Table 1: Number and Percentage of BPL Households Using the DKA that access to piped water is definitely a

Exclusion approach Inclusion approach Play-Safe Approach Restrictive Approach Number % Number % Number % Number %

165 35.18 302 64.39 343 73.13 125 26.65

Table 2: Number of Households in Each Social Group Possessing Certain Assets

Type Fridge Scooter Landline Colour TV Electricity Total

Casual labourers 4 (2.41) 20(12.05) 33(19.88) 64 (38.55) 138 (83.13) 167

Landless agricultural labourers 0(0) 4(4.25) 14(3.19) 33(35.11) 73(77.66) 94 have access to piped water seem to be

Table 4: Comparison of the Sensitivity of DKA to Alternate Inclusion/Exclusion Criteria

Exclusion Criteria 1: Exclude a Household if it owns any of the baseline assets. These numbers are the same as Table 1. Exclusion Criteria 2: Exclude on the basis of ownership of any of the baseline assets or a multiroom pucca house that is self-constructed. The numbers, expectedly, drop across the approaches. TV-less analysis (1 and 2): The numbers are recalculated here after doing away with the TV as an exclusion criterion. Here, the numbers rise because TVs are a fairly common asset. *Sample size is 249 households.

Table 5: Percentage from Disadvantaged Groups Excluded under Various Exclusion Criteria

EPW

NOTES

self-constructed.4 This, as Table 4 (p 258) shows, halves the number of poor if one uses primarily exclusion-based approaches, but maintains the hierarchy in numbers. Moreover, as Table 5 (p 258) shows, any exclusion-based approach would lead to leaving out significant number of deserving households from the disadvantaged classes.

Recalculating the numbers without using TV as an exclusion criterion leads to a significant increase in the number of poor, especially amongst those approaches that use exclusion criteria as the primary means of separating the rich

Data on amount of land owned was collected for three villages – a total of 249 houses. As Figure A3 (p 262) in the Appendix shows, the strong negative correlation between a household’s total points and land owned is once again striking. There is first a slight increase in the average amount of land owned, between points 0 to 1, and then the graph slopes downward in an almost linear fashion. Those with 0 points represent shopowners, goldsmiths, bank employees and hotel owners, several of whom have sold off their land and have turned to their professions full-time. This might explain the marginally lower

Table 6: Mean Scores and Standard Deviations for Socio-economic Groups in the Sample

Mean Score Standard Deviation Minimum Maximum

Illiterate adults 4.18 2.25 1 11

Landless agricultural labour households 6.16 1.62 4 11 own identification of poor households,

SWHH 3.10 2.34 0 8 obtained in a subsequent visit.

EPW

june 25, 2011 vol xlvi nos 26 & 27

whatever indicator one chooses. Only Table 7: A Comparison of Exclusion and Intersection Sets between the Restrictive Approach and the SCA

eight such houses own agricultural land. Similarly 290 households are classified as being non-poor by both approaches.

The exclusion sets are less straightforward. Of the 57 excluded under the restrictive approach (Table 7), 22 are

DKA Poor (Restrictive Approach) DKA Non-Poor (Restrictive Approach)

SCA poor (points>3) 73

57

SCA, non-poor (points <=3) 49

290

The second and the third cells give the exclusion sets – those households whose classification is in conflict. The fourth cell comprises all those households that both the methodologies classify as not poor. Sample size is 469 households.

Table 8: Households Revisited, Classification and Exclusion Sets

excluded solely on the basis of ownership Total number of Households Number of QP Number of Poor by the Number with Scores Number with Scores Restrictive Score above in Two Villages Restrictive Approach above Four above Three Approach, Three, SCA

of a colour TV – an exclusion criterion

Excluded Excluded

that, as previously mentioned, is question

147 60 52 31 47 18

able. Consequently, DKA comes across as too harsh on some kinds of households – of those excluded, 33 are at least doubly disadvantaged and six are landless and triply disadvantaged.8

On the other hand, there is a significant drop in the percentage of both landless agricultural labourers and casual workers in this excluded set (when compared to the intersection set) and the percentage of SC/ST households almost exactly halves. The average points per household, consequently, reduce from 6.5 to about

4.9. These factors indicate that perhaps indeed these houses are slightly better off than those in the intersection set.

The SCA, on the other hand, consistently misses out on asset-poor marginal farmers who own plots of land of barely half an acre. Agriculture as a means of income is growing increasingly infeasible in the district and could perhaps be even heading for a “total collapse” (Government of Karnataka 2008: 38). The hardest hit are the marginal farmers.

These figures were checked in a subsequent visit using intensive discussions with previously sampled respondents and key informants (including anganwadi and panchayat workers) and personal revisits to reconcile, if possible, the households that were incongruously classified by the two approaches, and also to identify deprived households that escaped the notice of both approaches. This was done in two villages, resurveying 147 of the 185 households in these villages. The result was a list of what may be termed the “qualitative poor” (QP); that is households that a majority of my respondents classified as poor or abjectly poor.9

The number of QP households is higher than those given by all other methodologies (Table 8). Indeed, this accounts for over 40% of the overall number of houses revisited. I would venture to suggest that there might be some upward bias in these numbers; when I personally visited households with conflicting status, I usually found that such households were not as poorly off as they were made out to be.

Table 8 also looks at households excluded in each approach vis-à-vis those that were classified as QP. A cut-off score of three under the SCA still results in over 40% of QP households being excluded. The restrictive approach excludes fewer QP.

The most glaring exclusion of QP housholds under SCA is Shambhu Naik. He is a marginal farmer whose plot of land is

NEW

Price for 5 year CD-ROM (in INDIA) Individuals - Rs 1500 Institutions - Rs 2500

negligible (0.5 acres). He lives with his sister in a thatched hut as his governmentprovided Ashraya house is still being constructed. His house has no electricity; he spoke of going days with little or no food when the monsoon failed. Every single respondent categorised him in the “abjectly poor” category. Yet, the SCA awards him a mere two points and he slips under the radar of the methodology.

The restrictive approach did not make conspicuous errors, at least as far as this sample is concerned. There are few unanimously poor households that the methodology leaves out. Furthermore, there is no

EPW

NOTES

common thread that binds the excluded households.

The revisit also drove home a few ground realities. Separating households that were, quite literally, sitting on the poverty line proved quite difficult. In the end, some will always be left out, irrespective of how accurate an approach one uses. Certain persons – for example, the lower-level staff in a panchayat office, or anganwadi teachers – seemed to have a clear grasp of local conditions. My own analytical findings were not always accurate, but these persons could well tell me, with consummate ease and quite some precision, where most people stood in the income spectrum. Their access to specialised knowledge was beyond any methodology’s or researcher’s scope.

Indeed, of all the QP households in the two villages, only one did not have a BPL card. This suggests, counter-intuitively, that the existing methodology worked pretty well. The BPL list was, in fact, drawn up by a process that could best be described as intuitive. The panchayat secretary and his assistant, ignoring official guidelines, would sit with the voter list, discuss, debate, visit houses and conjure up a list of poor. However, only 73% of the households revisited owned a BPL/Antyodaya Anna Yojana (AAY) card. This points to two lessons. One, if one wants to ensure that all poor households have access to a card, then overall BPL coverag e must be fairly high. Two, the importance of local knowledge cannot be exaggerated.

5 Ease of Implementation

During the course of the survey, it became amply clear that with respect to time taken to implement the survey, the SCA may be an improvement over practices currently in place, but the DKA is quite some distance ahead. As a surveyor, I found that the time taken for completion of the DKA questionnaire was much lower. On quite a few occasions, the inclusion and exclusion criteria were so evident that I need not have asked the questions. For example, in Ajri, I surveyed a household in the ST hamlet. Since STs are automatically included, that took care of all the inclusion-related questions. I was then ushered into a living room that had a TV.

If I was only interested in identifying households based on the DKA, I might as well have thanked the household head and walked out.

In terms of ease of understanding the criteria too, the SCA does not fare as well. Under the DKA, it is typically a matter of common knowledge whether a household has a baseline asset or satisfies an exclusion criterion. While outsiders might not exactly know all the assets a particular household owns or indeed, which inclusion criteria it satisfies, most of them would easily be able to identify one of each from a list of many.

In contrast, I found it rather difficult to explain to even well-read villagers the intricacies of the points system, and the different weights for distinct characteristics. For Hosadu, the final village I evaluated, I managed to arrange for someone else to carry out the study, so that I could mitigate any inherent surveyor’s bias. While he seemed quite at ease asking either of the two sets of questions, I found his DKA-related sheets to be more reliably filled.

Alternative Approach: Recognising that their method might be “too rigid, insofar as it does not allow for weighing and aggregation of different criteria, as in the scoring method” (2010: 61), Dreze and Khera propose an alternative binary scoring method. The essence of this method lies in awarding a single point for each of the inclusion criteria. Therefore, if a household satisfies only one of the inclusion criteria – say, it is an SC household – then it is assigned a score of

1. If a household is doubly disadvantaged

– say, it is an SC household and owns no agricultural land – it is assigned a score of two and so on.

The single, and significant, point of deviation from the usual binary scoring method is that it includes an additional asset-based criterion. If a household does not own any of the baseline assets as defined previously, it is awarded a point that adds to its overall tally. This is in recognition of the fact that the lack of ownership of assets significantly curtails a household’s freedoms.

Any household that has a binary score of at least two – i e, is doubly disadvantaged or worse – is included in the BPL list. Incorporating some form of aggregation, this proves advantageous since it is flexibl e and that inclusion [or exclusion] is not determined solely by a single characteristic.

Under this methodology, slightly fewer than 35% of the households are poor in the given sample. This number is comparable to other independent estimates. The Tendulkar Committee’s Report puts the figure for rural poverty in Karnataka at 37.5%. The Udupi Human Development Report puts the figure at 24% for the district and at 28% for Kundapura Taluk for the 1990s (Government of Karnataka 2008: 48).

This methodology also excludes fewer of the visibly needy. Of those excluded, none are landless agricultural labourers, only three are either SCs or STs and over 90% have at least one adult educated beyond class five.

6 Conclusions

The simplicity of the DKA is at the same time its greatest strength and weakness. Excluding or including households on the basis of a single characteristic in some cases requires such attributes to be very closely linked with poverty. While some exclusion criteria, like cars or fridges, are almost exclusively owned by the rich and therefore share a tight link with poverty, these are also quite narrow, excluding only a few who are well above the poverty line. As shown in this study, any attempt to broaden the criteria – by the addition of, say, TVs to the list

– would result in considerable dilution of this link and exclusion of arguably deserving households.

One would expect the obvious limitations of the SCA with regard to implementation to be overcome by its more comprehensive depiction of household indicators. That, however, does not occur. The SCA does no better and if anything, a little worse at identification and excludes a higher number of arguably deserving households than even the restrictive approach.

In purely comparative terms, the DKA seems to have an edge – while it does no worse in ranking households, it is far simpler to implement. To pick one of the four combinatorial approaches from the DKA is problematic and contextual. Any approach

References

Central Ground Water Board (2008): “Groundwater Information Booklet”, Government of Karnataka, Bengaluru.

Dreze, J and R Khera (2010): “The BPL Census and a Possible Alternative”, Economic & Political Weekly 45(9): 54-63.

Government of Karnataka (2008): Udupi Human Development Report (Bengaluru: Planning, Programme Monitoring and Statistics Department).

Appendix Table A.1: Differences in Ownership of Key Assets

Saxena, N C, Raghav Gaiha, N J Kurian, P Sainath, Harsh Mander, S E Seddey, M N Roy, K Raju, Manoj Kumar Singh, G S Sandhu, A K Misra, Amar Singh, Arun Kumar, Wilfred Lakka, Aruna Roy, S C Gautam and K L Datta (2009): “Report of the Expert Group on the Methodology for BPL Census”, Ministry of Rural Development, New Delhi. Accessed 20 June 2011: http://rural.nic.in/ ReportofExpertGroupChaired-Dr.N.C.Saxena. pdf

that gives primacy to exclusion criteria could potentially end up excluding deserving households; one that gives primacy to inclusion criteria would result in including at least 65% of the population, a fiscally burdensome proposition. The alternative approach looks promising, but has to be tested further.

Finally, the causes for the poverty of several households were occurrences in the immediate past – for example, advent of illness for the primary breadwinner or a poor monsoon or two. The indicators used to characterise their economic status do

not always capture this. Indeed, education of adult members above the age of 30, belonging to historically marginalised sec tion s or even asset ownership, especially land ownership, need not always reflect current economic status. This tension between proximate and distant causes of poverty is difficult to reconcile even theoretically, let alone in implementation.

Notes

1 Dreze and Khera (2010) prefer the term Social Assistance Base (SAB) list to BPL, but I use these terms interchangeably.

2 As the authors state, “[u]ltimately, both exclusion and inclusion errors may be important, but the main concern here is squarely with exclusion errors” (Dreze and Khera 2010:55).

3 Dreze and Khera recognise this fact: “In some areas, living in a pucca house is no indication of economic prosperity, e g, because stone houses are easy to build or the weather makes it hard to survive in a kaccha house” (2010: 56).

4 Using only a multiroom pucca house (and not the if clause that follows) as an exclusion criterion is not feasible since the bulk of government schemes in the region provide for two-roomed houses. Thus, several deserving poor would be eliminated. Dreze and Khera again anticipate this predicament: “some poor households live in pucca houses as beneficiaries of the Indira Awaas Yojana (IAY), a national programme for housing subsidies. In principle, one should consider ‘pucca houses other than IAY building’ as an exclusion criterion” (2010:6).

5 The cap is arbitrary in the sense that it is not based on any normative criteria and is to be seen only as a tool for exposition and comparison.

6 The analysis that follows treats SWHH like other households.

7 Thus implying that any household with points strictly greater than three is included in the BPL list.

8 Here, we do not, in the light of the previous analysis, consider SWHH as disadvantaged households.

9 The respondents were asked to classify original participants into four categories: (1) the abjectly poor (roughly coinciding with the bottom 15%),

(2) the poor (15-30%), (3) average households (30-50%), and (4) those above average (50% and above). In case of conflicting reports as to where a particular household stood, a simple majority rule was used to determine actual status. In nearly every case, respondents could easily point out those who were abjectly poor and those that were above average; those in between, however, were not as straightforward.

262

1

2

012112

100

80

60

40

20

Figure A2: Score Curves for Various Assets

0.9

0.8 07

0.6

0.5

0.4

0.3

0.2

0.1

0 -0.1

3

2.5 2

1..5 1

0.5 0