ArticlePDF Available

Identification of the Poor: Errors of Exclusion and Inclusion

Authors:
  • Centre For Multi-Disciplinary Development Research

Abstract and Figures

for the comments made at the seminars. Our sincere gratitude to P M Kulkarni and Biswaroop Das for their comments on an earlier draft of this paper. The usual disclaimers, however, apply. Motilal Mahamallik (moti_m13@rediffmail.com) and Gagan Bihari Sahu (gaganbs09@gmail.com) are with the Institute of Development Studies, Jaipur, and the Centre for Microfi nance Research, Bankers Institute of Rural Development, Lucknow, respectively. With the help of the 2004-05 National Sample Survey Organisation unit level consumption expenditure data, this paper tries to estimate the extent of inclusion and exclusion errors in the identification of below-the-poverty line households. In spite of continuous efforts towards improving the methodology of the BPL census, a significant difference between the "estimated" and the "identified" poor still persists. Against this backdrop, this paper attempts to develop an alternative method based on "vulnerability criteria" for the identification of the poor. Estimation shows that the prescribed criteria not only reduce the exclusion error significantly, but also suggest inclusion of a larger number of vulnerable households in the BPL list. Because of the data limitations, this exercise remains tentative. Nevertheless, it sets the possibility of further exploration to address the issue of large-scale errors.
Content may be subject to copyright.
SPECIAL ARTICLE
Economic & Political Weekly EPW february 26, 2011 vol xlvI no 9 71
This is a revised version of the paper which was earlier presented at t wo
seminars, “Social Exclusion in Contemporary India” and “Institutional
Aspects of Pro-Poor Policy: Revising the Indian Povert y Line” organised
by the Institute of Development Studies, Jaipur and t he Institute for
Social and Economic Change, Bangalore, respectively. We are grateful to
Suresh D Tendulkar, V M Rao, Gopal K Kadekodi, D Rajasekhar, Madura
Swaminathan, Manoj Panda, M R Narayan, Hima nshu, B P Vani and
K Navaneetham for the comments made at the seminars. Our sincere
gratitude to P M Kulkarni and Biswaroop Das for their comments on an
earlier draft of this paper. The usual disclaimers, however, apply.
Motilal Mahamallik (moti_m13@rediffmail.com) and Gagan Bihari Sa hu
(gaganbs09@gmail.com) are with the Institute of Development Studies,
Jaipur, and the Centre for Microfi nance Research, Bankers Institute of
Rural Development, Lucknow, respectively.
Identification of the Poor: Errors of Exclusion
and Inclusion
Motilal Mahamallik, Gagan Bihari Sahu
With the help of the 2004-05 National Sample Survey
Organisation unit level consumption expenditure data,
this paper tries to estimate the extent of inclusion and
exclusion errors in the identification of below-the-
poverty line households. In spite of continuous effort s
towards improving the methodology of the BPL census, a
significant difference between the “estimated” and the
“identified” poor still persists. Against this backdrop, this
paper attempt s to develop an alternative method base d
on “vulnerability criteria” for the identification of the
poor. Estimation shows that the prescribed criteria not
only reduce the exclusion error significantly, but also
suggest inclusion of a larger number of vulnerable
households in the BPL list. Because of the data limitations,
this exerci se remai ns tent ative. N evert heless , it se ts the
possibility of further exploration to address the issue of
large-scale error s.
Identifi cation of poor has been an important issue in India
since the 1980s, particularly within the context of ensuring
thembasic needs through a series of social assistance
schemes. The issue has drawn considerable attention because of
weak methodology and implementation failure, often leading to
wrong targeting. To provide minimum protection to the poor
through social assistance schemes, their “proper identi cation”
is necessary.
So far, criteria adopted for identifying the poor are non-
transparent, cumbersome and often non-verifi able (Alkire and
Seth 2008; Sundaram 2003). Besides, political interests and
power equations often in uence the identi cation process. The
vulnerable, powerless and forbidden households, instead of
being grouped under the “poor”, get shifted to the non-poor
categor y ( Hir way 2003). Irr espec tive of ado pt ing differ ent met h-
ods and parameters for identifying the poor in the last three sur-
veys (1992, 1997 and 2002), errors of exclusion and inclusion re-
main signifi cant. Ram et al (2009) estimated that 60% of the
households at the all-India level in the abject deprivation group
do not have a below-the-poverty line (BPL) card.1 Although a
positive association between povert y and landless-cum-near
landless, agricultural labour, scheduled castes (SCs) a nd s che d-
uled tribes (STs) households is well-documented in the develop-
ment l iterature (T horat and Ma hamall ik 2006), hous ehol ds from
such groups are found to be excluded from the BPL list (Swami-
nathan 2008).
Data from the 61st round of the National Sample Survey (NSS)
reveal that 15% of the non-poor households in the richest quartile
and 23.5% in the next quartile in rural areas possess either an
Antyodaya Anna Yojana (AAY) or BPL card. On the other hand,
51.4% households in the poorest quartile and 58.4% in the next
quartile do not possess either an AAY or BPL card. The former can
be termed as a Type-I error (i e, error of inclusion) and the latter
as a Type-II error (i e, error of exclusion). The fi rst includes the
non-poor in the poor category and the second shifts the poor to
the non-poor category.
The issue of “identi cation of poor” is becoming important due
to two reasons. First, a signifi cant proportion of households esti-
mated as poor by offi cial poverty estimates of Planning Commis-
sion is not identifi ed as poor as per the methods adopted by the
Ministry of Rural Development (MoRD). As a result, a signifi cant
difference has been observed between the “estimated” and the
“identifi ed” poor. Second, the issue remains unresolved even at
the methodological level as the criteria adopted are not directly
verifi able. So far, no satisfactor y methodology has been advanced
SPECIAL ARTICLE
february 26, 2011 vol xlvI no 9 EPW Economic & Political Weekly
72
to ensure inclusion of a majority of the poor. Consequently, many
poor households often do not hold either a BPL or an AAY card,
and, hence, remai n deprived of the benefi ts associated with such
ca rds. As s tat ed b y an exper t group, mos t po or a re of ten exc luded
from the BPL survey list because of their geographical isolation
and very marginal position in the social, economic and political
spheres. The prevalent view is that the exclusion error is a direct
function of the weak bargaining power of the poor as a collective
entity in the Indian democracy.
Er rors in ident ifyin g p oor g ive an impr essi on of a wea k coo rdi-
nation between the state and civil societ y at planning, implemen-
tation and monitoring process. Theoretically, the MoRD based on
some methodological ground should not exclude a higher propor-
tion of the estimated poor by the Planning Commission. This
raises a pertinent question as to how to minimise the gap be-
tween the numbers of “estimated” and “identifi ed” poor. It is
against this backdrop that this paper proposes an alternative
methodology for identifying the poor to minimise the level of
error of exclusion. It addresses these issues based on the 1997 and
2002 BPL survey2 and the unit level consumption expenditure
survey data (2004-05) of the NSS, and reviews methods imple-
mented and suggested for identi cation of the poor. It analyses
the extent of Type-I and Type-II errors based on criteria adopted
by the MoRD and Planning Commission, and suggests an alter-
native methodology for identifi cation of poor in rural areas. The
estimated outcome of the methodology is given in Section 4
followed by conclusions.
1 Why Identif ication of the Poo r Was So Poor?
In 1992, the MoRD used respondents’ self-reported income as t he
main parameter to identify poor households. However, the
underestimation of income by households has included more
than the expected number of poor in the list. As a result, the
generated BPL list looks bulky with a mix of poor and non-poor.
To overcome this, the 1997 BPL census used food expenditure
rather than income fi gure, in addition to “exclusion criteria”,3
and excluded the visibly non-poor. Subsequently, data on total
consumption expenditure (purchased from the market and home
grown) were collected by interview method from the remaining
families and the per capita consumption expenditure estimated
for each family treating all members as identical units. The per
capita consumption expenditure was then compared with the
state povert y line (estimated by the Planning Commission) and
the family counted in the BPL group if its per capita consumption
was within the norm set by the Planning Commission. However,
the exclusion criteria were too stringent. Poor families were
excluded, poverty lines were not available for all states and they
were not even uniform across states and district territories
(Mehrotra and Mander 2009; Alkire and Seth 2008). Besides,
there was no scope of allowing new households to be declared
poor in the interim period before the next survey was instituted
(Sun da ram 2003).
Considering the above limitations, the 2002 BPL census shifted
from exclusion criteria to a 13 criteria method based on socio-
economic indicators refl ecting the quality of life in rural areas.4
Each household assigned a score of 0-4 (based on their access or
ownership in an ascending order) for each of the 13 indicators,
depending on their response to the question. The scores of the ith
household on all these parameters were then summed to create
an aggregate score Si. Hence, the aggregate score of a household
was ranging from a minimum of zero to a maximum of 52 (sym-
bolically, 0 S
i 52). Finally, a household was categorised as
poor or non-poor based on the “cut-off” score decided by the
region and states. These cut-off score could vary across states
since the state governments were asked to restrict the percentage
of poor households equivalent to the estimated poor by t he Plan-
ning Commission for 1999-2000 with a plus/minus 10% margin.
In other words, the states were given the fl exibility of 10% mar-
gin to account for the transitory poor. According to the 2002 BPL
census, thus, household i is cons ider ed to be BPL if:
Si = Hij Sp
cut-off
whe re i = 1………….to n and j = 1 ………….13,
Si = aggregate score of the ith household,
Hij = the ith household on jth indicator, and
Sp
cut-off = State specifi c cut-off score.
Some state government raised objections against the ceiling
on the number of BPL households to be ident i ed with the appre-
hension that it may suppress the number of actual poor, and in
turn, reduce the fl ow of funds from the centre. T he forceful impo-
sition of ceiling can reduce the gap between volume of identifi ed
and estimated poor at the cost of exclusion of many actual poor
as evident from the signifi cant level of Type-I and Type-II errors.
The methodology to identif y the poor based on 13-point crite-
ria of 2002 faced severe criticism even before its implementation
(Sundaram 2003; Hirway 2003; Alkire and Seth 2008; Mehrotra
and Mander 2009; Himanshu 2008). The key criticisms relate to
(1) lack of clarity in the criteria, (2) methodological drawbacks in
scoring and aggregation, (3) data quality and corruption, and (4)
increasing probability of wrong selection. This paper primarily
focuses on the second criticism.
(1) Since the distance between response categories within each
dimension is not necessarily equal, treating t he ordinal responses
(0-4) as cardinal is misleading.
(2) As a one-point gain in one dimension can be compensated by
an equivalent decrease, in another dimension it would make it
irrelevant. For instance, the situation of a family eating only once a
day gets nullifi ed if it has quite a few items of clothing or is doing
well across any other dimension not as serious as not getting food.
(3) Equal weights of dimensions can be treated as a poor descrip-
tion of poverty. For instance, not having one square meal a day
throughout the year is treated equivalent to open defecation or
not possessing electrical appliances.
(4) No national poverty line is set. In practice, almost all states
and in some cases districts set their own poverty line across the
52-points scale, such that the number of BPL households is equiv-
alent or 10% more than the proportion declared by Planning
Commission for 1999-2000. Thus, households not declared as
BPL in their states might be considered as BPL if living i n a neigh-
bouring state.
(5) Though a cap in the statesBPL estimate to not exceed 10%
from the NSSO estimates of 1999-2000 was imposed for fi scal
reasons, it has been widely disputed across states.
SPECIAL ARTICLE
Economic & Political Weekly EPW february 26, 2011 vol xlvI no 9 73
(6) The poor often has no access to credit market because of their
inability to offer any acceptable collateral. But the highest score
of “4” has been assigned to the household who is not indebted.
Thus, the score attached to “ty pe of indebtedness” might have
ruled the poor out of the BPL category.
(7) With uncertain future, “preference of assistance” is meaningless.
People might have given wrong answers to get a favourable score.
2 Extent o f Type- I and Type-II Errors
Identifi cation of poor and distribution of AAY and BPL cards have
a great relevance at the policy level because once identifi ed as
BPL, households become eligible to obtain benefi ts from various
social assistance programmes implemented by the central and
the state governments. The 61st round of consumption expendi-
ture survey by NSS (2004-05) gives an insight into the magnitude
and nature of exclusion and inclusion faced by the households in
terms of availing these cards. The 2004-05 consumption expend-
iture sur vey has information on the possession of types of cards
by the households along with other related variables on socio-
economic and consumption expenditure. It is assumed that,
households possessing either BPL or AAY cards are identifi ed as
poor by the MoRD, whereas households reporting spending of less
than the poverty line xed by the Planning Commission for spe-
cifi c states and sectors are termed as poor. There is, thus, a differ-
ence between these two agencies on estimation and identi ca-
tion of poor. This discrepancy generates t wo unwarranted errors
which need to be addressed properly. At an all-India level, 70.5%
of rural households possessed no card or had an APL card, were
identifi ed as non-poor by the MoRD ( Table 1). N otab ly, o nly 39.6 %
of the rural households estimated as poor through the offi cial
poverty estimation method of the Planning Commission pos-
sessed either a BPL or an AAY card. This means that 60.4% of ru-
ral households, who were poor on the basis of consumption
expenditure, were not identifi ed by the MoRD as poor. In other
words, the magnitude of Type-II error is 60.4%. It was also esti-
mated that 26.3% of rural households belonging to non-poor
category as per the consumption expenditure method, were iden-
tifi e d as p oor b y t he MoRD. This implies that the volume of Type-I
error is 26.3%.
It appears that 29.5% households in rural India possessed
either a BPL or an AAY card which was 5% more than the Planning
Commission estimates of poverty in 2004-05. However, this was
not true for all states. In poorer states like Bihar, Uttar Pradesh,
Orissa, Jharkhand, Assam and Uttarakhand, the estimated
number of poor households was more than that of households
possessing card. For instance, in Bihar only 17.4% of households
had either BPL or AAY card whereas the estimated poor household
was 38.1%. The corresponding fi gures were 12.3% and 19.8% in
Assam, 16.4% and 28.6% in Uttar Pradesh, 25.8% and 40.8% in
Jharkhand, 25.7% and 35.7% in Uttarak hand and 44.4% and 45%
in Orissa. However, for states like Andhra Pradesh, Karnataka,
Gujarat, Kerala, Maharashtra and Himachal Pradesh, proportion
of households having either BPL or AAY card was signifi cantly
mor e c ompa red to the e st imated pe rcentage of po or households.
The consumption expenditure data based on 61st round of NSS
estimate that a signifi cant proportion of households falling below
the offi cial poverty line did not possess either a BPL or an AAY card
across states (Table 6, p 76). The highest proportion of such
households was 79.2%, in Punjab and the lowest fi gure of 28.4%
in Karnataka. However, in poorer states like Bihar, Uttar Pradesh,
Jharkhand, Rajasthan, Uttarakhand, West Bengal, Chhattisgarh
and Madhya Pradesh, the proportion of excluded poor households
varied from 77.6% to 51.2%. Thus, the degree of Type-II error
(those below the offi cial poverty line, but excluded) was quite
prominent in these poorer states. Data given in Table 6 also show
that the magnitude of Type-I error (inclusion of consumption non-
poor) varied from 55.9% to 9.3% across states. Notably, in case of
Andhra Pradesh and Karnataka, errors of inclusion of non-poor
households were larger than errors of exclusion of poor.
Figure 1 indicates variations in the share of total cards (BPL
and AAY) among poor and non-poor. Of the total card distributed,
only 32.2% was allotted to the “consumption poor” and the re-
maining 67.8% to the “consumption non-poor” households.
Around 68.8% of total BPL card was distributed among the non-
poor households, while the share of poor households was only
31.2%. More signi cantly, in case of the AAY, initiated to provide
foodgrains for the poorest among the BPL categor y households at
super subsidised prices, 58.2% of cards were distributed to non-
poor. Based on the information in Figure 1, it can be argued that
the distribution of BPL and AAY cards has gone in favour of the
no n-p oor c ompa re d to poo r ho useho lds . In this c ont ex t, R am et a l
(2009) are of the view that as the process of identifi cation as well
as distribution of BPL and AAY cards is often infl uenced by politi-
cally affl uent persons, it is the non-poor who benefi t more, irre-
spective of methods adopted in identifying the poor. Hirway
(2003) and Khera (2008) mention that outright corruption
ensures names of non-poor villagers in the BPL list.
3 Proposed Methodology
In order to address the methodological weaknesses in identif ying
the BPL households, this paper explores “vulnerability criteria”
(hereafter VC) approach, using the consumer expenditure survey
of NSS (61st round).5 This survey has information on the pos session
Table 1: Status o f Househol ds and Thei r Access to BPL C ard
Type of Ho use hol ds No of Hou seh old s Hav ing Tota l Num ber of
Either AAY or B PL Card Other or N o Cards Househo lds
Poor 1,42,42,308 (39.6) 2,17,02,062 (60.4) 3,59,44,370
Non-poor 2,99,56,449 (26.3) 8,40,70,915 (73.7) 11,40,27,364
Total 4,41,98,757 (29.5) 10,57,72,977 (70.5) 14,99,71,734
(1) Figures are estimated using weight and state specific poverty line; (2) Figures in the parenthesis
are percentage to respective total; (3) Missing numbers are excluded.
Source: Es timated f rom unit leve l record, Cons umption Ex penditur e Survey, NSSO, 20 04-05.
Figure 1: Distribution of BPL and AAY Cards among Poor and Non-Poor Households (in %)
70
60
50
40
30
20
10
0
Households Having
AAY Card
Households Having
BPL Card
Households Having
Either AAY or BPL Card
Poor Househol ds
Non-Poor
Households
41.8
31.2 32.2
Source: Sa me as Table 1.
67.8
68.8
58.2
SPECIAL ARTICLE
february 26, 2011 vol xlvI no 9 EPW Economic & Political Weekly
74
of various cards along with other socio-economic characteristics
of households such as landownership, occupation, social group,
demographic, education, etc. A household is defi ned as vulnera-
ble if it bears at least one of the following criteria: (1) households
do not own a dwelling unit; (2) households that do not own any
land and are not self-employed in non-agriculture and whose no
member is a regular salary earner; (3) where members of the
households primarily work as agricultural and other labour
having only homestead land with no regular salary earner;
(4) households that hold less than or equal to 2 ha of standard-
ised cultivable land6 with no regular salary earner and primarily
engaged in agricultural and other labour activities; (5) house-
holds belonging to scheduled caste and scheduled tribe;7 and
(6) households which spend less than Rs 216.29 per capita on
clothing during a year.
As a fi rst step, we assign a value of “one” when a household had
any of these possession and “zero” in an otherwise situation to
prevent complete substitutabilit y across the set of these criteria.
In the second step, the score of the ith household in all the six di-
mensions are added to arrive at an aggregate score. The aggre-
gate score expresses the extent of a household’s vulnerability by
the number of dimensions. The BPL survey by MoRD did not reveal
this because the same aggregate score could be arrived from any
combination of these dimensions. This is due to the fact that the
score for each dimension is not binary in nature.8 Most signifi -
cantly, an “union approach” has been applied to identify the vul-
nerable households, which means that a household is vulnerable if
exposed to at least one dimension. On the other hand, a non-vul-
nerable household does not score any value. What appears from
the above discussion is that the identifi cation of households for the
purpose of social support under the VC method is likely to include
both consumption poor as well as non-poor. Elsewhere, it is also
argued that the selection of BPL households should not necessarily
be based on a poverty line alone (Dreze and Khera 2010).
There is every possibility that few households, who deserve to
be included in the BPL list, might have left out by VC method on
account of data limitation. As a matter of moral concern, at
least households with (1) single woman member (in most cases
this would be either widow, divorced/separated, or unmarried),
(2) disabled person as a single breadwinner from casual works,
(3) no member above the age of 14 years age, (4) any member
works as a bonded labour, and (5) destitute characteristics and
households bearing similar attribute should be compulsorily
included in the BPL list.
4 Outcome of Proposed Methodology
While acknowledging the multidimensional nature of poverty,
households coming under VC should be identifi ed as poor by the
MoRD. Table 2 reports the extent of vulnerable households and the
coverage of poor and non-poor households, based on the con-
sumption expenditure within the VC. Evidently, 26.8% of house-
holds who are consumption poor can be captured if 3rd criterion is
adopted. The compulsory inclusion of SC and ST households,
irrespective of their economic status, shows inclusion of 46.5%
consumption poor into the BPL list.9 Similarly, 28.7% of con sum-
ption poor households can be included in the BPL list on the basis
of per capita expenditure on clothing per annum (6th criterion).
Signifi cantly, through the 1st, 2nd, and 4th criteria, 3.8%, 4.6% and
11.4% poor households can be included in the BPL list, respectively.
The proposed methodology identifi es 51.8% of total number
of households as vulnerable including 44.9% consumption
non-poor. Apparently, our estimate on vulnerable household
corroborates with estimation of poor by Saxena Committee
(2009) based on calorie intake. Of the total consumption poor
cardholders, the share of vulnerable poor constitutes 80.22%
whereas 62.6% of the total consumption non-poor cardholders
are found to be vulnerable (Table 2). A ll together, 68.3% of the
total cardholders are covered under VC. This suggests that, the
proposed methodology, based on VC covers a signi cant share of
identifi ed poor households.
The vulnerability of households by number of dimensions is
presented in Table 3. The second and third rows show the per-
centage of consumption poor as well as non-poor who can be
included in the list of MoRD as poo r throu gh r espec ti ve n umb er of
dimensions. For example, by following a single dimension (any
one) of the VC, it is possible to include 36.7% of consumption poor
households in the BPL list. Similarly, 26.8% of the consumption
poor households can be included in the BPL list by applying any
two dimensions of the VC. Apparently, only 26.2% of rural poor
households are not being captured by the VC. This can be termed
as “estimated Type-II error”.
By using “union approach”, the proposed methodology identi-
es 73.8% of consumption poor households for the entitlement
of either AAY or BPL card. Contrary to this, only 39.6% of
Tabl e 2: E xte nt o f Cov era ge of Poor Hou seh old s un der V uln era ble C rit eri a
Sl No Pa ram ete rs B ase d on Vu lne rab ili ty C rite ria Ide nti fie d Vul ner abl e
of the Households (in %)
Criteria Poor Non-Poor Total
1st House holds not having ow n dwelling unit (V ULN-1) 3.8 3.1 3.2
2nd Househ olds do not own any lan d and not self-emp loyed
in no n-agricultu re and no member is a re gular salary
earner (VULN-2) 4.6 3.6 3.9
3rd Agricultural or other labour households having
hom estead land only a nd no member is a regul ar
salary earner (VULN-3) 26.8 15.2 18.0
4th Households having 2 ha o f sta ndar dise cult iva ble
land and primarily engaged in agricultural or other
l abou r ac tivi ty w ith n o reg ular s ala ry ea rne r (VU LN- 4) 11.4 5 .9 7.2
5th Schedu led caste and sc heduled trib e
households (VULN-5) 46.5 27.9 32.3
6th Hou seho lds s pend ing l ess t han R s 216. 29 pe r cap ita
pe r year on cloth* (VULN -6) 28.7 6.9 12.1
All Total numb er of househol ds having at least on e or
more parameters 73.7 44.9 51.8
Perc enta ge of vuln erab le p oor h ouse hold s hav ing e ithe r
AAY or BPL card out of t otal AAY or BPL cardhol der
who are consumpt ion poor 80.22
Percentage of vulnerable non-poor households having
either AAY or BPL ca rd out of total AAY or BPL c ardholder
who are consump tion non-poo r 62.6
* The mean ex penditur e on clothing (a scending or der) of the fi rst quar ter of the samp le
households.
Source: Sa me as Table 1.
Table 3: Indi cators of t he Chosen Di mension s under Vuln erable Cr iteria
Number of Dimension 0 1 2 3 4
Poor househ old under numbe r of dimensions (%) 26.2 36.7 26 .8 9.3 0.9
Non-po or household un der number of dimen sions (%) 55.1 30.1 12.1 2.5 0. 2
Sou rces : Sa me as Ta ble 1.
SPECIAL ARTICLE
Economic & Political Weekly EPW february 26, 2011 vol xlvI no 9 75
con sumption poor households are distributed BPL/AAY card by
MoRD (Table 1). Since the proposed methodology able to identify
a signifi cant proportion of estimated poor households compared
to the BPL census, if employs it can reduce the extent of errors
of exclusion.
The data in Tables 2 and 3 indicate a close correspondence be-
tween the VC and household status (poor or non-poor). To con-
rm the magnitude of each indicator of the VC o n th e st atus o f the
househol d, a multiv ariate logis tic re gre ssio n a nalysi s w as car rie d
out. The general model is a binary choice model involving esti-
mation of the probability if “a household is poor or not” as a func-
tion of a vector of explanatory variables included in the VC. If P is
the probability of a household being poor, then
P = [1+e{–βX}]–1
where “
” is a vector of the unknown coef cients and “X” is a
vector of covariates that affect the probability of household
being poor. Thus, the general logistic model can further be
expressed as
k
j
j
i
i
e
X
X
P
P
0
1
log
Xij
The above express the log odds of a household being poor as a
linear function of the explanatory variables. We can interpret the
odds ratio [Exp ()] in terms of the change in odds, i e, if the
value is greater than 1, it indicates that as the predictor increases,
the odds of the outcome occurring increase. Conversely, a value
less than 1 indicates that as the predictor increases, the odds of
the outcome occurring decrease. The estimated results are out-
lined in Table 4. Notably, all parameters refl ect an expected
association and are statistically signifi cant.
The above analysis reveals
that there is a positive relation-
ship between households hav-
ing no dwelling unit and their
being poor. For instance, ceteris
paribus, the probability of such
households being poor is 1.31
times higher than those house-
holds having dwelling unit. As
expected, the landless house-
holds who are neither self-
employed in non-farm sector
and nor having any member as a
regular salary earner, the prob-
ability of their being poor is 1.24
times more compared to other households. Similarly, an agricul-
tural and other labour households having only homestead land
with no reg ula r s alary e arner i s m or e like ly to be po or, as evident
from the positive and statistically signifi cant coeffi cient of
VULN-3. Due to limited control over productive resources and
other social constraints, households from SC and ST category are
more likely to be p oor. Apparent ly, as capac it y to spend on clot h-
ing increases, the probability of being poor decreases. Generally
speaking, the results are consistent with theoretical expectation
and draw a plausible picture of the household being poor or not.
The signifi cant chi-square clearly shows that the estimated model
has a good fi t.
Notably, the “estimated Type-II error” (poor being excluded by
VC) is 26.2% (Table 3). The above fi ndings raise a pertinent ques-
tion, who are these excluded poor under VC? It is observed that
23.1% and 48.7% of such excluded poor households are self-
employed in agriculture and non-agriculture, respectively. This
su gge sts t hat suc h ho use holds s eem to h ave su ppr ess ed t heir c on-
sumption expenditure perhaps to enrol themselves in the BPL list.
Incidentally, out of total number of poor who are excluded under
VC, 70.2% of them are also unidentifi ed by the MoRD. A compari-
son between MoRD and VC suggests that, one unit increase in
consumption expenditure leads to more exclusion by the latter
criteria than the former (Figure 2).
Figure 3 shows the distribution of non-poor households having
access to BPL or AAY cards across the level of consumption
expenditure. The estimation
shows that a relatively higher
proportion of non-poor card-
holders are found at lower
expenditure level (within the
range of Rs 270.01-Rs 510)
under VC compared to the
MoRD method. Whereas, the
latter method includes a
higher percentage of non-
poor cardholders compared
to the former at higher
expenditure level (above
Rs 510.01). It implies that the
VC encourage inclusion of
less better-off and exclusion
of better-off non-poor house-
holds compared to MoRD.
The data in Table 5 de-
scribes three different groups of non-poor, viz, (1) already enlisted
under BPL census of MoRD; (2) as recommended by VC approach
Table 4: Resu lts of Logi stic Regre ssion
on Consumption Poor
Dependent Var iable: whether t he
househol d is poor (1 = Yes, 0 = No)
Variable s β coeff icient Exp (β)
Constant 1.66* (.05) 5.2 3
VUL N-1 .27* (.06) 1.31
VULN-2 .22* (.06) 1.24
VULN-3 .52* (.03) 1.67
VULN-4 .65* (.04) 1.92
VULN-5 .33* (.02) 1.39
VULN-6 -.01* (.00) 0.9 9
-2 Log likelih ood = 5948 6.33
R2 = .18 (Cox & Snell), .29 (Nag elkerke)
Model χ2(6) = 15256.538, Number of
observ ati ons = 78639 (un weig hte d samp le)
(1) * Significa nt at 1 % level. (2) Fig ures in
parenthe ses indica te Std Error.
1
3
5
7
9
11
13
15
17
270.01 -
320
320.01 -
365
365.01 -
410
410.01 -
455
455.01 -
510
510.01 -
580
580.01 -
690
690.01 -
890
890.01 -
1155
1155.01
and above
Level of Consumption (Rs)
Sou rces : Sam e as Ta ble 1 .
Figure 3: Distribution of Non-poor BPL or AAY Cardholders (in %)
Vulnerable Criteria
MoRD
Table 5: Share o f Non-Poo r Out of Total AAY
or BPL Card holders a cross Met hods
Level of Non-Po or Out of Total A AY or
Consumpt ion BPL Card Di stribut ed (%)
(Rs) MoRD Vulnerable Non-
Criteria Vulnerable
270.01 - 320 8.4 6.1 2.4
320.01 - 365 44.8 31.3 13.5
365.01 - 410 89.6 61.5 28.0
410.01 - 455 97.8 67.0 30 .8
455.01 - 510 99.7 62.9 36 .9
510.01 - 580 100 63.0 37.0
580.01 - 690 100 60.0 40.0
690.01 - 890 100 53.8 46. 2
890.01 - 1155 100 48.3 51.7
1155 .01 an d
above 100 45.7 54.3
Total hous eholds 67.8 42.4 25.4
Mean MP CE 58 4.4 555 633
(405.2) (347.4) (482.9)
Figure in parenthesis stands for Standard Deviation.
MPCF: Mont hly Per Capit a Expendi ure.
Sou rces : Sa me as Ta ble 1.
Figure 2: Proportion of Excluded Poor Based on MoRD and Vulnerable Criteria (in %)
51.1
55.8
68.0
y = 3.492x + 49.10
y = 6.690x + 6.243
0 -235
235.01 - 270
270.01 - 320
320.01 - 365
365.01 - 410
410.01 - 455
455.01 - 510
80
60
40
20
0
Linear (Percentage of Excluded Poor-MORD)
Linear (Perc entage of Lef t out Poor-Vul nerable Cr iteria)
Level of Consumption (Rs)
Sou rces : Sam e as Ta ble 1 .
y=6.6908x + 6.2433
y=3.4929x + 49.109
SPECIAL ARTICLE
february 26, 2011 vol xlvI no 9 EPW Economic & Political Weekly
76
for their inclusion in the BPL list; and (3) as suggested by VC ap-
proach for their exclusion from the BPL list. It is found that, the
percentage of non-poor cardholders increases with the increas-
ing MPCE class under MoRD method. Apparently, all households
who reported having more than Rs 455 MPCE, possessed either
BPL or AAY card. On the other hand, except the fi rst four slabs,
percentage of non-poor cardholders (out of total BPL and AAY
card distributed) decreases with the increasing MPCE class under
VC. This implies that the VC method excludes the non-poor in-
creasingly as the level of consumption expenditure goes up.
The data given in Table 5 also show that 67.8% of total BPL and
AAY cards are allotted by the MoRD to the consumption non-poor
households. Had VC been adopted while identifying poor house-
holds, such a fi gure would have come down to 42.4%. It implies
that the remaining 25.4% card is allotted to the non-vulnerable –
non-poor households, which constitutes 37.4% of Type-I error.
In the context of existing Type-I and Type-II errors, there are
certain justifi ed reasons to accept the inclusion and exclusion of
some non-poor households as suggested by the VC. First, such
method advocates for the inclusion of those non-poor households
who come under lower MPCE class and ask to exclude those
households who fall under higher MPCE class. Second, of the con-
sumption, non-poor households coming under the VC, 71.5% of
them primarily lead their livelihood from agricultural and other
labour activities. Third, the method strongly recommends to
exclude self-employed households, who fall under relatively
higher MPCE class. This suggests that the VC approach identifi es
relatively less well-off non-poor for their inclusion in the BPL
category against the counterpart MoRD methodology.
4.1 State Leve l Analysis
The extent of errors in identif ying the poor across states is
reported in Table 6. Following observations can be made from
this table. First, extent of Type-II error under MoRD method is
quite high and varies across states. For instance, the level of such
error is 79.2%, in Punjab, the highest, and 28.4% in Karnataka,
the lowest. Second, the magnitude of Type-II error is higher
compared to Type-I error in all states except Karnataka and
Andhra Pradesh under MoRD methodo logy. Third, for all the
states and union territories except Uttarakhand and Jammu and
Kashmir, the magnitude of Type-II error is less compared to
Type-I error under the VC method. Fourth, the extent of Type-II
error is less in VC against MoRD method. It implies that more
percentage of consumption poor households are being deprived
of from getting enrolled them into the BPL list under the latter
methodology compared to the former. Fifth, as the VC approach
suggests inclusion of consumption poor as well as border line
consumption non-poor, the level of Type-I error looks high.
As the VC propose 51.8% of rural households to be included in
the BPL list, it would be interesting to see how many of them are
Table 6: Exte nt of Error s at the State Leve l (%)
Major State s Type-II Er ror Type-I Err or
MoRD Vulnerable Criteria MoRD Vulnerable Criteria
Andhra Pradesh 38.5 18.0 55.9 51.8
Assam 75.5 33.0 9.3 43.6
Bihar 77.6 28.8 14.4 33.4
Gujarat 46.6 22.8 34.1 47.5
Haryana 65.4 25.2 16.6 35.7
Himachal Pradesh 51.5 27.4 14.1 38.8
Jammu and Kas hmir 41.8 52.4 22.0 26 .6
Karnataka 28.4 31.8 47.5 44.6
Kerala 53.3 22.8 27.5 41.6
Madhya Pradesh 51.2 20.7 26.8 47.6
Maharashtra 48.2 31.2 29.2 39.2
Oriss a 42.1 17.4 33.4 48.0
Punjab 79.2 9.5 11.3 51.1
Rajastha n 67.2 26.9 15.9 39.8
Tamil Nadu 70.8 18.7 18.1 53.6
Uttar Pr adesh 74.6 34.8 12.8 36.6
West Bengal 57.7 19.5 26.7 54.7
Chhattisgarh 51.4 20.5 34.1 6 0.7
Jharkhand 67.2 27.8 20.9 50.9
Uttarakhand 58.1 44.4 16.7 32.3
Other st ates and UTs 35.5 15.6 20.5 67.3
Total 60.4 26.3 26.3 44.9
Sou rces : Sa me as Ta ble 1.
SAMEEKSHA TRUST BOOKS
1857
Essays from Economic and Political Weekly
A compilation of essays that were f irst published in the EPW in a special issue in May 2007. Held together with an introduction by Sekhar Bandyopadhyay,
the essays – that range in theme and subject from historiography and military engagements, to the dalit viranganas idealised in traditional songs and the
“unconventional protagonists” in mutiny novels – converge on one common goal: to enrich the existing national debates on the 1857 Uprising.
The volume has 18 essays by well known historians who include Biswamoy Pati, Dipesh Chakrabarty, Peter Robb and Michael Fisher. The articles are grouped
under five sections: ‘Then and Now’, ‘Sepoys and Soldiers, ‘The Margins’, ‘Fictional Representations’ and ‘The Arts and 1857’.
Pp viii + 364 2008 Rs 295
Available from
Orient Blackswan Pvt Ltd
Mumbai Chennai New Delhi Kolkata Bangalore Bhubaneshwar Ernakulam Guwahati Jaipur Lucknow Patna Chandigarh Hyderabad
Contact: info@orientblackswan.com
SPECIAL ARTICLE
Economic & Political Weekly EPW february 26, 2011 vol xlvI no 9 77
already enlisted in such list
by MoRD. Apparently, 38.8%
of the total v ulnerable house-
holds had already possessed
either BPL or AAY card. In
other words, 61.2% house-
holds who are vulnerable,
need to be included (Table 7).
Th e size of v ulnerable hous e-
holds who are not covered in
the BPL census varies across
states bet ween 83.2% and
33.4%. More importantly, such
gures are quite high even
in poorer states like Uttar
Pradesh, Bihar, Rajasthan,
Jharkhand, Uttarakhand, West
Bangal, Chhattisgarh and
Madhya Pradesh.
5 Conclusions
This paper explores the pos-
sibility of a simple method for
identifi cation of households
to declare them as eligible to
avail of the benefi ts from
various social assistance schemes. Using the limited database, the
study underlined the importance of revising the existing
methodology that would give a better coverage of the actual
poor, who have been left out under the prevalence method of
MoRD. The proposed VC method attaches binary values to each
parameter (having a value of 1 in case the household af rms the
concerned parameter, otherwise zero) and uses the union
approach to identify the vulnerable households. With this, the
study tries to overcome some of the criticisms of exist ing met ho-
dology as discussed in Section 2.
The estimation based on the proposed methodology not only
reduces the number of unwanted households in the BPL list, but
also advocates a larger coverage of the vulnerable poor. The
“universal approach” is a welcome step on the ground of pro-
viding food security for all. However, there is no compelling
reason to subsidised privileged households. As the present exer-
cise of identifi cation of households is to declare them eligible for
various social assistance schemes, this identifi cation goes beyond
food security. More importantly, it is infeasible as well as irra-
tional to ensure benefi ts to all rural households associated with
the BPL card.
Since the vulnerable non-poor are relatively less well-off
compared to the non-vulnerable – non-poor, the estimated error
of inclusion is justifi able. It is worth mentioning here that as
the VC app roa ch co ver s a s igni fi cant proportion of households liv-
ing BPL, it allows them to reduce the gap between the estimated
and identi ed poor. It also suggests the possibility of withdraw-
ing cards (BPL or AAY) from non-vulnerable non-poor households
for redistribution among the actual poor as identifi ed by the
VC approach.
Tabl e 7: Vu lne rab le H ous eho lds a nd T hei r
Exten t of Coverage und er BPL Score M ethod
Major State s Vulnerable Non-Vulnera ble
Househo lds Household Having
Not Having Card as a Per Ce nt
Either BPL or to Vulnerable
AAY Card Household
(%) Withou t Card
Andhra Pradesh 36.7 108.6
Assam 83.2 10.6
Bihar 73.6 13.7
Gujarat 49.4 41.7
Haryana 70.5 24.0
Himachal Pradesh 76.3 22.1
Jammu and Kas hmir 68.0 78.1
Karnataka 33.4 118.4
Kerala 56.0 37.5
Madhya Pradesh 56.6 27.1
Maharashtra 53.0 52.2
Orissa 48.4 37.8
Punjab 79.5 2.3
Rajasthan 69.8 15.7
Tami l Na du 7 6.2 13 .9
Uttar Pr adesh 75.2 15.6
West Be ngal 62.3 19.7
Chhattisgarh 57.3 27.1
Jharkhand 68.8 17.6
Uttarakhand 63.1 42.0
Other st ates and UTs 67.5 9.8
Tota l 61. 2 29.5
Sou rces : Sa me as Ta ble 1.
Notes
1 Abject deprivat ion is a situat ion where a hou se-
hold does not have any adult l iterate member,
lives in a k achh a house in a r ural a rea and in a
kachh a or semi-pucca hous e in an urban are a, has
no land in rural area and no toilet facility in an
urban area, does not own any d rink ing wate r
facility or consumer durables such as bicycle, tel-
evisio n or radio or has elec tricity in t he house. For
detaile d infor mation on t his, see Sriniv asan and
Mohant y (2002) and Ram et a l (2009).
2 The distribution of cards based on 2002 BPL cen-
sus was incomplete in some states by 2004-05 be -
cause of a stay order issued by the Supr eme Court
in May 2003. T heref ore, 2004- 05 cons umption
expenditure survey includes information on pos-
session of r ation cards based on 1997 as well as
2002 BPL cens us (for more det ails see, Drez e and
Khera 2010).
3 A s et of fi ve questions, viz, whether (1) operating
more tha n 2 ha of land; (2) havi ng a pucca house
as defi ne d in the population c ensus; (3) a ny resi-
dent member having annual income more than
Rs 20,000 from sala ry or self-employment;
(4) the household ow ns the listed consumer du ra-
bles includ ing TV, refriger ator, ceili ng fan, motor-
cycle/sco oter and t hree-wheelers; ( 5) the house-
hold-owned farm equipment such as tractor,
power tiller, combined thresher/harvester – were
asked to eac h household i n the vi llage. I f house-
holds resp onded in the affi rmative to any of the se
questions, they were declared to bevisibly
non-poor”.
4 The 13 indicators inc lude size of la ndholdi ng,
type of house, ava ilabilit y of clothing per p erson,
food sec urit y, sanitation, literacy, possession of
consumer durables, me ans of livelihoo d, status of
household labour, status of c hildren between 6
and 14 years, ty pe of indebtednes s, reasons for
migrat ion and preference for assist ance.
5 As it is possible to obtain the same level of score
through combinations of dimensions, the revised
methodolog y suggested by th e Saxena Commit tee
(2009) could not overcome the most debated
issue of “complete substitutabilit y” across the m,
and hence, neglecte d to distingui sh between
severity of the problems. In addition to this,
agreei ng to the cap on incidence of pover ty im-
po sed b y eith er g ram p anch ayat s or bl ock s is l ike-
ly to restr ict the possibi lity of ide ntif ying t he
desired number of poor hous eholds.
6 The sta ndardised cu ltivable land of a hous ehold is
estim ated as follows: (Area u nder irrigated land
× 1.5) + area und er unirrigat ed land.
7 In India, exclusion revolves arou nd societal inst i-
tutions that excludes, disc riminate aga inst, i so-
late and depr ive some groups on the basis of thei r
group ident ity (Thorat and N ewman 2010). There
is ample ev idence that t hese groups are often dis-
criminated by the market forces. In this context, a
case ca n be made for compulsor y inclu sion of all
the SC/ST households in t he BPL list to ensure
minimum basic need.
8 Attac hed a value of ‘1’ or ‘0’ for the pres ence or
absent of an individual dimension.
9 A n expert comm ittee set up by the MoR D headed
by N C Saxena a lso sugge sts to inc lude all SC /ST
households i n the BPL list.
References
Alkire, Sabina and Suman Seth (2008): “Determining
BPL Status: Some Methodolog ical Improvement ”,
Indian Journal of Human Deve lopment, Vol 02,
No 2.
Dreze, Jean and Ree tika K hera (2010): “The BPL
Census and a Possible A lternat ive”, 27 February,
Vol 45, No 9.
Himan shu (2008): “What Ar e These Ne w Povert y
Estimates and Wh at Do They Imply”, Economic &
Political Week ly, 25 October.
Hirway, I (2003): “Identifi cation of BPL Households
for Povert y Alleviation Progr amme”, Economic &
Political Week ly, 8 No vember.
Khera, Reetik a (2008): “Access to the Targeted P ublic
Distr ibution Sy stem: A Cas e Study in R ajasthan”,
Economi c & Political Week ly, Vol 43, No 44,
1 November.
Mehrot ra, S and Harsh Ma nder (2009): “How to Iden-
tify the Poor? A Proposal”, Economic & Political
Weekly, Vol XLI V, No 19, 9 May.
Ram, F, S K Mohanty and U R am (2009): “Under-
standi ng the Di stribution of BPL Cards: Al l India
and Selected States”, Economic & Polit ical Weekly,
Vol XLI V, No 7, 14 Febr uary.
Saxena Com mittee R eport (2009): “Identifi cation of
BPL Household s in Rura l India”, Planning Com-
mission, Gover nment of India.
Srinivasan K a nd S K Mohant y (2002): “Deprivat ion of
Basic Amenities by Caste and Religion in India:
An Empirical St udy Using N FHS Data”, Economic
& Politi cal Weekly, Vol XXX IX, No 7, 14 Februar y.
Sundara m, K (2003): “On Identifi cation of House-
holds Below Poverty Line in BPL Censu s 2002:
Some Comments on the Proposed Methodology”,
Economi c & Political Week ly, 1 Marc h.
Swaminathan, Madhura (2008): “Public Distribution
System a nd Social Exclu sion”, The Hindu, 7 May.
Thorat, Su khadeo a nd Motilal Mahamallik (2006):
“Chron ic Poverty and So cially Disadv antaged
Groups: Ana lysis of Causes a nd Remedies”, CPRC-
IIPA Working P aper Series, No 30.
Thorat, Su khadeo a nd Katherine Ne wman (2010):
“Economic Discrimination, Concept, Consequences,
and Remed ies” in T horat, and Newman (ed.),
Blocked by Caste: Economic Discrimination in Modern
India (New Del hi: O xfor d Uni vers ity P res s).
... That the subsidized items do not reach the poor partly reflects poor targeting. The mechanism used is the BPL list, which has vast inclusion and exclusion errors, as numerous studies have shown (Sundaram, 2003;Hirway, 2003;Jain, 2004;Mukherjee, 2005;Jalan and Murgai, 2007;Alkire and Seth, 2008;Saxena, 2009;Jhabvala and Standing, 2010;Roy, 2011;Mahamalik and Sahu, 2011;Alkire and Seth, 2012). ...
Book
Would it be possible to provide people with a basic income as a right? The idea has a long history. This book draws on two pilot schemes conducted in the Indian State of Madhya Pradesh, in which thousands of men, women and children were provided with an unconditional monthly cash payment. In a context in which the Indian government at national and state levels spends a vast amount on subsidies and selective schemes that are chronically expensive, inefficient, inequitable and subject to extensive corruption, there is scope for switching at least some of the spending to a modest basic income. This book explores what would be likely to happen if this were done. The book draws on a series of evaluation surveys conducted over the course of the eighteen months in which the main pilot was in operation, supplemented with detailed case studies of individuals and families. It looks at the impact on health and nutrition, on schooling, on economic activity, women’s agency and the welfare of those with disabilities. Above all, the book considers whether or not a basic income could be transformative, in not only improving individual and family welfare but in promoting economic growth and development, as well as having an emancipatory effect for people long mired in conditions of poverty and economic insecurity.
... Although the coverage of food subsidies program is large in India, and become a major source of primary food for poor households, it is limited to rice and wheat. The exclusion and inclusion errors 10 and food leakages are prime reasons to hinder the calorie demand of calorie poor households (Mahamallik and Sahu 2011;Dreze and Khera 2015). Besides, there are also other cereal items that are necessary for a healthy life and not provided at the subsidies rate. ...
Article
This study examined predictors of households’ calorie demand using consumer expenditure survey data during the time frame of millennium development goals. It draws suggestions for achieving sustainable development goals to eliminate calorie-poverty. We used the log of per-capita calorie intake as the calorie demand. Endogeneity corrected quantile regression was applied to examine the distributional effect of predictors. Findings revealed calorie-monthly per-capita consumption expenditure (MPCE) elasticities were positively statistically significant across quantiles in rural-and urban-areas, but, contrary to traditional wisdom, elasticities are lower for calorie-poor than calorie-rich households. Dietary diversification of food items, relative food price, and share of medical-and education-expenditure were the main adverse drivers of calorie demand. Our results are robust to the under-reporting and measurement error. The policy implications are: (a) only focusing on pro-poor income enhancing strategies will not able to reduce calorie deprivation, it should be backed by imparting awareness about food choice and nutritional value of low price food items, (b) to implement necessary policy to maintain stable food inflation and effectively targeted food subsidy for calorie poor, (c) to adopt forward-looking medical-and education-policy such as free health and education facilities to all by enhancing public spending to revive the quality of public hospitals and educational institutions.
... This card, issued to households, has two main different categories: below poverty line (BPL) and above poverty line (APL). 14 Although it is fraught with design and implementation flaws, as discussed by Jain (2004) and Mahamallik & Sahu (2011), it is another relevant metric for disaggregating the sample. In total, 52% of sample households reported possession of BPL cards, 34% APL cards, and the remaining 14% reported that they did not have official documents to prove their poverty status. ...
Technical Report
Full-text available
The report looks at the impact of subsidies and reform to kerosene and liquefied petroleum gas (LPG) from a gender perspective across three countries (Bangladesh, India and Nigeria). The research was based around two overall research questions namely ‘How do existing kerosene and LPG subsidy policies affect the welfare, productivity and empowerment of women and girls in low-income households?’ and ‘How might the welfare, productivity and empowerment of women and girls in low-income households change as a result of specific, nationally relevant proposals for the reform of existing kerosene and LPG subsidies?’ These questions were explored using secondary data, household surveys across the three countries, reaching over 2,400 households, and focus group discussions. The questions were answered within the context of hypotheses made during the scoping phase and literature review for the research, based on a review of 28 reform episodes. This research attempted to answer the above questions in relation to an income effect, energy use effect and an energy supply effect, from a gender perspective.
... Reforms in Chhattisgarh are done on a continuous basis, and the government is committed to further improvements. The PDS system of Chhattisgarh is now one of the most efficient PDS systems in the country and has leakages under 3 per cent (Mahamallik and Sahu, 2011). This already is a role model for the country, but Dr Raman Singh and his team is not satisfied even after achieving so much. ...
Article
Dr. Raman Singh became the newly elected chief of a state of Chhattisgarh situated in the eastern part of India. Now, after getting power, Dr. Raman Singh established himself as a leader of masses through the series of reforms in the Public Distribution System (PDS) system, and this helped him establish his credibility as a leader who can bring his party to victory and fulfill the aspirations of the people of the state. When Dr. Raman Singh took over, he faced many challenges and the social sector was the one which required immediate attention. Confronted with many challenges, he decided to place the PDS in his top priority of reforms. The public distribution system was in a bad phase at that time, and the food grain supply intended for poor was lost in the middle before reaching to the intended beneficiaries. Having experiences in central government ministries as a cabinet minister, he knew that the task of reforming PDS system in the state is not going to be an easy task. The challenges faced were primarily from that of the reticent bureaucracy and dealing with them requires patience and skill of a transformational and motivational leader which Dr. Raman Singh possessed. There were the slews of technological and process reforms backed by organizational and social reforms which led to the complete transformation of the PDS system of state and made it a model example for all the others Indian states to follow and that too within the short time span of 6-7 years. These reforms happened in a very coordinated fashion, and the most important aspect of these reforms was the people and employee engagement. The reticent bureaucracy turned into highly motivated employees of the Chhattisgarh government and public of the state who never attached themselves with any government schemes, was directly involved in the PDS scheme and took it to their right rather than some privilege. The rest is all history and India witnessed the success of PDS in the state of Chhattisgarh.
... Poor people being excluded from official security nets are common and have been part of the policy debate in India. [4][5][6][7][8][9][10][11][12] Despite the fact that these errors occur in various targeted social sector programmes, policy planners in both developed and developing countries have continued 'targeting' and trying to refine the design of these programmes to improve the delivery of social services to the poor. The evidence from national-level surveys (Tables II to VII) suggests that the strategy of refining designs has not worked. ...
Article
Full-text available
This study examined the effect of out-of-pocket expenditure of patients on their illness and other household changes impacting on their well-being, and assessed the socioeconomic status and below poverty line (BPL) card status among patients at the All India Institute of Medical Sciences, New Delhi. We did a hospital-based cross-sectional study of 374 inpatients and outpatients. Among the 374 study subjects, more than 69% of poor did not possess a BPL card. On the other hand, 5.5% of the above poverty line patients among the respondents possessed a BPL card. Of those having BPL cards, 84.4% belonged either to the lower middle, upper lower and lower socioeconomic status categories. Our data suggest that the inaccuracies in providing BPL cards limit access of the genuine poor to healthcare. In the light of national-level surveys on accessing healthcare and out-of-pocket expenditure by patients there is a case against 'targeting' in the delivery of public health services and user charges have an adverse impact on access to healthcare by the poor.
... Although the coverage of food subsidies program is large in India, and become a major source of primary food for poor households, it is limited to rice and wheat. The exclusion and inclusion errors 10 and food leakages are prime reasons to hinder the calorie demand of calorie poor households (Mahamallik and Sahu 2011;Dreze and Khera 2015). Besides, there are also other cereal items that are necessary for a healthy life and not provided at the subsidies rate. ...
Article
Full-text available
This study examines food (in)security dynamics for India since the year 2000 using National Sample Survey data on household consumer expenditure. In the absence of true panel data, a synthetic panel approach has been applied to analyse food security transitions, in order to examine the movement of households both into and out of food security. The overall findings revealed that the maximum chronically food insecure households were 50.4 percent in rural and 46.1 percent in urban areas whereas the households which remained food secure were 39.8 percent in rural and 42.2 percent in urban areas. The maximum percentage of people who escape from food insecurity was 19.4 percent in rural and 18.6 percent in urban areas whereas the fall into the food insecurity in the rural was 29.2 percent and 30.2 percent in urban areas. The paper additionally evaluates food insecurity dynamics by income decile groups and different household identities, such as within the caste, religion, and gender of the households, which indicated that forward castes and Islam religion noticed a higher degree of food insecurity transition within the different caste and religious identities, respectively. Between the genders of the household head, the larger transition was observed for male and female-headed households, where male-headed shows inward and female-headed witnessed outward transition. The higher income groups are facing more inward and outward food insecurity transition in rural and urban areas. The results were robust to the under-reporting. This study provided a better understanding of food insecurity dynamics in India, which will help in effective policy design.
Article
This research is conducted to know and explain the trend of effectivity of CfB distribution in period of 2010-2014 in Maluku Province and Maluku Utara Province. The research show that the trend of effectivity of CfB distribution in period of 2010-2014 in Maluku Province was decreasing that is the mistargeted ratio in 2014 as 1,57% was higher than the mistargeted ratio in 2010 as 0,00% and the trend of effectivity of CfB distribution in period of 2010-2014 in Maluku Utara Province was increasing that is the mistargeted ratio in 2014 as 1,85% was much lower than the mistargeted ratio in 2010 as 23,54%.Keywords : Credit for business, economic development, productive sector, poverty, working capital.
Article
Governance of food distribution systems as part of social safety net programs continues to be a major concern for policy makers in developing countries. In this paper we assess how successful recent initiatives in governance of public food distribution (PDS) in the Indian state of Odisha have been in dealing with inefficiencies that have long plagued it. Analysis, based on data from the National Sample Survey Organisation (NSSO) and micro-level evidence obtained from Village Level Studies (VLS), indicate a marked revival in governance of food distribution in the state with several positive outcomes. These include much wider coverage of PDS from 6.4% of households in 1993–1994 to 58.3% in 2011–2012, notable improvements in cereal consumption at the household level across both rural and urban areas and improved nutrition of a large segment of the socially and economically weaker sections of society.
Article
Full-text available
In striving for "efficiency" by means of narrow targeting, households that should be entitled to basic food security through the PDS have been left out. During periods of high inflation in food prices, governments must provide a basic minimum quantity of food grain and other food items at low prices through public distribution systems to low-income, food-insecure, and vulnerable populations. In India, the ostensible purpose of the Targeted Public Distribution System (TPDS) was to take food to the poor; in practice, it has resulted in the large-scale exclusion of the poor and food-insecure from the public food system. Recent evidence from a report titled Public Distribution System and Other Sources of Household Consumption 2004-2005 (GOI, 2007), which presents data from the 61st Round of the National Sample Survey (NSS), establishes that targeting has led to high rates of exclusion of needy households from the Public Distribution System (PDS) and a clear deterioration of coverage in States like Kerala where the universal PDS was most effective. Let me illustrate with evidence from rural India. The Targeted PDS was begun in 1996. In March 2000, the prices of grain for above-poverty-line (APL) cardholders were hiked and the gap between prices for below-poverty-line (BPL) and APL households widened. In many States, APL prices of grain were close to market prices and, as a result, households with APL cards stopped buying grain from the PDS. The Antyodaya programme introduced a new category, the "poorest of the poor" for whom rice and wheat are available at even lower prices than for BPL households. In the present situation, a person who belongs to a household that has neither a BPL nor an Antyodaya card is effectively excluded from the PDS.
Article
Full-text available
In a modern market-oriented economy, possession of basic social and physical necessities of life can be considered the basis of a dividing line of different levels of deprivation. This paper, by using consecutive NFHS data (1992 and 1999), attempts to estimate levels of deprivation based on possessions at the household level of some basic amenities of life. It examines changes in levels of deprivation, categorised as 'abject deprivation', 'moderate deprivation', 'just above deprivation' and 'well above deprivation' across Indian states and also analyses changes in terms of caste and religion.
Article
Full-text available
Using the recent National Family and Health Survey-3 data, this paper examines the distribution of below poverty line cards. The possession of BPL cards by the households in various economic and social settings index is analysed. The findings suggest that about two-fifths of the BPL cards in India are with the non-poor households. On the other hand, in many of the states a majority of households in abject deprived groups do not possess a BPL card. The extent of misuse is higher in Andhra Pradesh, Kamataka and Kerala, while it is lower in Tamil Nadu. In economically weaker states like Orissa and Bihar, a higher proportion of non-poor households possess a BPL card.
Article
Full-text available
The World Bank's recent estimates of poverty in the developing world have led to an upward revision of the number of poor in the world by 400 million. These adjustments are made on the basis of the revision in purchasing power parity estimates as part of the International Comparison Program exercise. Using the same ICP exercise, the Asian Development Bank claims an even higher estimate of the poor in Asia. A proper examination of the underlying database and the methodology for estimating poverty across countries suggests that though these estimates are better than the earlier ones, the assumptions behind the adjustments and the quality of data obtained from the ICP limit the usefulness of such an exercise for cross-country poverty comparisons. For India, both these estimates suggest severe underestimation in the official ntimbers on poverty.
Article
Full-text available
to identify eligible beneficiaries. On the one hand, the data were of poor quality and coverage, and were influenced by corruption; on the other hand, the methodology suffered a number of flaws, particularly in treating ordinal data as cardinal, and allowing complete substitutability among all levels of achievement. In order to isolate and scrutinize the methodological points, this article uses the 2005-06 National Family Health Survey (NFHS) dataset, to generate two scores: a pseudo-BPL score, and our preferred score, which we call M0. M0 uses the Adjusted Head Count methodology developed by Alkire and Foster (2007). Not only is the newly proposed methodology able to identify the poor households more effectively, but it also provides additional insights for policy. For example, since the M0 measure can be broken down by dimensions, we can immediately show significant differences in the components of poverty in different states. Since liberalization in the early 1990s, India has enjoyed a strong rate of economic growth in terms of its Gross Domestic Product (GDP). Yet after over 15 years of growth, 46 per cent of the children in the country are still undernourished 2 (NFHS, 2005/06), about 50 per cent of the adult females are illiterate, 67 per cent of the population does not have access to improved sanitation, and 44 per cent of the population
Article
This paper explores the possibility of a simple method for the identification of households eligible for social assistance. In exploring alternative approaches for identifying a "social assistance base", of which the BPL list can be seen as a particular case, this note explores possible uses of simple exclusion and inclusion criteria. It first considers the possibility of a quasi-universal approach, whereby all households are eligible unless they meet pre-specified exclusion criteria. It then looks at various inclusion criteria for drawing up a SAB list. Finally, it explores four simple ways of combining exclusion and inclusion criteria to construct a SAB list. The intention here is to point to possible directions of further enquiry, including experimental applications of the suggested method, rather than to present definite recommendations. Whether any convincing method of selecting SAB households actually exists is an open question. Some of the findings here can be read as a reinforcement of the case for a universal approach. Indeed, the search for a "safe" way of excluding privileged households, without significant risk of exclusion for poor households, remains somewhat elusive.
Article
States and Union Territories have been asked, in the context of the Tenth Five-Year Plan, to launch the BPL (Below Poverty Line) Census 2002 as per the recommendations of the Expert Group on Identification of Households Below Poverty Line. Besides being used for the implementation, monitoring and evaluation of the poverty alleviation programmes of the ministry of rural development, the BPL-APL (Above Poverty Line) distinction is already being used for pricing grain under the public distribution system and is likely to be used by other ministries also for 'better targeting' of their programmes. Since expenditure of vast sums of taxpayers' money annually will thus be riding on the APL-BPL categorisation of households, it is imperative that the methodology proposed by the Expert Group underpinning the 2002 BPL Census is widely discussed and debated.
Article
This article examines the government of India's 1997 criteria for selection of households for below poverty line ration cards. The main conceptual problems are that the criteria are static and uniform across the entire country. Using primary data (collected in 2002) from 400 randomly selected households from eight villages of Rajasthan, the exercise here calculates the proportion of "wrongly exduded" (i e, who qualify according to government criteria but did not get a BPLcard) and "wrongly induded" households. Of the one-third of sample households that were classif ied as BPL, nearly a quarter have been wrongly induded. Besides, 44 per cent of the households which should have been counted as BPL were wrongly exduded. However, one must consider the appropriateness of the selection criteria along with these large selection errors.
Article
This paper provides a theoretical introduction to the study of discrimination with particular reference to the caste system. It sets the stage for the four empirical papers that follow, by highlighting the ways in which caste persists as a system of inequality that burdens the Indian economy with inefficiencies in the allocation of labour and other critical resources, reducing the full development of human capital in society. Far from disappearing as the economy modernises, discrimination remains a problem which, for reasons outlined here, is not amenable to self correction, but rather requires interventionist policies to remedy