Others

DroughtSudden health

problemCyclone/Drought

flood/hailstormSudden health Pest attack

problem

Pest attackLivestock

epidemicFire accidentDeath of other

family membersBad seed qualityDeath of HoHH

Bad seed quality Family division/Fire accident

divorceRobbery/violence

Death of HoHHFamily division/

Robbery/

violenceHuman epidemic

Figure 3: Percentage of Households Reporting Different Risk Figure 4: Percentage of Households Reporting Different Risk

Events (Karnataka) Events (All States)

35.00

40.00

35.00

30.00

30.00

25.00

25.00

20.00

20.00

15.00

15.00

10.00

10.00

5.00

0.00

0.00

Drought

Sudden healthproblemPest attackDeath of HoHHDeath of other

family membersLivestock epidemicHuman epidemicCyclone/flood/hailstormBad seed qualityFire accidentRobbery/violenceFamily division/

divorceOthers

DroughtSudden healthproblemCyclone/flood/hailstormPest attackLivestock epidemicDeath of other

family membersDeath of HoHHFire accidentBad seed qualityFamily division/

divorce

Human epidemicRobbery/violence

Others

years? (c) Who in the household participates/participated in these quartiles as compared to the bottom two quartiles (Table 3). programmes? The survey covers 25 safety net programmes; For example, a slightly lower percentage of households in the most of them being centrally-sponsored and funded schemes. poorest quartile are aware of the education-related programmes The 25 programmes cover four broad categories, viz, cash than the richer ones but the difference is not large. Most of transfer, in-kind transfer, work fare and subsidy based livelihood the differences (means) are not statistically significant. What programmes. However, only 13 programmes are covered for this suggests is that if the average awareness is high in the analysis because of limitations of sample size. We distinguish village for a specific programme, the awareness for all quartiles targeted and universal programmes among these. is likely to be high for that programme.

Across quartiles, differences in awareness are small for most Differences in awareness between villages: Our survey covers programmes. In general, for majority of the programmes, the 45 villages. We constructed an awareness index for all proawareness percentage is only slightly higher for the top two grammes at the village level, and used village-level factors as

Table 1: Percentage of Households Reporting Risk Events by the Type of Risk Events and by Quartiles

State Orissa Madhya Pradesh Q1 Q2 Q3 Q4 Total Q1 Q2 Q3 Q4 Total 1 2 3 4 5 6 7 8 9 10

Drought 12.39 33.93 30.36 34.82 27.84 22.03 16.22 40.35 50.00 32.17 Cyclone/flood/hailstorm 5.31 14.29 11.61 13.39 11.14 11.86 16.22 35.09 38.60 25.38 Pest attack 1.77 6.25 20.54 16.96 11.36 0.85 6.31 21.93 26.32 13.79 Bad seed quality 0.00 0.00 1.79 3.57 1.34 0.00 1.80 6.14 3.51 2.84 Livestock epidemic 6.19 13.39 11.61 10.71 10.47 0.00 7.21 6.14 7.02 5.03 Fire accident 0.00 2.68 0.00 0.89 0.89 3.39 0.90 6.14 3.51 3.50 Robbery/violence 0.88 0.89 0.89 0.89 0.89 1.69 0.90 0.88 0.88 1.09 Human epidemic 0.00 0.89 0.00 0.89 0.45 0.00 1.80 0.00 0.00 0.44 Death of HoHH 3.54 2.68 3.57 3.57 3.34 0.85 3.60 0.88 0.88 1.53 Death of other family members 4.42 5.36 7.14 3.57 5.12 1.69 2.70 3.51 4.39 3.06 Sudden health problem 27.43 29.46 25.89 29.46 28.06 27.97 24.32 17.54 11.40 20.35 Family division/divorce 0.88 0.89 1.79 0.00 0.89 2.54 2.70 2.63 1.75 2.41 Others 0.00 0.00 0.00 0.00 0.00 0.85 0.00 0.00 0.00 0.22

State Karnataka Total

Q1 Q2 Q3 Q4 Total Q1 Q2 Q3 Q4 Total 11 12 13 14 15 16 17 18 19 20

Drought 49.56 28.32 33.04 33.93 36.22 27.91 26.19 34.62 39.64 32.08 Cyclone/flood/hailstorm 0.00 0.88 2.68 3.57 1.78 5.81 10.42 16.57 18.64 12.83 Pest attack 2.65 0.88 8.04 15.18 6.67 1.74 4.46 16.86 19.53 10.62 Bad seed quality 0.00 0.00 0.89 2.68 0.89 0.00 0.60 2.96 3.25 1.70 Livestock epidemic 1.77 3.54 1.79 3.57 2.67 2.62 8.04 6.51 7.10 6.05 Fire accident 0.88 0.00 0.00 2.68 0.89 1.45 1.19 2.07 2.37 1.77 Robbery/violence 1.77 0.88 0.00 0.00 0.67 1.45 0.89 0.59 0.59 0.88 Human epidemic 2.65 1.77 1.79 2.68 2.22 0.87 1.49 0.59 1.18 1.03 Death of HoHH 4.42 3.54 4.46 2.68 3.78 2.91 3.27 2.96 2.37 2.88 Death of other family members 4.42 0.88 4.46 2.68 3.11 3.49 2.98 5.03 3.55 3.76 Sudden health problem 30.97 17.70 12.50 15.18 19.11 28.78 23.81 18.64 18.64 22.49 Family division/divorce 0.00 0.00 1.79 0.00 0.44 1.16 1.19 2.07 0.59 1.25 Others 0.00 0.88 0.00 0.00 0.22 0.29 0.30 0.00 0.00 0.15

Notes: Idiosyncratic risks include: fire accident, robbery/violence, death of head of household, death of others in family, sudden health problem, long-term health problems, family division/divorce. Covariate risks include: drought, cyclone/flood/hailstorm, pest attack, bad seed quality, livestock epidemic, human epidemic.

independent variables to see if these factors influence aware-Karnataka and Orissa, the corresponding participation rate is ness at the village level. The logit model results are presented more than 60 per cent. in Table 4. Factors (combined for all states) such as better Factors determining participation: Participation in the safety functioning of the panchayat raj institutions (PRIs), status nets depends on several factors such as ownership of assets, of women in the household, presence of an NGO in the vil-social characteristics like caste, occupation of the household,

lage and high (overall) level of education in the village have contributed positively and significantly to creating awareness of safety net programmes in sample villages. However, when disaggregated state-level PRI functioning indices are used as explanatory variables, relative to Karnataka, the coefficients for Orissa and Madhya Pradesh are either non-significant or have a negative sign. It is not clear why in Orissa (negatively related to PRI) and Madhya Pradesh PRI functioning has not led to better awareness and participation – we pursued this issue in our qualitative analysis (Section V). An interesting finding is that awareness of safety net programmes is low in wealthier villages. Understandably, the concern (and probably the demand) for safety net programmes seems to be high in relatively poorer villages. At the village level, awareness overall is better in Orissa and Madhya Pradesh than in Karnataka.

Awareness of the programme is a precondition for participation but it does not guarantee participation. This section looks at participation rates at the household level by quartiles based on wealth/assets and by social groups. All households (all states): The participation rates are more than 60 per cent for in-kind programmes like public distribution system (PDS), Integrated Child Development Scheme (ICDS), mid-day meal and free textbooks. The PDS showed the highest participation rate (70 per cent) among the 13 programmes considered in the study (Table 5). As expected, the participation rates of PDS are higher for below poverty line (BPL) households than for above poverty line (APL) households. On the other hand, the higher awareness rates notwithstanding, the participation rate is less than 10 per cent for programmes like the Antyodaya Anna Yojana (AAY), Sampoorna Grameen Rozgar Yojana (SGRY), food for work and Swarnajayanti Gram Swarojgar Yojana (SGSY). For pension schemes it is around 30 per cent. As for differences between the states, participation rates for all households are higher in eight programmes in Orissa than in Madhya Pradesh and Karnataka. It is a source of concern that only 32 per cent of BPL households participated in the PDS in Madhya Pradesh despite higher awareness. In

literacy, household size, sex of the head of the household, social capital of households, women’s autonomy and participation, village level characteristics like social and economic infrastructure, functioning of local councils like PRIs. The factors determining participation of households in the programmes are examined using multivariate analysis. We have selected 13 programmes, for which the sample size is sufficiently large. For each of these programmes we have estimated a logistic regression model to determine the factors explaining the participation of households:

where P takes value 1, if any member of the household participates in the programme and zero otherwise. H and V are household and village level variables. At household level, we have considered the following variables:

water user associations and women’s participation in the elections (village/block/district/state council and national elections). We expect similar participation in the safety net programme to increase with women’s participation index.

We used age of the head of the household (HoHH) and average land size as controls. At the village level we used the following controls:

relationship with participation in many programmes. For example, the SCs have significant and positive coefficient for eight out of 13 programmes, viz, IAY, rural scholarship, AAY, MDM, free text, free uniform, SGRY and food for work programmes. In comparison with scheduled castes, the households belonging to scheduled tribes have probability of participation in only fewer (four) programmes.

Table 6: Significance Levels of the Variables in 13 Safety Net Programmes

including such things as obtaining a BPL card. Ideally, the poor Table 7: Factors Determining the Contacts with Middle Men

should be able to gain access to BPL cards and other programmes without anybody’s help, but typically that did not appear to be case in villages we surveyed. So we collected the village-level information on the role of middlemen (or other “contacts”) in accessing the programmes, and computed a village-level index, and used various village-level factors to assess what factors influence the dependence of poor households on middlemen to access programmes. Results are reported in Table 7.

Female literacy and women’s general status in the household and households “trust” in public institutions significantly reduce the dependence on middlemen. The dependence on middlemen is higher in the relatively poorer state of Orissa than in the other two states.

By their very definition, the anti-poverty programmes are supposed to be targeted to poor households, i e, the households below the poverty line. We have divided all households into four quartiles in terms of the asset/wealth index. In order to gain insights into targeting efficiency, we look at the distribution of beneficiary households by quartiles. This is shown in Table 8. One would expect the beneficiary households to belong to the poorest quartile or the bottom two quartiles. We see a different picture. For IAY, widows pension, AAY, SGRY, FFW and rural education scholarship, the bottom three quartiles account for the bulk of participants. The beneficiaries are about equally distri buted across all quartiles for PDS, ICDS, MDM, free uniforms and free textbooks. A high proportion of beneficiaries of credit-based micro finance programme (SGSY) belong to the richest quartile. It is clear that with few exceptions, most anti-poverty programmes are reaching the poorest quartile as well as not-so-poor quartiles – thus confirming the widely acknowledged fact of poor targeting.

The distribution of beneficiary households by quartiles does not inform us about the actual benefits (kilos of grain, for example). We have tried to quantify the benefits and estimated distributive shares across quartiles for major safety net programmes. The results are discussed below. Public distribution system (PDS): The PDS benefits targeted to BPL families are supposed to go to the lowest two quartiles (particularly to the bottom quartile). Table 9 shows that the lowest quartile for all states gets around 30 per cent for rice and wheat and 36 per cent and 40 per cent for sugar and kerosene, respectively. The second quartile also gets more than 25 per cent for all the commodities. However, the share of the third quartile is high for wheat at 30 per cent and 25 per cent for rice. The richest quartile also gets 17 per cent for rice and less than 15 per cent for other commodities. The results can be interpreted in two ways. The programme is reasonably targeted to the poor as their share is higher than the top two quartiles. However, there is substantial leakage of benefits to the non-poor.

Statewise details show that the programme effectiveness, in terms of higher quantity of benefits reaching the poor, in Orissa and Madhya Pradesh is far better than Karnataka (Table 9). In Karnataka, the second and third quartiles receive most of the benefits; even the fourth quartile receives substantial benefits. By contrast, in Madhya Pradesh the benefits accruing to richest quartile are negligible for all commodities. Antyodaya Anna Yojana: As one would expect, this programme

fares better in terms of benefit incidence (Table 10). The lowest quartile for all states gets around 36 per cent to 44 per cent of the total. In the case of wheat, second and third quartiles get higher than the first quartile. The share of top quartile was less than 5 per cent for all commodities except rice. At the state level, the performance is the best in Karnataka followed by Orissa. In Orissa, all commodities except kerosene are well-targeted to the bottom two quartiles. In Madhya Pradesh, the third quartile also gets substantial benefits from all commodities. SGRY: Since wage employment may be expected to be selftargeted, the bottom two quartiles should participate disproportionately in the programme. We disaggregated the information also by gender (Table 11). The wage employment generated by SGRY for all states shows that the benefits are going more to the second quartile as compared to the first quartile. More than 80 per cent of the males in SGRY belonged to the second quartile. Interestingly, the targeting seems to be more effective for females than for males. Around 59 per cent of the females working in SGRY belong to the lowest quartile. Among the seasons, works during rabi and summer seem to be more pro-poor for females as seen by the higher share of lowest quartile. In the case of males, the share of the second quartile dominates in all the seasons. Clearly the programme is not attracting the richer quartiles, thus confirming some degree of self-selection. Food for work: The employment generated under food for work seems to be even more pro-poor than SGRY. The lowest quartile has a share of 43 per cent in the total employment generated under food for work (Table 12). In all the seasons, the share of the lowest quartile is the highest. Particularly, the share of the first quartile in summer season (when unemployment rates are high and the opportunity cost of labour is low), FFW employment is 68 per cent. About the benefits to males and females, the results are similar to those of SGRY. However, we have a counter-intuitive result: 28 per cent of the employment is generated by the households belonging to the rich quartile. The reasons for participation by the rich are not known.

To supplement the quantitative analysis, we conducted focus group discussions and also collected brief family histories of selected poor households who either attempted but failed or actually participated and benefited from safety net programmes. Ideally, a qualitative analysis should throw light on those aspects in which the quantitative analysis is either inadequate or throws some puzzles. For example, we found that in two important dimensions, awareness of programmes and the pathways in which households access programmes, the contribution of PRI institutions has been minimal except in Karnataka. In order to unpack some of these puzzles, we organised focus group discussions across the different stakeholders that included the target communities, lower level bureaucracy, NGOs and political leaders.

The respondents from the communities across the three states generally agreed that the functioning of safety net programmes depended a great deal on the institutions implementing the programmes including the local governing bodies (PRIs), elected representatives and grassroot level bureaucracy, and the relevant institutions/agencies channelling the funds, as well as “activeness” of target communities themselves. The latter depended on households’ literacy levels, cohesiveness among women, and the extent of caste, religious and class discrimination practices prevalent in the communities.

local bodies. Representatives of NGOs were of the view that political interference and bureaucratic hassles are the main hindrances for the effective implementation of programmes. Between Orissa and Karnataka states, focus group discussions revealed that in Karnataka, largely due to better functioning of PRI institutions and the grassroot level bureaucracy, targeted communities manage to access programmes better than in other states. However, the respondents from the sample communities of Karnataka have reported that the PRIs and the bureaucracy have neglected the poorest of the poor in providing access to the safety net programmes. This perhaps explains the lack of relationships of PRI functioning and access to programmes that are targeted to the poorest of the poor (such as IAY).

One important finding that emerged from focus group discussions is that better functioning of PRIs can lead to functioning of the grassroot level bureaucracy – as is the case in Karnataka. Respondents from bureaucracy across the states also have expressed this perspective. The relatively better status of decentralisation was evident from better empowerment of PRIs in Karnataka as compared to Orissa and Madhya Pradesh. From what one gathers from respondents in Madhya Pradesh is that the mere fact of decentralisation to empower the PRIs may not work in favour of vulnerable communities given the pervasive social discrimination practised.

The respondents have also reflected on the role of alternative institutions, viz, NGOs and women’s self-help groups in implementing the safety net programmes. The respondents from Karnataka state are not totally in favour of NGOs. They have preferred PRIs to NGOs. This indicates the trust of the targeted communities in PRIs. Moreover, this also indicates that the NGOs are crowded out if the PRI institutions function well. The NGOs are preferred in Orissa and Madhya Pradesh. Women’s self-help groups are preferred as alternatives to improve the implementation process of safety net programmes in all the states, by and large.

This study is based on household-level and village-level surveys on the profile of household risks and the functioning of safety net (anti-poverty) programmes in the states of Orissa, Madhya Pradesh and Karnataka. The survey instrument itself is concise and does not include a consumption module. In all 13 programmes covered in the study, which fall into four broad categories, viz, cash transfer, in-kind transfer, work fare and subsidy-based livelihood programme. The conclusions are summarised as follows:

(1) The evidence in three states reveals that sudden health risk dominates all idiosyncratic risks. Risk patterns vary by states, quartiles and social classes. The proportion of households that experience a health risk is much higher among the poorest quartile than for other quartiles, higher among scheduled castes and tribes than among other castes, and higher in Orissa than in other states. Weather-induced covariate shocks are present in all states and among quartiles, but the incidence is varied. The proportion of households affected is larger in Karnataka (which has a large percentage of dry land) than in other states. In general, a larger proportion of households in the upper two quartiles experienced weather risks, whereas a larger proportion of households in the poorest two quartiles experienced sudden health risks. The risk of livestock epidemic is largest in Orissa.

(2) Awareness is high for some of the in-kind transfer programmes like PDS and mid-day meals and cash transfer programmes like IAY and pension schemes as compared to workfare and subsidy-based livelihood programme. The PDS has the highest awareness, followed by mid-day meal and IAY. Differences in awareness across quartiles are small for most programmes. However, differences in awareness across social groups are significant for a few programmes. In particular, a significantly lower proportion of tribal households is aware of three critical programmes which include, surprisingly PDS, and also education-related safety net programmes. Awareness of programmes is better in Orissa than in other states. At the village level, factors such as status of women in the household, presence of NGO in the village, and high (overall) level of education in the village have contributed positively and significantly to creating awareness of safety net programmes in sample villages. As compared to Karnataka, the PRIs in the other two states are not playing an important role in creating awareness of programmes. An interesting finding is that awareness of safety net programmes is low in wealthier villages; presumably the demand for safety net programmes is high in relatively poorer villages.

accounts for the bulk of expenditure on safety nets in India. In general, the newly introduced cash-based, low budget programmes and workfare programmes seem to be better in terms of targeting outcomes reaching the poorest quartiles disproportionately, than the largest food-based and expensive programmes like the PDS.

(8) The qualitative analysis strongly complements the findings from the quantitative analysis. In particular, it reveals the role of caste discrimination in gaining access to programmes. In both Orissa and Madhya Pradesh, caste discrimination seems to be a major factor lowering the participation rates. Households of disadvantaged social classes do not seem to have much faith in the functioning of PRI institutions in Orissa and Madhya Pradesh. Particularly discriminated are tribal households who simply do not seem to avail of PRI institutions to air their grievances. It also explains why poor households in all three states seem to be using their networking ability to gain access to programmes. The qualitative analysis also shows that in places where caste discrimination is pervasive in PRI functioning, poor households seem to use NGOs and SHGs to access and benefit from programmes. Where PRIs are doing reasonably well, as in Karnataka, poor households do not seem to favour NGOs.

The study points to one gaping hole in the safety net policy/ programme framework. The prevailing safety net programmes do not seem to address the most dominant and pervasive risk of poor households, viz, exposure to serious health risk. Considering that the poor resort to coping strategies of borrowing and working extended hours by women (and possibly also children), episodes of serious illness in the household are the most likely cause of perpetual indebtedness and possibly also to poverty trap situations.

Communitywide weather-induced risks (drought, cyclone and/ or flood) are also experienced by households in all quartiles, but their direct incidence is highest among households in the relatively richer quartiles who happen to own land. Unlike in the case of health risks, there are workfare programmes (SGRY, FFW) in place to provide consumption-smoothing in the wake of such weather-induced risks. However, participation rates in these programmes is low – either because of inappropriate timing of programmes or due to sporadic or untimely release of funds for these programmes.

The risks that are best covered by programmes are those relating to household food security. PDS, MDM, and ICDS are all reaching a large proportion of poor households. But the problem is: they are also reaching a high proportion of households in the upper two quartiles – an immediate result of weak targeting enforced by BPL lists.

Homelessness is a risk that attracted policy attention several years ago when the programme of housing subsidy (IAY) was launched. But gaining “entry” into the programme would require households to possess enough “social capital”, i e, networking ability. Not surprisingly, the programme (which has a very large cash transfer) is attracting households in the upper quartiles. Qualitative analysis had shown substantive abuses and corruption in the actual implementation of this programme.

Recent programmes have begun to cover other risks, viz, the risk of old age poverty, poverty among widows and the risk of children dropping out school. The two pension programmes

– NOAP and widows’ pensions – seem to be reaching poor households, especially widows’ pension is strongly correlated with participation by widows from the poorest quartiles. The education programmes (free uniforms, textbooks, etc) on the other hand, show mixed results. These programmes seem to work well in Madhya Pradesh, but not in the other two states. The fact that the presence of educated individuals in the household is an important factor promoting children’s participation. It suggests that children belonging to parents with no education, especially belonging to scheduled tribes, have not been able to take advantage of this safety net. On the positive side, women’s empowerment seems to strongly influence participation in child-related safety nets. To the extent women are empowered by programmes such as self-help groups, children of deprived communities should be able to take advantage of these programmes.

The study has provided some important insights into the functioning of institutions and social capital. Karnataka clearly stands apart from the other two states in the functioning of PRI institutions, and their positive role in promoting awareness and participation of the poor. Where caste discrimination is pervasive, households do not seem to trust PRI institutions. Instead, they seem to rely on their social capital and caste networks to gain entry into programmes (and even in obtaining BPL cards). As the stranglehold of caste is eased over time, the PRI institutions may be expected to do better in reaching out to the poor.

Looking into the future, immediate policy initiatives need to focus on improving the productivity of existing programmes by encouraging PRI institutions at the village level to promote awareness of programmes among poor households and especially correct the observed serious exclusion errors (as evidenced by low participation of households belonging to scheduled tribes). At the same time the outreach of programmes to the poor have to be tightened via improvements in approaches to targeting (so as to avoid inclusion errors), launch new programmes to cover uncovered risks (especially health risks). Most importantly the PRI institutions should be made accountable for better functioning of safety net programmes with possible external oversight, and work towards synergy between programmes and policies launched by the centre and states so as to avoid duplication of efforts.

EPW

Email: profmahendra@gmail.com

[For comments and suggestions on earlier drafts, we are grateful to the participants of a seminar at CESS and to Philip O’Keefe, Mansoora Rashid and Puja Dutta. The views expressed herein are those of the authors and do not necessarily reflect the opinions or policies of the organisations to which they belong.]