ADBI Working Paper Series

AN ASIAN POVERTY LINE?

ISSUES AND OPTIONS

Stephan Klasen

No. 609
November 2016

Asian Development Bank Institute

Stephan Klasen is a professor of development economics at the University of Gttingen

in Germany.
The views expressed in this paper are the views of the author and do not necessarily
reflect the views or policies of ADBI, ADB, its Board of Directors, or the governments
they represent. ADBI does not guarantee the accuracy of the data included in this paper
and accepts no responsibility for any consequences of their use. Terminology used may
not necessarily be consistent with ADB official terms.
Working papers are subject to formal revision and correction before they are finalized
and considered published.

The Working Paper series is a continuation of the formerly named Discussion Paper series;
the numbering of the papers continued without interruption or change. ADBIs working
papers reflect initial ideas on a topic and are posted online for discussion. ADBI encourages
readers to post their comments on the main page for each working paper (given in the
citation below). Some working papers may develop into other forms of publication.
Suggested citation:
Klasen, S. 2016. An Asian Poverty Line? Issues and Options. ADBI Working Paper 609.
Tokyo: Asian Development Bank Institute. Available: https://www.adb.org/publications/asianpoverty-line-issues-and-options
In this publication, $ refers to US dollars.

Asian Development Bank Institute

Kasumigaseki Building, 8th Floor
3-2-5 Kasumigaseki, Chiyoda-ku
Tokyo 100-6008, Japan
Tel:
Fax:
URL:
E-mail:

+81-3-3593-5500
+81-3-3593-5571
www.adbi.org
[email protected]

2016 Asian Development Bank Institute

ADBI Working Paper 609

S. Klasen

Abstract
Given Asias record of rapid economic growth and the conceptual and empirical problems
of the current international income poverty line (dollar-a-day), this paper discusses whether
there is merit to develop an Asia-specific poverty line that addresses some of the
shortcomings of the dollar-a-day line and additionally considers Asias particular economic
situation. We consider various ways of creating an Asia-specific poverty line, including
an Asia-specific international income poverty line (using purchasing-power parity [PPP]
adjusted dollars) that is derived from Asian national poverty lines. We argue that there
can be some merit in developing an Asian poverty line and that, in the case of income
poverty, it would be best to ground such an Asia-specific poverty line in a consistent method
of generating national poverty lines using national currencies rather than generating a
PPP-adjusted poverty line in international dollars that is specific for Asia. It is important that
such a poverty line also considers relative poverty in its assessment to reflect the rising
aspirations of Asian societies, in line with suggestions made by Chen and Ravallion (2013)
on weakly relative poverty lines. In terms of multidimensional poverty lines, there is also
some merit in developing an Asia-specific multidimensional poverty index that takes into
account the specific living conditions of Asian societies.

ADBI Working Paper 609

S. Klasen

Contents
1.

INTRODUCTION ....................................................................................................... 1

2.

ADVANTAGES AND DISADVANTAGES OF AN ASIAN POVERTY LINE .............. 1

3.

OPTIONS TO CONSTRUCT AN ASIAN POVERTY LINE ......................................... 4

3.1
3.2
3.3
3.4

4.

An Absolute Income Poverty Line using the World Banks Methods .............. 4
A Weakly Relative Poverty Line using the World Banks Approach.............. 8
Using National Poverty Lines to Measure Poverty in Asia .............................. 9
An Asia-specific Multidimensional Poverty Index?........................................ 10

SOME TENTATIVE CONCLUSIONS....................................................................... 11

REFERENCES ................................................................................................................... 12
APPENDIX.......................................................................................................................... 14

ADBI Working Paper 609

S. Klasen

1. INTRODUCTION
Using the widely-used international income poverty line ($1.25 per person per day),
poverty in Asia has fallen dramatically in recent decades. In fact, the very rapid
progress on absolute income poverty reduction in Asia is largely responsible that
the first Millennium Development Goal (MDG) aiming to halve the incidence of
absolute poverty between 1990 and 2015 has been reached four years ahead of
schedule. This was achieved by particularly rapid progress in many populous Asian
economies (particularly, the Peoples Republic of China (PRC), Indonesia, Viet Nam),
overcompensating for much slower progress in poverty reduction in Sub-Saharan
Africa (Chen and Ravallion 2013).
Despite this progress, it is too early to declare victory on the poverty front in Asia for
various reasons. First, progress in poverty reduction remains fragile in many Asian
countries and the vulnerability to poverty remains high (Klasen and Waibel 2013,
2014). Second, there is the recognition that poverty captures more than a lack of
incomes, an issue covered by the literature on multidimensional poverty (e.g. Alkire
and Santos 2014; Rippin 2013). Progress in reducing multidimensional poverty in Asia
has generally been more uneven (although there are substantial uncertainties about
the data, particularly comparable data over time). Lastly, in many Asian countries
national poverty lines are substantially higher than the international $1.25-a-day
poverty line; in some, including PRC and India, they have been revised upwards to also
reflect the rising aspirations of the populations in these societies. At these higher (and
increasing) poverty lines, poverty is far from defeated.
Partly as a result of these factors, the Asian Development Bank (ADB) is considering
whether there is merit in developing an Asia-specific poverty line. In addition, it is
considering ways to derive such an Asian poverty line, closely related to the methods
developed and applied by the World Bank (Ravallion, Chen and Sangraula 2009) in
deriving the international $1.25-a-day poverty line but specific for the Asian situation.
This paper first discusses whether there indeed is merit in developing an Asia-specific
poverty line. We then discuss various options of developing such a poverty line,
considering income and multidimensional versions of such a poverty line. We argue
that there can be some merit in developing an Asian poverty line and that, in the case
of income poverty, it would be best to ground such an Asian-specific poverty line in a
consistent method of generating national poverty lines using national currencies rather
than generating a PPP-adjusted poverty line in international dollars (see also Klasen
2013a and Klasen et al. 2015). It is important that such a poverty line also considers
relative poverty in its assessment to reflect the rising aspirations of Asian societies
(see Ravallion and Chen 2011; Chen and Ravallion 2013). In terms of multidimensional
poverty lines, there is merit in developing an Asia-specific multidimensional poverty
index (MPI) that takes into account the specific living conditions of Asian societies.

2. ADVANTAGES AND DISADVANTAGES

OF AN ASIAN POVERTY LINE
Before discussing options to derive a poverty line for Asia, it is important to first discuss
whether it is indeed useful to develop such a line to begin with. We consider four
possible arguments for an Asia-specific poverty line. First, one could argue that
conditions in Asia are so different from other parts of the world that it justifies a different
poverty line, in the sense that, it would reflect these particular circumstances. For

ADBI Working Paper 609

S. Klasen

example, households tend to be smaller than in Africa, family ties are quite strong, and
the provision of public services by the state is substantial. To the extent that this is
the case, it might justify a lower poverty line, measured in terms of private per capita
incomes because fewer private incomes are required to achieve a certain level of
well-being. But it is not obvious that these apparent differences justify a peculiar
Asian poverty line because the heterogeneity within Asia in these economic and social
arrangements is very large. Also, one would first need to investigate the empirical
importance of these claims and their relevance to particular Asian countries before
one could draw any firm conclusions on this. It should also logically lead to different
poverty lines within Asia, depending on the particular circumstances. It would thus be
particularly difficult to use this argument as a motivation for a uniform income poverty
line appropriate for all of Asia.
A second argument relates to differences in levels and trends of economic
performance that ought to be reflected in the setting of a poverty line. Average incomes
in Asia are higher than in Africa (but lower than in Latin America and the Middle East)
and most economies in Asia have experienced rather rapid growth in the last three
decades. This might justify the use of an Asian poverty line that reflects the average
income level and, more importantly, reflects its rapid economic performance. We argue
below that such a poverty line should contain a relative element, i.e., increase with
rising prosperity in Asia. Of course the high heterogeneity in Asias income levels and
economic growth experience might be seen as a counterargument to a single and
uniquely Asian income poverty line. But to the extent that neighboring countries
benchmark their performance against each other, an argument for a unique line
reflecting these special features can be made.
A third argument is that an Asian poverty line would be more closely aligned with
national poverty lines in Asia and, thus, the disconnect between national and
international poverty measurement would be correspondingly smaller (see Dotter and
Klasen 2014a; Klasen 2013). This is essentially an empirical question. Clearly, the
current $1.25 international poverty line is only very loosely linked to Asian realities. The
only country included in the sample of the 15 poorest countries that were used to
derive the $1.25 the poverty line is Tajikistan (Ravallion, Chen and Sangraula 2009). In
fact, one can argue that the $1.25 poverty line is much more a reflection of national
poverty realities in Africa than in Asia (which includes only South, East, and Southeast
Asia; Central Asia is included with Europe here). This can also be seen in Figure 1
below. This figure shows the difference in the poverty headcount using the national
poverty line minus the headcount using the international poverty line. A negative
number means that the international poverty line is higher than the national poverty
line. This is the case in a number of Asian countries, including large ones (the size
of the bubble indicates the number of poor people there). For these countries, the
international poverty line appears rather high. In Europe and Central Asia, the
international poverty line is far too low. An extreme case is Tajikistan where the poverty
headcount using the national poverty line is 40 percentage points higher than using
the international poverty line (even though this country is included in the sample used
to calculate the international poverty line). Thus, the mismatch between Asian national
poverty lines and the international poverty line appears substantial, which would
argue for an Asia-specific poverty line. Of course, if an Asian poverty line was
directly grounded in country-specific poverty lines (in contrast to the international
poverty line that is an average of poverty lines across the world), the linkage between
national and pan-Asian poverty measurement could be even closer. This is an issue we
discuss below.

ADBI Working Paper 609

S. Klasen

Figure 1: Mismatch in Poverty Headcount by Region using National

and International Poverty Lines

Notes: NPL refers to national poverty lines; IPL to the international poverty line ($1.25 a day); the y-axis measures mean
consumption per capita per month, taken from the national accounts (NA). Europe includes Eastern European and
Central Asian countries; LAC refers to Latin American and Caribbean countries; and MENA refers to countries in the
Middle East and North Africa; where these regions refer to World Bank regional classifications.
Source: Dotter and Klasen (2014a).

A fourth argument relates to a multidimensional poverty measure. The most prominent

internationally comparable multidimensional poverty measure, UNDPs MPI (see UNDP
2010, Ch. 5; Alkire and Santos 2014), uses the same indicators and cutoffs across
the entire developing world. Due to differences in climate, economic and social
arrangements, social preferences, and the nature and state of public services, one
might argue that an Asian MPI should reflect this in terms of indicators and cutoffs. For
example, one might argue that the role of education as key to personal advancement is
seen as particularly important in Asian societies, and an MPI should reflect this by
giving education more weight and possibly argue for a higher cutoff. Of course, as
before, the heterogeneity within Asia is a problem for this line of reasoning. When
considering multidimensional poverty measures below, we revisit this issue again.
One should also mention that there are, of course, important disadvantages to
generating a continent-specific poverty line. The most important is that comparisons
across continents are difficult and untransparent both in terms of levels and trends.
To conclude this section, it is not obvious that a specific Asian poverty line is desirable.
The most compelling arguments are that it could reflect income levels and faster
economic progress better than a global measure; that it can be linked more closely to
national poverty lines in Asia; and that it might reflect uniquely Asian conditions and
settings in a multidimensional measure. But there are costs to it and this suggests that
one should not drop a global measure for an Asian one, but only treat an Asian poverty
line as complementary to a global assessment.

ADBI Working Paper 609

S. Klasen

3. OPTIONS TO CONSTRUCT AN ASIAN

POVERTY LINE
There are, of course, different options to generate an Asia-specific poverty line. First,
one needs to distinguish between an income and a multidimensional poverty line.
When constructing an income poverty line, we consider three options. The first option
is to mimic the estimation method of the World Bank of generating the $1.25-a-day
poverty line, but only using Asian countries in the estimation; the second option uses
the same set of countries to produce a weakly relative poverty line (Ravallion and
Chen 2011; Chen and Revallion 2013); and the third option grounds an Asian poverty
line in national poverty measurement (see Klasen 2013, 2013b). Thus, together with an
Asia-specific multidimensional poverty line, altogether four options are considered.

3.1 An Absolute Income Poverty Line using

the World Banks Methods
The World Bank has been generating an international poverty line since 1990
(World Bank 1990; Ravallion, Datt and van de Walle 1991). In 1990, it stood at $1.02
in 1985 PPP-adjusted dollars; in 2000, it was adjusted to $1.08 in 1993 PPP-adjusted
dollars (World Bank 2000; Chen and Ravallion 2001); and, in 2008, it was adjusted
to $1.25 in 2005 PPP-adjusted dollars (Ravallion, Chen and Sangraula 2009). Now,
the World Bank is working on adjusting its estimated poverty line using 2011
PPP-adjusted dollars.
The methods of deriving the international poverty line have essentially been the
same (although differing in some details of data used) and we focus on the latest
completed revision done in 2008. Ravallion, Chen and Sangraula (2009) explain how
the World Bank derives the international income poverty line using the following
steps. First, available national poverty lines for 74 developing countries are translated
into poverty lines expressed in PPP-adjusted international dollars in 2005 prices.
Then, these national poverty lines (expressed in 2005 international dollars) are lined
up against the log of consumption per capita (in 2005 international dollars) of the
74 countries (see Figure 2a below). They then observe that, below a certain threshold,
the relationship of per-capita consumption, national poverty lines are very similar,
indicated by the flat portion of the curve below, whereas above a threshold they rise,
less than proportionately, with mean incomes. The $1.25-a-day poverty line is then
simply the average poverty line of the flat portion.
For an Asian poverty line, one could simply limit the sample to 21 national poverty lines
from Asian countries, as done by ADB (2014) in an illustrative exercise reproduced
in Figure 2c below. As can be seen below, also here the relationship between log
per-capita consumption and the national poverty lines is also nonlinear, with a flat
portion below a certain threshold level of per capita incomes and an increasing portion
above that. The average poverty line of the flat portion in Asia extends substantially
further to the right. Whereas, in the global sample, only two Asian economies are in
included in the flat portion that constitutes the reference line for the poverty line, in the
Asian sample, nine countries are included in the reference group. This then also leads
to a substantially higher Asian poverty line of around $1.51. But this difference
between the international and the Asian poverty line is actually due to differences in the
estimation method between Ravallion, Chen and Sangraula (2009) and the estimate
produced by ADB (2014). Whereas the former estimate the relationship between the
level of the national poverty line and per-capita consumption from the national accounts

ADBI Working Paper 609

S. Klasen

(i.e., a linear model, in effect modeling the relationship in Figure 2b), ADB (2014)
estimates the relationship between national poverty lines and the log of per-capita
consumption (i.e. a log-linear model). As shown by Greb et al. (2012), using a loglinear model also leads to a larger reference group in the global model and a global
poverty line of $1.45 per day. Thus, using a sample of Asian countries does not lead to
a different poverty line from using the global sample if the same estimation method is
used. Nevertheless, the question arises which estimation method is to be preferred.
But both the estimations by Ravallion, Chen and Sangraula (2009) and by ADB (2014)
are problematic from a statistical point of view. In particular, in the linear model used by
Ravallion, Chen and Sangraula (2009), there actually is no statistical evidence of a kink
in the curve so that the kink is imposed on the data rather than observed (which is
visible from Figure 2b, see Greb et al. 2012 and Klasen et al. 2015). In the log-linear
model, the residuals are not normally distributed so that the inference, especially
regarding location and significance of the kink which separates the flat from the rising
portion, is problematic. As shown by Klasen et al. (2015), the preferred statistical
specification is actually a log-log model where there is statistical evidence for a kink as
well as normally distributed residuals. Using the latter model, a global poverty line of
about $1.21 is obtained, with a slightly larger reference group (of 19 countries).
Applying this to the estimation of the Asian poverty line, one could then take the
average of the Asian countries included in the reference group for the global line. That
would result in a poverty line of about $1.28 but only three Asian countries are included
in the reference group (Tajikistan, Nepal, and Bangladesh) which makes this approach
unreliable. If we instead estimated the linear or the log-log model using the 21 Asian
observations, this would yield an 'Asian' poverty line of $1.41$1.43; 1 this is driven
largely by the relative high national poverty lines in Tajikistan, Yemen, and Mongolia.
Also here, the reliability is an issue because a nonlinear threshold model is estimated
on just 21 observations where outliers and small data problems could have a large
impact; it is also likely that the switch to the 2011 PPP round would lead to a
substantial change in this estimate. Overall, we suggest that this method would not
generate a very reliable and robust estimate for an Asia-specific income poverty line.
Besides these estimation issues, there are more serious concerns and criticisms of this
entire approach which have been discussed extensively in the literature (e.g., Reddy
and Pogge 2009; Klasen 2013, 2013b; Klasen et al. 20015; Deaton 2010; Dotter and
Klasen 2014a). We highlight four of the most important issues that have been
discussed in the literature. First, this method is rather unstable and highly dependent
on the sample of countries included in the estimation and the PPP exchange rates
used. When, in 2008, the World Bank switched from using the 1993 PPPs and the
sample of countries used for estimating the poverty line. It led to the switch of the
international poverty line from $1.08 in 1993 dollars to $1.25 in 2005 dollars. Currently,
similar issues are arising with the new 2011 PPPs which could lead to serious
reassessments of poverty levels in the world and in different regions (Klasen et al.
2015). More seriously, the 2008 revision led to a massive upward shift in global poverty
for all years, e.g., from about 29% in 1990 to about 41% in the same year; thus, the
base year of the first MDG was changed substantially with a large impact on what
halving global poverty would mean. The pace of poverty reduction was, however, less
affected (Chen and Ravallion 2010). As shown by Deaton (2010) and Greb et al.
(2012), the main reason for the massive increase in levels of observed global poverty
was not the switch of the PPPs, but the switch in the sample of countries used to
estimate global poverty. Deaton (2010) additionally noted that the change in the
1

I thank Tatyana Krivobokova for providing these estimates, based on the methods described in
Klasen et al. (2015).

ADBI Working Paper 609

S. Klasen

sample led to some perverse effects. In particular, he noted the case of India. Whereas
India was part of the reference group of countries that made up the global poverty line
using 1993 dollars, high subsequent growth ensured that India was no longer in the
reference group in the assessment using 2005 dollars. Because Indias poverty line
is rather low, the exclusion of India from the reference group led to an increase in
the global poverty line, which, in turn, led to an increase in measured poverty in India
using that new line. In a sense, rapidly rising incomes in India have led to higher
observed poverty in India using the international poverty line, clearly a problematic
effect. In short, there appear to be substantial problems and uncertainties associated
with switches in PPPs and national poverty lines used to estimate the global poverty
line. The 2011 PPPs suggest that prior assessments of PPP-adjusted incomes
underestimated per capita incomes in PRC and India, and some other Asian
economies. If these are used to generate a new international poverty line, this could
have substantial implications for poverty in those countries, compared with other
regions, as well as on global poverty.
A second line of criticism relates to the use of PPPs more generally for this type of
assessment (Deaton 2010; Klasen 2013b; Reddy and Pogge 2009). One criticism is
that PPPs are generated to compare overall price levels, not price levels for the poor;
worse, they can be sensitive to changes in the price level for goods unrelated to the
poor (Reddy and Pogge 2009). Another criticism is that PPPs are only valid for a
particular benchmark year, but not over time. Thus, the question arises whether one
should use only one PPP benchmark year (as currently being done in the World Banks
approach to poverty measurement), or several benchmark years (as done for the Penn
World Tables that also use PPP-adjusted income data).
A third line of criticism is that the international income poverty line has limited relevance
for country-level poverty assessments because the difference between country-level
income poverty lines and the international income poverty line is substantial (Dotter
and Klasen 2014a). This point, which was already alluded to above, is nicely visible in
the estimation of the Asian poverty line below. As can be seen, the difference between
country-level poverty lines and the estimated Asian poverty line is substantial. In
Tajikistan and Yemen, poverty using the Asian poverty line is much lower than using
national poverty estimates whereas, in Nepal and India, it is much higher. In fact, there
is a clear regional pattern to the difference between national poverty lines and an Asian
poverty line. All South Asian countries are below the estimated line, i.e., poverty is
lower using national poverty lines than the international poverty line. The converse is
the case for all Western and Central Asian countries that were part of the former Soviet
Union. There national poverty lines are all above the line; thus, poverty is much lower
using the common Asian poverty line. This clear regional pattern appears problematic
and suggests substantial problems with one common Asian poverty line.
A fourth criticism of such an approach, closely related to the one just discussed, is the
increasing irrelevance of the $1.25 poverty line for an increasing number of Asian
countries. In many Asian economies, this poverty line is simply too low to be relevant
for policymakers there. In fact, several Asian countries, including PRC and India, have
recently increased their national poverty line to make it more relevant for national
policymaking. In this context, the question arises whether one should adjust the poverty
line because of increasing prosperity. This is precisely the suggestion by Ravallion and
Chen (2011) of a weakly relative international poverty line to which we turn below.

ADBI Working Paper 609

S. Klasen

To summarize, the case for an Asian poverty line using the World Banks method of
deriving the $1.25 poverty line is weak. It would not lead to a substantially different
poverty line, it is poorly linked to national poverty lines, it is unstable due to the link
to the PPPs and the estimation method, and it would be increasingly irrelevant for
fast-growing Asian economies.
Figure 2a: Threshold Model for log(Consumption)

Figure 2b: Threshold Model for Consumption

Note: Analysis based on Klasen et al. (2015). The data are from Chen, Ravallion, and Sangraula (2009).

ADBI Working Paper 609

S. Klasen

Figure 2c: National Poverty Lines from Asian Countries plotted against Mean
Consumption in International 2005 PPP

Note: PCE refers to per capita expenditures; all figures refer to dollars/capita/month.
Source: Elaboration of ADB (2014:8), based on data from Chen, Ravallion and Sangraula (2009).

3.2 A Weakly Relative Poverty Line using

the World Banks Approach
Ravallion and Chen (2011) proposed a weakly relative international poverty line. It can
be derived in various ways, but the easiest is to consider Figure 1a again. The
suggestion is that below for countries on the flat portion of the curve, the $1.25-a-day
should be the relevant line. For those on the ascending portion of the curve, the
poverty line should rise with the increase in mean income. It turns out that the best
empirical fit is that, above the threshold, the poverty line should increase by 0.33$
for every 1$ increase in per-capita consumption of a country above the threshold
(Ravallion and Chen 2011; Chen and Ravallion 2013). The elasticity of the weakly
relative poverty line is substantially below one (but increases with increasing incomes),
which distinguishes it from a purely relative line.
Such a weakly relative poverty line has several features that make it advantageous to
be used for an Asian poverty line (see Klasen 2013; Klasen et al. 2015). First, it adjusts
the poverty line automatically with increasing prosperity in Asia, thereby addressing
the problem of the increasing irrelevance of the very low $1.25-a-day poverty line.
As shown by Chen and Ravallion (2013), the weakly relative poverty line in East Asia,
for example, is about $2.34 in 2008, and, in South Asia, it is $1.94 (see Appendix
Tables A.1 and A.2, which are reproduced from that paper). Given that this poverty
line increases under-proportionately with mean income, it will still be the case that
distribution-neutral growth will lower weakly relative poverty, but will do so at a smaller
pace than when using a purely absolute line. As can be seen in the tables below, there
was still substantial poverty reduction in Asia using this approach, but poverty remains
a very serious issue.
Thus, the weakly relative poverty line has some advantages. At the same time, all the
other disadvantages of the World Banks method remain so that it is not clear whether
this is the best way forward. But it clearly seems to be superior to simply deriving an
Asian absolute poverty line using the World Banks method.

ADBI Working Paper 609

S. Klasen

3.3 Using National Poverty Lines to Measure Poverty in Asia

A third approach of setting an income poverty line is to actually coordinate a process in
Asia of setting national poverty lines using a standardized methodology. These national
poverty lines would be expressed in national currency but one could still aggregate the
poor across countries in a consistent fashion if the poverty lines were consistently
derived. This proposal was made by Reddy, Visaria and Attali (2008) and later by
Klasen (2013, 2013b) and Klasen et al. (2015). One advantage is that one would avoid
the problems associated with the PPP exchange rates. A second advantage is that
such a poverty line would be more closely linked to national poverty measurement and,
thus, would have a higher relevance.
At the same time, a range of questions would need to be addressed before such a
proposal could be implemented (see Klasen (2013b) for an extensive discussion). First,
how should such a poverty line be grounded? The most promising approach would be
to use the method most commonly used to set national poverty lines in developing
countries, the Cost of Basic Needs Method (Ravallion 1994). It involves first identifying
a reference group of households (which should be close to the poverty line) whose
spending pattern would be used to derive expenditure shares on a basket of goods and
services used to assess poverty. In a second step, the food expenditures in that basket
are turned into calories and then the basket is scaled up (or down) to reach the
required caloric norm for households. This basket (including non-food items) then
defines the quantities of food and non-food items to be consumed at the poverty line.
The cost of that basket then yields the poverty line. This poverty line is then updated for
price changes of goods included in the basket over the years. But over longer time
periods, the basket is adjusted to reflect changing expenditure patterns. In a rapidly
growing economy, it usually means that the basket changes by reducing the food share
and increasing more higher-quality goods. In this way, relative poverty considerations
can be brought in when the poverty basket is adjusted.
Although the methods are straight-forward and have been applied in many countries
(including in Asia), setting these poverty lines in a consistent fashion across countries
is challenging. The first-best option would be for participating countries to agree on a
consistent system of poverty measurement using this approach. It would ideally also
include coordinating household surveys so that the questionnaires are similar enough
that they can be used consistently. The model would be the System of National
Accounts where a similarly coordinated process of standard methods is accepted
across the world.
At the same time, it is unlikely that such a coordinated way to set national poverty lines
would be agreed upon quickly. In the meantime, a second-best option would be to use
existing household surveys from these Asian countries and apply consistent poverty
lines in these surveys, even if these lines are not the current approaches used by the
governments. In this way, one could demonstrate the feasibility of this approach and,
thereby, move the debate forward.
Thus this approach is promising but requires a longer-term process to implement it
fully. But, as suggested, a short-cut is possible and it is useful to illustrate the feasibility
of this approach.

ADBI Working Paper 609

S. Klasen

3.4 An Asia-specific MPI?

Lastly, there is the option to consider an Asia-specific MPI. It is widely recognized that
poverty is a multidimensional phenomenon. The challenge has always been to come
up with a set of indicators and weights that would allow for a consistent analysis of
poverty over time and across space. With the publication of the MPI in 2010 (UNDP
2010, ch.5), a first attempt to create such a comparable poverty measure was made. It
uses a so-called dual cutoff method proposed by Alkire and Foster (2011) where the
first cutoff defines whether a household is deprived in a particular dimension, and a
second cutoff defines whether a household has passed the threshold of deprivations to
be called multidimensionally poor.
Although there are many questions of details that still need to be addressed (Dotter
and Klasen 2014b), it now appears feasible to generate an Asia-specific version
of such an MPI. As discussed above, one would first need to think through why and
how an Asian MPI would have different indicators, cutoffs, or weights. This is not a
straight-forward question and has to deal with the great heterogeneity among Asian
countries. Although one may argue that, because of differences in climate, social
structures, or values in particular sub-regions (e.g., South Asia, Southeast Asia and
Central Asia), one could choose appropriate indicators, cutoffs, and weights to
generate MPIs for these different sub-regions, it would be hard to develop an MPI for
all of Asia. The only way out of this dilemma would be to initiate a process, possibly at
the level of ADB, to develop a common understanding for indicators, weights, and
cutoffs, although it is expected that such a consensus would not be reached easily.
A second way by which one could construct an Asian MPI that differed less
fundamentally would be to adjust cutoffs to better reflect the average performance
of Asian economies in these MPI indicators. For example, a cutoff of five years of
education of a single household member to render the entire household non-poor, as
currently done in the MPI, might be too low for many Asian countries. Thus, the idea
would be to move away from an absolute indicator of acute multidimensional poverty
to a (weakly?) relative one that considers the performance of Asian economies in
these indicators.
A third approach would be to change the weights used for an Asian-based
multidimensional poverty line. Here the results from the illustrative exercise of Pasha
(2014) are interesting. She uses Principal Components Analysis to derive statistical
weights for the indicators included in the MPI. Pasha (2014) indeed finds substantial
differences across countries in these weights. In India, the weights for child mortality
and nutrition are higher than in all the 22 countries included in her sample. Conversely,
the weights for education are rather low. The weights for standard of living are
very high (altogether 80%), with quality of floor and access to cooking fuel particularly
important. Using such country-specific weights would lead to quite different
multidimensional poverty measures and might provide interesting new insights. Of
course, it would lower the ability to compare levels and trends across countries. But
clearly this is an issue well worth exploring further.

10

ADBI Working Paper 609

S. Klasen

4. SOME TENTATIVE CONCLUSIONS

This discussion suggests that the case for developing an Asian poverty line is not
straightforward. In particular, we have argued that there are no good reasons to adjust
the World Banks $1-a-day approach to an Asian setting. Many of the problems of the
World Bank's international poverty line would carry over to its Asian version; in addition,
the database to estimate such a poverty line would be even smaller, leading to
questions of reliability and robustness. Also, the large heterogeneity in existing poverty
lines in Asia would militate against this proposal. A more promising option is to
consider a weakly relative Asian income or non-income poverty line that takes into
account the rapid growth in living conditions and aspirations in many of Asia's
economies. But many of the drawbacks of the current international poverty line would
carry over to the 'weakly relative' case. Even more promising could be a coordinated
process for setting national income poverty lines where national poverty measurement
is based on a common conception of poverty. This is a long-term agenda that would
need a great deal of coordination between Asian economies but it is well worth
pursuing this further. Another option would be the creation of an Asia-specific MPI,
maybe one that adjusts itself automatically to improving living conditions by adjusting
the cutoffs. But all of these proposals would have to be rigorously tested to see
whether they can be implemented and yield new insights that are not visible in current
approaches to poverty measurement in Asia.
The discussions about the changes in the international poverty line to reflect the results
of the 2011 PPP show the difficulty of maintaining a reliable, consistent, and robust
international poverty line (Klasen et al. 2015). Thus, it is all the more important to
consider alternatives.

11

ADBI Working Paper 609

S. Klasen

REFERENCES
Alkire, S., and F. James. 2011a. Counting and Multidimensional Poverty Measurement.
Journal of Public Economies 95(7): 476487.
Alkire, S., and M. E. Santos. 2014. Measuring Acute Poverty in the Developing World:
Robustness and Scope of the Multidimensional Poverty Index. World
Development 59: 251274.
ADB. 2014. Poverty in Asia: A Deeper Look. Manila: Asia Development Bank.
Chen, S., and M. Ravallion. 2001. How Did the Worlds Poor Fare in the 1990s?
Review of Income and Wealth 47(3): 283300.
. 2010. The Developing World is Poorer than We Thought, but No Less
Successful in the Fight against Poverty. Quarterly Journal of Economics
125(4): 15771625.
. 2013. More Relatively Poor in a Less Absolutely-poor World. The Review of
Income and Wealth 59(1): 128.
Deaton, A. 2010. Price Indexes, Inequality, and the Measurement of World Poverty.
American Economic Review 100(1): 534.
Dotter, C., and S. Klasen. 2014a. The (Ir-)Relevance of the International Poverty Line
for National Poverty Assessment. Mimeograph, University of Gttingen.
. 2014b. The Multidimensional Poverty Index: Achievements, Conceptual and
Empirical Issues. UNDP HDRO Occasional Paper, New York: UNDP.
Greb, F., S. Klasen, S. Pasaribu, and M. Wiesenfarth. 2011. Dollar a Day Re-Revisited.
Courant Research Center Discussion Paper No. 91. University of Gttingen.
Klasen, S. 2013a. Is it Time for a New International Poverty Measure? In Development
Cooperation Report 2013: Ending Poverty, edited by E. Solheim. Paris: OECD.
. 2013b. Measuring Levels and Trends in Absolute Poverty in the World:
Open Questions and Possible Alternatives. In Poverty and Social Exclusion:
New Methods of Analysis, edited by G. Betti and A. Lemmi. London: Taylor
and Francis.
Klasen, S., T. Krivobokova, F. Greb, R. Lahot, S. Pasaribu, and M, Wiesenfarth. 2015.
International Poverty Measurement: Which Way Now? Mimeograph, University
of Gttingen.
Klasen, S., and H. Waibel. 2013. Vulnerability to Poverty. London: Palgrave.
. 2014. Vulnerability to Poverty in South-East Asia: Drivers, Measurement,
Responses, and Policy Issues. World Development, DOI: http://dx.doi.org/
10.1016/j.worlddev.2014.01.007
Pasha, A. 2014. Regional Perspectives to the Multidimensional Poverty Index.
Mimeograph, University of Gttingen.
Ravallion, M. 1994. Poverty Comparisons, Fundamentals of Pure and Applied
Economics Volume 56, Chur, Switzerland: Harwood Academic Publishers.
Ravallion, M., S. Chen, and P. Sangraula. 2009 Dollar a Day Revisited. World Bank
Economic Review 23(2): 163184
Ravallion, M., and S. Chen. 2011. Weakly Relative Poverty. Review of Economics and
Statistics 93(4): 12511261.
12

ADBI Working Paper 609

S. Klasen

Ravallion, M., G. Datt, and D. van de Walle. 1991. Quantifying Absolute Poverty in the
Developing World. Review of Income and Wealth 37(4):34561.
Reddy, S., and T. Pogge. 2009. How Not to Count the Poor. In Debates on the
Measurement of Global Poverty, edited by S. Anand, P. Segal and J. Stiglitz.
Oxford: Oxford University Press.
UNDP. 2010. Human Development Report. New York: UNDP.
Reddy, S., S. Visaria, and M. Attali (2008) Inter-country Comparisons of Income
Poverty Based on a Capability Approach. In Arguments for a Better World,
Vol. II, edited by K. Basu and R. Kanbur. Oxford: Oxford University Press.
Rippin, N. 2013. Considerations of Efficiency and Distributive Justice in
Multidimensional Poverty Measurement. Ph.D. dissertation, University
of Gttingen.
World Bank. 1990. World Development Report 1990: Poverty. New York: Oxford
University Press.
. 2000. World Development Report 2000/01: Attacking Poverty. Washington DC:
The World Bank.

13

ADBI Working Paper 609

S. Klasen

APPENDIX
Table A.1: Average Relative Poverty Line by Region and Year
Region

Method 1
East Asia and Pacific
PRC
Eastern Europe and Central Asia
Latin America and the Caribbean
Middle East and North Africa
South Asia
Sub-Saharan Africa
Total
Total excl. PRC
Method 2
East Asia and Pacific
PRC
Eastern Europe and Central Asia
Latin America and the Caribbean
Middle East and North Africa
South Asia
Sub-Saharan Africa
Total
Total excl. PRC
Region
Method 1
East Asia and Pacific
PRC
Eastern Europe and Central Asia
Latin America and the Caribbean
Middle East and North Africa
South Asia
Sub-Saharan Africa
Total
Total excl. PRC
Method 2
East Asia and Pacific
PRC
Eastern Europe and Central Asia
Latin America and the Caribbean
Middle East and North Africa
South Asia
Sub-Saharan Africa
Total
Total excl. PRC

1981

Mean Poverty Line $/Person/Day at

1984
1987
1990
1993

1.33
1.25
4.05
4.32
2.42
1.27
1.55
2.00
2.29

1.34
1.25
4.21
4.25
2.56
1.27
1.55
2.01
2.30

1.35
1.25
4.35
4.07
2.40
1.30
1.53
2.00
2.28

1.39
1.25
4.21
4.00
2.41
1.35
1.51
2.00
2.28

1.43
1.25
3.78
4.28
2.47
1.38
1.49
1.99
2.26

1.36
1.26
4.19
4.51
2.84
1.30
1.71
2.09
2.41
1996

1.38
1.29
4.36
4.17
3.04
1.32
1.66
2.09
2.39
1999

1.41
1.33
4.49
4.36
2.84
1.35
1.65
2.12
2.41
2002

1.45
1.34
4.27
4.42
2.85
1.30
1.60
2.11
2.38
2005

1.54
1.42
4.76
4.47
2.98
1.38
1.54
2.18
2.46
2008

1.57
1.35
3.79
4.41
2.50
1.47
1.51
2.08
2.34

1.66
1.48
3.98
4.68
2.59
1.54
1.51
2.17
2.41

1.82
1.64
4.54
4.76
2.80
1.58
1.53
2.30
2.52

2.03
1.85
5.61
5.09
3.05
1.74
1.55
2.54
2.77

2.34
2.20
6.99
5.79
3.37
1.94
1.60
2.90
3.13

1.73
1.61
4.37
4.68
2.98
1.41
1.56
2.24
2.46

1.80
1.71
3.96
4.70
3.06
1.46
1.58
2.23
2.41

2.06
2.01
4.39
4.66
3.35
1.48
1.61
2.37
2.50

2.39
2.40
5.22
5.17
3.02
1.54
1.63
2.60
2.66

2.72
2.81
6.71
5.93
3.24
1.58
1.78
2.94
2.98

Source: Reproduced from Table 4 from Ravallion and Chen (2013).

14

ADBI Working Paper 609

S. Klasen

Table A.2: Weakly Poverty Measures for the Developing World, 19812008
Region

1981

% of Population below Relative Poverty Line

East Asia and Pacific
80.5
PRC
85.2
Eastern Europe and Central Asia
22.0
Latin America and the Caribbean
49.6
Middle East and North Africa
42.0
South Asia
64.0
Sub-Saharan Africa
62.3
Total
62.8
Total excl. PRC
54.6
Number of Relative Poor (in millions)
East Asia and Pacific
1,143.4
PRC
847.0
Eastern Europe and Central Asia
94.7
Latin America and the Caribbean
180.7
Middle East and North Africa
72.4
South Asia
594.7
Sub-Saharan Africa
248.1
Total
2,333.9
Total excl. PRC
1,486.9
Region
1996
% of Population below Relative Poverty Line
East Asia and Pacific
51.9
PRC
51.2
Eastern Europe and Central Asia
34.0
Latin America and the Caribbean
49.9
Middle East and North Africa
38.4
South Asia
58.0
Sub-Saharan Africa
66.6
Total
52.7
Total excl. PRC
53.2
Number of Relative Poor (in millions)
East Asia and Pacific
925.2
PRC
623.1
Eastern Europe and Central Asia
159.7
Latin America and the Caribbean
241.1
Middle East and North Africa
98.7
South Asia
752.0
Sub-Saharan Africa
400.6
Total
2,577.1
Total excl. PRC
1,953.9

1984

1987

1990

70.0
72.6
21.4
50.3
41.2
61.6
64.3
58.5
53.4

60.4
59.0
21.5
46.9
40.7
60.9
64.2
54.4
52.8

63.6
65.2
25.4
46.8
39.3
60.3
65.1
56.0
52.9

60.1
61.1
32.2
50.0
38.6
58.9
66.9
55.7
53.8

1,044.1
753.1
94.7
195.4
77.8
615.7
278.6
2,306.1
1,553.0
1999

946.2
639.8
97.5
193.6
83.9
653.4
302.8
2,277.3
1,637.5
2002

1,047.0
739.9
118.0
204.3
88.5
691.6
333.6
2,483.0
1,743.1
2005

1,031.8
720.0
150.8
230.0
93.3
720.3
371.4
2,597.6
1,877.6
2008

52.1
51.3
32.2
51.1
38.8
56.9
66.5
52.6
53.0

48.8
48.4
30.4
51.2
37.9
56.8
65.3
51.2
52.0

43.4
41.9
29.3
47.9
36.6
55.1
63.6
48.2
50.2

42.4
41.0
28.2
45.9
35.0
53.5
61.1
46.9
48.6

959.2
642.7
151.3
259.1
105.6
781.6
431.9
2,688.7
2,046.1

924.4
620.0
143.1
270.9
109.0
821.4
457.9
2,726.6
2,106.6

841.7
546.5
138.0
263.4
111.1
836.6
479.9
2,671.0
2,124.5

840.4
542.6
133.6
261.5
111.9
849.4
496.4
2,692.9
2,150.3

Source: Reproduced from Table 5 from Ravallion and Chen (2013).

15

1993