Global Poverty and the New Bottom Billion Revisited: Why are Some People Poor?

Full text

!
1
GLOBAL POVERTY AND THE ‘NEW BOTTOM BILLION’ REVISITED:
WHY ARE SOME PEOPLE POOR?
Andy Sumner, King’s College, London
18 February 2016
Abstract: This paper revisits the debate on the changing ‘geography’ or location of
global poverty. Specifically, most global poverty is concentrated in a set of populous
countries that have transitioned from low income to middle income countries. The
paper revisits the debate and argues that the shift in global poverty implies a
questioning of the dominant theory of absolute poverty in all but the world’s very
poorest countries: that is that poverty in developing countries is explicable at societal
level by insufficient public and private resources to address absolute poverty. Instead,
a structural theory it is argued - meaning here theory that takes account of questions
of distribution - is increasingly relevant to most, but not all, of global poverty. To this
end an indicative empirical example of resources nationally available to end extreme
poverty is explored in the form of the reallocation of public spending from regressive
fossil fuel subsidies to poverty transfers.

!
2
1. INTRODUCTION
A series of papers and books since 2010 have discussed these questions which relate
to the shifting location or ‘geography’ of global poverty.1 The shift in global poverty
is that poverty has ‘moved’ from low income to middle income countries and as a
result about a billion people or up to three-quarters of the world’s poor now live in
middle income countries. This raises various questions about the distributional
patterns of economic development and the effectiveness of growth in reducing
poverty as well as questions about the distribution of aid and the dominant country
analytical categories of low and middle income themselves. More fundamentally there
is a question, which is the focus of this paper, about whether the causes of global
poverty are in the process of changing and what that implies for theories of absolute
poverty in the developing world.2
Theorizing on the causes of absolute poverty in the developing world has
tended to underemphasize questions of national distribution, which has led to poverty
not being seen as being a structural outcome of specific patterns of growth and
distribution. This is, at least in part, due to a prevailing assumption that distribution is
not a central factor in explaining poverty if everyone is poor. If it is no longer the case
in many developing countries that everyone is poor then theories of poverty that relate
to distribution potentially have a new-found relevance.
This paper revisits the geography of poverty discussion and reviews the
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
1 See for range of discussions see: Alkire et al., 2011, 2015a; 2015b; Alonso 2012; Alonso et al., 2014;
Clarke and Feeny, 2011; Carbone, 2013; Edward and Sumner, 2012; Glennie, 2011; Haddad, 2012,
2014; Herbert, 2012; Kanbur and Sumner, 2012; Keeley, 2012; Koch, 2015; Mallet and Sumner, 2013;
Lundsgaarde, 2012; Madrueño-Aguilar, 2015; Ottersen et al. 2014; Poke and Whitman, 2011; Sumner,
2010, 2012; Vasquez and Sumner, 2013; 2015).
2 Thanks to Isa Baud and Laura Camfield for comments on an earlier version of this paper.

!
3
overall trends in, and changes in the distribution of, global poverty updating for the
new global poverty line and latest data. The paper then discusses theories of poverty
and argues that the changes in global poverty observed raise questions for theorizing
about the causes of absolute poverty in the developing world. Section 4 then explores
an empirical example of nationally available resources that could be reallocated from
a public ‘bad’, regressive fossil fuel subsidies, to poverty transfers. Section 5
concludes.
2. GLOBAL POVERTY SINCE THE COLD WAR
As noted at the outset, since 2010, a series of papers has discussed a shift in the
location or ‘geography’ of global poverty. The shift is quite simple: that the
distribution of global poverty has shifted from countries officially classified by the
World Bank as low-income countries towards countries classified as middle-income
countries (MICs). This has led to the following stylized fact: three-quarters of the
world’s poor live in middle-income countries. This amounts to about a ‘new bottom
billion’ of poor people who live in middle-income countries. This is in contrast to
Collier’s (2007) original ‘bottom billion’ which was the total population of the 58
poorest countries.3
Of course the world’s poor have not ‘moved’. What has happened is quiet
simple: global poverty is concentrated in a small number of populous countries (see
table 1 later) and some of those countries where many of the poor live experienced
striking increases in average incomes and poverty did not fall as much as one might
expect in absolute numbers. Certainly one could have expected a higher poverty
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
3 This is ‘new bottom billion’ is based taking a consumption poverty line of $2 (in 2005PPP) or $2.50
(in 2011PPP) or by taking multi-dimensional poverty (see Alkire et al., 2011; 2013; 2015).

!
4
efficiency of growth given the frankly staggering increases in income per capita.
Although there is no sudden change in countries (or people) when a line is crossed in
per capita income, higher levels of average per capita income do imply substantially
more domestic public and private (taxable) resources available for poverty reduction
and greater access to private capital markets to further expand resources. This shift
has implications for debates on the future of aid, its allocation and purpose (see for
discussion Glennie, 2011; Herbert, 2012; Kanbur and Sumner, 2012; Keeley, 2012;
Koch, 2015; Lundsgaarde, 2012; Ottersen et al. 2014; Poke and Whitman, 2011) and
the continued relevance of the country classification categories that are used in debate
over aid allocations and more generally (see for discussion Alonso 2012; Alonso et
al., 2014; Madrueño-Aguilar, 2015; Vasquez and Sumner, 2013; 2015). More
fundamentally, this paper argues there are questions about whether the shift implies a
need to revisit theories of the causes of global poverty (see for related discussions
Haddad, 2012, 2014; Sumner, 2010; 2012).
First, what has actually changed? Figure 1 shows the number of Low Income,
Lower Middle, Upper Middle and High Income Countries (respectively, LICs,
LMICs, UMICs and HICs), from 1987 when the World Bank classification began to
2013 (the most current classification as this is based on GNI Atlas per capita from two
years prior). Figure 1 shows the decline in the number of low-income countries (less
than $1045 GNI Atlas per capita in 2013) notably since about 2000. Prior to that the
number of low-income countries had been rising (partly due to the
transition/economic collapse following the end of the Cold War in Eastern Europe
and the breakup of the Soviet Union). The number of low income countries started to
drastically fall over the 2000s from approximately 60 low income countries at the turn
of the century to about 30 LICs today. The number of high-income countries (which

!
5
are countries with more than $12,000 GNI Atlas per capita in 2013) has doubled from
about 40 in 1990 to about 80 in 2013. The MICs are a heterogeneous group of
countries, which is not surprising given their average income ranges currently from
$1000 to $12000 per capita albeit, with a demarcation at $4000 to separate Lower
Middle Income and Upper Middle Income Countries. In contrast, the remaining LICs
are now relatively homogenous in terms of their structural economic characteristics
and a shared poor recent growth history and almost all are members of the UN
grouping of Least Developed Countries. The new MICs include many fast growing
‘emerging economies’ where growth with structural change is evident such as China,
India, Pakistan, Nigeria, Indonesia, Bangladesh, and Vietnam. These populous
countries are home to many of the world’s absolute poor. There are also a set of post-
Socialist countries or ‘bounce-back’ new MICs that experienced economic collapse
and have grown back since to MIC levels (eg. Albania and Ukraine). Finally, there are
various islands and countries with less than 10 million population (e.g. Mongolia,
Nicaragua and Bhutan).
Figure 1. Number of LICs, MICs and HICs, 1987-2015
Source: Data processed from World Bank (2015a).
0!
20!
40!
60!
80!
100!
120!
1985! 1990! 1995! 2000! 2005! 2010! 2015!
LICs!
LMICs!
UMICs!
MICs!
HICs!

!
6
Figures 2 and 3 show the proportions and absolute numbers of the population of the
developing world living in poverty taking the following consumption lines (in 2011
PPP): $1.90-a-day (the new World Bank poverty line, which is the adjustment of the
former $1.25 line in 2005 PPP and also the median of national poverty lines in LICs,
see for discussion Ferreira et al., 2015; Lolliffe and Prydz, 2015 and for a critique see
Lahoti and Reddy, 2015); $3.10-a-day (the new World Bank upper poverty line
derived on the same basis as the earlier median poverty line of all developing
countries, which was $2-a-day poverty line in 2005 PPP); $5-a-day (the current
median and mean of all developing country national poverty lines) and, $10-a-day (a
daily consumption associated with permanent escape from poverty in longitudinal
studies – see Lopez-Calva and Ortiz-Juarez, 2014). Of course this is all monetray
poverty which has to be taken with the usual set of caveats. One could also take multi-
dimensional poverty as the number of developing countries that have two data points
increases (see Alkire et al., 2015b). Figure 4 shows the distribution of $1.90 poverty
by type of country – low, middle and high income. What do the charts show? First,
figures 2 and 3 show that by whatever poverty line taken poverty has fallen.4
However, this is much less impressive when China is removed.
Including China, those living under $1.90 has fallen from 54.8% of the global
population in 1981 to 14.8% in 2012. Excluding China those living under $1.90 has
still fallen, from 33.1% in 1981 to 17.6% in 2012. Including China the population
below $1.90 more than halved from almost 2 billion people to about 0.8 billion
people. The data excluding China are less dramatic but significant: The population
under $1.90 stood at just under 1 billion in 1981 but fell to 700m people by 2012. It is
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
4 Estimates in this paper at $1.90 differ slightly from the ‘official’; World Bank estimates of 896.7
million (of Ferriera et al., 2015) and 902 million (of Cruz et al., 2015d) because estimates here do not
‘fill’ missing data with regional averages (see Ferriera et al., 2015, p. 28) nor extrapolate to specific
years but take the closest available year. Estimates here are 795.8m.

!
7
worth noting that data before 1990 is based on few surveys and emphasis should be
placed on post-1990 estimates.
Figure 2 Percentage of population of developing countries who consume under $1.90,
$1.90-$3.10, $3.10-$5 and $5-$10 per day and above $10 per day (2011PPP), 1981-
2012
Source: Data processed from World Bank (2015b).
Figure 3 Percentage of population of developing countries excluding China who
consume under $1.90, $1.90-$3.10, $3.10-$5 and $5-$10 per day and above $10 per
day (2011PPP), 1981-2012
Source: Data processed from World Bank (2015b).
0%!
10%!
20%!
30%!
40%!
50%!
60%!
70%!
80%!
90%!
100%!
<$1.9! $1.9-$3.1! $3.1-$5! $5-$10! >$10!
0%!
10%!
20%!
30%!
40%!
50%!
60%!
70%!
80%!
90%!
100%!
<$1.9! $1.9-$3.1! $3.1-$5! $5-$10! >$10!

!
8
Figure 4. Distribution of population of developing countries below $1.90 by
LIC/LMIC/UMIC, 1985-2012
Source: Data processed from World Bank (2015b). Note: Classification begins 1987
which is based on GNI pc in 1985 so graphs start in 1985
Figure 4 then shows the global distribution of absolute poverty by $1.90. The share of
global poverty accounted for by China has fallen from approximately 40% of global
poverty in the early 1980s to about 10% in 2012. In contrast, the share of global
poverty accounted for by India has remained about the same. The proportion of global
poverty in LICs and UMICs has remained largely the same. It is in ‘other LMICs’
(meaning excluding India) where the proportion of global poverty has increased. This
includes a small set of populous countries notably Bangladesh, Nigeria, Pakistan,
Indonesia amongst others. In short, the world’s poor are concentrated in a relatively
small set of populous countries, most of which have grown substantially over the last
generation. In these countries poverty in absolute numbers has not fallen as much as
one might expect and certainly not as much as in China where most of the global
gains for poverty reduction have been made.
These lines for LICs/MICs are all constructs, of course, and the results are a
0%!
10%!
20%!
30%!
40%!
50%!
60%!
70%!
80%!
90%!
100%!
1985!
1986!
1987!
1988!
1989!
1990!
1991!
1992!
1993!
1994!
1995!
1996!
1997!
1998!
1999!
2000!
2001!
2002!
2003!
2004!
2005!
2006!
2007!
2008!
2009!
2010!
2011!
2012!
China! India! LICs! LMICs!(excl.!India)! UMICs!(excl.!China)!

!
9
function of where the lines are drawn. The income lines used for low and middle
income per capita are arbitrary in the sense where you cut the line determines the
number of people and countries above and below. However, in this discussion we are
not concerned per se about the precise numbers in each ‘band’ of consumption. The
thesis of the paper is the overall shift from the late 1980s from a world of poor people
in poor countries to a world of poor people in fast growing countries.
We can also consider the sensitivity of the country thresholds (see figure 5).
Whether revising or updating the LIC/MIC threshold would make a difference would
depend on how much it was revised by. Nigeria, India and Pakistan had experienced
their GNI per capita rise by 2014 to $3000, $1600, and $1400 GNI Atlas per capita
respectively (and $5400, $5400, and $4800 in GNI PPP per capita) which is
substantially above the $1045 GNI Atlas per capita threshold. The shift in global
poverty is not a product of the world’s poor living in countries who have only just
crossed the threshold in average income per capita into the MIC group. One would
need to at least double (or triple) the LIC/MIC threshold to make much of a
difference, as that would push India (and Nigeria) back under the threshold. One
would need to increase the threshold by four fold or six fold to bring Indonesia and
China respectively back into the LIC group of countries. There is one exception –
Bangladesh which has just crossed the LIC/MIC threshold this year.

!
10
Figure 5. Country density curve for low and middle income countries by GNI Atlas
per capita, 2013 (plots proportional to population size)
Source: Data processed from World Bank (2015b).
There are, however, all sorts of important questions about the setting of the line for
the threshold of LIC to MIC (and HIC) and indeed the classification by income is
under review at present given it was established in the late 1980s (see discussion in
Sumner, 2012). For example, one question is if GNI per capita ought to be in PPP$,
with the large caveat that PPPs for the poorest countries are subject to considerable
contention, as this would make comparisons over time and across countries stronger
(see Edward and Sumner, 2014). More importantly, multi-dimensional approaches to
clustering developing countries by more than income per capita suggest that, as is
well known, income per capita is not always a good correlate to other dimensions of
development and there is no unequivocal linear development pathway from low to
high income country (See Vázquez and Sumner, 2014). That said, the current LICs do
fall into one clustering of countries which is that group with the poorest
characteristics across various economic, social and political dimensions. In contrast,
200! 2,000!
GNI!Atlas!per!capita!(log.!scale)!
LICs!
LMICs! UMICs!

!
11
the current MICs are much more heterogeneous (see Vázquez and Sumner, 2015, p.
14-17). In short, if one wants a crude classification the income per capita country
classification used by the World Bank does not do too bad a job in separating the very
poorest countries from other countries that no longer ‘stuck’ at the bottom.
What may be of more significance is that the current LIC/MIC threshold has
both symbolic and real impacts. It is seen by many donors and politicians as a
symbolic line to cross. It does – in general – lead to greater access to private capital
markets. It is - broadly speaking - worth in real terms what it was worth in 1990.
LMICs are also considerably more developed by a range of indicators than the
remaining LICs. Most importantly the aspects that differentiate LMICs from the
remaining LICs are growth prospects and structural economic characteristics. The
remaining LICs are much more homogenous as a group. Not only do the remaining
LICs have weak growth history suggesting weak growth prospects, they also face the
structural economic handicaps that characterise the LDC classification such as
literacy rates and an export structure dominated by primary goods. In fact almost all
of the remaining LICs are now LDCs.5
Indeed extrapolating GNI Atlas per capita for 2010-14 might suggest most of
the remaining LICs will be LICs for some considerable time to come suggesting ODA
commitments will be essential. Some remaining LICs might graduate by 2050 whilst
others will take to 2100 or beyond (see Figure 6).
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
5 Some MICs are also LDC but this LDC MIC is largely small island economies. The UN LDC is
based on a methodology that combines human assets (including nutrition, child mortality, school
enrolment, and adult literacy), economic vulnerability (measures of the instability of agricultural
production, population displaced by natural disasters, instability in exports, and the share of agriculture
in GDP and exports), proxies for economic ‘smallness’, ‘remoteness’ and GNI (Atlas) per capita. The
main problem of the LDC category is that it is somewhat static. Guillaumont (2009), among others, has
argued that the graduation criteria make it very difficult for countries to ‘graduate’ as the conditions for
exit are difficult to meet.

!
12
Figure 6. LICs with data that could be expected to cross LIC/MIC line in the next 100
years based on 2010-2014 per capita growth
Source: Data processed from World Bank (2015c).
Another logic for the LIC/MIC threshold is that under which it was originally
developed. Historically it was established based on the relationship between GNI
(Atlas) per capita and other variables such as poverty. Taking a whole-of-society
wellbeing measure (rather than poverty which is not the whole of society but a
proportion of population given that poverty levels are generally lower now than in the
late 1980s as a proportion of population at whatever poverty line), life expectancy
would suggest that approximately $1000 per capita is the absolute minimum threshold
for a country to achieve the average societal wellbeing of an ‘advanced’, meaning
OECD nation. It is the absolute minimum threshold for a country to achieve an OECD
comparable average life expectancy level. The lowest life expectancy in an OECD
country is currently 75 years in Turkey (in 2013). One new MIC, Vietnam, did
achieve a life expectancy similar to that of Turkey, 75 years, at just $1000 GNI
(Atlas) per capita in 2008 as it crossed the threshold into MIC status (and GDP PPP
2010!
2020!
2030!
2040!
2050!
2060!
2070!
2080!
2090!
2100!
Cambodia!
Zimbabwe!
Tanzania!
Rwanda!
Nepal!
Mozambique!
Burkina!Faso!
Benin!
Congo,!Dem.!Rep.!
Liberia!
Niger!
Projected!year!of!LIC!graduation!

!
13
per capita was $4,000).6 In short, life expectancy close to the lowest in the OECD has
been reached at about $1000 GNI per capita per year but this is – of course - by no
means guaranteed. Turkey with the same life expectancy had a GNI (Atlas) per capita
of almost $11000, close to the HIC threshold (and $19000 in GDP PPP per capita).
When another new MIC, Nigeria, crossed the LIC–MIC threshold of $1000 in 2008 it
had life expectancy of just 50 years. In 2013, Nigeria and Vietnam, both had virtually
the same income per capita (in both GNI and GDP PPP) but life expectancy remains
50 years in the former and 75 years in the latter. Furthermore, this discussion is based
on the mean life expectancy and life expectancy is distributed unequally within
societies. One would readily need the median not the mean but such data is not
readily available. Looking across developing countries the $1000 GNI (Atlas) per
capita line is associated in 2013 with a life expectancy of about 62 years (using a
logarithmic regression or just below 60 years using a exponential regression model)
which is just below retirement age in many OECD countries.
In sum, as arbitrary as the LIC/MIC threshold may seem in the first instance, it
has some underlying logic in at least two senses (discounting for now the life
expectancy discussion): First, the remaining LICs are now those with poor historic
growth records and structural handicaps that shape the LDC (as most of the remaining
LICs are). Second, crossing the threshold does lead to reassessment of credit
worthiness by the credit rating agencies and thus – in principle – access to private
capital markets. It also has symbolic value in the sense that some political freedom
follows through a changing relationship with aid donors as the only source of
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
6 In contrast, Turkey with the same life expectancy had a GNI (Atlas) per capita of almost $11000,
close to the HIC threshold (and $19000 in GDP PPP per capita). Further, when another new MIC,
Nigeria, crossed the LIC–MIC threshold of $1000 in 2008 it had life expectancy of just 50 years. In
2013, Nigeria and Vietnam, both had virtually the same income per capita (in both GNI and GDP PPP)
but life expectancy remains 50 years in the former and 75 years in the latter.

!
14
development finance, which is often accompanied by conditionalities. The more
significant point is that the remaining 31 low income countries appear to be stuck at
the bottom for the foreseeable future, based on their recent growth history, suggesting
there is a difference between those countries joining the ‘middle’ and those left
behind.7
3. REVISITING THEORIES OF ABSOLUTE POVERTY
To date, global poverty has been considered explicable with reference to a theory of
absolute poverty which is largely based on insufficient domestic public and private
resources. Does the shift in global poverty described above question the value of the
dominant or orthodox theory of global poverty? If (almost) everyone is poor then
distribution is an irrelevant variable in explaining poverty. If it is no longer the case
that almost everyone is poor then there is a question mark over existing theories of
poverty in the developing world. The thesis argued henceforth is that increasingly
absolute poverty in the developing world will be explicable not by the lack of
resources at societal level, but by the distribution of those resources, and thus issues
of political economy and governance of growth and public finances. This resonates
strongly with Sen’s discussion of the causes of famine.
One could say orthodox theories of absolute poverty have tended to date to
focus on issues of insufficient resources at various analytical levels in developing
countries. One can refer to these types of theories of poverty henceforth as ‘material’
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
7 Although it is the case that some LDCs are actually MICs and this somewhat undermines the sense of
the LDCs being the poorest countries across a set of dimensions if some are, at least in income per
capita terms, not among the poorest LDCs that are MICs are small population or small-island
developing states which ought to be considered separately due to specific macroeconomic
vulnerabilities of such economies.

!
15
theories of poverty. One can argue that theories of absolute poverty have tended to
overemphasize the micro-level to the detriment of the macro-level and often describe
the symptoms of individual poverty rather than the underlying societal causes. To date
poverty has been largely measured and defined in terms of a set of deficits such as in
the quantity or quality of income, nutrition, health, and education or a poverty line set
based on those. However, none of these are underlying causes of poverty in
themselves. Deficits in well-being describe the immediate consequences of poverty
rather than present a theory on the causes of poverty. Furthermore, any theory of
poverty implies a different type state response or ‘public responsibility’ in terms of
the social policy or the welfare regime or more generally the governance of growth
(the management of growth processes and distribution of opportunities and benefits
from growth and economic development). For example, if poverty is structurally
related to distribution of wealth, income, opportunities, and labour markets then there
is a substantial role for a state to govern growth and redistribute with an
interventionist welfare regime.!
Material theories of poverty might also include theories based deprivations or
deficits of something—typically productive and human assets and livelihood
opportunities related to those assets and vulnerabilities or hazards faced or exposed to.
Such theories are discussed implicitly in the well known reviews of Ravallion (2013)
as well as those of Ruggeri et al. (2003) and Stewart et al. (2007). In such theories the
poor have few private assets and/or limited access, entitlements or claims to public or
common assets. Thus, people are poor as they have few assets from which they can
extract income and consumption. Such theories are largely individual-based—at
which level these theories are logically consistent—rather than taking societies as

!
16
their unit of analysis. In some cases there may be household, village, or higher units
of assessment of assets and livelihoods.
In contrast to the above, a heterodox or alternative group of theories which
could be labelled as structural, distributional, or relational theories of poverty may be
increasingly relevant given the shift in global poverty outlined earlier. Such theories
have so far been less systematically applied at a macro or societal level to developing
countries under the assumption that they are less of relevance if (almost) everyone is
poor. Important examples of approaches to poverty which have sought to bring in
aspects of a more structuralist approach include: the livelihoods literature which has
emphasised asset accumulation at a micro-level (see for example, Krishna, 2006;
Moser, 1998; 2006); the development of multi-dimensional poverty measures
including household assets (e.g. Alkire and Foster, 2010; Alkire, et al, 2015);
distributional questions raised in participatory poverty assessments (e.g. Narayan et
al., 1999); the literature on the concept of human wellbeing which has expanded the
lens of poverty research to include relational (social and personal) aspects of human
wellbeing (see for example, Gough and McGregor, 2007) and related theorising
related to ‘welfare regimes’ and the the set of policies and institutions that support
welfare improvement (see Wood and Gough, 2006); the attention to assets which is
also central to research on the intergenerational transmission of poverty and broader
research on poverty dynamics (see for example, Bird 2007; Hulme and Shepherd
2003; Kabeer, 2003). Most notably, the empirical survey of longitudinal datasets in
Dercon and Shapiro (2007) that draws out the causes of remaining in, or escaping
from, poverty which include changes in economic and social assets as well as
structurally based factors such as social exclusion or discrimination, and being located
in remote or otherwise disadvantaged areas. A question then follows as to how such

!
17
factors are distributed across any given society. Sen (1983) for example argues that
although in terms of absolute deprivations, capabilities are likely to be set across
societies, poverty relates to the society in the sense that the resources required to
expand capabilities are dependent on what is available in a given society. Townsend
(1979) too identified that theories of poverty require attention to the analysis and
distribution of resources, the patterns of production and distribution, the forms of
consumption that different resources generate, the social classes that influence
relationships in the system, and the over-representation of minority groups among the
poor. Although such theories have in general been well applied to OECD countries
there has been much more limited application in developing countries to date. Of
course Townsend himself made use of such theories in his own research (see for
example, Townsend and Gordon, 2002; Gordon et al., 2003) and one further example
of a structural theory of poverty is that of Harriss-White (2005), who argues in favour
of shifting from theories based on individual deprivation to theories based on an
explanation of the unequal distribution of power, wealth, and opportunity, and thus
the social processes, structures, and relationships that lead to poverty and its
reproduction.
Structural, distributional, or relational theories are based on the structural
position of the poor within the distribution of wealth and income and their labour-
market position. The poor’s hierarchical location in the social structure determines the
choices people have and their consequences. The social structure continually recreates
a population of poorer people because income/consumption levels at the lower end of
the distribution start low and – on average - grow at the rate of the mean rather than
the rate of growth of richest fractiles of the population. Even if the
income/consumption at the bottom of society were to rise faster than at the top of

!
18
society, given the low starting point it would take a long time to make any large
difference to the level of absolute inequality. In terms of inequality, the stability of the
social structure exists as people at each level use their resources to protect their
advantage and pass that advantage on to their children in the intergenerational
transmission of inequality. Furthermore, the poor operate in informal, volatile, and
insecure labour markets such as seasonal agricultural wage labour or informal urban
service sectors leading to large fluctuations in income/consumption at different points
in time. Access to public entitlements and public goods and assets may be haphazard
and mediated via the non-poor who may have perceptions about the ‘deserving’ and
‘undeserving’ poor and/or demand informal payments to allow the poor access to
their state entitlements which diminish their net value to the poor. In short, in such
theories poverty is caused by structural factors such as the distribution of wealth and
thus income, the distribution of education and human capital, and the related
stratification of labour markets (the existence of lower and more highly rewarded
labour markets, which may also be characterized by uncertainty, informality and
differing prospects to raise incomes). Poverty is also caused by discrimination and
prejudice faced by the poor as a result of perceptions of hierarchy and status which
condition inequality and resource access in terms of class, gender, ethnicity, sub-
national geography, and age. Such theories may extend poverty into concepts of
social exclusion, which is a framing that allows for deprivations and wealth to co-
exist. It also explores the processes that generate each of these, as well as the
exclusion of some groups from the benefits of economic growth (see for discussion
Hills and Stewart, 2005). Such arguments are theoretical. How would one assess if a
country were able to end extreme poverty based on national resources? One empirical
avenue would be the reallocation of regressive public spending to poverty transfers.

!
19
Such transfers have received significant attention in Latin America in particular but
also elsewhere and are often referred to as ‘cash tarnsfers’. Such transfers need to be
funded from reallocation of public spending (or external resources in the from of aid).
One source of current regressive public spending is fossil fuel subsidies. In the
followings section it is discussed if and where the reallocation of this public ‘bad’
might end extreme poverty.
4. AN EMPIRICAL EXPLORATION OF THE CAPACITY FOR NATIONAL
REDISTRIBUTION
It follows that social and political structure are inter-related; as is the extent to which
re-distribution of resources is included in policies such as social protection and
insurance and preferential access to education and employment by the reallocation of
other public spending (including who pays tax and how much). Take for example
fossil fuel subsidies: Post-tax fossil-fuel subsidies in developing countries in 2011
amounted to $895 billion in current dollars (or almost two trillion in 2011 PPP
dollars) (Clements et al., 2013). Such subsidies largely benefit the upper middle
classes and elite and if redirected could cover the cost of the total poverty gap via
redistributive social transfers in many MICs. Arze del Granado et al. (2012) in a
sample of twenty developing countries during the 2005–9 period, including several of
the new MICs such as Indonesia, Sri Lanka, India, and Ghana, find that, on average,
the richest 20 per cent of households gain six times more from such subsidies than the
poorest 20 per cent of households. The former capture, on average, 43 per cent of the
total subsidy value, the latter capture just 7 per cent. It is worth noting that the
distributional impact of the subsidies does vary by product. For example, gasoline is

!
20
the most regressive and kerosene the least. Subsidies to natural gas and electricity are
highly regressive. What if these subsidies were reallocated to redistributive social
transfers, with compensation for the poorest for the impact of the subsidy loss to
them? The administrative costs of new or expanded social transfer programmes
would, potentially, be covered by the administrative costs formerly associated with
the fuel subsidy programme.8 Indonesia in 2015 did exactly this: a long enduring fuel
subsidy that had grown to a post-tax subsidy that Clements et al., (2013) estimated at
5.4 per cent of GDP was drastically reduced and social programmes in health,
education, and in-cash payments expanded, as well as commitment to infrastructure
programmes.
At a global level one can make estimates using the dataset of Clements et al.
(2013) who provide a conservative data set on fossil fuel subsidies by their
components – petroleum products (gasoline, diesel, and kerosene), electricity, natural
gas and coal – as a proportion of GDP for each country.9 If we convert all subsidies
into 2011PPP dollars and produce estimates with and without full compensation for
the poorest quintile in line with the average welfare impact in the Arze del Granado et
al. (2012) study, we find that the $1.90 poverty gap is covered by the post-tax fossil-
fuel subsidy in many MICs where global poverty is concentrated (see Table 1). Some
caveats are important though: The calculations here are intended as indicative. Even
though the cost of subsidies is conservatively estimated, oil prices are falling at least
temporarily. It would though seem unlikely that oil prices will remain so low in 5-10
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
8 There are caveats, not least that food prices may rise due to the removal of the subsidy due to
transportation costs though the poorest could be compensated for these too.
9 Alternative estimates by Coady et al., (2015, p. 19) argue the data in Clements et al., (2013) is too
conservative and provide substantially higher estimates that include pricing of the externalities of fuel
consumption (e.g. costs of health impact of emissions, global warming, congestion and so forth). They
assume that the market price of energy undercharges for damages resulting from fuel consumption and
that leads to a larger implied subsidy.

!
21
years time. Estimates of Clements et al., (2014, p. 42) take petroleum prices for
2000-2011 and coal and natural gas prices for 2007-2011. This would imply Crude oil
prices at an average of approximately $52/bbl; Coal at $92/mt and natural gas at
$6/mmbtu. In years of higher energy prices relative to 2011 the estimates here will
underestimate the poverty gap covered and vice versa. There are further
methodological issues on the quantification of subsidies (for discussion on measuring
fossil fuel subsidies see in particular, Kojima and Koplow, 2015). There is a wide
array of policies from which producer/consumer fuel subsidies might arise: grants,
transfers, tax credits, trade restrictions, price controls, amongst others. The data used
here is based on the ‘price-gap approach’. This calculates the difference of retail
prices to ‘reference prices’. The main contention is which reference price to use when
comparing fuel producing versus fuel importing countries.
Although energy prices have been falling, given the exercise is indicative not
precise the argument holds that substantial national resources lie in subsides that
could be reallocated. Or one might say did the world just miss the opportunity to end
global poverty until fuel prices return to their 2011 levels which were well below the
peak of 2013-2014. The estimates should thus be viewed as indicative of substantial
domestic resources now available to some developing countries. Redistributing the
fossil fuel subsidy spending might, in practice, also not be easy for the following
reasons: Compensation may need expanding beyond the poorest quintile; as noted, the
removal of the subsidies may raise transportation costs and thus prices of other goods.
In short, the purpose of this exercise is solely to show there are potentially sufficient
public resources at a national level - in principle - to end much of global poverty. This

!
22
is a relatively new phenomenon - that most countries may have the public resources to
cover the poverty gap - even if their reallocation is not necessarily easy.10
What does the data say? If we focus on the 20 developing countries that
contribute more than 1 per cent to the global total poverty gap we find that just two
countries do not have fossil fuel data to make an assessment (Burundi and
Uzbekistan). Eight LICs and ten MICs account for eighty per cent of the global
poverty headcount and the global poverty gap at the $1.90 poverty line. Those ten
MICs alone account for two-thirds of the global poverty headcount but slightly less of
the global poverty gap. The global poverty headcount is distributed one third in sub-
Saharan Africa, one third in South Asia, a fifth in East Asia and the remainder
elsewhere. However, half of the total global poverty gap is in sub-Saharan Africa
alone and a quarter in South Asia.
We find, surprisingly, even some LICs would have a substantial proportion of
their total poverty gap (TPG) covered by fossil fuel subsidies. Ethiopia and Uganda
have a third of their TPG covered and Tanzania almost a half at the $1.90 poverty
line. Compensation for the poorest makes some difference but not a large amount. In
contrast, in MICs that have more than one cent of the total global poverty gap, the
fossil fuel subsidies would cover the $1.90 total poverty gap. Exceptions are Zambia
which does have almost three quarters coverage and Kenya that has only a fifth
coverage of the TPG). Overall, the data show seventy per cent of global poverty at
$1.90 would be covered (with and without compensation) by the reallocation of fossil
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
10 As recently as the early to mid-2000s, estimates of redistributive capacity, suggested that such
redistribution would not cover the poverty gap unless the marginal tax rates (MTRs) were exorbitant
for most developing countries. Ravallion (2009a) taking survey data for the early to mid-2000s,
produced estimates for the $1.25 and $2 poverty gap (2005 PPP) and the necessary taxation to cover it.
Ravallion estimated the MTRs for the ‘rich’ (which he defined as those earning more than $13 per day
or living above the US poverty line) required in order to end poverty in each country. He argued that
MTRs over 60 per cent would be prohibitive. While the MTRs needed to end poverty are less than 10
per cent in many of the ‘old’ MICs or UMICs, in many new MICs or LMICs they would have to be
much higher (see for estimates, Ravallion, 2009a, pp. 30–2).

!
23
fuel subsidies to poverty transfers even with compensation for the poorest quintile of
the population. The reallocation of fossil fuel subsidies is though just one example of
a source for transfers to end poverty. One could further explore other public ‘bads’
though one would have to make value judgments on what those are and what
minimum spend might be appropriate.
Table 1. Estimates of the proportion of total extreme poverty gap covered by fossil fuel subsidies in
countries contributing more than 1% to total global poverty gap and in aggregates, 2011
Proportion of global poverty
accounted for
Proportion of the poverty gap
covered by fossil fuel subsidies
LICs
% of total
global poverty
headcount
% of total
global poverty
gap
Without
compensation
for poorest
With
compensation
for poorest
Burkina Faso
0.9
1.1
12.6
11.7
DRC
5.2
9.0
4.2
3.9
Ethiopia
3.0
2.8
34.2
31.7
Malawi
0.5
1.7
6.9
6.4
Madagascar
1.1
3.0
6.5
6.0
Mozambique
1.8
2.7
24.5
22.8
Tanzania
1.7
2.3
47.6
44.2
Uganda
2.2
1.2
29.9
27.8
MICs
Bangladesh
6.7
5.8
226.1
209.8
Brazil
1.1
2.0
129.4
120.1
China
15.0
12.1
2016.5
1871.3
India
26.4
18.0
690.3
640.6
Indonesia
3.9
2.4
2089.4
1939.0
Kenya
1.4
1.6
21.3
19.7
Nigeria
8.7
12.1
115.0
106.7
Pakistan
1.4
0.7
3120.0
2895.4
Philippines
1.2
0.9
211.8
196.6
Zambia
0.9
1.5
72.6
67.4
Regions
Sub Saharan Africa
38.7
51.4
78.9
73.2
East Asia and the Pacific
21.0
16.3
2060.3
1912.0
South Asia
34.9
24.8
650.8
603.9
Income groups
LIC
32.0
39.7
48.6
45.1
LMIC
49.2
43.3
777.8
721.8
LMIC minus India
22.9
25.3
840.2
779.7
UMIC
18.7
17.0
3768.3
3497.0
UMIC minus India
3.8
4.9
8131.3
7545.9
All developing
100.0
100.0
995.1
923.4
Source: Data processed from Clements et al., (2014) and World Bank (2015c). No fossil fuel data for
Burundi (LIC, 1.1% of total global poverty gap at $1.90) and Uzbekistan (MIC, 2.5% of total global
poverty gap ay $1.90).
!

!
24
5. CONCLUSIONS
Is poverty the characteristic of an individual or a society? This distinction is important
as the difference between individual (or household) analysis and social structure is a
central contention in theories of poverty. An individual might be able to increase their
education, get a job, and possibly move out of poverty, as some theories emphasize.
However, individuals cannot change the unemployment rate or education
opportunities across a society, nor the ‘welfare regime’ or the ‘growth regime’ (the
macro-economic policy orientation), which are a product of, amongst other things, the
form of economic development pursued in a country. One can argue that poverty in
developing countries has become disconnected from the processes of accumulation
and economic development which are visible in many MICs. Harriss (2007) provides
one of the most thought provoking expressions of this in that explanations of
individual deprivation ignore the study of social relations and inequalities in income,
expenditure, wealth, and ultimately power and governance structures that determine
the welfare regime of country. It is not that this view is not visible but that it has been
side-lined by a dominant narrative on the causes of global poverty which emphasises
a lack of public and private resources in developing countries.
In conclusion, much of global poverty is concentrated in a small set of
populous countries that have transitioned from low to middle income per capita.
Economic growth since the Cold War has expanded national resources in developing
countries, and it can be argued that consequentially global poverty has become less
about lack of resources and more about questions of national inequality, issues of
social policy, patterns of economic growth and economic development, and the form
of economic development pursued. Certainly, at the end of the Cold War, few if any

!
25
developing countries had domestic public resources to end national poverty. In
contrast, the expansion of domestic public as well as private resources as a result of
substantial economic growth has led to most global poverty being concentrated in
countries that already have, or will have in the foreseeable future, sufficient resources
to end absolute poverty. The thesis of this paper has been that this shift in global
poverty implies that the dominant or orthodox theory of global poverty – that poverty
is explicable at the societal level by insufficient resources - requires, at the least,
questioning. It is argued that increasingly absolute poverty in the developing world
will be explicable not by the lack of resources at societal level, but by the distribution
of those resources and thus issues of political economy and governance of growth and
public finances.

!
26
REFERENCES
Alkire, S., J. Roche, E. Santos, and S. Seth. (2011). Multidimensional Poverty 2011,
Oxford: Oxford Poverty and Human Development Initiative (OPHI).
Alkire, S., Roche, J., Seth, S. and Sumner, A. (2015a) Identifying the Poorest People
and Groups: Strategies Using the Global Multidimensional Poverty Index. Journal of
International Development 27 (2015), 362-387
Alkire, S., Roche, J., and Vas, A. (2015b) Changes over time in multi-dimensional
poverty: Methodology and Results for 34 countries. Oxford: Oxford Poverty and
Human Development Initiative (OPHI).
Alonso, J. (2012) From aid to global development policy. UNDESA Working Paper
No. 121. New York: UNDESA.
Alonso, J., Cortez, A., and Klasen, S. (2014) LDC and other country groupings: How
useful are current approaches to classify countries in a more heterogeneous
developing world? UNDESA Working Paper. New York: UNDESA
Clarke, M. and Feeny, S. (2011) Old challenges and new opportunities for the MDGs:
now and beyond 2015. Journal of the Asia Pacific Economy 16(4): 509-519
Carbone, M. (2013) International development and the European Union's external
policies: changing contexts, problematic nexuses, contested partnerships. Cambridge
Review of International Affairs 26(3): p. 483-496
Cruz, M., Foster, J., Quillin, B., and Philip Schellekens1 (2015) Ending Extreme
Poverty and Sharing Prosperity: Progress and Policies. World Bank Policy Research
Note. Washington DC: World Bank.

!
27
Edward, P. and A. Sumner. (2014). ‘Estimating the Scale and Geography of Global
Poverty Now and In The Future: How Much Difference Do Method and Assumptions
Make?’. World Development 58: 67–82.
Ferreira, F. et al., (2015) A Global Count of the Extreme Poor in 2012
Data Issues, Methodology and Initial Results. World Bank Policy Research Working
Paper. Washington DC: World Bank.
Glennie, J. (2011) The Role of Aid to Middle-Income Countries: A Contribution to
Evolving EU Development Policy, London: Overseas Development Institute.
Haddad, L. (2012) Five Assumptions of Dominant Thinking in International
Development. Development (2012) 55(1), 34–44. doi:10.1057/dev.2011.106
Haddad, L. (2014) Equity: Not Only for Idealists. Development Policy Review 33(1):
5-13
Herbert, S. (2012) Reassessing aid to middleincome countries: the implications of the
European Commission’s policy of differentiation for developing countries. ODI
Working Paper. London: ODI.
Kanbur, R., and A. Sumner. (2012). ‘Poor Countries or Poor People? Development
Assistance and the New Geography of Global Poverty’, Journal of International
Development 24.6: 686–95.
Keeley, B. (2012) OECD Insights From Aid to Development The Global Fight
against Poverty. Paris: OECD.
Koch, S. (2015) From poverty reduction to mutual interests? The debate on
differentiation in EU development policy. Development Policy Review 33 (4), 479-
502
Lundsgaarde, E. (2012) The Domestic Politics of Foreign Aid. London: Routledge

!
28
Madrueño-Aguilar, R. (2015) Global Income Distribution and the Middle-Income
Strata: Implications for the World Development Taxonomy Debate. European Journal
of Development Research
Poke, N. and Whitman, J. (2011) The Millennium Development Goals and
Development After 2015. Third World Quarterly 32(1): 181-198.
Ottersen, T., Kamath, A., Moon, S., and Røttingen, JA. (2014) Development
Assistance for Health: Quantitative Allocation Criteria and Contribution Norms.
Chatham House Working Paper. London: Chatham House.
Sumner, A. (2010) Global Poverty and the New Bottom Billion: What if Three-
Quarters of the World’s Poor Live in Middle-Income Countries? Institute of
Development Studies (IDS) Working Paper 349. Brighton: IDS.
Sumner, A. (2012). ‘Where do the Poor Live?’, World Development 40.5: 865–77.
Vasquez. S., and Sumner, A. (2013) Revisiting the Meaning of Development: A
Multidimensional Taxonomy of Developing Countries. Journal of Development
Studies 49(12): 1728-1745
Vasquez. S., and Sumner, A. (2015, forthcoming), Is the ‘Developing World’
Changing? A Dynamic and Multidimensional Taxonomy of Developing Countries.
European Journal of Development Research.
World Bank (2015a) Country and Lending Groups. Downloaded from:
http://data.worldbank.org/about/country-and-lending-groups
World Bank (2015b) PovcalNet. Downloaded from:
http://iresearch.worldbank.org/PovcalNet/
World Bank (2015c) World Development Indicators. Downloaded from:
http://data.worldbank.org/data-catalog/world-development-indicators