Annual GrowthCreated with Sketch. Climate RishCreated with Sketch. double-chevronE-mailCreated with Sketch. FacebookCreated with Sketch. GDPCreated with Sketch. GovernanceCreated with Sketch. Index@2xCreated with Sketch. IntensityPovertyCreated with Sketch. Page 1Created with Sketch. %UPCreated with Sketch. SafetyCreated with Sketch. SocialAssistCreated with Sketch. SocialAssistCreated with Sketch. swipe-iconTOTALPOPCreated with Sketch. TOTALUPCreated with Sketch. TwitterCreated with Sketch. Uplift_Logo_Reversed_Swooth_TagCreated with Sketch. Uplift_Logo_Swoosh_NoTag (2) (1)Created with Sketch. WomenInequalityCreated with Sketch. Map_Icon@2xCreated with Sketch.

Accelerating the End of Ultra-Poverty

Our Methodology

How we built the Synthetic Index

The Objective

As the international community monitors the number of people living in extreme poverty to ensure reductions continue at a pace to reach the year 2030 target of ending extreme poverty, there is also a need to understand whether a subset of the extreme poor are being left behind. Recent publications suggest this might well be the case.

To cite a few: the Brookings Institution suggests there is likely to be more than 100 million people left below the extreme poverty line in 2030. (1) Development Initiatives of the UK stated: “There is a risk even in the best-case scenarios (that) growth will fail to lift millions of people out of extreme poverty by 2030 unless growth rates are implausibly high or growth becomes far more inclusive”. (2)

These people still mired in extreme poverty need to be targeted for particular assistance, but, again in the words of Development Initiatives, this “requires knowing who the poor are, where they live, and how deep their poverty level is”.

Our methodology was designed with these specific questions in mind.

The Options

There are two ways to analyze ultra-poverty: utilizing ultra-poverty levels, or examining those segments of the population living in ultra-poverty.

In the first case, one can solely monitor, for example, the Human Development Index of a country and see how it evolves over time. This provides an idea of the average situation and evolution of the country, but it does not allow us to address our overall concern, which is to determine whether a segment of the population is being left behind.

In the second case, one can select a certain poverty threshold and see who lives below that. The threshold can be relative; i.e., expressed in terms of percentage. Development Initiatives publishes a report on that segment of the population living below the 20th percentile of poverty (referred to as “P20”). This segment of 1.5 million people, however, does not necessarily capture people who are in extreme poverty, much less in ultra-poverty (we will see later that the people in ultra-poverty are closer to being the poorest quintile of those living in the poorest quintile).

The more commonly-used threshold is absolute and it is expressed in monetary terms. The World Bank has defined the threshold of extreme poverty as being US$1.90/day in purchasing power parity (PPP). and it produces a Poverty and Shared Prosperity Report (3) as well as a data visualization (4) that summarizes the number of people living under this threshold in each country. The international community is monitoring this indicator in assessing progress toward the relevant 2030 Sustainable Development Goal. If we wanted to look at the subset of the extreme poor facing the greatest challenge, we could examine the population living, say, under US$1.30/day PPP. The World Bank’s PovCalNet allows the production of such data. (5)

Aside from some methodological limitations inherent with this approach (there are several countries for which the data is absent or very old), this method has one key downside; it focuses exclusively on one aspect of poverty, the monetary dimension. But no single dimension can capture the multiple aspects that constitute poverty. Along with income, measures that point to poor health, lack of education, an inadequate living standard, and disempowerment can help us to understand the multidimensional aspects of poverty better than any one indicator. We have chosen a multidimensional approach for several reasons:

  1. Income alone is too narrow. For instance, economic growth has been strong in India in recent years and yet childhood malnutrition has remained at nearly 50 percent – one of the highest rates worldwide.
  2. The way people experience poverty is multidimensional. They describe their poverty as including poor health, low education, violence, etc.
  3. The more policy-relevant information we have, the better equipped we are to have an impact. For example, a different strategy is needed to help people lacking education than might be applied to an area in which most people are deprived of housing.
  4. Furthermore, an increase in income does not automatically produce better access to basic services and to opportunities.

In response to similar concerns, the Oxford Poverty and Human Development Institute (OPHI), in collaboration with the United Nations Development Programme (UNDP), has developed and tracked for many years a global Multidimensional Poverty Index (MPI) based on deprivations in health, education, and livelihood conditions. By mapping outcomes for each individual or household against the criteria being measured, this method can capture both the percentage of people who are poor and the overlapping deprivations they face.

As summarized in Table 1, the MPI uses information from 10 indicators that are classified under three dimensions: health, education, and living standards. Each member of a household is identified as deprived or non-deprived in each indicator based on a deprivation cut-off. Health and education indicators reflect the deprivations of all household members. Each person’s deprivation score is constructed based on a weighted average of the deprivations the household experiences using a nested weight structure; equal weights across dimensions, and an equal weight for each indicator within dimensions.

Table 1: The dimensions, indicators, deprivation cutoffs, and weights of the global MPI (6)


Specifically, to analyze a household’s poverty level, OPHI determines whether they are deprived in each of 10 indicators: years of schooling, child school attendance, child mortality, nutrition, electricity, improved sanitation, improved drinking water, flooring, cooking fuel, and asset ownership. The two education indicators and the two health indicators each receive a weighting of one-sixth (i.e., a combined weighting per dimension of one-third). The six living standard indicators each receive a weighting of one-eighteenth for a combined weighting of one-third. A weighted poverty deprivation score of 33.3 percent or more identifies households as multi-dimensionally poor. (7)

OPHI identifies those people deprived in 20.0-33.3 percent of weighted indicators as “vulnerable to poverty” while those deprived in 50 percent or more of the weighted indicators are identified as in “severe poverty”.

Calculating the Number of People in Ultra-Poverty

We chose to describe a household as in ultra-poverty if it had a weighted rate of deprivation of 60 percent or more, based on the following:

  • Intuitively, we wanted to capture those who were not just at the 50 percent deprivation mark but who were clearly beyond this level.
  • Considered in detail, the 60 percent weighted deprivation corresponded to what we regarded as ultra-poverty:
  • Either a household fully deprived in at least two out of three dimensions (health, education, and livelihood), or
  • The household was at least half deprived in all three dimensions, with one dimension measured as nearly completely deprived.
  • Just as importantly, if not more so, under this definition those people in ultra-poverty were – in large part – a subset of the extreme poor (those living under US$1.90/day PPP).

The relationship between our definition of ultra-poverty and extreme poverty (as defined by the cutoff of US$1.90/day) warrants explanation. As can be seen from Table 2, there are 394 million people living in ultra-poverty worldwide, slightly more than half the number of people in extreme poverty. However, 47 million of those are actually living on more than $1.90/day but remain in ultra-poverty on a multidimensional basis. If we look at the prior survey data of those living under $1.90/day, it becomes apparent that while the country has reduced the number living under $1.90/day that has not translated into improvements outside of monetary gains. This requires further examination but it might suggest that gains in reducing monetary poverty have not yet translated into progress across multidimensional indicators, such as access to schools and health care.

Overall 47 million ultra-poor have incomes above US$1.90/day; i.e., almost 12 percent of the ultra-poor. Additional remaining ultra-poor could also technically have incomes above US$1.90/day but this is unlikely given their levels of deprivation.

Table 2. Ultra-poverty and populations living below US1.90/day 14 Countries

Had we chosen a level of 50 percent of deprivation (the severely poor), there would have been nearly as many severe poor as people in extreme poverty, and there would be less than 50 percent overlap in many countries, which means we would not be talking about a subset of the extremely poor more likely to be left behind.

At the 70 percent level of deprivation, the numbers became roughly half of what they were under our definition of ultra-poverty. These would have included a higher proportion of people unable to work because of age or disability and so might have been dismissed as a marginal – rather than central - problem in the poverty elimination effort. Such an approach also would have led to numbers too small to monitor effectively in many countries.

To carry out the calculations regarding ultra-poverty, for each country, we went into the OPHI country briefing web page (8) and used the percentage corresponding to 60 percent and higher deprivation among the multidimensional poor in Figure H and we multiplied it by the country’s rate of multidimensional poverty found in Table 1.1 of Global MPI 2017. (9)

We used the 2014 population data contained in Table 1.1. to derive our ultra-poverty population data.

Building a Synthetic Index to Assess a Country’s Prospects of Eliminating Ultra-Poverty

Having determined the levels of ultra-poverty across various countries, we found it useful to ask whether there is, in each country, an environment conducive to eliminating ultra-poverty. We asked, are there specific factors likely to influence the pace of any process aimed at alleviating poverty such that one country has a better likelihood of eliminating ultra-poverty by 2030 than another? How can these factors best be summarized?
We looked at the challenges that countries confront in helping people move and stay out of poverty and proceeded to develop a synthetic index (a composite index combining various sub-indices) to gain insight into the scope of the challenge each country faces in eliminating ultra-poverty. 

The Process

In assessing the prospects of countries eliminating ultra-poverty, we went through the following steps:

We developed a list of questions that we felt would help us to ascertain whether a country’s environment was more or less conducive to eliminating ultra-poverty:

  • Are the very poor already being reached by safety nets? 
  • Is the country generally making progress on the socioeconomic front? 
  • Does poor governance plague the country? 
  • To what extent are security of the country and personal safety of the population at risk? 
  • Do gaps in social inclusion, including economic and financial exclusion as well as community participation, cohesion, solidarity, and integration, present a challenge? 
  • Is the ultra-poverty so deep that, even with some improvement, many of those living in ultra-poverty would remain in ultra-poverty? 
  • Has the country been addressing the particular issues faced by women living in poverty?
  • Does climate change jeopardize any future gains? 
  • Is the country generally financially resilient?   

For each of these questions, we identified the most appropriate sub-index from existing data or indices already in the public domain.

Where one data source or index appeared insufficient to capture the breadth of the issue, we identified a second data source or index and sought to combine the two.

Where the data was not on a 0-100 scale, we standardized the sub-index and scaled it so all sub-indices could reflect a normalized measurement.

We then convened a ‘virtual’ panel of experts familiar with initiatives aimed at assisting people in ultra-poverty and asked them to vote on the relevance of each sub-index (proposing a weighting of zero if they thought the sub-index was irrelevant or inadequate, a weighting of one when they thought the sub-index was relevant, and a weighting of two if the sub-index was not only relevant but deserved additional weight). These experts were also asked to identify additional or more appropriate indicators, if known. The weightings were averaged for each sub-index and we used the average weightings to aggregate all sub-indices and derive a synthetic index that provided an insight into the challenges faced by countries on their way out of ultra-poverty. 

Identifying and Weighting the Sub-indices

“Are the very poor already being reached by safety nets?” To identify a sub-index to answer this question we consulted World Bank data. (10) Under “Social Assistance”, we took the average of the columns “1.25 Per individual Per Transfer” and “Poorest Quintile Pre-Transfer”, which display, for each country, the coverage of social assistance programs among people living on US$1.25/day (measured prior to the receipt of social assistance) and among members of the poorest quintile, as calculated prior to the receipt of social assistance. For countries where only one of the two data points was available, we used that data point. For countries where there were no data points, the indicator was deemed to be zero given that the coverage was probably close to zero or the targeting toward the poor was so weak as to warrant a score of zero. (Assigned Weighting: 1.4)

“Is the country generally making progress on the socio-economic front?”

The best sub-index in our estimation was progress in the annual Human Development Index. We took the average annual growth for 2010-2015 (11). We assigned a score of zero to the country with the lowest growth rate and a score of 100 to the country with the highest growth rate. The other countries were assigned a score in proportion to their proximity to the top and bottom countries. (Assigned Weighting: 1.27)

“Does poor governance plague the country?”

We used the Governance sub-index score from Legatum (12) which measures a country’s performance in three areas: effective governance, rule of law, democracy, and public participation. We did not assign any score for countries where there was no data. (Assigned Weighting: 1.33)

“To what extent are security of the country and personal safety of the population at risk?”

We used the corresponding Safety and Security sub-index score from Legatum (13) which ranks countries based on personal safety and national security. We did not assign any score for countries where there was no data. (Assigned Weighting: 1.2)

“Do gaps in social inclusion, including economic and financial exclusion as well as community participation, cohesion, solidarity, and integration, present a challenge?”

We used the corresponding Social Capital sub-index score from Legatum (14), which measures the strength of personal relationships, social network support, social norms, and civic participation. We utilized social capital as a proxy for measuring social inclusion, an important factor in curbing ultra-poverty that goes beyond economic and financial exclusion to capture elements of community participation, cohesion, solidarity, and integration. We did not assign any score for countries where there was no data. (Assigned Weighting: 1.4)

“Is the ultra-poverty so deep that, even with some improvement, many of those living in ultra-poverty would remain in ultra-poverty?”

For each country, we referred to the OPHI country briefing (15), and used Figure H to obtain the percentage of multidimensional poor who are poor at the 70 percent threshold and we divided it by the percentage of the multidimensional poor who are poor at the 60 percent mark. We then multiplied the result by 100. For instance, for a country with a resulting computation of 40 percent, it was given a score of 40. We made adjustments to reflect that, for other indicators, a higher score meant better prospects for those in ultra-poverty, whereas a higher intensity of ultra-poverty represented the opposite. Accordingly, we designed a “shallowness of ultra-poverty” indicator by subtracting the intensity of ultra-poverty from 100. (Assigned Weighting: 1.53)

“Has the country been addressing the particular issues faced by women living in poverty?”

We looked at two sources to determine whether women were receiving key services and were confronting several other barriers. For the first indicator, we looked at the percentage of deliveries taking place in health institutions for mothers living in the poorest quintile. (16) The second indicator, the Social Institutions and Gender Index (SIGI), is produced by the Organization for Economic Co-operation and Development based on the obstacles experienced by women in the areas of the family code (e.g. early marriage), restrictions to physical integrity (e.g., reproductive autonomy), bias for sons (e.g., fertility preference), access to resources/assets (lands and financial resources) and exercise of civil liberties (access to public space and political voice). Given the index measures limitations, it had to be inverted so that high scores could reflect better prospects for the ultra-poor.

Due to its very specific nature, we assigned only a weighting of one-third to the first component, whereas the SIGI-derived component received a weighting of two-thirds. Finally, the weighted average was multiplied by 100 to obtain a scale of 0-100, as with the other indicators. (Assigned Weighting: 1.13)

“Does climate change jeopardize any future gains?”

We turned to two sources.  Firstly, the climate risk index of Germanwatch (17), which measures for weather-related events of the past 15 years:

  • Number of deaths resulting from the events
  • Number of deaths per 100 000 inhabitants
  • Sum of losses in US dollars in PPP
  • Losses per unit of gross domestic product (GDP)

We also looked at the Notre-Dame Gain Vulnerability Index, which focuses on climate change from the angle of vulnerability; the degree of exposure for each country, as well as how vulnerable their populations would be to climate events. The index is also produced in the form of a GDP-adjusted vulnerability, which measures the distance (positive or negative) of a country’s score to the expected score of a country with the same level of GDP per capita. We utilized this format.

We built a 0-100 index of vulnerability in the same manner as in the preceding cases, and inverted the score so a lower vulnerability can yield a higher index (the same treatment as applied to the intensity of ultra-poverty indicator). The Germanwatch index is also rescaled to a 0-100 range and the scores derived from Germanwatch and Notre-Dame Gain Vulnerability Index are averaged to provide the final climate risk indicator. (Assigned Weighting: 1)

“Is the country generally financially resilient?”

Here we started with 2015 GDP per capita data. (18) Of the 56 countries examined, the country with the highest GDP per capita was assigned a score of 100 and the lowest given zero. The other countries were assigned a score in proportion to their relative distance to these two countries; for instance, a country one-third of the way up from the lowest country was assigned a score of 33.3. (Assigned Weighting: 1.2)

Methodological Limitations

Ideally, in developing this synthetic index, we would have looked for experimental or quasi-experimental evidence of the extent to which these factors make it more difficult to reduce ultra-poverty. No such studies exist and conducting one would require considerable resources and time; time that those people living in ultra-poverty simply cannot afford. We chose to proceed in a less rigorous but more expeditious way. 

The large majority of the experts that we consulted had experience working with families living in ultra-poverty and were familiar with the challenges that keep families in deep poverty or reduce the velocity of their escape from poverty. These experts had significant experience in varied Asian, African, and Caribbean contexts but could not claim to have direct knowledge of all the various factors in all the country contexts. Therefore, at times, the expert opinion may be more akin to a perceptions survey, much like a business climate or outlook survey.

There are two points to note in this respect. First, it is worthwhile to consider that the expert weightings gave us, for just about every country, a synthetic index that is very similar to what a synthetic index with equal weighting would have shown. Country rankings are also very similar. This means that country scores and rankings based on the synthetic index appear to be largely unaffected by variations in expert opinions. It is unlikely that a larger and more diverse and perhaps more knowledgeable group of experts would have significantly changed scores and rankings.

Second, it is also worth noting that the synthetic index tracks (and is correlated to) the prevalence of ultra-poverty in each country. This is a form of common sense validation of plausibility. Indeed, many of our sub-indices did not change significantly over the short-term (the condition of women and the threats of climate change, for instance, evolve slowly).  This means that the synthetic index of a few years ago was not far off from the current synthetic index, and if this recent synthetic index was sound, we should find a lower percentage of ultra-poverty (and conversely) and this is exactly what we found.  There was good correlation between the synthetic index and the prevalence of ultra-poverty.

It is important to emphasize that there is no sub-index that measures the political will and implementation capability of a country to eliminate this type of poverty, which are key factors for many experts.


  1. The Final Countdown: Prospects for Ending Extreme Poverty by 2030(Interactive)”, Brookings Institution, April 24, 2013.

  2. Investments to End Poverty: Real Money, Real Choices, Real Lives”, (London: Development Initiatives Ltd., 2013).

  3. Taking on Inequality”, Poverty and Shared Prosperity, 2016, (Washington DC: World Bank, 2016).

  4. Poverty & Equity Data”, World Bank, 2017.

  5. PovcalNet”, World Bank,

  6. Oxford Poverty and Human Development Initiative (2017). “Lesotho Country Briefing”, Multidimensional Poverty Index Data Bank. OPHI, University of Oxford.

  7. Oxford Poverty and Human Development Initiative (2017).

  8. MPI Country Briefings 2017”, Oxford Poverty & Human Development Initiative, University of Oxford, last updated December 2016.

  9. Global MPI 2017.

  10. Table 2: Coverage of Social Assistance & Social Insurance Programs by Quintile of per Capita Welfare”, World Bank.

  11. Table 2: Trends in the Human Development Index, 1990-2015”, Human Development Reports, United Nations Development Programme.

  12. The Legatum Prosperity Index 2016”, Legatum Institute Foundation, 2017.

  13. The Legatum Prosperity Index 2016”, Legatum Institute Foundation, 2017.

  14. The Legatum Prosperity Index 2016”, Legatum Institute Foundation, 2017.

  15. MPI Country Briefings 2017”, Oxford Poverty & Human Development Initiative, University of Oxford, last updated December 2016.

  16. An index with this data is produced by UNICEF.

  17. Sönke Kreft, David Eckstein, and Inga Melchior, “Global Climate Risk Index 2017: Who Suffers Most from Extreme Weather Events? Weather-related Loss Events in 2015 and 1996 to 2015”, (Bonn, Germany: Germanwatch e.V., November, 2016).
  18. GDP per Capita, PPP (Current International $)”, World Bank, 2017.