Should we care equally about poor people wherever they may live?


This page in:

Not so long ago, those countries designated as “low-income countries” (LICs) in the World Bank’s World Development Indicators accounted for the bulk of the world’s poor, such as by the $1.25 a day standard. Today many very poor people live instead in what are called “middle-income countries” (MICs).  The change seems dramatic. Almost all (94%) of those below $1.25 a day in 1990 lived in LICs. By 2008 the proportion was down to 26%, with the rest in MICs. Andy Sumner attracted much attention to this aspect of how the global profile of poverty has changed in his paper “Where do the Poor Live?.” Amanda Glassman, Denizhan Duran and Sumner dub this emergence of large poverty counts in MICs as the “new bottom billion.”

There has been much discussion about the implications of this change for overseas development assistance (ODA) and development policy more broadly. In particular, there have been calls for concentrating ODA on the LICs, assuming that the MICs can now look after their own poor.

But we need to look more closely at this “LIC-MIC” distinction, to understand why we have seen this change in the global poverty profile, and what relevance it might have for development policy.

An influential but aging classification of uncertain origin
The World Bank assigns “LIC” and “MIC” status based on countries’ gross national income (GNI) per capita, using the so-called “Atlas method.” (This is based on a moving average of official exchange rates adjusted for inflation relative to the G5 countries. It is unclear why the Bank does not use its own PPP exchange rates for this purpose.) As best I can determine, the graduation threshold for going from LIC to MIC, as given here, was officially set in 1988 at the value of the Bank’s “Civil Works Preference” (CWP), which had been set at a GNI of $200 per capita in 1971. The threshold has only been updated over time for global inflation.

So this widely-used income threshold for becoming a “MIC” has been essentially fixed in real terms for over 40 years! Knowing this, it can hardly be surprising that the share of the poor (and, indeed, of the whole population) in LICs has declined over time. Naturally, the fact that the two most populous countries, China and India, recently crossed the threshold magnifies the effect dramatically.

And yet the relevance of this old MIC threshold to the development aid debate is far from clear.  CWP status means that eligible domestic contractors from that country can be given preference in evaluating civil works bids for World Bank projects under international competitive bidding. It is unclear why the MIC threshold was tied to the CWP. It just seems to have been a convenient number on hand at the time. It is unclear why “arcane concerns with civil works preference” (as Charles Kenny has put it) are salient to development assistance choices.

Putting aside these issues, the question remains as to why we should treat equally poor people differently, depending on differences in the assigned status of the countries they live in.  The “new bottom billion” are not really so new. People living below $1.25 a day in “middle-income” India and China are not some new set of families that have recently emerged. Roughly speaking, they are just as poor as before their countries graduated to MIC status. Why should we treat them any differently?

In response, one might argue that past economic growth points to lower expected future poverty, and that external development assistance would thus be better placed where there has been little or no such growth. (We can probably put aside theoretically possible but practically doubtful concerns about perverse incentive effects.) But then we still don’t have a rationale for using this “LIC-MIC” distinction. Would not this logic apply with equal force within these categories? We can measure growth at country level.

A deeper argument might be found in the fact that the nation state has a clear role in efforts to fight poverty, and states differ in their capacities for doing this well. The task of fighting poverty in MICs might then be seen as largely a matter of domestic policy—that MICs should typically be left to look after their own poor, with ODA focusing more on LICs. 

But why is the fact of crossing this arbitrary old threshold of relevance to the capacity of countries for looking after their own poor? More careful treatments of the topic acknowledge MIC heterogeneity, though the nature of the differences relevant to aid allocations is rarely articulated. In a briefing note for the Center for Global Development, Andy Sumner points to India as an example of a MIC that has (he claims) “little need for ODA” despite having many very poor people. But it is not clear why he puts India in this category. (In fairness to Sumner, his recent paper, “Poor Countries or Poor People?” with Ravi Kanbur, provides a nuanced discussion of the reasons for caution in arguing that MICs can be safely left to tend to their own poor. But this does not tell us whether India, or some other MIC, does or does not need external assistance for fighting poverty.)

Some evidence on domestic capacity for redistribution
To inform discussions of aid policy, a better approach is to look instead at more direct measures of domestic capacity for fighting poverty. There are many aspects of that capacity one might focus on. (The World Bank’s allocations of its concessional lending use numerous indicators of domestic capacity and the quality of policies and institutions.) However, since much of the recent debate has concerned domestic capacity for directly intervening to help poor people, it is of interest to look more closely at this aspect. Is average income an adequate proxy for domestic capacity for fighting poverty through income redistribution?

In thinking about how one might address that question, we can reasonably assume that aid donors will not be indifferent to how incomes are distributed above the $1.25 poverty line when assessing a developing country’s capacity for redistribution to its own poor. It is clearly not acceptable to say that a country has a high capacity if (given its income distribution) redistribution would require putting almost all the tax burden on people living just above $1.25 a day. Some countries have greater affluence—with more people living well above the poverty line—than others, and this must be brought into the picture.

To help identify a defensible measure of domestic capacity for redistribution, it can be assumed that citizens of a rich country would find it intrinsically (ethically or politically) unacceptable to expect a developing country to address its poverty problem by taxing people who would be considered poor in the rich country. This leads us to focus on the capacity for redistribution from the “rich” to the “poor” within developing countries, with middle-incomes untouched.

Several arguments support this characterization of the capacity for redistribution. Popular judgments about “inequality” appear to give greater weight to redistributions from the rich to the poor than redistributions amongst the middle class or from the middle class to the poor. There are also instrumental arguments that point to the expected role of the middle class as agents of progress and also the likely incentive effects on those near the poverty line. One can also question the political feasibility within a given developing country of asking middle-income groups to shoulder the burden of poverty relief. Redistribution from the “rich” to the “poor,” without involving middle-incomes, can also emerge as a public choice equilibrium. This would require the usual conditions for the median voter theorem—namely that the issue to be voted on is one dimensional and the utility function is single-peaked in that dimension—plus the assumption that the decisive voters care sufficiently about poverty and are concentrated around the median, which is (invariably) below the mean.

Following this idea, the “capacity for redistribution” of any developing country can be measured by the marginal tax rate (MTR) on those who are not poor by rich-country standards that is needed to cover the domestic poverty gap—the aggregate deficit of poor people relative to the poverty line. I have proposed such a measure, in “Do Poorer Countries Have Less Capacity for Redistribution?.” The “non-poor” by Western standards are identified as those living above the US poverty line of $13 a day in 2005. And the poverty gap is judged by the $1.25 a day standard.  So an MTR of (say) 25% means that a tax of $1 for each $4 of extra income above $13 a day would generate sufficient revenue to cover the poverty gap relative to $1.25 a day. Figure 1 shows how my calculated value of this marginal tax rate varies across countries, according to their mean consumption. (Whether one uses mean consumption or mean income is unlikely to make much difference for this purpose.)

This suggests that there is a positive correlation between domestic capacity for redistribution (as indicated by a low required MTR) and average income. I find that for most (but not all) countries with annual consumption per capita under $2,000 the required tax burdens are prohibitive—often calling for marginal tax rates of 100 percent or more. By contrast, the required tax rates are very low (1% on average) among all countries with consumption per capita over $4,000, as well as some poorer countries.

Figure 1: Marginal tax rates on the non-poor by US standards needed to cover the poverty gap by poor country standards across developing countries
Note: MTR’s are truncated at 100%. Source:
However, notice how much the marginal tax rates vary across countries at a given mean consumption (or income). For countries in the region $1,000-2,000 per year, the tax rates vary from close to zero to 100%. Average consumption or income is clearly not a good proxy for a country’s capacity for fighting poverty through internal redistribution.

It is instructive to see where India is found. My calculations indicate that the tax on the non-poor (by Western standards) needed to cover India’s poverty gap would be truly prohibitive. Not even a tax rate of 100% would be sufficient!  Indeed, appropriating all of the incomes of those living in India above the US poverty line would cover only a modest fraction of the country’s aggregate poverty gap.

This simple arithmetic leads me to question whether India can yet be considered to have ample capacity for eliminating poverty by domestic redistribution. By contrast, China could generate revenue sufficient to cover the poverty gap with a more modest (though hardly small) marginal tax rate of 37% on incomes of the non-poor by US standards. (Although in some provinces of China, the required tax rates could well be much higher.)

In conclusion
The global community would not need to worry so much about the many very poor people living in “middle-income countries” if it could be reasonably confident that any country attaining such a designation is capable of tackling its own poverty without external help. But that is plainly not the case. Merely knowing whether a country’s average income puts it above or below the arbitrary historical threshold used to delineate “low-income” from “middle-income” countries tells us rather little about key aspects of the country’s capacity for fighting poverty, including through domestic redistribution.  

Future success against poverty will depend heavily on the policies adopted by governments in developing countries, at all income levels. While there is scope for better redistributive policies even if they can’t reasonably be expected to solve the poverty problem, supporting economic growth will continue to play an important role in fighting poverty. Domestic policy changes will often be needed to assure higher growth. And in many countries, those changes will have to include effective efforts to redress the pervasive inequalities that constrain the economic opportunities of poor people. 

While such domestic policies will continue to play a crucial role in fighting poverty, external development assistance can still help, including in the so-called “middle-income countries.” This includes assistance for developing the analytic and administrative capacity for designing and implementing better domestic policies.

And, by the way, is it not time for these arcane income thresholds for “graduating” from “low-income” status to be laid to rest?



Martin Ravallion

Martin Ravallion, Edmond D. Villani Professor of Economics, Georgetown University

Andy Sumner
November 08, 2012

Martin, I think I agree with all of the above. How often do you here that?

Plenty more in IDS papers I've done that I think covers much of what you note too.

Having spent a lot of time on this my conclusion is indeed the same as yours - time to dump the LIC/MIC threshold.

You could come up with an alternative way of thinking on relative capacity to end poverty?

And here's just some food for thought on thinking about alternatives:…

Andy Sumner
King's College London

Rob Yates
November 08, 2012

As a number of Latin American countries have shown, providing universal access to free, publicly funded social services can be an effective redistibutive policy especially when backed up with targetted cash transfers. Judging by this it looks as if the Indian Government is also thinking along similar lines:…

November 09, 2012

I think the Bank's own mission statement -A World without poverty- answers the question in the title. Whether the Bank will successfully adapt to this new situation where there "bottom million" are scattered in different countries is another story.

Also, that "graduation" threshold is silly and should be put to rest, together with the term "graduation" itself. The Bank should have a more continuous treatment of counties that need not be linear... which calls into question the separation between IDA and IBRD. But maybe that's too much to ask?

Demberel sambuu
November 10, 2012

Suppose a country wants to move toward more inclusive and sustainable growth path by declaring many things on this and that.suppose this county's committed to this but as this is new path there will be a strong need of political and social awareness to make politicians,businesses and people to support it.but it is not possible with the same measurement at the macro level as gdp growth etc how this country will measure it's performance on a.m.areas?

Lucio Vinhas de Souza
November 14, 2012

Very convincing and solid argumentation, Martin, which provides a useful contribution to development aid debate.


Eric V. Swanson
November 15, 2012

Martin adopts the US poverty line as a standard against which to measure countries' capacity to redistribute from rich to poor on pragmatic grounds: it might be easier to explain to a US citizen why a country that cannot reach even the US poverty line needs some further assistance. But the result of his own analysis is to leave a large group of undifferentiated countries "censored" at the 100 percent line. And it may be no easier to engage a US or European legislator in a discussion over setting a floor of $13 a day for the incomes of developing country citizens. As Branko Milanovic notes in his recent working paper (WPS 6269) “Those who are considered nationally poor in the United States or the European Union have incomes which are many times greater than the incomes of the poor people in poor countries and moreover often greater than the incomes of the middle class in poor countries. “ Martin’s analysis seems to overlook his oft cited finding that above an absolute minimum (currently estimated at $1.25 per person per day in 2005 prices)national poverty lines rise with income. Taking this line as an international standard, it would be interesting to recalculate the capacity of countries at each income level to redistribute income to the poor. In other words, assess their effort by their own standards, not those of rich countries. Given the nature of regression lines, we would expected to find as many countries above as below it and only a few topped out at 100 percent. But the size of the gaps might reveal interesting information about the poltical economy of each country. Does this help us to reset the income clusters currently used by the World Bank? I'm not sure that it does. That requires another discussion, which is already underway.

Martin Ravallion
November 15, 2012


Thanks for the comment. There seems to be some confusion about the purpose of my calculations. No doubt my exposition is at fault. Let me clarify three key points.

First, I am doing these calculations to illustrate a more general point: that there is great heterogeneity amongst countries near the low-middle boundary in terms of their ability to deal with poverty through domestic redistribution. I argue that the way I have measured this makes sense—that its assumptions are defensible. I do not claim it is the only way. But is suffices to demonstrate my key point: any use of these income graduation thresholds in discussing development assistance is questionable, given that heterogeneity.

Second, I am not saying that one should use my measure of capacity for redistribution to “reset the income clusters used by the World Bank.” The role of my measure is to point out that those clusters carry little weight for the policy discussions that have relied on them. They should be handled with greater care. I agree that a major re-think of these income clusters is needed. I understand that these thresholds have been reviewed a few times over the last 40 years but each time they have survived un-changed. Maybe a deeper re-think is called for this time.

Third, my measure already does what you ask for, namely to assess “the capacity of countries at each income level to redistribute income to the poor” (your words) as judged by poverty lines used in developing countries. The domestic poverty line is taken to be $2 a day, which is the average poverty line of developing countries. So I am using “their own standards” as you want. The US poverty line of $13 is NOT the assumed floor for incomes in developing countries as you say. Rather it is the point above which hypothetical taxes are levied on the citizens of developing countries to finance transfers to those living below $2 a day. Here the logic is that donors would not (presumably) want people who would be judged poor in the donor country to finance redistribution to even poorer. Yes, one could use the (typically higher) poverty lines found in Western Europe. But using the US line seems adequate for the purpose of the exercise.

I hope this is now clearer.


Milind Sohoni
November 15, 2012

Well, India has a small Income Tax/GDP ratio of about 4-5 %, which is very low. Overall tax/GDP ratio is also very low. So, this is not a fair comparison. About 100 individuals in India corner about 25% of GDP so actually a broader tax coverage on just the top hundred will be enough to pay for our poor.
The article seems to deflect attention from the really high gini coefficients for India.

November 15, 2012

Helping the developing countries - or the low and middle income countries - is not a matter of redistributing wealth. There is not a limited quantity of wealth on the planet that we have to move around. The genius here is to help them, educate them, empower them to create their own wealth. The developed nations were not just lucky - they had a culture that drove wealth creation. Some of that is fading - but the developing nations need to invest in wealth creation - not redistribution.

And their wealth creation will enrich all of us in the end.

Seth Kaplan
November 18, 2012

As I write here:…

a better system ought to depend on many more factors than just income, more akin to what the United Nations uses now than the World Bank system. These should include a variety of social, economic, and political factors, such as the average level of human development, the quality of public services, macroeconomic conditions, the degree of export diversification, and the extent of social conflict. Income would still matter but much less than it does now.

I agree with Andy that dividing all countries into separate clusters makes more sense, though my formulation would differ substantially from his.

Martin Ravallion
November 19, 2012


Thanks for the response. I am not surprised you agree, as your work has influenced my thinking on this issue more than anyone. I hope we can both influence others to abandon this dubious LIC-MIC distinction in the aid and poverty debate, and think more deeply about the issues.


Matt Berkley
November 19, 2012

I am not an economist: formally, I studied philosophy, ancient history and literature, then experimental psychology.

I have attempted to understand the reasoning in Dr Ravallion's and Dr Chen's methodology papers underlying statements on global poverty since 1981, and the reasoning behind Dr Ravallion's proposals for the future.

The purpose of this comment is to ask how the likely extent of the potential problems below is known to or inferred by macroeconomists to be small, or alternatively how the uncertainties about them are taken into account.

To the points in the email below I add two more:

1. Since the FAO adjust for rising food needs per person due to rising adult-child ratios, it is not clear why the World Bank does not do likewise. Other things being equal there would be an element of spurious progress.

2. It is not clear why the Bank has not factored in numbers of AIDS deaths, especially in sub-Saharan Africa, into an assessment of the progress of poor people since 1981.

I am unaware of anyone other than myself who raised this seriously as a general problem for international goal setting or economic theory - that macroeconomic figures look better if some people die - in the roughly two decades between the discovery of AIDS and the Millennium Declaration of 2000.

Following my raising the problem with Professor Morduch, then of Princeton, his colleague Angus Deaton raised it as a serious issue in a paper of December 2000 commissioned by the Chief Economist of the World Bank on how to monitor progress for MDG1.

I subsequently explained the problem to, among others, Professors Sachs and Kanbur, three senior employees at the World Bank, officials at the OECD and the UK Department of International Development, my member of Parliament and through him the UK's Governor of the World Bank Clare Short.

Professor Sachs and Professor Kanbur, and the incoming Chief Economist of the World Bank Francois Bourguignon, later co-wrote papers based on this idea - that economists aggregating outcomes count those who die as well as those who stay alive. As I understand it, the papers have been accepted by other senior economists as important.

This comment is to ask Dr Ravallion to clarify the two issues above and the four in the email below of 2007 for non-economists.

Thank you.


Email of 18 May 2007:

Dear Dr Ravallion

I am a member of the public with an interest in international development. I would be grateful for any comments you might have on the importance or otherwise of the following.

I worked in Bangladesh in the mid-1980s, and lived with an illiterate family. A theme at the time was landlessness due to population rise: parents' land would be divided among several offspring.

Apart from the general problem that income or spending do not directly measure assets, it seems to me that if people move to cities for work, they may begin paying rent, having previously lived on family land.

It was not clear to me from Absolute Poverty Measures whether this kind of thing - changes in consumption needs - was taken into account, or whether the new urban/rural adjustments were solely for prices.

Secondly, in respect of economies of scale, a plausible scenario might be that economic growth coincides with smaller household units, as single people travel to cities for work. This would be additional to considerations about economies of scale related to children's needs and smaller family sizes as birth rates fall.

Thirdly and fourthly, it is not clear to me how surveys have generally dealt with expenditure on personal debt, or costs which are sometimes the subject of public concern and whose incidence varies across countries and times, such as charges for water, medicine or schooling.

I would be grateful for any light you might be able to shed on these issues.

Thank you.

Yours sincerely,

Matt Berkley.

Tim Bodin
November 28, 2012

Perhaps i am missing something critical, but isn't it true that a threshold of poverty in the US of $13/day when translated nominally to a low income country leads to a living standard much above US poverty levels for PPP, relatively lower local prices for non-tradables? If so we still have definitional work to do. Very interesting nonetheless.