate of exposed households.proportion (0 to 1)
derived.overlap_impliednumberThe same standardised share for the overlap-population estimate (its mu0 plus the estimate); not an observed rate.proportion (0 to 1)
derived.share_inside_overlapnumber1 minus the share of people outside the overlap population (summary.mpi.outside_weighted_share).proportion (0 to 1)
derived.rr_ceilingnumber1 / mu0 of the primary estimate, the largest risk ratio the comparison rate allows.ratio
derived.relative_increasenumberThe risk ratio (summary.mpi.evalue_rr) minus 1, the difference as a share of the comparison rate.proportion
derived.placebo_as_extremewhole numberPlacebo estimates at least as far from 0 as the real one.placebo estimates
derived.robustnessobjectRange of the robustness estimates.not applicable
derived.robustness.minnumberSmallest published robustness estimate.as in the estimates' scale (proportion)
derived.robustness.maxnumberLargest published robustness estimate.as in the estimates' scale (proportion)
derived.leave_one_zone_outobjectRange of the leave-one-zone-out estimates.not applicable
derived.leave_one_zone_out.minnumberSmallest published leave-one-zone-out estimate.as in the estimates' scale (proportion)
derived.leave_one_zone_out.maxnumberLargest published leave-one-zone-out estimate.as in the estimates' scale (proportion)
derived.smallest_groupwhole numberThe smallest group (exposed or comparison) behind any published estimate.households or people
derived.benchmarksobjectThe widest range the three observed-covariate benchmarks (head's education, place, sector) move the primary estimate to.not applicable
derived.benchmarks.minnumberLowest of the benchmarks' lower bounds.as in the estimates' scale (proportion)
derived.benchmarks.maxnumberHighest of the benchmarks' upper bounds.as in the estimates' scale (proportion)

Table downloads (CSV and JSON)

Under every data table on the site is a line "Download this table: CSV, n KB · JSON, n KB". There is one pair of files per table, so there are hundreds; all share one structure, described here once. They sit under data/downloads/: main/ for the main page, poverty/ for the poverty page and states/<state>/ for each state page (for example states/akwa-ibom/), one <table>.csv and one <table>.json for each table.

The fields of every table's JSON download
NameMeaningUnits
tableThe table's id, unique on its page (also the file name).not applicable
nameThe table's short name, as in the download links.not applicable
captionThe table's caption on the page.not applicable
pageAddress of the page that shows the table.not applicable
surveyThe survey (or source) behind the figures.not applicable
yearThe survey's year or years.not applicable
sourcesCitations of the sources, as a list.not applicable
termsThe reuse terms. No licence is granted for these files; reuse is subject to the source terms.not applicable
valuesHow the numbers relate to the page (unrounded, and how words and floors are written).not applicable
columnsThe columns, in order.not applicable
columns[]One column.not applicable
columns[].labelThe column heading, as in the CSV's header row (with the unit added where the heading lacks it).not applicable
columns[].type"text", "number" or "integer".not applicable
columns[].decimalsDecimal places the page shows (number and integer columns).not applicable
columns[].unitThe unit, where the column has one (e.g. "%").not applicable
rowsThe rows, one value per column; null where the page shows nothing.not applicable
rows[]One row.not applicable

The API

An optional, read-only JSON API serves some of these files at https://curb-cut-api.onrender.com. Its fields are the files' fields, described above, inside a small envelope (a count, the units and the source). Its interactive documentation is at curb-cut-api.onrender.com/docs. This site never calls it: every figure is in the files themselves. It runs on a free plan, so the first request after a quiet spell can take about a minute.

API endpoints and the files each one serves
EndpointWhat it returnsFiles
/statusWhether the data are loaded, from where, and each file's row count and SHA-256.dhs2024_state_prevalence.parquet, dhs2024_national_zone.parquet, prevalence_comparison.parquet, states.json, poverty.json
/statesThe 37 states with their facts and DHS 2024 headline.states.json
/states/{state}One state, by the site's slug or its ISO 3166-2 code.states.json
/prevalenceDHS 2024 cells by state, zone or nation (the headline by default).dhs2024_state_prevalence.parquet, dhs2024_national_zone.parquet
/compareThe comparison rows, in the site's order.prevalence_comparison.parquet
/povertyThe pre-registered poverty results (note, estimates, summary).poverty.json

Other files

The site also serves these, which are not data tables: