Chindarkar, Namrata Ravindra; Chen Jie, Yvonne (2014). 'The Effect of Social Network on Intra-household Resource Allocation – A Natural Experiment in Gujarat, India' Paper presented at the annual conference of the HDCA, 2-5 September 2014, Athens, Greece.

The role of women in agricultural households and determinants of their well-being within the household has always been of interest to economists. In the unitary household model, household resource allocation and labor supply is determined by a single household welfare function (Becker 1965). The collective bargaining model, on the other hand, assumes that household member reaches a cooperative solution of resource distribution through a bargaining process (Chiappori 1988, 1992, 1997). The actual outcome is determined by individuals' control over potential income, resources or other environmental parameters (Browning et al. 1994, McElroy and Horney 1981). Empirical evidence from developing countries has demonstrated that non-earned incomes, including controls over household assets, land titles and dowries, affect intra-household resource allocation. These findings provide strong support for the cooperative bargaining household model.

Theoretical literature has also shown that non-human capital controlled by women, such as strong kinship or social network support, enhances their bargaining power in the household (Schultz 2001, Folbre 1998). However, very little empirical evidence is available to support this theory.

This paper aims to estimate the social network effect on intra-household bargaining process. We study the problem in the context of a program that provides access to non-farm employment opportunities in Northern Gujarat, India. As part of the collective action and water management initiative, the Self Employed Women's Association (SEWA) has been training women in this region to become repair technicians for government handpumps. As a result, members of SEWA formed social networks that have various levels of access to this non-farm employment opportunity. We compare the outcomes among these women in order to assess the effect of social network on intra-household resource allocation.

In many parts of rural Gujarat, access to drinking water remains constrained, owing largely to unsteady supplies. In many villages, handpumps are the sole source of water during the dry season. The Gujarat Water Supply and Sewerage Board (GWSSB) has increasingly found it difficult to maintain the 10,000 handpumps across the state. As an effort to provide better repair services and guarantee on-time water delivery, GWSSB brought in private and NGO contractors to maintain handpump at designated sub-districts. These private and NGO vendors put in a bid for a repair contract every year. Service contract is awarded to a bidding vendor through a competitive tendering process once a year. Once a contract is awarded, the contractor is responsible for all handpump repair requests within that sub-district. SEWA has been participating in the tendering process since 2004. SEWA and GWSSB provide regular training for women that anticipate becoming a repair technician. GWSSB reimburses all costs associated with the repair and pays these repair technicians a fee for the work performed. In order to minimize transportation costs, women living in the same sub-district as the contract location are more likely to be called upon to perform the repair job.

The empirical difficulty of assessing the relationship between social network and intra-household allocation lies in the fact that formation of social networks in rural households could be endogenous. It could be jointly determined with decisions of marriage, consumption, and labor supply. In the context of this study, it may be that once a woman decides to become a member of SEWA she is more likely to rope in other women from her family and extended family. In this study we overcome the endogeneity issues by exploiting the exogenous variation in the program participation as a result of the location of the service contract. The random assignment of contract location by GWSSB allows us to design a quasi-experimental framework for the study. Women technicians were assigned to different levels of social network depending on the distance between their households and the service location.

The organization structure of the repair program exogenously induces three level of social network access among SEWA members. The first level consists of women that went through the training and are currently working as repair technicians. These women belong to a core group of technicians that regularly perform repair jobs together. They also receive regular support from SEWA in the form of other resources such as transportation and equipment cost. The second level consists of women that went through the training but are not actively participating in the repair activities due to distance. These women have received adequate training to perform the job and have access to the field coordinator as well as other women in the group. However, their social network support and access to SEWA resource is limited comparing to the first group. The third group is SEWA women members who have never undergone the training and therefore never participated in the training program.

Data for this study were collected by the authors in two sub-districts in Sabarkantha, Gujarat namely Bayad and Dhansura. As indicated in Figure 1, these two adjacent sub-districts share similar climate and ground water levels. SEWA has successfully obtained service contract in Bayad in 2013, but not in Dhansura. A total of 225 women and their husbands were surveyed during winter 2013.  Among these women, 118 went through the training program provided by SEWA. A wide range of outcome variables were included in the survey instruments, such as household and individual-level consumption and investment. For each woman, we also surveyed their husband or adult son for intra-household decision making and other family relationship questions. As a measure of the non-monetary effect of the program participation, we also examine women's empowerment level through a series of questions that assess their level of aspiration and subjective well-being. A summary table of total surveyed individuals is provided below.

Key testable hypotheses are:

i. women who are currently working or used to work as handpump repair technicians are more likely to influence consumption and investment decisions within the household as compared to those who have never been trained or worked

ii. women who are currently working or used to work as handpump repair technicians are more likely to have higher aspirations and greater perceived well-being as compared to those who have never been trained or worked