undp-hdr

The statistics of the Human Development Report of the United Nations Development Programme (UNDP), as its Human Development Report Office (HDRO) publishes them at https://hdr.undp.org/data-center/documentation-and-downloads: the Human Development Index (HDI) and the other composite indices (the Gender Development Index GDI, the Inequality-adjusted HDI IHDI, the Gender Inequality Index GII and the Planetary pressures-adjusted HDI PHDI) with their components, for 195 countries and territories and 11 aggregates, 1990 to 2023. Beside them are the statistical annex tables of the report and the global Multidimensional Poverty Index (MPI) tables, which HDRO produces jointly with OPHI (University of Oxford), unchanged.

Licensed under Creative Commons Attribution 3.0 IGO (see Licence).

HDRO recalculates the whole series every year. In its words: "The entire series of Human Development Index (HDI) values and rankings are recalculated every year using the same the most recent (revised) data and functional forms. The HDI rankings and values in the 2014 Human Development Report cannot therefore be compared directly to indices published in previous Reports. Please see hdr.undp.org for more information." (quoted as published, with the report year of the original wording). Values and ranks from different reports are not comparable; compare years within one release's time series only.

Releases

One subset per table and release. A release is named after the directory of the report on hdr.undp.org (.../2025_HDR/... is hdr2025); a new report will be added beside it, not over it.

subset	rows	made from
hdr2025.composite_indices_long	200,004	raw/hdr2025/HDR25_Composite_indices_complete_time_series.csv
hdr2025.composite_indices	206	the same CSV

hdr2025 is the Human Development Report 2025, whose data HDRO updated on 6 May 2025. It was the current release when the files were fetched on 2026-10-04; HDR 2026 data had not been published. The MPI tables in raw/hdr2025/ are the MPI 2025 update of 17 October 2025, filed with the release that was current when they were fetched.

All the raw files are HDRO's own, byte for byte, in raw/hdr2025/; its MANIFEST.json gives their URLs, sizes, sha256 and Last-Modified.

file	what it is	here as
HDR25_Composite_indices_complete_time_series.csv	every composite index and component, 1990 to 2023	raw, and the two Parquet subsets
HDR25_Composite_indices_metadata.xlsx	the codebook of that CSV, and the recommended citation	raw only
HDR25_Statistical_Annex_HDI_Table.xlsx	annex table 1, HDI	raw only
HDR25_Statistical_Annex_HDI_Trends_Table.xlsx	annex table 2, HDI trends	raw only
HDR25_Statistical_Annex_IHDI_Table.xlsx	annex table 3, IHDI	raw only
HDR25_Statistical_Annex_GDI_Table.xlsx	annex table 4, GDI	raw only
HDR25_Statistical_Annex_GII_Table.xlsx	annex table 5, GII	raw only
HDR25_Statistical_Annex_PHDI_Table.xlsx	annex table 7, PHDI	raw only
HDR25_Statistical_Annex_Tables_1-7.xlsx	the six annex tables in one workbook	raw only
2025_gMPI_Table1and2.xlsx	MPI tables 1 (national) and 2 (subnational)	raw only

The xlsx files are not converted. They are tables laid out for the printed report: a title row, two to four rows of headings and units, rows that name a group of countries in the middle of the data, a column of footnote letters beside many value columns, and notes below the table. The codebook sheet of the metadata workbook has a heading row between its sections. None of them is a table of one header row and one record per row. Most values of the annex tables are also in the time series CSV; some are not, for example the income shares and the inequality-adjusted component indices of table 3 and the GNI per capita rank minus HDI rank of table 1. The MPI is not in the CSV at all.

Files and columns

composite_indices (wide)

The CSV as it is: one row per country or aggregate, 1,112 columns in the order of the CSV. The first four label the row; each of the other 1,108 is one indicator in one year, named <indicator>_<year>, for example hdi_1990 or gni_pc_f_2023. The type comes from the column name, never from inference:

columns	type
iso3, country, hdicode, region	VARCHAR
hdi_rank_2023, gii_rank_2023, gdi_group_2023, rankdiff_hdi_phdi_2023	INTEGER
every other <indicator>_<year>, for the 36 indicators of the codebook below	DOUBLE
a column that fits none of these (there is none in hdr2025)	VARCHAR

An empty cell is NULL. Each value was checked against a regular expression (-?[0-9]+ for INTEGER, -?[0-9]+(\.[0-9]+)? for DOUBLE) before the cast, and a value that did not fit would have stopped the build.

composite_indices_long

The same values, one row per filled cell of the 1,108 value columns: 1,108 x 206 = 228,248 cells, of which 28,244 are empty and left out.

column	type	from
iso3	VARCHAR	the row's iso3
country	VARCHAR	the row's country
region	VARCHAR	the row's region (NULL where the CSV is empty)
indicator	VARCHAR	the column name before its last underscore
year	INTEGER	the four digits after it
value	DOUBLE	the cell

The rule: a column name <indicator>_<year> is split at its last underscore when four digits follow it, so hdi_1990 is hdi and 1990, gni_pc_f_2023 is gni_pc_f and 2023, and hdi_rank_2023 is hdi_rank and 2023. value is DOUBLE for every indicator; the ranks and groups are small integers and a DOUBLE holds them exactly. hdicode is not repeated here; join the wide table on iso3 for it. Rows are sorted by indicator, iso3 and year.

Indicators

From the codebook sheet of HDR25_Composite_indices_metadata.xlsx (the names as written there):

index	indicator	full name	years
HDI	hdi_rank	HDI Rank	2023
HDI	hdi	Human Development Index (value)	1990-2023
HDI	le	Life Expectancy at Birth (years)	1990-2023
HDI	eys	Expected Years of Schooling (years)	1990-2023
HDI	mys	Mean Years of Schooling (years)	1990-2023
HDI	gnipc	Gross National Income Per Capita (2021 PPP$)	1990-2023
GDI	gdi_group	GDI Group	2023
GDI	gdi	Gender Development Index (value)	1990-2023
GDI	hdi_f, hdi_m	HDI female, male (value)	1990-2023
GDI	le_f, le_m	Life Expectancy at Birth, female, male (years)	1990-2023
GDI	eys_f, eys_m	Expected Years of Schooling, female, male (years)	1990-2023
GDI	mys_f, mys_m	Mean Years of Schooling, female, male (years)	1990-2023
GDI	gni_pc_f, gni_pc_m	Gross National Income Per Capita, female, male (2021 PPP$)	1990-2023
IHDI	ihdi	Inequality-adjusted Human Development Index (value)	2010-2023
IHDI	coef_ineq	Coefficient of human inequality	2010-2023
IHDI	loss	Overall loss (%)	2010-2023
IHDI	ineq_le	Inequality in life expectancy	2010-2023
IHDI	ineq_edu	Inequality in eduation (sic)	2010-2023
IHDI	ineq_inc	Inequality in income	2010-2021, 2022-2023=2022
GII	gii_rank	GII Rank	2023
GII	gii	Gender Inequality Index (value)	1990-2023
GII	mmr	Maternal Mortality Ratio (deaths per 100,000 live births)	1990-2019, 2020-2023=2020
GII	abr	Adolescent Birth Rate (births per 1,000 women ages 15-19)	1990-2023
GII	se_f, se_m	Population with at least some secondary education, female, male (% ages 25 and older)	1990-2023
GII	pr_f, pr_m	Share of seats in parliament, female (% held by women), male (% held by men)	1990-2023
GII	lfpr_f, lfpr_m	Labour force participation rate, female, male (% ages 15 and older)	1990-2023
PHDI	rankdiff_hdi_phdi	Difference from HDI rank	2023
PHDI	phdi	Planetary pressures-adjusted Human Development Index (value)	1990-2023
PHDI	diff_hdi_phdi	Difference from HDI value (%)	1990-2023
PHDI	co2_prod	Carbon dioxide emissions per capita (production) (tonnes)	1990-2023
PHDI	mf	Material footprint per capita (tonnes)	1990-2023
additional	pop_total	Population, total (millions)	1990-2023

The codebook has years 1990-2023 for every indicator not listed with another range, and the CSV has a column for each of those years even where the index itself starts later: the IHDI columns run from 2010 only.

-- DuckDB, straight off the Hub: HDI of Japan, Norway and Niger over time, HDR 2025
SELECT iso3, country, year, value AS hdi
FROM 'hf://datasets/yuiseki/undp-hdr/parquet/hdr2025/composite_indices_long.parquet'
WHERE indicator = 'hdi' AND iso3 IN ('JPN', 'NOR', 'NER')
ORDER BY iso3, year;

-- the wide table, countries only, ranked by HDI 2023
SELECT iso3, country, hdicode, hdi_rank_2023, hdi_2023, le_2023, gnipc_2023
FROM 'hf://datasets/yuiseki/undp-hdr/parquet/hdr2025/composite_indices.parquet'
WHERE iso3 NOT LIKE 'ZZ%'
ORDER BY hdi_rank_2023;

Things to know

• The rows are not all countries. 195 have a three-letter country code in iso3; 11 are aggregates whose iso3 begins with ZZ: the four HDI groups (ZZA.VHHD, ZZB.HHD, ZZC.MHD, ZZD.LHD), the six developing regions (ZZE.AS to ZZJ.SSA) and the world (ZZK.WORLD). Leave out iso3 LIKE 'ZZ%' to keep countries only; the aggregates have no hdicode and no region.
• hdicode is Very High (74), High (50), Medium (43), Low (26) or "Other Countries or Territories" (2: the Democratic People's Republic of Korea and Monaco, which have no HDI). region is the UNDP developing region (SSA 46, LAC 33, EAP 26, AS 20, ECA 17, SA 9); it is empty for the 44 countries outside them, as well as for the aggregates.
• Some values are not observations. The codebook gives mmr as "1990-2019, 2020-2023=2020" and ineq_inc as "2010-2021, 2022-2023=2022": for every country, mmr of 2021, 2022 and 2023 equals mmr of 2020, and ineq_inc of 2023 equals that of 2022. Only the 11 aggregates differ between those years.
• Ranks share places: hdi_rank_2023 has ties (Norway and Switzerland are both 2), and 193 rows have one.
• pr_f and pr_m add to 100 wherever both are filled.
• Of the 228,248 value cells, 28,244 are empty. In 2023, hdi is filled for 193 countries, gdi for 184, gii for 172, ihdi for 169 and phdi for 156; in 1990, hdi for 141 countries.
• The CSV is Windows-1252, not UTF-8. Only two rows are outside ASCII (Côte d'Ivoire and Türkiye); the Parquet holds them as UTF-8.
• The annex xlsx hold the same 2023 values as the CSV without the CSV's rounding (0.97199999999999998 in the xlsx against 0.972 in the CSV for Iceland's HDI).
• Many components come from other agencies (UNDESA, UNESCO, the World Bank, the IMF, the ILO, the IPU, WHO and others, listed in the Technical Notes of the report); the indices are HDRO's calculations from them.

Checked before upload

scripts/02_verify.py in the source repository checks, with Python's csv module reading the CSV in Windows-1252 and DuckDB reading the Parquet:

• every raw file has the size and sha256 in MANIFEST.json, and every xlsx is a sound zip
• the CSV has 206 rows; composite_indices has 206 rows and the CSV's 1,112 column names in the same order, each of the type the rule gives
• all 229,072 cells of composite_indices equal the CSV's, compared one by one: text as text, numbers as Python's int and float read them, an empty cell as NULL
• composite_indices_long has 200,004 rows, one per filled value cell, and the same value for every (iso3, indicator, year), with the row's country and region
• a DuckDB with extension autoloading turned off opens both files
• no row group of either file has a bloom filter

Licence

hdr.undp.org, Terms of use (https://hdr.undp.org/terms-use, read on 2026-10-04):

Creative Commons Attribution 3.0 IGO All materials provided on this website are copyrighted under the Creative Commons Attribution 3.0 IGO license. This International Governmental Organizations license was developed under the auspices of the World Intellectual Property Organization (WIPO) and has been adopted by a number of other international organizations, such as CERN, ILO, OECD, WHO, UNESCO and the UN.

You are free to: Share - copy and redistribute the material in any medium or format Adapt - remix, transform, and build upon the material for any purpose, even commercially.

Appropriate Credit You must give appropriate credit, provide a link to the license, and indicate where changes were made. You may do so in any reasonable manner, but in no way that suggests the licensor endorses you or your use.

(The separators in the original are dashes; they are written - here.)

The same page asks that the materials be used in line with the principles of the United Nations Charter, and, under "Latest available Data", that data incorporated in internal servers be the latest available and follow its updates. This dataset names its release (hdr2025) on every subset so that a newer release is not mistaken for this one.

The legal code is at https://creativecommons.org/licenses/by/3.0/igo/legalcode. The card declares license: other with license_name: cc-by-3.0-igo, which names this licence exactly; CC BY 3.0 IGO is not CC BY 3.0 (unported) and differs from it in, among other things, its dispute resolution by mediation and arbitration and its preservation of the privileges and immunities of the licensor.

This dataset is not made or endorsed by UNDP or HDRO. What was changed: the time series CSV was decoded from Windows-1252, typed and written as Parquet in two shapes (wide and long), as described above; no value was changed. The raw files are unchanged. LICENSE beside this file says the same.

The report itself (the PDF) carries "All rights reserved" in its imprint; it is not here. The time series and the metadata carry no licence text of their own, so the site's terms apply to them.

Citation

The citation HDRO recommends, from the recommended_citation sheet of HDR25_Composite_indices_metadata.xlsx (the separator in the original is a dash):

Source: UNDP (United Nations Development Programme). 2025. Human Development Report 2025 - A matter of choice: People and possibilities in the age of AI. New York.

The global MPI tables are produced jointly by HDRO and OPHI (Oxford Poverty and Human Development Initiative, University of Oxford).