What the Human Poverty Index Actually Measures
The Human Poverty Index was UNDP's attempt to quantify deprivation rather than average achievement. Unlike the HDI which looks at means, HPI looked at shortfalls. It measured how many people were falling behind on basic human development, not how well the average person was doing. I first ran into this when a colleague at a World Bank project asked why our country's HDI ranking looked fine but the ground reality told a different story. The numbers didn't match the poverty maps. That's when I learned that HPI captures something HDI glosses over entirely.
How to Calculate the Human Poverty Index
The formula uses three dimensions with specific indicators. For HPI-1, which applied to developing countries, you take health deprivation (percentage of people not surviving to age 40), knowledge deprivation (adult illiteracy rate), economic survival (combined measure of no access to safe water and proportion of population below income poverty line), and sometimes a fourth indicator depending on the year. You normalize each indicator, take the geometric mean, and round to one decimal place. The result ranges from 0 to 100. In practice this means you're downloading life expectancy at birth data, adult literacy rates, and access to water statistics from UNESCO, WHO, and national surveys. The data lag is usually 3 to 5 years because these indices don't get updated annually. If you need current numbers, you're mostly working with estimates. I spent about two weeks once trying to get HPI values for a cluster of landlocked African countries. The problem was that some national surveys used different poverty lines based on local food baskets rather than the international $1.90/day standard. The UNDP methodology required adjusting these to a common baseline, and the adjustment factors varied by source. I ended up using the country-specific poverty headcount ratios from the World Bank's PovcalNet instead, applying the conversion factor of roughly 1.5 to account for purchasing power differences. That saved me about 10 hours of back-and-forth with data requests.
Common Pitfalls People Miss
Here's something most tutorials don't mention. HPI penalizes countries differently depending on their population distribution. A country with a large rural population that lacks safe water will score worse not just because water access is low, but because the indicator weights rural deprivation more heavily when the calculation is done at the national level. The index doesn't separate urban from rural in its standard published form, so you lose that granularity. Another issue: the health component uses "not surviving to age 40" which sounds arbitrary but was chosen deliberately. It captures early mortality from preventable causes in developing contexts. In practice this means countries with high maternal mortality or child mortality get hit hard even if their life expectancy at birth looks decent on paper. The two measures diverge significantly in sub-Saharan Africa where HIV/AIDS and infectious disease create a gap between average life expectancy and the proportion dying before 40. I've also seen analysts make a mistake combining HPI-1 and HPI-2 data. HPI-2 was the version for OECD countries using different indicators like long-term unemployment and households with insufficient assets. You can't mix these datasets in a single analysis without adjusting for the different indicator sets. I watched a researcher do this in a conference presentation and the entire comparison fell apart within five minutes of Q&A.
Get the Full Details
When HPI Fails Completely
The index breaks down in very small populations. Countries under 1 million people have unreliable estimates because the sampling error swamps the signal. Somalia and South Sudan in particular had wildly unstable HPI readings year over year due to survey gaps and displacement complicating data collection. It also doesn't capture inequality within poverty groups. A country where 30% of people are extremely poor and another where 30% are moderately poor will have the same HPI score. The index treats all deprivation as equal once the threshold is crossed. This matters because the policy implications of severe versus mild deprivation are completely different, but HPI can't distinguish them. UNDP retired HPI after the 2010 Human Development Report and moved to the Multidimensional Poverty Index. MPI uses overlapping deprivations across health, education, and living standards at the household level rather than aggregate national indicators. If you're starting a new analysis today, you should probably use MPI instead. It requires more granular data from DHS or MICS surveys but gives you household-level results that are actually useful for targeting interventions.
Where to Get the Data
Historical HPI values are archived in the UNDP's Human Development Report database at hdr.undp.org. The data goes back to 1997. For newer work, grab the MPI datasets from the Oxford Poverty and Human Development Initiative website at povertylinks.org or from national DHS survey microdata through the ICPSR archive. The DHS program offers free access to their survey data if you register as a researcher. If you're doing this for academic work, the Stata commands are available through the MPIdb package. It handles the weighting and aggregation automatically. I switched to that about four years ago and it cut my processing time from a couple days per country cluster to about 20 minutes. The old manual method of pulling indicators from five different databases and normalizing them myself was tedious and error-prone. The Human Poverty Index Meaning is straightforward once you understand what it's designed to show, but the execution has enough traps that even experienced researchers make mistakes. The key is knowing which version applies to your study, whether you should be using HPI or MPI instead, and that the data you're working with actually meets the quality standards for cross-country comparison.
