Can Producer Associations Improve Rural Livelihoods? Evidence from Farmer Centres in India

Abstract

Rural producer associations are considered a potential community-driven solution to the problems of smallholder agriculture. This article evaluates the impact of organising female farmers into producer associations in Gujarat, India. The initiative provided training, information, access to inputs, risk mitigation, and market linkages. Over 18 months, the programme weakly increased members’ non-farm income and access to output markets. It had stronger impacts on members’ awareness and utilisation of financial services. Impacts were heterogeneous, varying by pre-existing socioeconomic conditions. These findings suggest that producer associations can lower transaction costs for smallholders, but that poverty alleviation may be a longer-term prospect.

Acknowledgements

This evaluation was funded by a grant from the Tecovas Foundation, whose financial support is gratefully acknowledged. The authors thank the directors and staff of the Self-Employed Women’s Association and the Global Fairness Initiative for their partnership and assistance in facilitating this evaluation. Surveys for this study were conducted by IMRB’s Social Research Institute, New Delhi. The authors received comments from Alessandra Delgado, Sonalde Desai, Reema Nanavaty, Vijayendra Rao, Aalap Shah, Nisha Shah, Caleb Shreve, and two anonymous referees. All errors and omissions are the authors’ own.

Notes

^1. Eighty-one per cent of all landholdings in the country are under five acres (National Sample Survey Organisation, 2003). 70 per cent of farmers operated less than one hectare of land in 2003 compared with 56 per cent in 1982 (Government of India, 2008).

^2. Exact estimates of spending on these organisations are very difficult to find, but numerous case studies of collaboration between multilateral organisations and local NGOs are found in Rouse (2006), World Bank (2008) and Spielman and Pandya-Lorch (2009).

^3. A typical SHG consists of 10–20 poor women from similar socioeconomic backgrounds who meet once a month to pool savings and discuss issues of mutual importance. Facilitators typically oversee the operations of the group and ‘link’ women to formal institutions such as banks or government programmes. India now has more than 7 million groups, with 97 million members (National Bank for Agriculture and Rural Development, 2011).

^4. For example, we do not find significant differences across the four districts based on levels of female representation in the Gujarat Vidhan Sabha (legislative assembly). Over the past two legislative assembly sessions (2007 and 2012), the cumulative percentage of female members from these districts is: Surendranagar (8.3%); Anand (7.1%); Patan (14.3%); and Mehsana/Kheda (6.3%). See http://www.gujaratassembly.gov.in/index.asp.

^5. Our basic results are robust to the exclusion of these two marginal groups of individuals (64 non-members living in SEWA villages plus 29 SEWA members in villages where SEWA is not present, or 93 individuals) as well as to the exclusion of these 10 ‘mixed’ villages from the sample. These tests are available upon request.

^6. Of these women, 484 had been members of SEWA for more than one year, 47 had been members for 9–12 months, and 26 had been members for 6–8 months.

^7. We do not include marital status. Of surveyed women, 90 per cent are married, a proportion that does not vary between SEWA members and non-members.

^8. We use the binary indicator (landless, landholding) rather than the actual plot size, given problems with plot size measurement in areas where boundaries are not well-established. Note that landholding serves not only as an indicator of farm holdings, but also identifies cultivators from tenant farmers or labourers (that is, those with zero land ownership).

^9. Houses that have walls and/or a roof made of material such as unfired brick or clay, bamboo, mud, grass, reeds, thatch or loosely packed stones are treated as kutcha.

^10. We use linear estimation of binary variables (rather than logit or probit) in order to facilitate interpretation of the magnitude and statistical significance of the interaction effect, given that marginal values of interaction terms used binary regression would need to be computed using the estimated cross-partial derivative rather than the coefficient of the interaction term (see, for example, Norton, Wang & Ai, 2004).

^11. None of the independent variables in the matching function can be considered endogenous to the intervention, including residence in a kutcha house. The programme did not make available any funds that could have been used to improve housing quality over the 18 months of this intervention. Moreover, the timeline used to assess programme outcomes is too short for women to have used additional wages or income gains from the agricultural support programme to finance housing improvements.