Nepal Provinces Ranked: MPI Poverty Rate and HDI by Province

Development in Nepal is deeply uneven across its seven provinces, and two authoritative datasets make the gap concrete. By the Human Development Index (HDI), which combines life expectancy, education and income, Bagmati Province is the most developed province in Nepal with an HDI of 0.669, followed by Gandaki (0.621). By the Multidimensional Poverty Index (MPI), which counts overlapping deprivations in health, education and living standards, Bagmati again ranks best, with only 7.0 percent of its people multidimensionally poor.

The poorest province in Nepal depends on which measure you use. Karnali Province is the poorest by multidimensional poverty: 39.5 percent of Karnali residents are MPI poor, by far the highest rate in the country. Madhesh Province (Province 2), however, records the lowest HDI at 0.519, making it the least developed province on the human-development scale. Because these two provinces top the deprivation tables in different ways, Nepal's development debate often names both Karnali and Madhesh as the country's most disadvantaged regions.

This page brings the two datasets together into a single ranked comparison the way government reports rarely do side by side. The MPI figures come from the Nepal Multidimensional Poverty Index 2021, prepared by the National Planning Commission (NPC) with the Oxford Poverty and Human Development Initiative (OPHI), UNDP and UNICEF, using 2019 survey data. The provincial HDI figures come from Nepal's National Human Development Report 2020 (NPC/UNDP), based on 2019 estimates.

The Human Development Index is a summary score from 0 to 1 that averages three dimensions of well-being: a long and healthy life (life expectancy at birth), knowledge (expected and mean years of schooling) and a decent standard of living (gross national income per capita). Nepal's national HDI was 0.602 in 2019, placing the country in the 'medium human development' band. All seven provinces also fall inside the medium band, but they range from Bagmati's 0.669 down to Madhesh's 0.519.

The Multidimensional Poverty Index takes a different, complementary approach. Instead of averaging economic aggregates, it counts the specific, overlapping deprivations a household actually experiences across 10 indicators grouped into three equally weighted dimensions. A person is identified as multidimensionally poor if they are deprived in at least one-third of the weighted indicators. The MPI is reported as a value (0 to 1) that equals the incidence of poverty multiplied by its intensity.

Two numbers matter most when reading MPI tables. Incidence (H) is the share of people who are poor, and intensity (A) is the average share of deprivations those poor people face. Nepal's national MPI in 2019 was 0.074, with an incidence of 17.4 percent and an intensity of 42.5 percent. Because MPI value combines both, a province can have a lower headcount but a higher overall MPI if its poor are more intensely deprived, which is exactly what separates Madhesh and Sudurpashchim in the rankings below.

The Nepal MPI 2021 report ranks provinces by their MPI value, and the spread is dramatic. Karnali Province sits at the bottom with an MPI value of 0.169 and an incidence of 39.5 percent, meaning nearly two in five Karnali residents are multidimensionally poor. At the other end, Bagmati Province has an MPI value of just 0.028 and an incidence of 7.0 percent. In other words, a person in Karnali is more than five times as likely to be multidimensionally poor as a person in Bagmati.

A crucial nuance sits in the middle of the table. By raw incidence, Sudurpashchim (25.3 percent poor) actually has a slightly higher headcount than Madhesh/Province 2 (24.2 percent). Yet Madhesh ranks as the second-poorest province by MPI value (0.109 versus Sudurpashchim's 0.105), because the poor in Madhesh face a higher intensity of deprivation (45.0 percent, the highest of any province) than the poor in Sudurpashchim (41.3 percent). This is why headline 'poorest province' claims should always specify the measure.

The picture changes again when you count people rather than percentages. Because Madhesh and Lumbini are populous, they carry the largest absolute numbers of poor: about 1.30 million multidimensionally poor people live in Madhesh and roughly 0.96 million in Lumbini, even though their poverty rates are far below Karnali's. Nationwide, the report estimates that about 5 million Nepalis were multidimensionally poor in 2019.

Nepal's National Human Development Report 2020 assigns each province an HDI, and the ranking closely tracks the poverty picture but is not identical. Bagmati leads at 0.669, powered by high educational attainment and per-capita income concentrated in the Kathmandu Valley (whose sub-regional HDI reaches 0.622). Gandaki follows at 0.621, reflecting strong human-development outcomes and high remittance-supported incomes in the central hills.

The middle of the HDI table is tightly bunched: Koshi Province (Province 1) scores 0.580, Lumbini 0.563 and Sudurpashchim (Sudur Paschim) 0.547. The bottom two provinces are Karnali at 0.538 and Madhesh (Province 2) at 0.519, the lowest in the country. Notably, Madhesh ranks last on the HDI even though its multidimensional poverty rate is lower than Karnali's, largely because low literacy, low mean years of schooling and modest incomes drag down its human-development score.

Every province still sits inside the 'medium human development' category (roughly 0.55 to 0.70 nationally), but the roughly 0.15-point gap between Bagmati and Madhesh is large in HDI terms. The report also shows a wide urban-rural divide, with urban Nepal at about 0.647 against rural Nepal at 0.561, and a hill-versus-lowland gap, hills at 0.623 against the Terai and mountains near 0.56.

Put the two indices together and one province stands clearly at the top. Bagmati is Nepal's most developed province on both scales: it holds the highest HDI (0.669) and the lowest multidimensional poverty rate (7.0 percent). As the home of Kathmandu, the federal capital, and the country's densest concentration of jobs, schools, hospitals and infrastructure, Bagmati is the reference point against which other provinces measure themselves. Gandaki is a consistent second on both measures.

At the bottom, the title of 'poorest province' is genuinely split. Karnali is the poorest by the multidimensional poverty rate, a remote, mountainous province where deprivations in schooling, sanitation, cooking fuel and nutrition overlap heavily for a large share of households. Madhesh, by contrast, is the least developed province by HDI, held back by the lowest literacy and schooling outcomes in the country despite its fertile Terai farmland and comparatively better connectivity. Sudurpashchim rounds out the three most disadvantaged provinces.

The gap is stark in human terms: the incidence of multidimensional poverty in Karnali (39.5 percent) is more than five times that of Bagmati (7.0 percent), and Bagmati's HDI exceeds Madhesh's by about 0.15 points. Reducing this east-west and hills-versus-Karnali divide is a central goal of Nepal's federal fiscal transfers and targeted programmes, which channel extra resources to Karnali, Madhesh and Sudurpashchim.

The provincial gap is not random; it reflects geography, connectivity and history. Karnali's very high multidimensional poverty is tied to extreme remoteness, harsh mountain terrain, thin road networks and long distances to schools and health posts, all of which push up deprivations in schooling, child nutrition and access to clean cooking fuel and sanitation. When households are deprived on many indicators at once, both incidence and intensity rise together, which is why Karnali tops the MPI.

Madhesh's challenge is different in character. Its multidimensional poverty rate (24.2 percent) is well below Karnali's, and its Terai location gives it flatter land and better transport links, yet it records the country's lowest HDI. The binding constraint there is human capital: low literacy, low mean years of schooling and modest incomes weigh heavily on the education and income components of the HDI. Madhesh also has the highest intensity of poverty among the poor (45.0 percent), meaning those who are poor tend to be deprived across many indicators simultaneously.

Bagmati and Gandaki, meanwhile, benefit from concentrated urban economies, higher schooling, larger remittance flows and better service delivery, which lift both their HDI and their poverty performance. Across all provinces, the biggest single contributor to the MPI is deprivation in years of schooling, followed by nutrition and household assets, underlining education as the most powerful lever for closing the provincial development gap.

The direction of travel is encouraging. Nepal roughly halved its multidimensional poverty incidence from 30.1 percent in 2014 to 17.4 percent in 2019, and every province recorded gains. The MPI 2021 report highlights that some of the poorest provinces reduced poverty fastest in relative terms, which helped narrow, though far from close, the gap with Bagmati and Gandaki. Reductions in cooking-fuel and housing deprivation were especially widespread.

More recent subnational HDI estimates suggest continued, broad-based improvement while preserving the same ranking. Independent estimates (for example, the Global Data Lab subnational HDI, and figures compiled in reference databases) put Nepal's national HDI around 0.62 in the early 2020s, with Bagmati still highest and Madhesh still lowest, and Sudurpashchim continuing to edge above Karnali. The east-to-west and Karnali-versus-rest pattern has proven durable across every dataset.

A few caveats are worth stating. The official MPI 2021 figures are anchored to 2019 survey data (the Nepal Multiple Indicator Cluster Survey, NMICS 2019) and predate the COVID-19 pandemic, so real-time conditions may differ. The NHDR provincial HDI values reflect 2019 estimates. The MPI is also distinct from Nepal's consumption-based poverty rate, which is measured separately through the Nepal Living Standards Survey; the two answer different questions and should not be conflated.