Facilitating risk-sharing through self-help groups in Bihar

While the self-help group (SHG) programme in Bihar has improved access to low-cost credit for women, this article evaluates whether it improves risk-sharing by examining differences in village-level variance of consumption growth. It finds that improvements in risk-sharing occurred only in blocks with significant numbers of pre-existing SHGs. This suggests the importance of the programme’s administrative capacity in the form of a ‘community cadre’, that comprises members of existing SHGs and is responsible for the creation of new groups.

Understanding the impact of a group savings and lending programme on risk-sharing is an important topic from both a theoretical and policy perspective. In our recent research (Attanasio et al. 2023), we identify a frequently overlooked value of savings and lending programmes: namely, their role in reducing households’ vulnerability to risk. Incorporating this value can significantly alter the programme’s cost-benefit ratio. From a policy perspective, this study’s setting helps us understand the nature of risk-sharing in rural communities and identify existing imperfections that might prevent full risk-sharing. While most studies focus on the role of ‘demand’ side determinants such as the socioeconomic characteristics of the region, our research is among the few that emphasises that policy can play a significant role in enhancing the quality of these groups and hence their insurance value.

Implementing the programme at scale

We study the impact of Jeevika, a women’s self-help group (SHG) programme established in Bihar as part of India’s National Rural Livelihoods Mission (NRLM). A Jeevika SHG comprises between 10 and 15 women who reside within the same residential neighbourhood of a village. The core function of each group is to collect savings from individual members and use these to support intra-group lending. These internal funds are subsequently augmented by government grants and bank borrowing. The interest rate on SHG loans of 2% a month is significantly lower than that charged by informal lenders. Decisions regarding loan amounts and who gets a loan are made by group members. The programme operates at scale, targeting coverage of all poor women, with the definition of target households varying across states. Bihar, in particular, uses a broad definition that resulted in the coverage of 12 million households by 2021 (68% of the number of rural households in the 2011 census).

Implementing the programme at scale requires significant administrative capacity at the block level. As in other at-scale programmes, the NRLM addresses this by phasing its growth across blocks and villages in a state. Jeevika thus began in a set of Phase 1 blocks in 2006 and spread to Phase 2 blocks in 2012. Within blocks in each phase, the rollout of the programme is also staggered across gram panchayats¹.

Additionally, NRLM pioneered an innovative programme to ensure the growth of administrative capacity as it scaled. Specifically, it recruited a ‘community cadre’ from existing SHG members, trained them extensively and tasked them with the responsibility of creating new SHGs and monitoring their performance. Recognising constraints on women’s mobility, community cadre members were primarily deployed within a block.

Block-level organisation and variations

Our analysis builds on a prior cluster randomised evaluation of the programme by Hoffmann et al. (2021) that randomly divided panchayats drawn from 16 blocks in seven districts into treatment and control samples. In treated panchayats, the Jeevika programme was initiated in late 2011, following the completion of a baseline survey; in control panchayats, implementation occurred after an endline survey in late 2014. We combine the household surveys with data from an extensive follow-up survey in 2019 of sample SHGs, and from the government’s Management Information System (MIS) which provides information such as the year of formation and membership for the census of SHGs.

The first evaluation of Jeevika found that the programme improved household access to low-cost credit and consequently significantly lowered informal market interest rates (Hoffmann et al. 2021). Despite this, its impact on consumption and other household outcomes was small. One explanation is the considerable differences in socio- and agro-economic conditions across blocks. These differences can cause the impact of the programme to vary significantly across blocks, with the estimated average impact masking this variation.

Importantly, study blocks also varied in their past exposure to the programme. Some blocks were in the Phase 1 set, which commenced implementation in 2006 – well before the beginning of our study – while others were included in Phase 2 of the programme, which started in 2012, at the same time as our study². Though the study sample was drawn from panchayats in which implementation was yet to be rolled out, the inclusion of Phase 1 and Phase 2 blocks resulted in significant block-level variation in the number of SHGs in existence at the start of the study. For instance, while Phase 1 blocks had 186 SHGs per 100,000 population in 2010, this ratio was just 0.26 SHGS per 100,000 people in Phase 2 blocks.

The programme’s dependence on a block-level community cadre for administration meant that differences in the number of existing SHGs translated into significant differences in administrative capacity. This suggests a corresponding variation in the quality of SHGs– even those formed at the same time. Differences in SHG quality, in turn, affect its ability to share risk, a point made in the literature that explores the relationship between risk-sharing and a group’s ability to enforce contracts (Ligon et al. 2002, Albarran and Attanasio 2003, Abraham and Laczó 2018). Well-functioning SHGs accumulate more resources and hence provide the promise of greater financial resources to support intra-group lending in future periods. This higher expected future value gives members the incentive to honour their current contractual obligations, resulting in greater loan activities and hence risk-sharing in current periods.

Methodology

Most studies of risk-sharing test the implication that, with perfect risk-sharing, changes in individual consumption should be unaffected by idiosyncratic income shocks once changes in the aggregate or total income of the groups are controlled for. However, in our study as in many others, reliable data on idiosyncratic income shocks is unavailable. We therefore implement an alternative test of risk sharing proposed by Attanasio and Székely (2004), which in turn builds on work by Deaton and Paxson (1994) and Albarran and Attanasio (2003). Specifically, it uses the prediction that, under a set of assumptions, perfect risk-sharing arrangements imply no variance in consumption growth within the group. Correspondingly, we test for risk-sharing by examining whether the village-level variance of consumption growth between the two survey periods (2011 and 2014) is lower in treatment villages.

Allowing treatment effects to differ across Phase 1 and Phase 2 blocks captures differences in their administrative capacity but also underlying differences in economic conditions: Phase 2 (late entry) blocks lie in Bihar’s Kosi region which is characterised by significantly higher poverty. To separate the influence of administrative capacity from economic conditions, additional regression specifications allow the impact of treatment to vary by the number of pre-existing SHGs at the start of the experimental study, as well as by region (Kosi versus non-Kosi).

Findings

We find a significant effect of the programme on risk-sharing, but only in Phase 1 (early entry) blocks: Phase 1 treatment villages are characterised by significantly lower variance in consumption growth than those in Phase 2 blocks– despite programme entry in these two sets of treatment villages occurring at the same time. Strikingly, we find that this difference is fully explained by differences in the number of pre-existing SHGs in the block in which the panchayat lies, and not by underlying differences in socio- and agro-economic conditions.

Using data from the 2019 survey of SHGs, we show that SHGs in Phase 1 and Phase 2 blocks are similar in terms of socioeconomic attributes such as their size and in the education and caste composition of their members. Despite this similarity, they differ significantly in their quality, as measured by the Panchasutra score, an index of SHG performance which grades performance against expected norms in five areas (savings, lending, participation in group meetings, default on loans, and the maintenance of books of account).³ For our study, differences in loan activity are particularly significant.

Building on the theoretical literature described above, we explicitly test the effect of SHG quality on risk sharing using the measure of quality emphasised in this literature, which is namely the expected amount of (future) SHG resources. Using MIS data, we calculate the expected savings of all village-level SHGs in 2015 (at the end of the study period) as the total of the expected savings of each SHG in the village at that time. This measure reflects the number of SHGs in a village and their year of formation, and so reflects the speed of the programme’s growth within a village. This in turn reflects the programme’s pre-existing administrative capacity within the block. We show that this measure of SHG quality does indeed increase with the number of pre-existing SHGs in the block and that it explains the reduction in the variance of consumption growth in Phase 1 treatment villages.

Conclusions and policy implications

To the best of our knowledge, our study is amongst the first to document that well-functioning SHGs facilitate risk sharing, with an emphasis on well-functioning. We find that improvements in risk-sharing occurred only in blocks with significant numbers of pre-existing SHGs. A unique feature of the programme tying administrative capacity to the number of pre-existing SHGs, provides an explanation for this result. We show that the difference in the number of SHGs explains the greater impact of treatment on risk-taking in early-exposure Phase 1 blocks. Extending our results, we find that this difference also explains variation in the growth of financial resources by village SHGs, and that the expectation of higher resources in the future facilitates risk sharing.

Our research carries strong implications for policy. In particular, it emphasises the importance of administrative capacity for a programme’s success and shows how variation in capacity affects the programme’s effectiveness. In the NRLM, SHGs formed in the early stages of the programme’s expansion provide the personnel to support the programme’s subsequent expansion in late-entry panchayats in the same block. Our results show that late-entry panchayats in Phase 1 blocks benefitted significantly from this policy, while the poor performance of SHGs in Phase 2 blocks reflects the absence of this advantage.

Notes:

Gram panchayats are groups of 2-4 villages that form the lowest rung of each state’s decentralised hierarchy of local government institutions
In Jeevika’s Phase 1 blocks, the programme had started in some panchayats, but not in a sample of untreated panchayats. Then, for the purposes of this study, the World Bank team worked with Jeevika to determine the rollout to uncovered panchayats in these blocks. The remaining panchayats were randomly assigned to treatment and control. Consequently, Jeevika started operating only in these treatment panchayats.
SHGs’ Panchsutra score is internally used within the programme to measure their quality. The 2019 survey independently constructed this score for each SHG, using the same outcome variables and definition as used by the programme.