ISSN (Print) - 0012-9976 | ISSN (Online) - 2349-8846

A+| A| A-

'New' Lists for 'Old': (Re-) constructing the Poor in the BPL Census

This paper aims to understand the implications of implementing the Saxena Committee's recommendations in respect of identifying the poor in India. Relative to the one currently in use, the application of the proposed methodology appears to be more beneficial in general to social groups such as scheduled tribes, most backward classes and mahadalits, as well as those landowning households that might suffer from specific debilitating conditions. However, in some cases it is less sensitive to Muslims, non-mahadalit scheduled castes and agricultural labourers. These observations are based on the results of a census survey covering 4,500 households in 18 rural wards of Bihar and West Bengal. By comparing the subset of households classified as poor according to the 2002 and the 2009 methodologies, the paper analyses "moving in" and "moving out" of poverty lists according to occupational categories, caste groups and landowning profile of the poor.


‘New’ Lists for ‘Old’: (Re-)constructing the Poor in the BPL Census

Indrajit Roy

This paper aims to understand the implications of implementing the Saxena Committee’s recommendations in respect of identifying the poor in India. Relative to the one currently in use, the application of the proposed methodology appears to be more beneficial in general to social groups such as scheduled tribes, most backward classes and mahadalits, as well as those landowning households that might suffer from specific debilitating conditions. However, in some cases it is less sensitive to Muslims, non-mahadalit scheduled castes and agricultural labourers. These observations are based on the results of a census survey covering 4,500 households in 18 rural wards of Bihar and West Bengal. By comparing the subset of households classified as poor according to the 2002 and the 2009 methodologies, the paper analyses “moving in” and “moving out” of poverty lists according to occupational categories, caste groups and landowning profile of the poor.

I thank Sabina Alkire for her guidance and advice on the paper, and on the methodology in particular. I also benefited from email conversations and discussions with Jean Drèze and Santosh Mehrotra. Constructive feedback from an anonymous referee is most appreciated. Financial support from the Oxford Poverty and Human Development Institute made the survey possible. Discussions with participants at the British Academy-University of Oxford Conference on “Poverty in South Asia” (28-29 March 2011), especially with Dipankar Gupta, have been valuable. Suggestions from Nandini Gooptu on the politics of poverty have also been helpful. All errors remain my own.

Indrajit Roy ( is pursuing doctoral research at St Anthony’s College, University of Oxford, UK.

1 Introduction: A Tale of Two Methods

ince at least 2008, there has been vigorous debate within Indian and international policymaking and advocacy circles about revising the mechanisms to identify the “poor” in India, so that social assistance may be better targeted. In August 2009, the government made public the Report of the Expert Group that was advising the Ministry of Rural Development on the methodology for conducting the new Below Poverty Line (BPL) Census for the Eleventh Five-Year Plan, which outlined the contours of the revised scheme for identifying those eligible for social assistance (GoI 2009). The proponents of the 2009 methodology argued that it represented an advance over previous methodologies for the following reasons: (1) The criteria proposed by it were more transparently verifiable. (2) The use of a combination of nominal and ordinal data, which allowed it to automatically exclude and include households from BPL lists, and also rank them so that different cut-offs could be applied to them. (3) It was far more sensitive to vulnerable households because of the in-built bias in its scoring in favour of such groups, as for example agricultural labourers, destitute households, mahadalit castes, women-headed households and those comprising entirely of aged persons.

Perhaps its strongest argument was that it suggested a way to identify the most vulnerable households without necessarily enhancing the overall resource outlay. Indeed, it was designed to make the task of identifying the poor more efficient, its proponents arguing that the indicators proposed had “the best chance of being accurately executed by government staff” (GoI 2009: 26, emphasis added). It was predicated on a deceptively simple premise – the inclusion of the poorest and most deserving households in any listing of BPL households and, correspondingly, the automatic exclusion of the “undeserving” poor. Its advocates argued further that the application of this methodology would lead to effective targeting of the poor, unlike the previous exercises where the methodology was not sensitive enough to either ensure that “deserving” groups were included, or to prevent the inclusion of obviously “undeserving” groups.

The motivation for the government’s endeavours emerged from the mounting criticism levelled against the previously used methodology, mooted in 2002 and implemented in 2006.1 The “2002 methodology” as we refer to it hereafter, consisted of a survey and measurement system that assigned “scores” to each household on the basis of 13 questions. In addition to ownership of consumer durables and means of production (which were staple in the two census surveys pre-2002), the survey now included

may 28, 2011 vol xlvi no 22

questions on food security, sanitation, literacy status, means of livelihood, status of labour including women and children, indebtedness, and migration. The response to each question was scored on a scale of 0 to 4 and thereafter aggregated. Households could score anywhere from 0 to 52.

State governments were free to establish their own cut-offs, based on local criteria and considerations, subject to the Expert Group’s stipulation that the application of such standards should ensure that the total number of poor persons identified in that state/union territory did not exceed the estimates prepared by the Planning Commission by more than 10%. The panchayats were tasked with the responsibility of overseeing the surveys, collecting the data and reviewing it in the light of the cut-off scores established by the state government. The Bihar government incorporated the BPL schedule with no changes and set the cut-off at 13, meaning all the households who scored 13 and less (where the denominator is 52) were to be eligible for assistance to schemes for the BPL population. West Bengal, however, incorporated an amended version of the BPL schedule. The state’s technocrats dropped two questions – one, seeking information on sanitation; and two, measuring the respondents’ preference for assistance. They added a question that sought information on “special difficulties”, which was designed to ascertain if a given household comprised aged people living on their own, disabled persons,

Table 1: Scoring Scheme for Different BPL Methodologies

persons suffering from debilitating illness, or was headed by a woman. Thus, the West Bengal schedule had a total of 12 questions. The state government also reset the scoring scheme to a 1-5 scale, making the total marks obtainable by any household 60. The cut-off was set at 33, meaning that households would be considered as BPL if they scored 33 or less.

However, from the moment it was launched, the 2002 methodology met with a barrage of criticisms that only intensified in subsequent years (Alkire and Seth 2008; GoI 2009; Hirway 2003; Jain 2004; Mehrotra and Mander 2009; Sundaram 2003, among others). The criticisms of the methodology ranged against its cardinal treatment of ordinal data, assumption of complete substitutability across dimensions, and equal weighting of the 13 dimensions. In addition, observers raised doubts about its logistical practicability, policy relevance and immunity from corruption. They also questioned the appropriateness and adequacy of several of the indicators, which, they argued, did not adequately capture the complexity of the poverty question, especially the capabilities space. Indeed, the proponents of the 2009 methodology referred to several of these criticisms in order to justify their own recommendations for an overhauled methodology, which may well be the methodology adopted by the government in the very near future.

The 2009 methodology, as mentioned, combines the principles of automatic inclusion and exclusion with its own scoring mechanism

BPL 2002 Methodology BPL 2009 Methodology

Recommended by Planning Commission and Implemented in Bihar Automatic Exclusions (GoI, op cit: 21-22)

0 1 2 3 4

No operational Less than 1 1-2 hectare 2-5 hectare 5+ hectare Families who own double the land of the district average of the landholding hectare unirrigated/ unirrigated/ unirrigated/ agricultural land per agricultural household

unirrigated/ 0.5-1 hectare 1-2.5 hectare 2.5+ hectare Families who have three of four wheeled motorized vehicles

0.5 hectare irrigated irrigated irrigated irrigated

Houseless Kachha Semi-pucca Pucca Urban type Families who have at least one mechanised farm equipment

< 2 sets of clothing 2-4 sets >4 & <6 >6&<10 10+ Families who have any member drawing a salary of INR10,000

<1 square meal Normally one, 1 square meal per 2 square meals, Enough food Income tax payers per day but occasionally day throughout with occasional

less than 1 the year shortage

Open defecation Group latrine with Group latrine with Clean group latrine Private latrine Automatic Inclusions (GoI, op cit.: 26) irregular water regular water with regular water supply and regular sweeper

Ownership of no Own any one item Own any two Any three or all items All consumer durables Designated primitive tribal groups consumer durable or any one of designated

expensive items

Illiterate Highest literate Highest literate Highest literate Highest literate is Designated discriminated groups, such as mahadalits studied up to completed is graduate/ post-graduate/ primary secondary diploma professional graduate

Bonded labour Female and Only adult females Adult males only Others Single women-headed households Child Labour and no child labour

Casual labour Subsistence Artisan Salary Other Households with disabled person as breadwinner cultivation

Children not going to Going to school Going to school and Households headed by a minor school and working and working not working

Loan for daily Loan for production Loan for other Borrowing only No indebtedness Destitute households consumption from purpose from purpose from from institutional and possess assets informal sources informal sources informal source agencies

Migrate for casual Migrate for seasonal Migrate for other Non-migrant Migrate for other Homeless households work employment forms of livelihood purposes

Preference for Preference for Preference for Preference for Preference for Any member of the household is a bonded labourer assistance: wage assistance: assistance: training assistance: housing assistance: loan/ employment self-employment andskill-upgradation subsidy


may 28, 2011 vol xlvi no 22

Table 1: Scoring Scheme for Different BPL Methodologies (Continued) Scoring System Used in West Bengal Scoring (GoI, op cit: 27)
1 2 3 4 5 4 3 2 1 0
No operational landholding Less than 1 hectare unirrigated/0.5 hectare irrigated 1 -2 hectare unirrigated/0.5-1 hectare irrigated 2-5 hectare unirrigated/1-2.5hectare irrigated 5+ hectare unirrigated/2.5+ hectare irrigated SC/ ST Denotified Tribe/MBC Muslim1/ OBC
Houseless Kachha Semi-pucca Pucca Urban type Landless agricultural labourer Agricultural labourer with some land Casual worker/self-employedartisan/ fisherfolk
< 2 sets of clothing 2-4 sets >4 & <6 >6&<10 10+ No adult (30+ ) has studied up to class V
<1 square meal per day Normally one, but occasionally less than 1 1 square meal per day throughout the year 2 square meals, with occasional shortage Enough food Any member of family has TB, leprosy, disability, mental illness
or HIV
Ownership of no consumer durable Own any one item Own any two Any three or all item All consumer durables or any one of designated expensive items Household headed by an old person aged 60
Illiterate Highest literate studied up to primary Highest literate completed secondary Highest literate is Highest literate is graduate/ diploma post-graduate/professional graduate
Bonded labour Female and child labour Only adult females and no child labour Adult males only Others
Casual labour Subsistence Cultivation Artisan Salary Other
Children not going to school and working Going to school and working Going to school and not working
Loan for daily consumption from informal sources Loan for production Loan for other purpose from purpose from informal sources informal source Borrowing only from institutional agencies No indebtedness and possess assets
Migrate for casual work Migrate for seasonal Migrate for other employment forms of livelihood Non-migrant Migrate for other purposes
Special difficulties: Destitute elderly Special difficulties: Suffering permanent disability Special difficulties: Special difficulties: No such difficulties woman-headed household withhousehold debilitating illness, where cost of treatment >
household income

(1) The Expert Group remained divided over what score to assign to Muslim households and left it to the government. In our survey, we assigned them a score of ‘1’, as recommended by some

members of the Group.

(GoI 2009). The criteria for automatic exclusion from poverty lists include landholding (but, unlike earlier methods, juxtaposed vis-à-vis the quality of land and structure of landholding in the district or block), income and income-tax payment status, and other (supposedly) visibly verifiable indicators such as ownership of mechanised equipment and three-to-four wheel automobiles (ibid: 21-22). The criteria for the automatic inclusion, on the other hand, includes such diverse attributes as belonging to so-called primitive tribal groups and mahadalits; households headed by women, minors and disabled individuals or comprising a member who may be bonded; as well as destitute and homeless households (ibid: 26).

If a given household falls outside the inclusion/exclusion criteria, as is more likely, it is scored on the basis of five attributes, such as caste, occupation, education, health status of members and age of the household head. This scoring mechanism comprises both binary scoring (where the presence of a povertyenhancing attribute2 is scored 1 and its absence 0) as well as continuous scoring (while there is no standardisation, the maximum score a household can attain is 4).3 The committee recommends that the responses be aggregated; and those with higher scores be considered poorer than those with lower scores.

I will focus on the specific survey instruments of the 2009 methodology (as well as the 2002 methodology) rather than on the broader questions of institutional implementation. There are several further recommendations made by the committee to enable policymakers target the poor more effectively, and these include allocation of quotas and sub-quotas within the gram panchayats,4 so as to minimise political competition among elected representatives. While there is much to be said about the feasibility or otherwise of this recommendation, as indeed Drèze and Khera (2010) have argued, I concentrate on the impact of the survey questions on the classification of households as being “poor” or not.

2 Motivation, Arguments, Methods and Caveats

In this paper, I scrutinise the claim made by proponents of the 2009 methodology that it is more sensitive to “poorer” groups than the previous methodologies. I do this by comparing the

may 28, 2011 vol xlvi no 22

subset of households classified as poor according to the 2009 methodology with the subset of households classified as poor according to the 2002 methodology. To that end, I present results of a census survey of households in 18 wards of four gram panchayats in two districts – Maldah and Araria, in West Bengal and Bihar respectively. Data analysis reveals that though there is an overlap between the categories of the “poor” generated by the two methodologies, there are also considerable dissonances between the two categories, especially in terms of occupational, landowning and caste-specific attributes of households.

My overall argument is this. Although the application of the 2009 methodology appears to be more beneficial to social groups such as scheduled tribes (STs), most backward classes (MBCs) and mahadalits, as well as those landowning households that might suffer from specific debilitating conditions, it is less sensitive to Muslims and non-mahadalit scheduled castes (SCs), and to agricultural labourers in Bihar. While this in no way detracts from the argument of its sympathisers that the 2009 methodology demonstrates a more nuanced appreciation of the complexity of the poverty question – focusing on the web of adverse economic and social relations that constitute it rather than limiting its focus on the possession of specific assets or commodities – it does need to be borne in mind that it is not unproblematic either.

An important distinction may be drawn. Criticisms against the 2002 methodology have often tended to focus on corruption, i.e, to describe what happens when the 2002 methodology is subverted by local elites and political bosses and used as reward/ punishment instruments. This is important, but distinct, from my concern, which focuses on understanding what happens when the methodologies are implemented in accordance with the way they should be. Given the well known corruption and distortions in the actual implementation of the 2002 methodology and in the distribution of the BPL cards, I consciously avoid using data on BPL card-holders in this paper.

As this was a household census, all households in purposively sampled wards of the selected districts (Maldah and Araria) were covered. The survey schedule included the 13 questions from the 2002 methodology as well as the considerations set out by the 2009 methodology. In addition, there were other questions drawing on the multidimensional poverty indicators (MPI) that have been developed by the Oxford Poverty and Human Development Initiative for the recently-launched UNDP-authored Human Development Report (Alkire and Santos 2010) as well as the Destitution Questionnaire developed by the Right to Food campaign in India (RTF 2002).5

The survey schedule was administered by trained field investigators. Each response was scored, based on the scoring scheme outlined by the Planning Commission. The response code structure and the scoring system were harmonised, to the extent possible, in order to facilitate a relatively swift, but accurate, completion of the exercise. Based on these scores, each household was scored using the scoring structure of both the 2002 and the 2009 methodologies. In the cases where households were automatically excluded or included, they were labelled as such, and assigned a unique numeric code that facilitated comparison on statistical packages.

Economic & Political Weekly

may 28, 2011 vol xlvi no 22

The classification of households as “poor”/“non-poor” was the next and final step in the data collection. Notwithstanding the proposal by Drèze and Khera (2010) to distinguish poverty lines from social assistance, I continue to use the more familiar imagery of a poverty line and households being above and below it. To make this classification, I identified the households whose names were eligible to be on the BPL list under the 2002 methodology, or, simply, the “BPL 2002 Poor”, on the basis of their scores making it past the officially identified BPL threshold, i e, if they scored less than 33 out of 60 in West Bengal and 13 out of 52 in Bihar.

The identification of households eligible to be on the BPL list under the 2009 methodology – the “BPL 2009 Poor” – was slightly more tricky, given that disagreements continue on the issue of fixing poverty quotas. However, I used the state-level figures recommended by the Tendulkar Committee (GoI 2009: 29) to determine the cut-offs for the poverty lines under the 2009 methodology in the respective states/districts. I believe using the estimates of the Tendulkar Committee to set the 2009 BPL methodology cut-off will make my findings more policy-relevant.6

For the Maldah localities, the cut-off of 7 classified 27% of all households as poor, while a cut-off of 6 categorised 45% households as poor. Given the estimation of head-count ratio in rural West Bengal by the Tendulkar Committee at 38.2%, I chose the cut-off that would generate slightly higher numbers of poor households (that accounts for both the Committee’s estimation and a 10% increase to accommodate the numbers of the transitory poor), thereby settling for the cut-off of 6. Similarly, in Bihar, the cut-off of 5 classified 50% of all households as poor, whereas the cut-off of 4 classified 62% of all households as poor. Given the above considerations, I applied the cut-off of 4 to count the numbers of the poor in Bihar.

In the next section, I present data on the overall socioeconomic characteristics of the households in the survey area, including results from cross-tabulating the population deemed to be poor under the two methodologies. In Section 4, I report changes in the classification

Figure 1: Typologies of the Poor Studied in of households as poor/non-This Paper


poor by discussing the oc


cupational, landholding and social attributes of those who are classified as poor and non-poor. To substantiate my

X X X2002 Non-BPL BPL

arguments, I will focus on specific kinds of “poor households”, which are presented in Figure 1. In Section 5, I draw attention to the policy ramifications.

3 Socio-economic Characteristics

The study universe comprises 4,520 households, covering nearly 21,000 individuals. Nearly 42% of the surveyed population were below 18 years of age, 47% were women, 27% were SC, 14% were ST (mostly Saotal), 32% Muslim and 18% were backward castes. In West Bengal, the SCs comprised mostly communities classified as Rajbonshis, but who referred to themselves as Polias, and increasingly as Desiyas. In Bihar, where the SCs have been bifurcated by the order of the state government into scheduled castes and mahadalits,7 ostensibly to better target communities that were subjected to the worst forms of deprivations, the survey population included a substantial proportion of both administrative categories. Of the respondents in Bihar, 26% were mahadalits, a large majority of whom were Musahars. Similarly, nearly a quarter of the respondents were MBCs, the so-called Annexure 1 castes often referred to in terms of their location at the bottom of the Sudra/ Other Backward Classes heap (Guru and Chakravarty 2005). It has been reported that their representation in the bureaucracy has been even less than what has been achieved by dalits (Prasad 2005).

As is well known, caste operated on a day-to-day level among Muslims as well. The West Bengal government categorises the Shershabadiyas as OBC Muslim, although in the sociological literature, they have been described as arzal, a description that has led activists to refer to them as dalit Muslims. The socially dominant Sheikh Muslims looked down upon them. There were reportedly no commensal or connubial relations between the two social groups. In Bihar, caste consciousness among Muslims was even stronger. Here, the Kunjra/Sabzifarosh, Nuniya, Mansothi castes, which the state government has categorised as OBC Muslim, predominated numerically among Muslims. These groups are more commonly referred to as Pasmanda Muslims in activist circuits. The STs comprised overwhelmingly of Saotals in both the states, although they represented only a fraction of the population in Bihar.

Table 2: Ascriptive Status and Livelihood Profile of Households

Ascribed Identity Destitute Agricultural Casual Self- Salaried Other Total
Labourer Worker employed
Livelihood Profile: West Bengal
General Muslim 9 65 465 25 20 96 681
OBC Muslim 8 60 338 4 0 63 476
Scheduled caste 5 150 281 7 4 51 508
Scheduled tribe 2 250 307 3 3 35 600
Total: West Bengal 24 533 1,418 53 27 254 2,310
Livelihood profile: Bihar
Scheduled caste –
excluding mahadalit1 2 39 104 6 13 4 168
Mahadalit 3 212 357 6 3 2 583
General Muslim 1 2 19 1 0 0 23
OBC Muslim 4 78 153 14 1 6 256
Other Backward Classes 0 135 133 52 8 25 352
Most Backward Castes 8 172 295 35 4 5 519
General caste Hindus 1 79 47 77 20 45 269
Total: Bihar 19 738 1,115 199 51 88 2,210

1 Essentially, this category only includes the Dusadh (Paswan) caste as all the other scheduled castes have been classified as mahadalit by the Bihar state government (see news item at http:// Source: Survey data and own calculations.

Among the most striking observations is that a large majority of households in West Bengal and slightly over half the households in Bihar derived their household income from casual work. This trend has been observed as far back as the early 1990s by Mendelsohn (1993). That this has only strengthened may be borne out by Chatterjee’s recent assertion that the only thing that “peasants” in rural India seem to want is not to be peasants (2008: 59). Often (though not always), at least one member of such households “worked” in towns and cities as far and widespread as Bhilwara, Ludhiana, Jalandhar, Chandigarh, Chennai, Pune, Mumbai, Ranchi, and Delhi. They did a variety of “jobs” in industries as diverse as construction (residential complexes, telephone towers), brick kilns, transport and carriage, including physically lugging goods from and to carriages and warehouses or factories, driving cycle and auto rickshaws – and if they were fortunate – working in factories or offices in nearby towns such as Old Malda, Naldubi and Narayanpur (in the case of the Malda households) and small government offices or institutions in Bhargama and Raniganj (in Araria). Earnings were erratic but on the whole better than what they would earn as agricultural labourers. Although they were economically better-off than agricultural labour households, the very nature of their work meant that family members were separated from one another for extended periods, causing stress, trauma and a gradual dissolution of familial ties.

Disaggregating the occupational profiles by ascribed groups reveals that more households tended to eke out their livelihoods through casual work outside their villages than through agricultural labour inside it. OBC Muslim households in both West Bengal and Bihar (Shershabadiya in the former, 72%; and Kunjra/ Sabzifarosh/Mansothi in the latter, 60%), and Dusadh (SC, 62%), mahadalit (61%) and MBC (57%) households in Bihar were more likely to contribute to the ranks of “casual workers”. While in West Bengal, there was no social group that appeared to be disproportionately concentrated in agricultural labour to non-agricultural casual work, in Bihar, OBCs and general caste Hindus seemed to be.

Agricultural labour households comprised nearly a quarter of the surveyed population in West Bengal and a third of the population in Bihar. Following the Saxena Committee’s recommendation (GoI 2009: 7), a household was treated as deriving its income from agricultural labour, even if it engaged in both agricultural labour and non-agricultural casual work. In such households, incomes were lower than among “casual worker” households, although they were often cushioned by the network of social support available in their localities. However, the extent to which this support was deemed to be of much use to lower-income households was called into question by several of the households themselves, who frankly admitted that their lives were characterised by humiliation, often discrimination, and exploitation – though not the brutal violence of the elite classes as was the case two or three decades ago. They “envied” casual workers,

Figure 2: The Poor


1085 1225 1036 1274 466 1378 832

2002 BPL Measure 2009 BPL Measure

2002 BPL Measure 2009 BPL Measure

Bihar (N = 2210)

West Bengal (N = 2310)

ů; Ϳ

Ś; Ϳ

BPL poor BPL non-poor

especially those who had the opportunity to work outside the village, especially their supposed “freedom” from feudal obligations. Ironically, poorer classes often commiserated with the rural elite, arguing that since agricultural productivity was itself declining, and farmers themselves faced declining incomes, how could they possibly pay agricultural labourers fair wages. “When the mahajan is so poor, what can be expected of the garib janata?”, the women from these classes commented in the presence and within earshot of their employers, part-serious and part-joking.

may 28, 2011 vol xlvi no 22

4 Findings of the Survey

The combination of the 2002 methodology and the official cut-offs produced rather interesting results. In the West Bengal wards, 47% of the surveyed households were below the poverty line under this methodology – the “BPL 2002 Poor”. In contrast, only 21% of the households in the Bihar wards, operating at a cut-off of 13, were BPL. Local politicians and ordinary people widely complained that the cut-off had been set so low that it almost appeared the state government wanted to eliminate all poverty using statistics as a weapon. The bulk of the total BPL 2002 Poor in this study were contributed by West Bengal, comprising 70% of the total BPL households. On the other hand, 45% of the households in West Bengal were “BPL 2009 Poor”, as against 62% of households in Bihar.

The intersection of the two methodologies produced four groups of households. There were those who “remained” on the poverty lists, i e, these house-

Figure 4: The Livelihoods Profile of the Poor

programmes. In Bihar, there is an absolute increase in the numbers of the poor under the 2009 methodology (Figure 3). However, here too, as in West Bengal, significant changes may be observed. Around 21% of the BPL 2002 Poor (99 of 466 households) moved out of the poverty list, while 73% of the BPL 2009 Poor households (1,011 of 1,378 households) moved into the poverty list.

4.1 Livelihoods of the Poor

One of the more striking observations from the data on occupational groups is that more agricultural labourers are included in the ranks of the poor under the 2009 methodology than under the 2002 methodology. However, the share of agricultural labourers in the total poor increases only in West Bengal, and not in Bihar. As a corollary, while the numbers and proportion of the casual worker households among the poor in West Bengal declined, they experienced an increase in Bihar.

holds were classified as poor 831

by both the methodologies. The second group refers to those


who “moved out of” poverty lists and the third comprises

422 409 342

those that “moved into” poverty lists. The second group is of in

195186 165 147

terest because they have been


5538 7 226 17 18 6



availing of the social protec

2002 Poor Remain in’ Move into’ Move out of’ 2009 Poor


tion net hitherto, but are likely

(n = 1085) poverty list poverty list poverty list (n = 1036) (n = 629) (n = 407) (n = 456)

to lose it with the application


West Bengal (N = 2310)

Agricultural labourers

Subsistence cultivators and self-employed workers

of the 2009 methodology. The

third group is of interest be

cause they will be eligible to avail of the social protection net The configuration of the poor agricultural labour households hereafter, but have so far been excluded from it. Finally, there are undergoes a transformation as well. With the application of the those who “remained out of” poverty lists, i e, they were considered 2009 methodology, 11% of the agricultural labour households non-poor by both methodologies – this group remains marginal to classified as poor according to the 2002 methodology were demy core argument and is ignored in the rest of the paper. classified from being poor (they would move out of poverty lists),

I will now examine the extent to which these categories map and ceased to receive BPL-related services. On the other hand, onto one another. From Figure 3, it is clear that the West Bengal 57% of all agricultural labour-based households classified as poor localities actually experience a slight decline in the numbers of according to the 2009 methodology had not been recognised as the poor. Even more interesting are the changes in the composition being poor according to the 2002 methodology. These houseof the poor. As many as 42% of the BPL 2002 Poor (456 of 1,085 holds would now be eligible to receive BPL-related benefits. households) “moved out of the poverty lists” when the 2009 The situation vis-à-vis casual labour households is exactly the methodology was applied. Conversely, 39% of the BPL 2009 Poor obverse. Nearly half8 the casual worker households on the 2002 (407 of 1,036 households) moved into the poverty list, meaning poverty list moved out of it with the application of the 2009 they were hereafter entitled to the government’s social assistance methodology. Of the BPL 2009 Poor casual worker households,

nearly a third9 moved into the list as a result of the changes in the

Figure 3: Mapping the Poor

1011 methodology. This reveals that notwithstanding an apparent bias

of the BPL 2002 methodology in favour of casual workers, it had

so long excluded as many as 30% of casual worker households

from the poverty lists and consequently from social protection.

The Bihar results seem to mirror those of West Bengal; however, an important exception needs to be noted. While agricultural labourers comprised 43% of the poor according to the 2002 methodology, under the 2009 methodology, they comprised 37% of the poor. This finding is interesting because it demonstrates

Move into poverty list that, notwithstanding the apparent bias in favour of agricultural

ZRemain in poverty list

West Bengal (N = 2310) Bihar (N = 2210) 629 407 456 818 367 99 733








29 4057








2002 Poor Remain in’ Move into’ Move out of’ 2009 Poor

Ζ Ζ ŝ

(n = 466) poverty list poverty list poverty list (n = 1378) (n = 367) (n = 1011) (n = 99)

Bihar (N = 2210)

Casual worker


Salaried employment and othres

DMove out of poverty list

Remain out of poverty list labourers in the 2009 methodology, agricultural labour households

may 28, 2011 vol xlvi no 22

are actually less represented among the poor when this methodo-overall poor declined with the application of the 2009 methodology is applied. As I will show subsequently, this has to do with logy. Nearly 80%14 of OBC Muslim households in West Bengal occupational patterns among the Musahars and the way the “move out of” the poverty list with the application of the 2009 2009 methodology “manages” their poverty. Nonetheless, as a methodology. Only 19 households “move in”, with the consecorollary, the proportion of casual worker households to the total quence that both the proportion of OBC Muslims classified as poor increased marginally from the BPL 2002 list to the 2009 poor and their proportion to the total poor according to the 2009 one, from 56% to 58%. methodology plummeted. In Bihar, OBC Muslim households fared

The movement of agricultural labour households across poverty only slightly better as more households were included in the povlists shows trends similar to what can be seen in West Bengal, erty list as per the 2009 methodology. With its application, the albeit exaggerated – while nearly a fifth10 of the households representation of OBC Muslim households on the poverty list exmoved out of the BPL 2002 list, nearly 70%11 of the households were perienced a net gain (over half moved out, but significant numincluded in the BPL 2009 list. However, in a trend reflecting the bers were added). Despite this, their contribution to the “poverty reverse of what is seen in West Bengal, 22%12 of the casual worker list” mirrored that of West Bengal. households moved out of the 2002 list, while 70% households The application of the 2009 methodology appears to have moved into the 2009 list13, implying that 70% of these house-benefited STs as the data from West Bengal indicates. In direct holds had hitherto been excluded from the social protection net. contrast with the experience of the Muslims, across both castes

and states, comparatively fewer households among West Bengal’s

4.2 Caste Profile of the Poor STs “moved out” of poverty (10% of the ST BPL 2002 Poor,15 com-The two methodologies interacted differently with the different pared with 80% among West Bengal OBC Muslims). ascriptive groups. The BPL 2002 method did not explicitly favour The two methodologies treated the SC populations across the (or discriminate against) any ascriptive group, limiting its ambit to two states very differently. In both the states, more SC households largely economic and social welfare considerations. The 2009 were classified as “poor” according to the 2009 methodology. methodology, in contrast, explicitly introduced an “affirmative That was the only similarity between the SCs in West Bengal and action” bias. Not only that, it also weighed in favour of certain Bihar, where the state government’s introduction of the mahahistorically oppressed groups by recommending their automatic dalit category and its recognition as a putative category by the inclusion on the BPL list. However, the intersection of the provisions Saxena Expert Group for the purpose of enumerating the poor for automatic inclusion of certain groups, assigning scores to certain complicates the scenario. In West Bengal, the application of the occupational categories, and the need to rationalise numbers with 2009 methodology increased both the proportion of SC households scarce resources produced some perhaps unintended consequences. on the poverty list as well as their overall contribution to the ranks

Given the livelihood profiles of most Muslims, both the propor-of the poor; although compared with ST households, more SC tion of Muslim households on the BPL list and their ratio to the households moved out of the poverty list, many more households moved into it as well. In Bihar on the contrary, the application of

Figure 5: The Social Profile of the Poor


the 2009 methodology resulted in similar movements among the SC households (primarily Dusadh) as among Muslims.

The newly-created constituency of mahadalits significantly increased their representation on the ranks of the BPL 2009 Poor, which is hardly surprising, given the explicit bias of the 2009 methodology in their favour. Upon its application, an overwhelming number of the mahadalit BPL 2002 Poor “remained in” the poverty list even after the application of the 2009 methodology (as with the STs in the West Bengal localities), but in much larger

Scheduled caste Scheduled Tribe

numbers – only 11 of these households “moved out of” the poverty

Figure 6: The Ascriptive Status Profile of the Poor in Bihar

list. At the same time, what the application


of the 2009 methodology did was to add a whopping 350-odd mahadalit households to the category of the poor, so that nearly 90% of all mahadalit households16 are now included in this category. Expectedly, 67%17 of these households had so far been outside the ambit of the social protection provided by the state through the enumeration of BPL households. However, interest

= = Move into’ poverty ' = Move out of’ poverty =

2002 Poor (n = 466) Remain in’ poverty = 2009 Poor (n = 1378) list (n = 367) list (n = 1011) list (n = 99)

ingly and somewhat surprisingly, mahadalits

Biahar (N = 2210)


now contributed less (albeit marginally) to

'General Muslim

Scheduled caste-excluding mahadalits Scheduled caste-mahadalits only

the population of the poor than under the

OBC Muslim

Other Backward Classes

Most backward classes

2002 methodology.18

ΖΖ'2002 Poor (n = 1085) Remain in’ poverty Move into’ poverty Move out of’ poverty 2009 Poor (n = 1036) list (n = 629) list (n = 407) list (n = 456) West Bengal (N = 2310) 299 247 222 300 66 94 195 269 19 31 159 192 233 153 27 31 85 125 354 OBC Muslim-Badiya General Muslim-Sheikh
36 0 182 71 43 115 70 171 31 33 109 44 16 348 87 138 315 29 0 11 40 10 6 51 16 118 171 424

may 28, 2011 vol xlvi no 22

This bias in favour of mahadalits explains the conundrum posed Table 3: Ownership of Means of Production (%)

by the reduced representation of agricultural labourers on the Bihar BPL 2009 list alluded to earlier. Why, despite the 2009 methodology appearing to favour these households, are they comparatively less represented on this list? Agricultural labourers contribute one-third of the total population of the Bihar localities, but less than two-fifths of the population of the BPL 2009 poor. Casual workers on the other hand account for about half the population of these localities, but nearly 60% of the population of the BPL 2009 poor. The issue turns on the “automatic

inclusion” criteria of specified ascriptive identities, which tends towards certain occupations more than others. Mahadalits increasingly seek employment opportunities out of agricultural labour and migrate to towns and other rural areas – over 60% of all mahadalit families in the surveyed localities have at least one family member working in factories, construction sites or other urban areas, whereas only about 35% households depend primarily upon agricultural labour for income. The Saxena Committee justifiably privileges the inclusion Figure 7: Landownership Profile of the Poor of these castes on the BPL list. Inevitably, the numbers of casual workers represented on the BPL 2009


list registered an apparently steep increase.

The social group that perhaps most phenomenally benefited from


the application of the 2009 methodology was the MBC in Bihar. They increased their share in the poor


from 24% in the 2002 list to 30% in

104 10180 51

the 2009 list, in which over 80% of

18 9

all MBC households were included. This is an interesting outcome of the 2009 methodology – it does not priv

ilege MBCs the way it privileges mahadalits, yet households from these castes have tended to be better represented on the poverty list generated through this methodology. Their political antagonists, the OBCs, also registered an increase in their presence on the 2009 poverty list, from 9% to 12%. With this, nearly half of all OBC households were classified as poor.

4.3 Landownership

Given the overwhelmingly rural and agricultural context of our survey sites, the ownership of landholdings is a crucial indicator of economic (as well as social and political) status, and has been recognised as such by both the methodologies. The patterns of landholding in our study sites in the two states are clear from Table 3.

Over three-quarters of the surveyed households in both the states reported that they owned no land. These landless households comprised the bulk of the BPL lists generated by the two methodologies in both states. However, the application of the two methodologies affected the prospects of inclusion of these households in the BPL lists differently. In West Bengal, the number of landless households on the poverty lists actually reduced with the application of the 2009 methodology; landed households

Economic Political Weekly

may 28, 2011 vol xlvi no 22

West Bengal Bihar

No landholdings 1,702 (74) 1,700 (77)
1 hectare unirrigated or 0.5 hectares irrigated 294 (13) 402 (18)
More than 1 and less than 2 hectares unirrigated
or 0.5 to 1 hectare irrigated 213 (9) 61 (3)
More than 2 and less than 5 hectares unirrigated
or 1 to 2 hectare irrigated 62 (3) 33 (1)
5 hectare unirrigated or 2.5 hectare irrigated 36 (2) 1 (0.05)

Source: Survey data and own calculations.

actually increased their presence on the poverty list, from 11% of the poor under the 2002 methodology to 23% under the 2009 methodology. Does this undermine the claims of the Saxena Committee that the methodology proposed is more sensitive to vulnerability and marginalisation?

The story, however, is more complex, as is rural poverty, and the Saxena Committee correctly (even if at times imperfectly) perceives poverty in relational rather than merely commodity-fetishistic terms. A scrutiny of the profile of the 61 landowning households





that find themselves on the BPL 2009 poverty list in West Bengal reveals that 41 households are led by single women and 13 by individuals with debilitating health conditions. At least one member in eight of these landowning households is bonded to servitude. The livelihood profile of the landed BPL 2009 poor households reveals that nearly 70% are agricultural labourers who also engage in some subsistence cultivation, whereas the primary source of income for the remaining households is from casual work, primarily in the construction and manufacturing sector. About 47% of these 239 households have seen at least one member migrate to as far away as Delhi, Mumbai, Chennai and even tier-II towns such as Pune, Allahabad and Ranchi. Those who have not seen any members migrate depend on jobs in Malda town and its peripheral areas such as Naldubi, Narayanpur and Nawabgunj.

A similar dynamic animates the way in which the BPL 2009 methodology interacts with the landowning poor in Bihar. Nearly 15% of these families were mahadalit, which resulted in their automatic inclusion in the poverty lists. Over 95% of the remaining landowning families reported that they derived their household income from agricultural labour or from casual work in urban areas. Interestingly, only about 20% landless families “move out of” the


=2002 Poor (n = 1085) ' = i ' = ' = Remain in’ Move into’ Move out of’ poverty list poverty list poverty list (n = 629) (n = 407) (n = 456) =2009 Poor (n = 1036) =2002 Poor (n = 466) ' = Remain in’ poverty list (n = 367) 0 ' = Move into’ poverty list (n = 1011) ' = Move out of’ poverty list (n = 99) =2009 Poor (n = 1378)
l =West Bengal (N = 2310) i =Biahar (N = 2210)

No land

1 hectare unirrigated/0.5 hectare irrigated

=1 hectare unirrigated/= 0.5 hectare irrigated








24 9







poverty list when the 2009 methodology is applied to the localities in Bihar, in sharp contrast with the West Bengal localities where as much as 45% households make a similar transition.

5 Implications

I have shown, on the basis of select data, that the application of the 2009 methodology in enumerating the poor households in rural India has tremendous significance. It excludes many households who have been protected by the social assistance net on account of being “below poverty line”. At the same time, it also adds many households that have so far been deprived of their entitlements. In this paper, I have aimed to describe some of these changes without necessarily passing judgment on the nature of these changes. However, as a social observer, I wish to draw attention to what I perceive are the implications of these changes for certain groups of poor people.

5.1 Interpreting the Findings and Policy Sensitivity

First, I will draw attention to key criticisms that may be levelled against the 2009 methodology. The most important of these criticisms is that it continues to rely, even if in combination with other methods, on aggregating scores obtained by individual households along five dimensions. Thus, the primary criticism of the 2002 methodology, namely, that it is cardinalises ordinal data (Alkire and Seth 2008: 7; Sundaram 2003: 899) continues to hold against the scoring section of the 2009 methodology. As Drèze and Khera argue (op cit: 56), households obtaining one of ten scores (or even being either automatically excluded or included) means very little by itself and can be interpreted in many ways. For instance, a household obtaining a score of 7 (to use their example), or even a household being automatically included or excluded, can have different deprivations than other households with identical scores or conditions.

This brings us to the next aspect of our critique of this methodology, that it is perhaps not as policy-sensitive as such exercises should be. It seems content to serve a minimalist purpose, which is to identify households for the provision of BPL cards, which entitles the holders to a comprehensive package of services. The scores do not lend themselves to any meaningful policy intervention, which ought to be central to any exercise that claims to make targeting more accurate and efficient.19 An agricultural labourbased SC household comprising only of old people could obtain the same score either by its members having never gone to school (over 30 years ago) or by a member being physically challenged. Both scenarios are hardly comparable.20 Similarly, a mahadalit household and a homeless household are both “automatically included” in the BPL list, but how can this information be meaningfully interpreted to ensure that the poverty of the two different households (caused by two very different set of circumstances) will be addressed, reduced and alleviated?

Although the proponents of this methodology correctly argue that poverty has several socio-economic and cultural antecedents, they pay less attention to those aspects that provide the basis for the state to intervene and target its resources. For instance, arguing in favour of mahadalit castes being automatically included in the BPL list, the Expert Group members justly make the case that these castes have been historically marginalised and have not benefited from even the affirmative action policies of the state. Defendants would argue that since the BPL list is only a list that identifies the “social assistance base” to make BPL cards available,

may 28, 2011 vol xlvi no 22

it is perfectly acceptable for entire groups of historically marginal-open to participatory verification and monitoring. The interpretation ised communities to be included. But “being mahadalit” by itself of the results lends itself not only to targeted social assistance and cannot be equated with “being poor” (or mahadalit-ness with effective policy intervention, but also makes it possible to measure poverty) and doing so is a deeply offensive construction.21 changes in poverty rates and associated social assistance. It cannot be equated with, say, the head of the household being disabled, or destitute. As a Musahar leader in one of the surveyed Social Conflict areas commented, “We are not poor because we celebrate Dina The final and the most far-reaching criticism relates to the politi-Bhadri [a festival unique to Musahars]. We are poor because we cal genies unleashed by the application of the 2009 methodology. are alienated from the primary means of production, namely, That of course is not the fault of its authors, and certainly not land. It is that question that needs to be addressed.” Moreover, if of those groups that will (or will not) benefit, but of a policy indeed the methodology is to be used to make mid-course correc-framework that privileges targeted allocation of resources rather tions or review the social assistance base, it is not clear what rel-than universal coverage of social assistance. Notwithstanding evance such data will have, since the Musahars (or for that mat-the improved targeting of poor households by the 2009 methodoter Saotals/Santals and Shershabadiyas) are not likely to morph logy, two things stand out clearly. One, given the abysmally low their identities into something else that will make them non-poor cut-offs in states like Bihar, exclusion of significant groups of poor (caste Hindus for instance?). Mahadalits in this calculation are people, (as we have already pointed out), will continue. Two, again doomed to remain classified as “poor”, till such time as the state as a consequence of narrow targeting ranges, the introduction of government declassifies them as such. new methodologies will see major shifts in the populations being

In this context, the 2009 methodology may benefit from the classified as “poor” to being “non-poor” and vice versa. As we insights provided by the Multidimensional Poverty Index (MPI). As have seen, given the occupational profile of the different commuthe authors of the MPI, Alkire and Santos comment, “The MPI reveals nities, this has the effect of moving more OBC Muslims and nonthe combination of deprivations that batter the household at the mahadalit SCs out of the BPL list, and of moving more STs, OBCs, same time” (2010: 7). At the core of this approach is the measure of MBCs and mahadalits into the BPL list. Communities adversely household poverty along three dimensions, namely, education, affected are unlikely to give up their existing entitlements, while health and standard of living. The education dimension comprises those who sense they will benefit will increase their pressure on indicators on years of schooling and child enrolment. The health these entitlements. Both trends will exacerbate identity-based dimension is made up of indicators on child mortality and nutrition. conflict. If the new methodology (indeed, any new methodology, The standard of living dimension is further made up of indicators including one that may be based more explicitly on the MPI) is such as electricity, sanitation, access to drinking water, floor type, implemented within a narrowly targeted policy framework, it is cooking fuel used, and ownership of assets. Each of the dimensions likely to result in large-scale social conflict that the political society is equally weighted. The data comprises nominal variables, and is at all levels may find difficult, if not impossible, to manage.

Notes 11 350 out of 514 agricultural labour BPL 2009 poor Chatterjee, P (2008): “Democracy and Economic households. Transformation in India”, Economic Political

1 The four-year delay was due to a stay order passed by India’s Supreme Court on a writ petition filed

12 57 of 259 casual worker BPL 2002 poor households. Weekly, 19 April. by the People’s Union for Civil Liberties, who al-13 609 of 811 casual worker BPL 2009 poor households. Dreze, J and R Khera (2010): “The BPL Census and a leged that it would reduce the number of persons 14 233 of 299 OBC Muslim BPL 2002 poor households.

Possible Alternative”, Economic Political Weekly, 45(9): 54-63

eligible to social assistance, causing several poor 15 31 of 300 ST BPL 2002 poor households. people to lose their entitlements. Government of India (2009): Report of the Expert

16 545 of 583 households.

Group to Advise the Ministry of Rural Develop2 For example, affliction of a household member 17 348 out of 519 BPL 2009 poor households.

ment on the Methodology for conducting the with TB or leprosy.

18 519 of 1378 households, or 38%.

Below Poverty Line (BPL) Census for Eleventh 3 Specific characteristics within each attribute are

19 I am grateful to Sabina Alkire for alerting me to Five-Year Plan, Ministry of Rural Development, scored higher than others, on the argument that

this point. New Delhi, accessed at

these correlate with poverty, such as being scheduled 20 Drèze and Khera make a similar argument (Drèze Guru, G and A Chakravarty (2005): “Who Are the caste is scored a higher mark than being OBC and

and Khera 2010). Country’s Poor? Social Movement Politics and being agricultural labourer gets a household a

21 This is not to contest the fact that the bulk of the Dalit Poverty” in R Ray and M Kaznetstein (ed.),

higher mark than being casual labour.

mahadalits live in poverty compared with other Social Movements in India: Poverty, Power and 4 I am grateful to Jean Dreze for drawing my attencastes, or that they, along with the Saotals and Politics, pp 135-60 (Oxford: Rowman and Little

tion to this limitation

Muslims account for an overwhelming majority of field Publishers). 5 Again, I am grateful to Jean Dreze for pointing me

the poor, as indeed we have also shown through Hirway, I (2003): “Identification of BPL Households to this resource. our data. But recognising that point is conceptu-for Poverty Alleviation Programmes”, Economic 6 I would like to thank an anonymous referee for ally and empirically distinct from arguing that Political Weekly, 14 November. helping me make this argument. their caste/religion makes them poor in the way

Jain, S (2004): “Identification of the Poor”, Economic

7 Twenty-two SCs have been categorised as maha-that landlessness, homelessness or no access to Political Weekly, 20 November. dalit, including Chamar, Pasi, Musahar, Dom, health or protected drinking water sources does –

Mehrotra, S and H Mander (2009): “How to Identify Ram, Dhobhi, and others. The only caste to not be they are problems of a different “order”, which

the Poor? A Proposal”, Economic Political Weekly,

categorised as mahadalit is the Dusadh caste, a may not be comparable.

44(19), pp 37-44.

typical surname of the caste being Paswan. In this Mendelsohn, O (1993): “The Transformation of Austudy, given that the 2009 methodology treats the thority in Rural India”, Modern Asian Studies, 27, two categories differently, I have applied this bi-REFERENCES pp 805-42 furcation, without necessarily endorsing it.

Alkire, S and M E Santos (2010): “Acute Multidimen-Prasad (2005): “MBCs: Political Tsunami”, The Pioneer,8 409 of 831 casual worker BPL 2002 poor housesional Poverty: A New Index for Developing 19 December.


Countries”, OPHI Working Paper 38. Sundaram, K (2003): “On the Identification of 9 186 of 608 casual worker BPL 2009 poor households.

Alkire, S and S Seth (2008): “Determining BPL Status: Households Below Poverty Line in BPL Census 10 40 out of 204 agricultural labour BPL 2002 poor Some Methodological Improvements”, Indian 2002”, Economic Political Weekly, 1 March,

households. Journal of Human Development, 2(3): 407-24 pp 896-901.

Economic Political Weekly

may 28, 2011 vol xlvi no 22 91

Dear Reader,

To continue reading, become a subscriber.

Explore our attractive subscription offers.

Click here

Back to Top