CAN PEER AND NEIGHBOURHOOD EFFECTS IMPROVE MATERNAL HEALTH OUTCOMES IN RURAL BIHAR? ASSESSING THE ROLE OF FORMAL AND INFORMAL NETWORKS

Prof. Mousumi Dutta (Professor, Economics Department, Presidency University)

Prof. Zakir Husain (Professor, Economics Department, Presidency University)

Dr. Saswata Ghosh (Health Specialist, Center for Health Policy, Asian Development Research Institute, Patna

PROJECT SUMMARY

In economics there is a large body of literature on networks. A network comprises of nodes (agents), and links between nodes (Goyal, 2007). Much of economic literature concerns itself with formation of such connections, and their effect on behavior. Network effects exist if an agents’ behaviour is affected by that of other agents belonging to his/her reference group. Although Leibenstein (1950) had shown that such interdependencies may exist in consumption (bandwagon and snob effects), research on network effects started only in the 1970s, with Rohlfs (1974) noting that the utility of a user of a communications service increases as other consumers use the service as well.

In the 1970s and early 1980s, research on demand externalities was conducted in the context of the telecommunications industry (Rohlfs, 1974; Oren and Smith, 1981). A particular focus of interest was the optimal pricing strategy to ensure that product diffusion is supported and a critical mass of users is achieved (Dybvig and Spatt, 1983). In 1985, seminal papers by Katz and Shapiro (1985), and Farrell and Saloner (1985) started an avalanche of theoretical work on network effects. Since the mid-1990s, a variety of industries have also been covered by empirical studies on network effects (Shy, 2001; Shapiro and Varian, 1999; Rohlfs, 2001).

Over time, the concept of strategeic interdependence between agents was extended as researchers observed that diverse economic, political and social activities are undertaken within and by organized groups of individuals. Such activities have increasingly become the subject of intensive research in economics using game theoretic methods and borrowing the concept of social networks from sociologists. Examples include students influencing each other’s school performance (Sacerdote 2001; Zimmerman 2003), effects of colleagues on employees’ motivation and conduct at the workplace (Guryan et al. 2009; Falk & Ichino 2006), use of social contacts in labor markets (Granovetter, 1994a), choice of pensions (Duflo and Saez, 2003), role of trust and social capital (Dasgupta and Seralgeldine, 1999), farmers learning from each other about how to use and benefit of new technologies (Bandiera & Rasul 2006; Conley & Udry 2001; Conley & Udry 2010; Duflo et al. 2008), patterns of crime (Glaeser, Sacerdote and Scheinkman, 1996), rates of capital investment (Banerjee and Munshi, 2004), and spillovers between firms and financial institutions (Leary & Roberts 2014; Kaustia & Rantala 2015). In recent years, network analysis has also been applied in the field of health (Alexander, 2001; Gross et al., 2002; Kwait et al., 2001; Morris, 2004; St. Clair et al., 1989; Valente et al., 2001; Valente et al., 2004). In contrast to the case of network industries, where profit-maximizing firms own and control the functioning of their network, there is no single entity who owns the above types of network. The focus of economists was, therefore, to develop a framework where individual entities create their own links with others, which shapes the structure of economic and social interaction.

Questions addressed in the literature on networks include:

1. How do network relationships form between individuals?

2. What is the architecture of networks that are likely to form when individuals have the discretion to choose their connections?

3. How are such network relationships important in determining the outcome of interaction?

4. How efficient are the networks that have been formed?

5. How does efficiency of network depend on the way that the benefit from outcome is allocated among the individuals?

6. What is the relation between equilibrium and socially desirable networks?

The study proposes to focus on the second and third question, in the context of maternal health care seeking behaviour in Bihar.

Bihar is a state with a long history of poor maternal health indicators. At the onset of the National Rural Health Mission (NRHM; subsequently expanded to cover urban areas and renamed as National Health Mission) in 2005, Bihar was listed as one of the High Focus states, and special attention was paid on improving health indicators. The latest round of National Family Health Survey (NFHS, 2015-16), however, reports that 12.2% of women aged 15-19 years were pregnant and had given birth to a child at the time of the survey (India: 7.9%). The study also reports that 58.3% of pregnant women were suffering from anaemia (India: 50.3%). The percentage of women receiving ante natal care (ANC) was very poor — only 14.4% (India: 51.2%) of pregnant women received at least 4 checkups, and only 3.3% (India: 21%) received full ANC services. Institutional delivery took place for 63.8% cases (India: 70%). Even among the urban educated and affluent section, maternal health indicators are substantially lower than that of similar socio-economic groups in other states. Improving maternal health outcomes in accordance with the Sustainable Development Goals (SDG) is, therefore, a major challenge before policy makers in Bihar.

The proposed study argues that supradyadic social networks (maps studying links between multiple persons, which can be either ego-centric or socio-centric in their nature) can be utilised to disseminate information about good practices, thereby encouraging optimal maternal health seeking behaviour and improving maternal health outcomes.

The role of networks in modifying health care seeking behaviour by disseminating information has been particularly stressed in recent studies in public health. Studies have established that social relationships can have a significant positive impact on physical and mental health; the role of such relations in improving maternal health outcomes in countries like the USA, Italy, Bangladesh, and Ethopia, has also been demonstrated widely.

Peer and neighbourhood effects can amplify the impact and outreach of existing intervention strategies. In particular, among poor households in developing countries, for whom the adoption of health products and services often remains sub-optimally low, social networks can play a crucial role as peers, kin and neighbours often represent the only source of information on which decision making is based.

Friedson (1960) argues that size of the network, strength of relational ties, density of the network, dispersion of the network, and, diversity of the network are important attributes of the network determining its impact on decision-making and choices. The proposed study will examine whether, in addition to the above factors, 'formal' networks existing between members of Self Help Groups (SHGs) or Micro-Finance Institution (MFI) promote optimal health care utilization, compared to ‘informal’ relations based on kinship, daily contact, and residential proximity. Our hypothesis will be that formal networks, vis-a-vis informal networks, will have a greater impact on maternal health care seeking behaviour. Further, we argue, such networks are an important source of information for women whose partners have migrated, further limiting their sources of information about optimal health practices and Government incentives. This is an important issue, given the high rate of out migration of males from Bihar (52.61 lakh persons, comprising 6.3% of the population of Bihar, in 2001), working as daily workers in other states.

MAIN RESEARCH QUESTIONS

In recent years, policy makers are gradually realising that MFIs and SHGs — originally established with the objective of mobilising women from poor households to improve their economic conditions — may also be used as an instrument for improving maternal and child health outcomes. Despite the growing importance of this trend, there are few empirical assessments of such attempts to influence members’ adoption of the provided health services. The social mechanism underlying the change — how the ‘social fundaments’ embedded in the SHG/MFI networks explain the normative impact of peer effects — is another neglected area of research. We have not been able to trace any Indian study on these aspects, particularly in the state of Bihar, where horizontal caste has created deep fissures within society.

The study proposes to address these lacunae and assess whether the peer effects operating through MFIs and SHGs can influence maternal health care seeking behaviour. The research questions of the proposed study are:

[1] Do supradyadic social networks improve maternal health outcomes?

[2] What are the attributes of social networks that encourage health outcomes? Apart from size, strength, density, dispersion, and diversity of the network we will also consider the nature of nodes that exert significant influence on choices, viz. formal (nodes providing advise and/or assistance to the respondent are members of SHGs or MFI) or informal (nodes providing advise or assistance are either kin or neighbours).

[3] What is the mechanism through which social networks modify decision-making and behaviour?

[4] Do networks play a more important role in families where the male members are absent due to migration?

The study will identify the sources providing information on maternal health care services. We will examine the ratio of such sources belonging to the informal domain (comprising of kin and neighbours) to sources belonging to the formal institutional domain (consisting of MFI/SHG members). Another novelty of this study is that it will identify the nodes that are relatively more influential in modifying maternal health related choices and behaviour, and trace the source of their power to transform choices.

An important attribute of the proposed study is its location in Bihar. This is a state whose social structure is deeply divided by caste-based identity. It is important to study whether the peer effect is restricted within caste groups even in formal institutional structures like MFI/SHGs, or whether they can transcend caste barriers in clusters of women from poor households. This will be an important extension to the studies on 'homophily' in social networks — the tendency of people to form ties to similar others (McPherson, et al., 2001). Whereas existing studies have focussed on attributes like political leanings, social class, attitudes, personality, age, religion, education, and occupational prestige, the role of caste in forming social networks has not been fully explored in the literature.

The study will, therefore, add to the existing stock of knowledge in Social Network Analysis (SNA) theory in public health by helping us to understand the mechanism through which such peer effects operate, how they can be sustained, and why some networks are more effective than others in producing behavioural changes.

POLICY RELEVANCE

Investing in the health of women and children result in significant and long lasting economic and social benefits to society. Both the Millenium Development Goals (MDGs) and Sustainable Development Goals (SDGs) have emphasized in improving MCH outcomes. Unfortunately, only 15 developing countries had attained the MDG target, while 88 countries (including India) are seriously off-target. Considerable regional variations in attaining MCH-related targets within India is also observed, with states like Bihar lagging behind others. On the demand side, lack of information and rigid attitudes are important barriers to adoption of optimal maternal health care services.

Experience with public health interventions has demonstrated that changing the knowledge, attitudes, and practices of target population groups is possible. However, research has also shown that such effects are typically modest and short-lived. The proposed study, based on social network analysis, is expected to address the challenge of permanently modifying behaviour as follows:

• Understanding specific mechanisms causing the behavioural change: Social networks provide the means to understand mechanisms for behavior change. Our study is based on the assumption that people learn, contemplate, acquire information, try, and ultimately adopt new behaviors in the context of their interpersonal relationships. The more complex or challenging the behavior change, the more people rely on their social networks at each stage of change. So analysing social networks and their role in disseminating information about maternal health care practices will help us to understand the mechanisms underlying behavioural change.

• Sustaining change after the resources are withdrawn, or the programme ends: Generally, funding for the services and the change agent training disappears or gets reduced over time. Withdrawal of external support signals to the community that the problem is no longer important, and the effort in sustaining the change need not be continued any more. Furthermore, it is difficult for people to sustain their own personal behaviour change, and many people relapse to their original behaviours after completing a behaviour change intervention. By deploying social network techniques in the behavior change program, sustainability can be achieved because the change is based on local members who remain embedded in the community even after the program is over. The process of change, therefore, is independent of external interest and support, and can be maintained based on local efforts.

• Scaling interventions to have more pervasive social impacts, and/or allow the intervention to be replicated in other backward regions: If MFI/SHG-based peer effects can be used to modify health-related choices an decision-making, policy makers may also contemplate scaling up such intervention strategies to modify behaviour in other domains. Broader social issues related to women empowerment and social change can be addressed. On the other hand, if the mechanism of behavioural change can be identified, social engineering can be employed in other regions to introduce appropriate transformative components into the community structure. This will allow the replication of intervention strategies in other regions.

The study also fits in with IGC 's philosophy of engaging with policy makers to ensure sustainable growth by providing local solutions that strengthen the local institional capacity.

POTENTIAL POLICY IMPACT

In India, a few women in the locality are utilised as core members to form MFIs/SHGs. Such women are typically more empowered, knowledgeable of government schemes, have contacts with local officers, serve as officials in several MFIs/SHGs, and have access to multiple ego-centric networks (such members are located at the hub of a wheel, with the rim delineating other members, and spokes the connecting links between them).

Educational intervention through such dominant members will enable a large number of women to be targeted simultaneously, increase acceptability of the message, and ensure that the information provided is considered to be authoritative.

Further, members of the network could support and reinforce the woman in her efforts to integrate the recommendations of the health care provider into her daily routine.

Finally, providers could assess whether the woman is receiving conflicting information or recommendations from other members of her daily environment and take necessary steps.

This makes it more likely that the intervention will be accepted in the short run, and sustained in the long run.

POLICY STAKEHOLDER ENGAGEMENT

Stakeholders comprise of, at the grass root level, Accredited Social Health Activists (ASHAs), SHG members and women who are currently pregnant or have delivered in the year preceding the study. Before administering the questionnaire, the Investigator will explain the objectives of the study and why it is important. Subsequently, we propose to hold Focus Group Discussion with some of the stakeholders. This will enable us to obtain grass root level feedback from beneficiaries/target groups and potential catalytic agents (ASHAs and peers) about the mechanism for behavioural change through peer and neighbourhood effects, socio-cultural impediments to behavioural change, the possible components of an effective intervention strategy, the potential for its sustainability in the long run, and feasibility of replicating and/or scaling up the strategy in other regions. Such information should greatly assist us in incorporating ground level realities when suggesting policy inputs.

At the policy level, officials of the Health & Family Welfare Department are the stakeholders. We propose to engage with them as follows:

[a] Start of study: Liaison meeting will be held to brief state-level officials about the proposed study, its objectives and the policy relevance of the study. Feedback from the officials will also be sought. Meetings will be held seeking assistance and facilitate survey at block level. Subsequently, we will hold meetings at the district level and block level. We will discuss, with the district- and block-level officials, about the need to improve MCH in Bihar, the objectives of the proposed study, methodology of the study, the policy potential of the study, and nature of assistance sought from them.

[b] After completion of the study: Debriefing meetings will be held at district and state-level to disseminate findings of the study. The meetings will also discuss issues like sustaining behavioural change, replicating the strategy in other regions, and scaling up the intervention strategy to have a wider impact.

RESEARCH DESIGN

The study will employ mixed methods — that is, an effective mix of quantitative and qualitative research methods.

A. DATA COLLECTION:

The quantitative data will be collected on the basis of a primary survey to be undertaken in Bihar. The approximate sample size of 2,400 women will be recruited from women in the age group 15-40 years, who had delivered a child in the two years preceding the survey. The sample will be selected as follows:

[i] Districts will be ranked on the basis of human development indicators (such as percent of literate women, percent of male non-agricultural worker, and percent of SC/ST population) and tercile classes created.

[ii] Two districts will be selected randomly from each tercile class. In each district, the blocks will be ranked on the basis on human development indicators, rates of out migration (using figures from the 2001 Census, because 2011 data is still not available), and number of existing SHGs/MFIs.

[iii] Two blocks will be randomly selected from each district, with one block being taken from the top four blocks, and the other block from the bottom four blocks.

[iv] Ten villages (forming the Primary Sampling Unit) will be selected from each block using the Probability Proportional to Size sampling method (PPS).

[v] In each village we will prepare two household lists of potential respondents—one comprising MFI/SHG members, and the other comprising non-members. Systemic random sampling will be employed to select 10 members from each list.

This survey design will allow us to derive sampling weights, making our data more representative. This will enable us to generalise our results for the entire state of Bihar. Secondly, the hierarchical structure of the data will allow us to employ multi-level econometric models in our statistical analysis. Such models enable researchers to make inferences about a population of groups, and separate out effects due to observed and unobserved group characteristics.

The survey will be outsourced to a field agency from Bihar, identified by Centre for Health Policy (CHP). The field survey team will be trained before the field survey by the Principal and Co Investigators. One Research Assistant will also monitor the survey. In addition, a mid-term review will be undertaken to identify and correct errors in data collection. This will help in improving capacity for undertaking field surveys at the local level.

The field survey will be based on a questionnaire designed to collect information on the socio-economic characteristics of the respondent (control variables), maternal health care services (dependent variable), migrant status of respondent’s husband, and the network(s) to which the respondent belongs (predictor variables). It will be pre-tested using a sample of 100 respondents. Before administering the questionnaire, investigators will obtain verbal informed consent from the study participants by reading out a statement mentioning why such study is being conducted and guaranteeing that the information provided by the participants will be kept confidential and will be used only for research purposes. Moreover, in the statement of informed consent the volunteer nature of participation in the study will be emphasized.

In order to control data quality, Principal and Co-Investigators (PI and CIs) will monitor and cross-check data collection activities during the field survey through periodic field visits. This will be in addition to the monitoring of the field survey by a Research Assistant recruited from Bihar. The data will be edited at the field level by the data-editors. Mismatched and unmatched events will be verified by the supervisors before final acceptance. Such rigorous procedure of quality control will ensure minimization of missing or spurious information. Data will be entered in a well-designed CSPro package through double-entry procedure to minimize human error. The entered data will be cleaned by identifying mismatched and unmatched entries and verified from the original questionnaire.

B. ECONOMETRIC ISSUES RELATED TO CAUSALITY

Most papers on social interactions have used models in which the outcome of each individual depends linearly on own characteristics, on the mean outcome of reference group and on mean characteristics of peer group. An advantage of this model is that it is naturally related to the standard simultaneous linear model (Moffitt, 2001). In a pioneering study, however, Manski (1993) identified some econometric issues in using linear models to understand the mechanism underlying peer effects [see also Moffitt (2001) and Blume et al. (2011)].

Manski (1993) distinguishes between:

1. Endogenous effect (influence of peer outcomes): the propensity of an individual to behave in some way varies with the behaviour of the group

2. Exogenous (or contextual) effects (influence of exogenous peer characteristics): the propensity of an individual to behave in some way varies with the exogenous characteristics of the group, and

3. Correlated effects: individuals in the same reference group tend to behave similarly because they are alike or face a common environment.

The proposed study is interested in whether an individual’s adoption of the health services is influenced by her peers’ adoption decision. There are three econometric challenges in identifying the first (peer) effect using linear models.

Firstly, peer groups do not form randomly, but are themselves the result of individual choices. For examples, in Bihar, caste may play a major role in dictating the persons with whom social connections will be formed. This may confound the estimation of peer effects as the similarity in socio-economic characteristics, and not social interaction, may cause the peers’ outcomes to be correlated.

Secondly, peer groups may be exposed to the same environmental shocks, which may affect the outcomes, even if the peer group members did not interact with each other. Health shocks in a neighborhood, for instance, could make people simultaneously decide to undergo medical check-ups, even without having directly influenced each other in their decisions.

Thirdly, as peers may influence each other simultaneously, it is impossible to determine whether the peers’ behavior is the cause of an individual’s behavior, or vice versa. Even in the absence of correlated effects, simultaneity in behaviour of interacting agents introduces a perfect collinearity between the expected mean outcome of the group and its mean characteristics. Such reverse causality, referred to as the ‘reflection’ problem by Manski (1993), prevents researchers from distinguishing between the endogenous and exogenous effects as there are “social multipliers”.

In the recent years, an increasing number of studies have employed randomized controlled trial designs to study peer effects on the uptake of health innovations as they allow solving some of the aforementioned challenges (Sacerdote, 2001; Zimmerman, 2003; Kremer & Miguel 2007; Oster & Thornton 2012; Dupas 2011a). Commonly, these designs vary the access to an intervention randomly within networks to study how the exogenously created variation in the number of adopters or peers with access influences the individual adoption decision. Although methodically rigorous, such designs may face practical limitations if a treatment cannot be varied randomly within networks. Instrumental variable methods have also been used in several studies variables (Gaviria and Raphael, 2001; Hanushek et al., 2003; Evans, Oates and Schwab, 1992; Graham and Hahn, 2005).

A recent study (Hoffman, 2017) tackles these econometric challenges with an innovative instrumental variable (IV) identification strategy:

“The approach, which was first proposed by Bramoullé et al. (2009, see also Lin (2010) for other contributions), exploits variation in the network structure and the existence of partially overlapping peer groups for the unbiased estimation of peer effects. An overlap in peer groups exists if not all actors in a network are connected with each other and if some individuals are part of different peer groups. In this setting, exogenous characteristics’ of second order peers, i.e. the friends of an individual’s friends who she is not personally connected with, can be used to instrument the behaviour of the directly connected peers. The idea of the approach is that second order peers can only influence an individual by influencing her peers’ behaviour in form of contextual effects allowing to break the reflection problem” (Hoffman, 2017: 3).

We propose to adopt a similar strategy (use characteristics of second order peers as instruments) in our analysis. Following existing studies, we further propose to control for a wide set of background variables and neighbourhood fixed effects to minimise the potential bias due to environmental shocks.

C. DATA ANALYSIS

The data will be analysed using SPSS and STATA. The analysis will start with a descriptive analysis of the data: bivariate analysis will be used to undertake a preliminary test of our hypotheses.

In the second stage of our analysis we will use econometric methods. Given the hierarchical nature of our data, we will use clustered regression models. Our dependent variables will be:

1. Contraception method being used before pregnancy

2. Whether first ANC visit was in first trimester of pregnancy

3. Whether respondent availed of at least four ANC check ups

4. Whether respondent was given 2 Tetanus Toxoid injections

5. Whether respondent was given 100 IFA tablets to prevent anaemia

6. Whether respondent delivered in an institution

7. Whether delivery was assisted by skilled person

8. Whether respondent availed of postpartum check up within 48 hours of delivery

9. Whether respondent was given advice on postpartum contraception method

Most of the variables are binary in nature (with the exception of “Contraception method being used before pregnancy”). We will, therefore, use logit/probit models, in general. In the case of the first variable we will estimate a multi-nomial logit/probit model.

The main predictor variables are: mean outcome of reference group, characteristics of the network, and whether the husband of the respondent has migrated outside Bihar in search of work.

Since our hypotheses relates to the importance of MFI/SHG membership, we will obtain information on whether the network contains members of MFI/SHG to which the respondent belongs to, and (if so) how many. Based on literature survey we will also identify other features of networks that are important influences on health seeking behaviour. Such variables will include the size of the network (the number of people with whom the reference individual maintains contact), the strength of relational ties (the emotional intimacy and the frequency of interpersonal contact between the reference individual and each member of his/her network), the density of the network (the extent to which the members of an individual's social network know and contact one another independently of the reference individual), the dispersion of the network (ease with which network members can make face-to- face contact), and the diversity of the network (characterised as the number of different types of social relationships an individual maintains).

As control variables we will take age, education, occupation, number of siblings, socio-religious identity, number of children, and decision-making power of the respondent. We will also collect information on assets owned by the respondent’s family and use the data to create an Asset Index using polychoric Principal Component Analysis. Among other relevant data are household size, and education and occupation of respondent’s husband. Some relevant supply side variables will also be incorporated.

Since both first stage and second stage models will be discrete choice models, the control function approach (Arrelano, 2008) will be used. It means that we will substitute standardised residuals, and not predicted values, from the first stage model into the second stage model. [1]

C. UNDERSTANDING MECHANISM FOR CHANGE:

In the qualitative part of our survey, we will undertake one focus group discussion (FGD) in each block. The participants will be women from the sampling frame, with about half being members of MFIs/SHGs. We will also ensure that there is representation of women whose husbands have migrated outside Bihar in search of livelihood. Each FGD group will comprise of 8-10 women. Sample size will be about 96-120.

In addition, we will visualize networks and analyze their nature. This part of the analysis will be undertaken as follows:

• We will ask a randomly chosen woman to identify her sources of information and assistance in availing maternal health services.

• Using the snowball sampling method, we will next go to those persons who had helped the starting person, and ask about their sources of information and assistance (if any). (Snowball sampling is preferred to random sampling as the latter tends to generate disconnected graphs, particularly when the number of links per node is few).

• This will be repeated, so that the selected nodes grow like a snowball and selected nodes are all connected.

• We will collect data on the nature of assistance provided, relationship between the agents, and whether they were MFI/SHG members.

• After completing one such cycle, we will choose a different starting node (respondent), and trace the social network to which she belongs in a similar manner.

• This will be repeated for four respondents (starting nodes) in the same site to trace four social networks.

• Two of the respondents, from whom we start tracing networks, will be MFI/SHG members, while the remaining two will be non-members.

• We will undertake this exercise in one village in each district; the village will be chosen from the list of villages covered in the primary survey on the basis of random sampling method.

• Since snowball sampling is employed it is difficult to pre-determine the sample size for network analysis. However, we expect a sample size of about 240 persons.

The data so collected will be employed to visualise and analyse the nature of social networks using appropriate software (like Gephi or Cytospace). The nature of the network will also be analysed mathematically, focusing on the nodes from different perspectives such as how closely related individuals are, who is the connector or hub in the network, who has best visibility of what is happening in the network, and what are the distances and similarities of individuals from each other. We will also identify influential nodes within the network using concepts like degree centrality, closeness centrality and betweenness centrality (Freeman, 1979; Marsden, 2002).

D. DISSEMINATION AND FEEDBACK:

We will dissemination our findings after completion of analysis. One meeting will be held in Patna, with state-level officials and local think thanks. In addition, we will also hold meetings in each district head quarters with local level officials and SHG/MFI members. The objective will be to disseminate results and obtain feedback from all stakeholders. This should help us in the preparation of final report and policy brief.

E. DELIVERABLE OUTPUTS:

Among the deliverable outputs are:

• Periodic progress reports, reporting on work done in trimester

• Blog posts disseminating project objectives, views expressed by stake holders, case studies, non-technical discussion of results, and policy notes

• Policy brief

• Final report

• Academic output, in the form of seminar presentations and publications

A time line for the execution of the study, undertaking of field survey and data collection, and delivery of output is attached as a separate pdf file.

F. ALLOCATION OF RESPONSIBILITIES WITHIN RESEARCH TEAM:

Principal Investigator: Will meet State-level officers before survey starts, assist in training of field survey team, undertake FGD, undertake data analysis, draft report and maintain blogs, and disseminate results.

Co-Investigator 1: Train field survey team, supervise field survey, prepare data entry package, supervise data cleaning/editing process, undertake analysis of social network data, assist in drafting report, take part in dissemination activities.

Co-Investigator 2: Identify field survey agency, undertake initial liaison with state-level officials, prepare survey design, supervise field survey, assist in data cleaning/editing, assist in data analysis, assist in drafting report, take part in dissemination activities.

One Research Assistant (RA) will be employed in the institute of Principal Investigator who will undertake data cleaning/editing, assist in data analysis, and help in drafting of report and maintaining blogs. The possibility of enrolling this Research Assistant in the doctoral programme at Presidency University will be explored.

The second RA will be recruited from Bihar, and will be responsible for supervising the field survey. He/She will also assist the research team during the FGDs and collect the data for social network analysis.

The entire team (PI, CIs and RA) will be responsible for preparing academic papers for presentation in international and national seminars, and eventual publication in an internationally reputed journal.

REFERENCES

Alexander, C., Piazza, M., Mekos, D., & Valente, T. W. (2001). Peer networks and adolescent cigarette smoking: An analysis of the national longitudinal study of adolescent health. Journal of Adolescent Health, 29, 22-30.

Arellano, M. (2008) “Binary models with endogenous explanatory variables”, Available at: http://www.cemfi.es/~arellano/binary-endogeneity.pdf, [Accessed on 01/04/2013].

Bandiera, O. & Rasul, I. (2006) Social Networks and Technology Adaption in Northern Mozambique. The Economic Journal, 116 : 869–902.

Banerjee, A. & Fudenberg, D. (2004) Word-of-mouth learning. Games and Economic Behavior, 46(1): 1–22.

Bramoullé, Y., Djebbari, H. & Fortin, B., 2009. Identification of peer effects through social networks. Journal of Econometrics, 150(1), pp.41–55.

Conley, T. & Udry, C. (2001) Social Learning through Networks. The Adoption of New Agricultural Technologies in Ghana. American Journal of Agricultural Economics, 83 (3): 668–773.

Conley, T.G. & Udry, C.R. (2010) Learning about a new technology: Pineapple in Ghana. American Economic Review, 100 (1): 35–69.

Dasgupta, P., and I. Serageldin (1999) Social Capital: A Multifaceted Perspective. Washington, DC: World Bank Publications.

Duflo, E., and Saez, E. (2003) The role of information and social interactions in retirement plan decisions: evidence from a randomized experiment. Quarterly Journal of Economics, 118: 815–42.

Duflo, E., Kremer, M. & Robinson, J. (2008) How high are rates of return to fertilizer? Evidence from field experiments in Kenya. American Economic Review, 98(2): 482–488.

Dupas, P., 2014. Short-Run Subsidies and Long-Run Adoption of New Health Products: Evidence From a Field Experiment. Econometrica, 82(1), pp.197–228.

Dybvig, P.H. and Spatt, C.S. (1983) Adoption externalities as public goods. Journal of Public Economics, 20 (2), 231–247.

Evans, W.M., Oates, W.E. and Schwab, R.E. (1992) Measuring peer group effects: Studying the behavior of teenage behavior. Journal of Political Economy, 100(5): 966-991.

Falk, A. & Ichino, A. (2006) Clean Evidence on Peer Effects. Journal of Labor Economics, 24(1): 39–57.

Farrell, J. and Saloner, G. (1985) Standardization, compatibility, and innovation. Rand Journal of Economics, 16 (1), 70–83.

Freeman, L.C. (1979) "Centrality in social networks: conceptual clarification", Social Networks, 1:215–239.

Friedson, E. (1960) Client Control and Medical Practice,American Journal of Sociology, 65(4): 374-382.

Gaviria, A. and Raphael, S. (2001). “School-based peer effects and juvenile behavior,” Review of Economics and Statistics, 83(2): 257 - 268.

Glaeser, E., Sacerdote, B. and Scheinkman, J. (1996) Crime and Social Interaction. Quarterly Journal of Economics, 111: 507-548.

Goyal, S. (2007) Connections: An introduction to the economics of networks. Princeton University press: Princeton.

Graham, B.S. and Hahn, J. (2005) Identification and estimation of the linear-in-means model of social interactions. Economic Letters, 88: 1-6.

Granoveter, M. (1994) Getting a Job: A Study of Contacts and Careers. Northwestern University Press: Evanston, IL.

Gross, C. P., Cruz-Correa, M., Canto, M. I., McNeil-Solis, C., Valente, T. W., & Powe, N. R. (2002) The adoption of ablation therapy for Barrett’s esophagus: A cohort study of gastroenterologists. American Journal of Gastroenterology, 97 : 279-286.

Guryan, J., Kroft, K. & Notowidigdo, M.J. (2009) Peer effects in the workplace: Evidence from random groupings in professional golf tournaments. American Economic Journal: Applied Economics, 1 (4): 34–68.

Hanushek, E., Kain, J., Markman, J. and Rivkin, S. (2003) Does Peer Ability Affect Student Achievement? Journal of Applied Econometrics, 18: 527-544.

Hoffman, R. (2017) Following the peers: The role of social networks for health care utilization in the Philippines . Working Paper, No. 17/2017, Vienna Institute of Demography: Vienna.

Katz, M.L. and Shapiro, C. (1994) Systems competition and network effects. Journal of Economic Perspectives, 8 (2), 93–115.

Kaustia, M. & Rantala, V. (2015) Social learning and corporate peer effects. Journal of Financial Economics, 117(3: 653–669.

Kremer, M. & Miguel, E., 2007. The Illusion of Sustainability. Quarterly Journal of Economics, 122(3), pp.1007–1065.

Kwait, J., Valente, T.W., & Celentano, D.D. (2001) Interorganizational relationships among HIV/AIDS service organizations in Baltimore: A network analysis. Journal of Urban Health, 78: 468-487.

Leary, M.T. & Roberts, M.R. (2014) Do Peer Firms Affect Corporate Financial Policy? Journal of Finance, 69 (1):139–178.

Leibenstein, H. (1950) Bandwagon, snob, and Veblen effects in the theory of consumers’ demand. Quarterly Journal of Economics, 64 (2), 183–207.

Lin, X., 2010. Identifying Peer Effects in Student Academic Achievement by Spatial Autoregressive Models with Group Unobservables. Journal of Labor Economics, 28(4), pp.825–860.

Mcpherson, M., Smith-Lovin, L. & Cook, J.M., 2001. Birds of a Feather: Homophily in Social Networks. Annual Review of Sociology, 27(2001), pp.415–444.

Manski, C.F. (1993) Identification of endogenous social effects – the reflection problem. Review of Economic Studies, 60 (3), 531–542.

Marsden, P.V. (2002) "Egocentric and socio-centric measures of network centrality", Social networks, 24: 407-422.

Morris, M. (ed.) (2004) Network epidemiology: A handbook for survey design and data collection . Oxford University Press: New York, NY.

Oren, S.S. and Smith, S.A. (1981) Critical mass and tariff structure in electronic communications markets. Bell Journal of Economics, 12 (2), 467–487.

Oster, E. & Thornton, R., 2012. Determinants of Technology Adoption: Private Value and Peer Effects in Menstrual Cup Take-Up. Journal of the European Economic Association, 10(6), pp.1263–1293.

Rohlfs, J. (1974) Theory of interdependent demand for a communications service. Bell Journal of Economics, 5 (1), 16–37.

Rohlfs, J. (2001), Bandwagon Effects in High-Technology Industries , The MIT Press, Cambridge, MA.

Sacerdote, B. (2001) Peer Effects with Random Assignment: Results for Darmouth Roommates. Quarterly Journal of Economics, 116(2): 681-704.

Shapiro, C. and Varian, H.R. (1999), Information Rules: A Strategic Guide to the Network Economy, Harvard Business School Press, Boston, MA.

Shy, O. (2001), The Economics of Network Industries, Cambridge University Press, Cambridge.

St. Clair, P.A., Smeriglio, V.L., Alexander, C.S. and Celantano, D.D. (1989) Social Network Structure and Prenatal Care Utilization. Medical Care, 27(8): 823-832.

Valente, T.W., Mouttapa, M., & Gallaher, M. (2004) Social network analysis for understanding substance abuse: A transdisciplinary perspective. Substance Use & Misuse, 39, 1685-1712.

Valente, T.W., Watkins, S., Jato, M. N., Van der Straten, A., & Tsitsol, L. M. (1997) Social network associations with contraceptive use among Cameroonian women in voluntary associations. Social Science and Medicine, 45, 677-687.

Zimmerman, D. (2003): Peer Effects in Academic Outcomes: Evidence from a Natural Experiment. Review of Economic s and Statistics, 85: 9-23.