WPS7996 Policy Research Working Paper 7996 Who Should Be at the Top of Bottom-Up Development? A Case Study of the National Rural Livelihoods Mission in Rajasthan, India Shareen Joshi Vijayendra Rao Development Research Group Poverty and Inequality Team March 2017 Policy Research Working Paper 7996 Abstract It is widely acknowledged that top-down support is essen- to manage, while the Government of India centrally man- tial for bottom-up participatory projects to be effectively aged other sub-regions. The study finds that the nature of implemented at scale. However, which level of government, top-down management had a large bearing on the nature national or sub-national, should be given the responsibility and quality of local-level facilitation. Centrally and locally to implement such projects is an open question, with wide managed facilitators formed several groups with similar variations in practice. This paper analyzes qualitative and financial performance. But centrally managed facilitators quantitative data from a natural experiment in the state of formed groups that were less likely to engage in collective Rajasthan in India, where a large national flagship project action, be politically active, and engage with other civil that mobilized women into self-help groups for micro-credit society organizations. These results raise important ques- and created a women’s network for other development tions on how responsibilities for participatory development activities was implemented in two different ways. Some projects should be devolved, and how the nature of manage- sub-regions were given to the state government of Rajasthan ment affects the sustainability of bottom-up interventions. This paper is a product of the Poverty and Inequality Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at vrao@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Who Should Be at the Top of Bottom-Up Development? A Case Study of the National Rural Livelihoods Mission in Rajasthan, India Shareen Joshi (Georgetown University) Vijayendra Rao (World Bank) JEL Codes: D70, O10, D72 Keywords: Community Driven Development, Decentralization, Self-Help Groups, India Acknowledgements: The authors are indebted to Radha Khan for supervising the qualitative fieldwork, Ritwik Sircar and Anirvan Chowdhury for outstanding research assistance, SRI- IMRB for collecting the quantitative data for this study, the Government of Rajasthan for sharing valuable data, PRADAN for valuable initial conversations, and the staff of PRADAN, PEDO, SERP and the Rajasthan SHG federations for their support. The authors are grateful to Georgetown University and the South Asia Food and Nutrition Security Initiative for their financial support. For excellent comments and conversations we thank Radha Khan, Sanjay Sharma, Irfan Nooruddin, Milan Vaishnav, Anders Olofsgard and participants of the India Politics Workshop in Washington DC. Introduction Devolving funds and management functions to local communities is an important, and heavily funded, tenet of development assistance (Mansuri and Rao 2012, Smoke, Loffler and Bossi, 2013).1 The practice relies heavily on the principle of subsidiarity, i.e., the idea that local institutions are in the best position to meet community needs because they are close to the people, have good knowledge about local preferences and have strong incentives to promote accountability, reduce corruption and increase the efficiency of resource allocation within their boundaries. While the logic of this approach has broad appeal, one of the key findings of recent studies is that community-level participatory processes function best when they have the active support of the “top”—project implementers who provide technical advice and political support, and who facilitate processes of mobilization, monitoring and adaptive learning (Fox 1993, Mansuri and Rao 2012). In other words development, managed by and accountable to local citizens, works best in a “sandwich.” This however, raises the important question of exactly what constitutes the “top”? Most large-scale community development projects are implemented by governments and their partners. Governments, however, are usually structured into decentralized hierarchies (Bardhan, 2002; Treisman, 2007). The development literature provides little insight into the level of government—central, provincial or local—that should be responsible for implementing such projects at scale. The principles that should guide the devolution of the implementation of local level participatory projects remain unclear. The literature on devolution and subsidiarity would argue that the management of a community- level project should, optimally, be the responsibility of the lowest level of government capable of managing it. In large federal countries, this economic logic can be supplemented with a cultural argument. Sub-national entities such as states and provinces have distinct cultures and notions of sub-nationalism, and a sub-national identity. These constitute symbolic public goods than can facilitate collective action, help build civic capacity and promote economic development (Rao, 2008; Singh, 2016). Sub-national governments are, thus, best suited to implement participatory development projects because they are better able to culturally match (Cornell and Kalt, 2000) the needs of local communities with the design of the intervention and the process of implementation. On the other hand, a counter-argument can be made that central management works best, particularly in countries where provincial and local governments do not have the capacity to 1Devolution is typically defined as a process of transfer of political, administrative and fiscal management powers between central government and lower levels of government, primarily operating at city and region levels (Potter, 2001). 2 implement community-based projects (Bardhan, 2002). Implementing complex projects in low- capacity environments requires the establishment of effective support and management information systems. The large cost of setting up such systems makes central management more feasible than decentralized management. Central management can also help break through cultural norms, such as extreme patriarchy, that are not conducive to development objectives, by translating lessons from one part of the country to another. This question of which level of government should manage community level projects hinges largely on who is best suited to managing the complex task of facilitation. Facilitators are arguably the most important agents of government in community development projects for they are responsible for promoting participation as well as for implementing day-to-day management at the community level: mobilizing citizens, informing them about the working of the program, and facilitating the process of local implementation (Mansuri and Rao, 2012; Perry et al., 2014). 2 Since facilitators are almost always low-level salaried bureaucrats accountable to higher levels of government, their deployment and management can be the key to an effective community- development intervention (Barron, Woolcock and Diprose, 2011; Baiocci, Heller and Silva, 2011). Yet, the vast literature on decentralization and devolution is relatively silent on the relationship between facilitators and governments. This paper examines these issues by focusing on the case of the Indian National Rural Livelihoods Mission (NRLM). This program builds networks of women’s self-help groups in rural areas with the twin objectives of alleviating poverty and empowering women. The large $5 billion program was funded directly by the center with support from the World Bank. For the most part state governments implemented it, but management of some blocks (administrative areas below the district level) was with the center.3 In the state of Rajasthan, a natural experiment in the implementation of the program provides an opportunity to understand how centrally managed community development projects compare in their quality of implementation with projects managed at the state level. We compare “Resource Blocks” which are directly managed by the center, with “Intensive Blocks” which are managed by the Rajasthan state government. We examine the contrast, both in terms of processes and outcomes. We focus on comparing how central and state management affects the work of local facilitators, their attributes, activities and performance across the two types of management strategies, by using qualitative, survey and administrative data collected from villages in the different blocks. 2 Mansuri and Rao (2012) review the literature on facilitation, documenting its central role in community driven projects. Perry et al. (2014) review the history of community health care worker programs in low-, medium- and high-income countries. They argue that in the past three decades, community health workers have played a critical role in helping health systems achieve their potential, regardless of a country's level of development. 3 The program has been modified considerably over the last five years and now gives state s considerable leeway in how it is managed. 3 The next section of the paper describes the NRLM in detail, which is followed by a context- setting section on Rajasthan. We then describe our research methods, followed by a discussion of our findings. We conclude the paper by examining the implications of our findings for policy. The Context National Rural Livelihoods Project The National Rural Livelihoods Mission (NRLM), launched in 2011, is one of the flagship programs of the Ministry of Rural Development4 of the Government of India. Aiming to reach 350 million rural poor in 13 states of India, it is one of the largest community-based projects in the world. The NRLM seeks to create and empower Self-Help Groups (SHGs). SHGs are “membership- based organizations”, i.e., organizations whose members provide each other with mutual support while attempting to achieve collective objectives through community action (Chen et al. 2008). A typical Indian SHG consists of 7-12 poor women from similar socio- economic backgrounds who meet once a month to pool savings and discuss issues of mutual importance. The government regards this program as not only an investment in extending credit to impoverished rural populations, but also an investment into community empowerment. The focus on women is a deliberate attempt to address the wide gender gap and provide resources directly to women (Planning Commission, 2011). 5 The NRLM is modeled after the Andhra Pradesh District Poverty Initiatives Project, also known as Velugu (light) (Galab and Rao 2003). Launched in 2000, Velugu was a joint initiative by the State of Andhra Pradesh and the World Bank in six backward districts of Andhra Pradesh. The program was run by the Society for the Elimination of Rural Poverty (SERP), an autonomous state institute staffed by professionals from both public- and private-sector organizations. While the program’s core involved microfinance and SHG-Bank linkages, it also included additional features such as the establishment of grassroots institutions, the provision of community investment funds, and the training of women’s SHGs to address social problems such as child labor, dowry and gender inequality. The NRLM is implemented at the grass-roots level by a network of community coordinators. These individuals are generally professionals who are rigorously trained and required to live in a village for a month prior to placement. The Andhra Pradesh program has been widely touted as 4 The National Rural Livelihoods Mission is funded from two sources, a $1 billion loan from the Government of India, and a $4.1 billion budget from the Government of India. The World Bank funded component of the project is focused on India’s poorest states (including Rajasthan) and is known as the National Rural Livelihoods Project (NRLP). For the sake of brevity we will refer to project as the NRLM in this paper, though our focus is technically on the NRLP. 5 See http://rural.nic.in/sites/downloads/latest/NRLM_23122010.pdf. 4 increasing incomes, reducing poverty and improving women’s participation in household decisions and civic engagement (Aiyar et al, 2007), and there is some evidence to show that it has been effective to some extent in achieving these goals (Deininger and Liu, 2009).6 SERP is now a formal partner of the Government of India in implementing the NRLM program in numerous states of India. SERP community coordinators travel to states all across the country and mobilize women into SHGs. Politics and Civil Society in Rajasthan Implementation of large-scale poverty alleviation programs in the state of Rajasthan is affected by the state’s distinctive and cohesive sub-national identity (Singh 2016). The landlocked, arid and sparsely populated state has a long history of feudal rule, with rulers well-known for their rivalry and competition.7 As a result, the state has remained quite isolated from the political, economic and social forces at work in the rest of the country; it was a late democratizer and its pace of development lagged behind other states for the first three decades of independence (Tudor and Zeigfeld, 2014; Singh 2016). Rajasthan’s politics and culture coalesced during the National Emergency of 1975-77 when the pre-colonial political structures were effectively harnessed by local elites to generate new forms of political competition (Tudor and Zeigfeld, 2014). A new generation of citizens emerged and local parties were strengthened (Singh 2016).8 This contributed to the development of distinctive Rajasthani politics, civil society and cultural identity. One of the manifestations of this was the rise of new NGOs which emerged particularly in rural areas. Sewa Mandir in Udaipur, Urmul in Bikaner, the Social Work Research Center in Ajmer, and the Gramin Vigyan Vikas Samiti in Jodhpur are the most well-known examples. Most of these NGOs were initially set up to complement the state in the delivery of public services (Singh 2016). 9 In subsequent years, however, they broadened their focus and became important 6 Deininger and Liu (2009) use data from household surveys from 2004 and 2006 to document three main impacts: increases in social capital and economic empowerment, nutritional improvement (despite persistent drought at the time), and an increase in consumption for participants of new groups. The findings did not find increases in income or assets. There was however, evidence of spillovers within the communities in which SHGs were formed. 7 Under British rule, the region was indirectly governed through the Rajputana Agency, a collection of 18 princely states and 2 chiefdoms (Gupta and Bakshi 2008). 8 Singh (2016:104) describes many strategies used by the BJP government to promote the develo pment of a pan- Rajasthani identity throughout the 1990s. She notes that the speeches of leaders were replete with references to heroes of the past, highlighting the culture of valor and independence. The government also promoted tourism, supported cultural fairs and events, and demanded constitutional recognition of the Rajasthani language. 9 Under the fifth five year plan (1980-85), the government provided Rs. 500 million to support NGOs. In 1980, an important amendment to the Income Tax Act allowed corpo rations to deduct donations to NGOs from their taxable income (Kudva, 2005). 5 project implementing partners for international aid organizations (Bhargava, 2007).10 These efforts gained recognition as the pace of economic and social development in Rajasthan intensified. Between 1991 and 2001 literacy rates for females doubled from 20% in 1991 to 44%, and literacy rates for scheduled castes doubled from 26% to 52% (Census of India, 1991 and 2001). These achievements were met with national as well as international acclaim and insights from several Rajasthani programs shaped national policies such as the Sarva Shiksha Abhiyan at the national level. Over time, some NGOs expanded the scope of their activities even further to include political mobilization of the rural poor (Bhargava, 2007). 11 A particularly notable NGO in this regard is the Social Work and Research Center (SWRC) in Tilonia, Rajasthan (John, 2003).12 Offshoots of this organization eventually spread all over the state of Rajasthan. PRADAN (Professional Assistance for Development Action) was established in Alvar in 1987, and also noteworthy was PEDO (People’s Education and Development Organization). A distinctive feature of PEDO and its offshoots is the focus on promoting sangathan (organization), sangharsh (agitation) and samrachna (constructive work) across the state (Rajvansi, 2007: 326). While the SWRC has retained a broad focus on economic self-reliance and environmental issues at the local level, some of the offshoots have addressed other issues. Both PRADAN and PEDO for example, have focused on women. They favored the creation of new institutions and platforms such as women’s self-help groups, cooperatives and collectives. 13 The efforts of these groups have been successful enough to influence national legislation. The National Right to Information Act, passed in in 2005, makes it mandatory for the government to share information with any citizen of India who requests it. The law was largely shaped by citizens’ movements, particularly under the leadership of Aruna Roy’s Mazdoor Kisan Shakti Sangathan (MKSS), a spinoff of the SWRC in Southern Rajasthan. Similarly, the Mahatma Gandhi National Rural Employment Guarantee Scheme (MGNREGS) was also significantly influenced by citizens’ movements in Rajasthan, as well as NGOs that had considerable experience with drought-relief programs in the state (Menon, 2008). The NRLM in Rajasthan The NRLM operates in 17 of the poorest districts of Rajasthan and is locally known as the Rajasthan Rural Livelihoods Project or RRLP. The program began in 2011. By 2013, delays in 10 The Shiksha Karmi program for example, featured the recruitment and training of local teachers by NGOs. The Lok Jambish program scaled up community involvement in education through the process of mapping and micro- planning at the village-level and was again led by NGOs. 11 Some prominent examples are the Jangle Jamin Jan Andolan led by Astha; Movement for Water Conservation by Tarun Bharat Sangh; and Right to Food by Akal Sangharsh Samiti. 12 Headed by Bunker Roy, the SWRC was established in Rajasthan in 1972. It began with a focus on water and irrigation but soon its mission expanded to empowerment and sustainability. It rose to prominence in the 1990s. Roy has been named one of the 50 environmentalists who could save the planet by the Guardian and one of the 100 most influential people in the world by TIME magazine. 13 As of March 2013, Pradan worked with 18,736 SHGs across 7 states, representing a total membership of 252,070 rural poor women. These SHGs have mobilized a total savings of 1230 million rupees. 6 project rollout and implementation had resulted in significant risks to the project (World Bank, 2013).14 To gain momentum, two strategies were simultaneously promoted. The two models were as follows:  The centrally managed “Resource Block Strategy”: External organizations fro m other parts of India were invited into the state to organize the poor. In February 2013, the program first signed an agreement with the Society for the Elimination of Rural Poverty (SERP) from the state of Andhra Pradesh to work intensively in 10 blocks for 3 years. Facilitators, or Community Resource Persons (CRP) in this arm of the RRLP were intended to be from outside the state. Teams were required to work for 2 years without interruption. Each external CRP team was to work for one month in one cluster and take one month gap/break after each round. The CRPs teams were to be fully supported by the Rajasthan government.15 For the rest of this paper, we refer to these facilitators as “External Facilitators”.  The locally managed “Intensive Block Strategy”: Experienced local SHG members, with significant experience of organizing in their communities, were invited to continue existing work in affiliation with the state and to scale up their work. These local CRPs were asked to form new SHGs, strengthen pre-existing SHGs, identify and train bookkeepers for SHGs, identify and train potential CRPs from the local area, etc. Many of these local CRPs were, and still are, in fact closely aligned with reputable NGOs. As a result, these NGOs provide considerable support to the RRLP, often providing staffing and assisting in trainings. They also provide important support to project facilitating teams in the intensive blocks. PRADAN and PEDO are particularly important since they have a long history of successfully organizing women in rural Rajasthan and constructing federations. In January 2013, the Government of Rajasthan formed an agreement with six federations to implement the RRLP program. Facilitators were invited to participate in a 4-day long training workshop. By April 2013, 28 CRP teams were deployed in 7 blocks.16 For the rest of this paper, we refer to these facilitators as “Internal Facilitators”. These are two entirely different development strategies. In the resource blocks, the facilitator is an “outsider” from Andhra Pradesh. In intensive blocks, on the other hand, the facilitator is a 14 The official Implementation Status and Results Report, written in Feb ruary 2013, classifies the “Overall Risk Rating” of the project as “Substantial” and mentions that the RRLP was at least a year behind schedule and had only spent 0.54% of its allocated funds in the first 16 months of its operation (India - Rajasthan Rural Livelihoods Project (RRLP): P102329 - Implementation Status Results Report: Sequence 05, World Bank, 2013). 15 In the 1st phase, SERP deployed 20 PRPs on 23 February, 2013 and 20 CRP teams on 9 March, 2013. These have been deployed in 5 blocks of 5 RRLP districts, namely Baran, Churu, Jhalawar, Kota, Tonk. The rotation was to last until April 7, 2013. The next batch of 22 teams of CRPs and PRPs was deployed by the SERP in the month of May 2013 and would include blocks in Dungarpur as well as Udaipur districts. 16 The blocks were Bari and Bashedi of Dholpur district, Dausa block of Dausa district, Taranagar of Churu district , Simalwada block of Dungarpur district, Kolayat of Bikaner district, and Sarada of Udaipur district. 7 local woman who has worked with local SHGs for a long time. In addition, the distinction between these two types of facilitators goes far beyond the difference in facilitator appearance. They represent different paradigms of state-backed community development programs. In the resource blocks, a team of visitors from Andhra Pradesh is viewed as a representative of the Government of India whose task is to link poorer women to the state. They are viewed as bringing experience, training and services to the village with clear goals for the short- and long- term.17 In Intensive Blocks however, a team of local women, provided with little visible support from the government, are viewed as community members building on existing networks and capacities for long-term change. Their relationship to the state is not obvious – they are building on the Rajasthani sub-national tradition of social mobilization through non-state actors. Their goals, and their alignment with state development plans, are much less explicit. The decision to have two separate development strategies within a single state was unexpected and sudden. It emerged in January 2013, two years after the program began and when it was 18 months behind schedule. The initial plan, formulated in 2011, featured a phased rollout to approximate a randomized control trial: All blocks in poorer districts in Rajasthan were ranked in order of their Human Development Indices. Within each district certain blocks were randomly assigned to the NRLM program for immediate rollout. All other blocks were supposed to serve as controls until 2016. A baseline survey, conducted in 2011, ensured that the treatment and control blocks were balanced along the lines of key indicators (as will be seen later in this paper). However, when the program fell behind schedule and there were significant delays in implementation the plan had to be modified. Resource Blocks and Intensive Blocks were allocated randomly within the treatment group, and NGOs were then invited to submit applications to partner with the government in the Intensive blocks. This was done through announcements in newspapers and regional workshops in the summer of 2012. In early 2013, the lists for Resource blocks and Intensive blocks were made public. The intervention started soon afterwards -- SERP staff arrived in Rajasthan to implement programs within six weeks of this announcement. Our surveys were conducted a year after these announcements giving the program time to take effect. Data and Research Methods Our comparative mixed-methods approach gathered both qualitative and quantitative data from Resource Blocks and Intensive Blocks. The qualitative data were gathered between July and 17 The National Mission Management Unit (NMMU) of the NRLM had planned in 2012 to create 100 community managed and community driven intensive Resource Blocks across the country. The goal was to develop, through an intensive social mobilization process, a few blocks (2-6 blocks in NRLP/NRLM states) as role models. In the design of this strategy, it was emphasized that experienced project professionals , drawn from resource organizations , would receive considerable support from the cluster, block, district and s tate level to spearhead social mobilization and institution building. 8 October 2013. Field trips were made to Jaipur, as well as to districts where the RRLP program was being actively scaled up during the months of our study: Dausa, Tonk, Alwar and Dungarpur.18 A series of in-depth individual interviews were conducted with various stakeholders including the facilitators from four organizations, facilitators who were working for the RRLP under direct contracts, women participating in SHGs, and RRLP project staff who were involved in all aspects of the program’s design. At RRLP headquarters, we interviewed the project director, the deputy director, the monitoring and evaluation specialist, the Resource Block coordinator and the Intensive Block coordinator. Within the Intensive Blocks, we interviewed the senior staff of the NGOs who were present in the district offices. Within the Resource Blocks, we interviewed the most senior project officer in the state of Rajasthan, who was in charge of coordinating SERP activities along with the government. 19 Three facilitators at each NGO were randomly selected from project rosters to be privately interviewed. All interviews were conducted in a secluded area, respondents were assured that all information would remain confidential, and they had the opportunity to terminate the interview at any time. The open-ended interviews were subsequently recorded (in audio files), then transcribed and translated. The insights achieved during the qualitative analysis were then used to design a quantitative survey of facilitators.20 In May 2014, we interviewed 349 RRLP facilitators from the same four districts studied in the qualitative phase. Within the blocks we aimed to interview all active PRPs in the month of our survey. Interviews were conducted by the Indian Marketing Research Bureau (IMRB) at the RRLP field offices. The study was intentionally designed as a comparison of CRPs from SERP versus local NGOs, (PRADAN, PEDO and federations of SHGs). In Dungarpur, we interviewed 120 CRPs from PEDO. In the three districts of Dausa, Alwar and Dholpur, we interviewed CRPs from PRADAN as well as local federations. In each of these cases, as an incentive to complete our survey, we provided PRPs a small remuneration (Rs. 100) for travel costs. Despite these efforts however, we were unable to interview approximately 10-15 18 We chose the four blocks in our survey because this is where the key organizations were scaling up their efforts at that particular time. Balance tests of these areas suggest that they were similar to each other, and also similar to other RRLP districts of the state. This was checked extensively prior to launching the survey in May 2014. 19 We originally planned to interview the SERP workers back in their home state of Andhra Pradesh, after they returned from deployment. An entire research team at IMRB was hired and trained to make this possible. On the Friday evening before the interviews began, however, we were informed that the senior officials of SERP had decided against conducting the survey. Since the study was already underway, we decided to wait till the next cohort of facilitators arrived in Rajasthan and interview them there. We received permission from the Government of Rajasthan for our study, but given the late hours kept by the facilitators we had to conduct man y of our interviews by phone. 20 This was done in a two-step process. In July 2013, the authors and the research team prepared preliminary questions in a structured questionnaire for facilitators with open -ended responses possible for many questions. The interview transcripts were read, analyzed and compiled into a report that was circulated for deliberation within the research team. This was then used to refine and revise the survey questionnaire, which was then translated into Hindi and Telugu by the survey firm. In March 2014, the questionnaire was pretested by the authors on facilitators from SERP as well as PRADAN. Further edits were made prior to training the survey enumerators. 9 CRPs in each of these districts, primarily due to idiosyncrasies in schedules and other obligations on the day of the survey. We interviewed 40 CRPs from SERP out of a total of 120 who were actively working in Rajasthan at the time of the survey.21 We supplement this qualitative and quantitative data with two other sources of secondary data that are intended to provide perspectives on the operational aspects of the program as well as its impact. A survey of RRLP districts: A baseline survey of RRLP districts was conducted in early 2012 with the intention of evaluating the impact of the program. For the purposes of this paper we focus on the “treatment” areas where RRLP was scheduled to work in order to check if Intensive and Resource blocks had statistically similar characteristics prior to the entry of NRLM facilitators. This sub-sample consists of 3,852 households with 6 villages sampled per block in each of the 32 RRLP treatment blocks, scattered across 17 districts.22 Administrative data: Aggregate block-level outcomes are obtained from the Rajasthan government’s online progress report. We examine block-level outcomes of SHG performance as well as outcomes of individual SHG members to compare the performance of Intensive and Resource blocks. A comparison of individual SHG members in Intensive and Resource blocks is done using a survey of 302 women, split across the two types of blocks in October 2014. Analysis We should first note that the two types of blocks were largely well matched prior to the start of the project. Using the baseline household survey data, we can compare pre-program differences between the Resource Blocks and Intensive Blocks. These results are presented in Table 1. The Resource Blocks appear to have fewer SC/ST (Scheduled Caste/ Scheduled Tribe) members and slightly higher rates of home ownership, but most differences between the two groups are small and not statistically significant, particularly when it comes to existing levels of SHG membership. Our qualitative research suggests that RRLP staff and program directors had some knowledge and familiarity with NGOs that were working in at least two of the intensive blocks. However, the operational scale of these institutions was small, and since the selection of blocks for the two 21 We originally sent survey enumerators to Andhra Pradesh to interview th e SERP facilitators while they were home and taking a break from their rotations in Rajasthan. The head of the organization however, denied our surveyors access and refused to allow any surveys in the state of Andhra Pradesh. We were able to complete our surveys with tremendous difficulty only when the women went back to Rajasthan. The Government of Rajasthan allowed the surveys and overruled the NGO’s directives. The busy schedules of the SERP facilitators however, made it impossible to interview all women on a particular day – the division of labor between the women required some women to be constantly working on either cooking, cleaning, washing clothes, going out into the village to conduct mobilizations, and/or record key data and report this to headqua rters. 22 2,914 control households in 6 villages sampled per block in each of the 19 non-RRLP blocks in the same districts as the treatment villages were also surveyed, but we do not analyze the control data in this paper. 10 types of programs was done randomly, they were quite comparable. In our conversations with the director of the SERP program as well as with the SERP facilitators, we found that they were keen to work on their own, but did not always speak the local language fluently.23 As a result, they were generally given strictly defined and narrow tasks to perform in a specific amount of time, with clear deliverables. A Comparison of Facilitators ’ Characteristics We consistently find a distinct difference between the observable characteristics of external and internal facilitators. External facilitators were younger and more educated than local facilitators (Table 3, Panel A). They had fewer children, were less likely to own land and were less reliant on agriculture as a source of income (Table 3, Panel A). Our qualitative observations of these facilitators confirmed these patterns. We observed that the facilitators from Andhra Pradesh spoke up to three languages (Hindi, Telugu and English) and talked, freely and unprompted, about their own personal path out of poverty and their professional experiences. They presented themselves as a group of disciplined anti-poverty professionals whose best qualifications were their own personal journeys of transformation. In rural Rajasthan they stood out for their distinctive way of talking and dressing as well as for their confident demeanor. This was not seen in the case of the local facilitators, who appeared to fit in much more easily with the prevailing norms. In our survey data, however, we do not find an “experience gap”: both external and internal facilitators seemed to have similar number of years of experience (Table 3, Panel A). There were however, significant differences in reported income (Table 3, Panel A). According to official policy, external facilitators were paid Rs. 1,000 per day, and worked for 30 - 45 days at a stretch, earning more than Rs. 30,000 - Rs. 45,000 in a single rotation with the RRLP. They completed 4 rotations in a year, thus earned up to Rs. 180,000 annually (approximately $3,000, using market exchange rates at the time). Internal CRPs on the other hand, were paid Rs. 750 per day and worked for about 90 days in a year, earning approximately Rs. 68,000 (approximately $1,133, using market exchange rates at the time). In other words, internal CRPs earned about a third of the earnings of the external CRPs. In our conversations, SERP management and government officials justified the higher salaries of the external facilitators on the grounds that they had more experience and were leaving their homes and communities to come and work in another state. To test this assumption, we perform a Mincerian regression of monthly earnings, (i.e., a regression of earnings on education, experience and observable attributes [Table 4]). This specification, widely used by labor economists, allows us to examine whether the income “premium” observed among SERP staff in the descriptive statistics may be driven by their observable professional attributes. In the results in Table 4, we find no evidence that the monthly 23In these instances, they were paired with individuals from within the SERP team who had a stronger grasp of Hindi. 11 earnings of the external facilitators are explained by their education, experience or stated responsibilities: the coefficient for “SERP” is positive and statistically significant. The magnitude of the coefficient is consistent with the higher income we observe in the descriptive statistics, suggesting that the income advantage of this group is attributable to unobserved characteristics and attributes. Interviews with field workers, RRLP staff and senior administrators of the program suggested that the higher salaries were intended to compensate “outsiders” for the inconvenience of traveling very far from their home state to live in the difficult conditions of Rajasthan. The wage premium can also be explained as a “reward” to the external facilitators for being success stories and role models who escaped poverty through SHG participation. Comparing Cultures of Program Implementation Training and Responsibilities Both sets of facilitators had the same job requirements. All facilitators were required to mobilize village women, establish and co-opt SHGs, train them, hold meetings, follow-up with group members, maintain records, and provide advice wherever needed. The training processes for the two types of facilitators were, however, very different. External facilitators came into the project with a high degree of disciplined training. In most cases, they had been taught professional methods of mobilization and had applied these methods in areas other than their own state of Andhra Pradesh.24 The external CRPs were also provided training by the Rajasthan government prior to taking their field appointments. As illustrated in Table 3, Panel B, they were well versed in mobilization methods such as social mapping and participatory rural appraisals. They also worked in much smaller groups, conducting mobilization with a total staff of approximately six people, while others local facilitators used double that number. When we visited the external CRPs, we observed stacks of manuals and guide-books in their living quarters. We also noted that they had a kind of cadre-like discipline, with coordinated responses to our questions, using almost identical language to describe their activities and life stories. Training for internal facilitators was much more variable. Our interviews with the staff of the local NGOs suggested that the quality of training was largely dependent on the relationship between the state and the particular NGO with which the facilitator was aligned. One NGO in our sample, PRADAN, took a leadership role in the training and their facilitators were taught resource mapping as well as PRA and other mobilization techniques. Another NGO in our sample, PEDO, had a very different working relationship with the project and a different 24In our interactions with SERP facilitators we directly observed that they had access to a set training materials and methods, which they brought with them from Andhra Pradesh. 12 methodology of work. Consequently, it did not adopt the PRA techniques or use resource mapping exercises.25 Interestingly, there was a bigger difference between insiders and outsiders in the knowledge of methods rather than the actual use of methods. In our fieldwork we noted that the external facilitators did not always use the methods that they were trained in. Rather they used discretionary logic to identify the marginal and excluded, drawing on their personal experience of poverty in their home state of Andhra Pradesh. This was because the SERP facilitators believed that they were intrinsically better equipped to identify the poor due to their high level of empathy. Table 3, Panel C confirms this: outside facilitators are more likely to have knowledge of methods such as wealth ranking, poverty mapping and PRA analysis, but the difference in the use of such methods is much narrower. Strategies of Mobilization Mobilizing the poor in rural Rajasthan is a challenging task. Villages are often far apart and the hot desert climate limits intra-village movement, particularly during the warmest hours of the day. Mobilization involves many conversations with locals, both in groups and individually. Women need to be convinced to set aside their time to participate in-group activities and step out of their homes. Local facilitators were able to use their personal networks, their personal stories and their past experience of working in rural Rajasthan to create interest in the SHG program. However, since they were working from within the system, they needed to be careful to respect local norms. They did not, for example, work at night or travel unescorted. They were very conscious of Rajasthan’s feudal culture, as well as the risks they would face as women working in villages other than their own. Given that they were familiar with the places where they were working, the Rajasthan government expected them to make their own arrangements for food and accommodation. The government largely took a back seat to their operations and did not actively prepare the village or the community for their visits. External facilitators operated very differently. The full force of the machine of government was mobilized to assist their work. Their visits were announced ahead of time, and powerful elites in the village prepared for their arrival. As shown in Table 3, Panel C, the external facilitators were 40% more likely to report that the village was prepared for their arrival, almost 70% more likely to report that the local Panchayat announced their visit ahead of time, and 30% more likely to hold their meetings in a public location. As a result, they were able to form groups more quickly than their local counterparts. They were 7% more likely to report having targets for the number 25Compared to all other facilitators, PEDO facilitators are 17-30% less likely to report either familiarity with professional methods or use of such methods. 13 of people they had to mobilize, and the number of groups that they had to create. They also achieved these targets more quickly than their local counterparts (Table 3, Panel C). In our fieldwork, we found that external facilitators used their status as “outsiders” to create new systems of working. A salient aspect of this is that they often chose to work late at night, sometimes holding meetings until 2AM, almost unaware of the dangers of doing so.26 We observed that the Rajasthani culture of hospitality towards outsiders often served to benefit the external facilitators. There was a great deal of curiosity about them, and their distinctive language, appearance and diet, made them highly visible in a way that local women were not. In our qualitative work, we found mixed opinions about whether this was an advantage or disadvantage, but most agreed that external facilitators found it easier to get the attention of villagers and develop an initial interest in the SHG program. Differences in Impact Self-Reported Measures of Impact In our facilitator survey, we asked respondents to provide examples of the impact of their mobilization. Specifically, we asked them to describe an example of collective action the women had taken that made them most proud (as mobilizers). 27 The response was left open-ended and we post-coded the responses into the following groups: woman-specific issues (dowry, domestic violence, women’s credit groups, child -care issues, etc.), political issues and issues related to public service delivery. To test the hypothesis that external facilitators induce collective action differently than internal facilitators, we ran the following regression: = 0 + 1 + 2 + where is a measure of collective-action outcomes reported by facilitator f in block b, is a dummy variable that takes value 1 if the SHGs in the block are being facilitated by external facilitators, is a vector of controls for facilitator f (this includes her age, education, experience, job-description and salary), and is the error term. We used the logit model to estimate this question, with local facilitators as the base (excluded) group. Given the results of our qualitative fieldwork, we expected external facilitators to report more examples of female empowerment as defined by individual stories about women using the SHG platform to secure financial independence, gain bargaining power in their households and 26 Their supervisors told them that this was ‘free’ time, and that village women would u nderstand that these meetings were for their benefit and so would be willing to temporary relax prevailing social norms. A Rajasthan government official told us that he once had a 2AM call from a SERP facilitator asking for transportation because she was standing alone at a bus stop and needed to get back to her lodging site. 27 The question was worded as follows: “Provide any example of an SHG woman engaging in collective action?” 14 improve their voices in the community by taking a stand against alcoholism, domestic violence, etc. However, given the close working relationship with the government, and the time pressure to meet narrow NRLM goals while in Rajasthan, we expected fewer examples of collective action on political issues or public service delivery—those take longer to develop. Previous work on SHGs in Rajasthan found that it took more than two years to see improvements in water and village infrastructure (Deininger and Liu, 2009; Desai and Joshi, 2013). Results of logit regressions are presented in Table 5. This table presents the odds ratios of the coefficients from each of the individual regressions. We note that—as predicted—external facilitators are considerably less likely in their responses to provide examples of collective action related to politics or the delivery of public services, and these estimates are statistically significant at the 1% level. Specifically, once the other variables in the model are held constant, the relative odds of external facilitators witnessing collective action on issues related to politics and public service delivery are lower by a factor of 0.05 and 0.02 respectively. The relative odds of collective action on women-specific issues is higher for external facilitators, but the coefficient is not statistically significant at the 1% level. We can conclude from this that even though external facilitators are quite successful in mobilizing women and forming groups, their alignment with the priorities of the state and their narrow focus on group formation may limit their effectiveness in mobilizing women toward deeper political and social change. Local facilitators were better able to connect women to other political organizations and enable them to achieve broader goals. Differences in Outcomes One of the most basic questions facing the RRLP is whether the strategy used to mobilize the poor has an impact on the women who were targeted by the program. Moreover, it is important to know whether the differences in the types of facilitators actually translated into differences in outcomes. To answer this question, we examine two types of data: block-level outcomes reported by the project, and results of a small survey of SHG women in the two types of areas. Given that the two areas were for the most part identical prior to the rollout of the RRLP, and that the program was rolled out in the two areas with almost the same intensity, we can, to some extent, interpret these estimates as causally related to the mobilization effort.28 We present two sets of results for the block-level outcomes. The first is aggregate data taken from the website of the RRLP program. (The raw data are reported in the Appendix.) Analysis of these data suggests that Intensive and Resource blocks performed very similarly along a number of different metrics: both types of facilitators were able to set up approximately 4-5 SHGs per 28 Official data are collected by a strong team led by a statistician, Mr. Hardeep Chopra, who has no vested interest in either block. He was brought into the project right at the time that the Resource and Intensive blocks were being decided in 2012. 15 village, and reach about 53 households per village. In both cases, fewer than 2% of SHGs had bank accounts, were credit-linked, had a prioritization plan, and had received two tranches of money. Next, we examine outcomes reported by SHG women using data from an “Intermediate Outcome Survey” conducted by the Government of Rajasthan for 302 women from 150 SHGs from 6 blocks of 4 districts in October 2014. The sample included 102 women in Resource Blocks (i.e., external facilitators) and 200 women in Intensive Blocks (i.e., internal facilitators). The survey asked them about their SHG participation, financial transactions and participation in other community activities. The survey is admittedly imperfect for measuring program impact because of its small size, its lack of a comparable baseline and possible Hawthorne effects,29 but our focus is on comparing differences between the two types of blocks. We have already established that the two types of blocks were similar prior to the introduction of RRLP. Therefore, even if women participating in the survey were screened or pre-selected based on their good standing in the program, and may have provided answers that would resonate positively with the enumerators, both types of blocks would be similarly affected by these limitations. Therefore we can, cautiously, infer differences about the program’s implementation in the two types of blocks from the differences in their response. To test the hypothesis that differences in the style of facilitation led to differences in women’s benefits from the RRLP program, we run the following regression model: = 0 + 1 + 2 + + Where is a measure of an outcome of SHG participant i in block b, is a dummy variable that takes value 1 if the SHGs in the block are being facilitated by external facilitators, is a vector of controls for woman i (this includes age of her SHG, her self-reported pre- program income and whether or not she had an income source prior to the program), is a block fixed-effect and is the error term. We examine several outcome variables, each considered in a separate regression:  Following RRLP rules: This is a dummy variable that takes value 1 if a woman reports that she follows the standard RRLP rules and codes for SHGs,30 and 0 otherwise.  Total individual savings: Measured in rupees, this is the amount of money that an individual reports having saved within the SHG. 29 The survey was conducted by the same government officials who manage the pro ject within the blocks. It is thus entirely possible that SHG women gave biased responses, in favor of showing that the program was working (in anticipation of further opportunities for financial or other benefits). 30 This is known as the panchsutra in the RRLP manual. 16  Interest rate: Measured in percent, this is the rate of interest that an individual reports on her largest loan.  SHG member is linked to a higher level institution (example: federation): This is a dummy variable that takes value 1 if a woman reports participating in a higher level community organization than her SHG, for example, a federation of SHGs, and 0 otherwise.  Number of GS meetings: This is the number of gram sabha (village group) meetings that the SHG member reports having participated in during the past year.  Access to MGNREGS: This variable takes value 1 if the woman reports that she has access to the Mahatma Gandhi National Rural Employment Guarantee Scheme, a program that provides rural Indians who are eligible with 100 paid days of employment a year.31  Access to PDS: This variable takes value 1 if the woman reports that she has access to the Public Distribution System, a program that provides rural Indians who are eligible with subsidized access to basic food items such as wheat, rice and sugar.  Access to scholarships for household members: This variable takes value 1 if the woman reports that she has access to the scholarship facility, a program that provides rural Indians who are eligible with free access to educational services.  Number of CDO meetings: The number of community development organization meetings attended by a woman or her SHG representative. The results of these regressions are presented in Table 7. We note several interesting patterns. First, women who are in SHGs facilitated by outsiders are 4% more likely to be following exact RRLP rules and protocols (Table 7, column 1). However, they achieved approximately Rs. 500 less in savings, which is 25% of the average of Rs. 1958 for the whole sample (Table 7, column 2). They are more likely to have taken out informal loans, which include internal loans from the SHG itself (Table 7, column 7). Thus, local facilitators seem to have been more effective than SERP facilitators in achieving a financial impact. However, externally facilitated SHG women attend two more Gram Sabha meetings than their counterparts who are facilitated by local women, and are 65% more likely to participate in the MRNREGS program. These are significant effects, both in magnitude and in their statistical power. We infer from these results that women who are led by external facilitators are able to effectively access some very basic opportunities offered by the state. Our qualitative data suggest that this is largely because the external facilitators leverage their status as “outsiders” and their strong backing from the state to insist upon new opportunities for women in Rajasthani villages. While there are concerns about the sustainability of these effects, it is noteworthy that we see 31 Both the MGNREGS and the RRLP programs are largely targeted to women below the poverty line. 17 such pronounced effects can be seen within just a few months from the start of the program’s operations. Interestingly however, SHG women who are led by local facilitators are much more likely to report an awareness of their other entitlements: they are 18% more likely to participate in, or send an SHG representative to, participate in local community meetings, particularly on issues of livelihoods and economic issues. They are also 20% more likely to be aware of opportunities for scholarships for household members. They are also 18% more likely to be linked, either directly or indirectly (via an SHG representative) to community institutions such as SHG federations. In summary, our results suggest that central and state-led implementation of the NRLM had different effects. External facilitation was much more expensive for the state, since professional outside facilitators require higher levels of compensation and incur significant travel costs. But, external facilitators were able to use their position as community outsiders to form new groups, and implement the program closer to the manner envisioned by NRLM’s central management. Local facilitators were much less expensive to hire and required less state coordination, and their main role was to link existing groups and networks to the operations of the state. They were, however, much more likely to have an economic impact, and were better able to link participating women to local political, economic and social opportunities in their communities. In sum, therefore, there is no question that local facilitation offered more value for money than external facilitation. CONCLUSION This paper discusses an important but under-appreciated topic in participatory development: which level of government is best suited to manage large-scale participatory projects. The subject is important because the success of participatory projects hinges on the effectiveness of implementation by facilitators at the level of neighborhoods and villages. Both central, and decentralized management have arguments that can be made in their favor. Centralized implementation allows for greater control over management and information systems, and allows experienced facilitators from one part of the country to transfer knowledge and capacity to another, less experienced, part of the country. And centralized implementation can help shift persistent norms, creating disruptive social change by transmitting ideas from part of the country to another. The arguments for decentralized implementation are similar to those made more generally for decentralized government: local facilitation is best supported by the level of government that is most proximate to local conditions in order to minimize information asymmetries, allow better monitoring, and hold facilitators more accountable for their actions. The level of the state that 18 implements local development should also be able to manage implementation that fits the local culture. Sub-national governments are more likely to have a cultural match with the “symbolic public goods” (Rao, 2008) being created at the local level by facilitators and thus mobilize collective identities with greater ease. To test these ideas we use a natural experiment from the National Rural Livelihoods Mission in India that separated blocks in each state into centrally or sub-nationally managed blocks known as Resource Blocks and Intensive Blocks. One method featured external facilitators who were actively backed by the state. The other featured local women who were actively backed by local civil society. Both arms of the program aimed to organize women into SHGs that would then link to the state to receive credit. We examine this in the state of Rajasthan using mixed methods to contrast two approaches of organizing poor rural women within the confines of a single poverty alleviation program. At first glance, the two strategies seem to accomplish similar goals in that they are both able to form a comparable number of groups with similar financial performance. But there are some important differences in what the groups actually do. Groups formed by outsider facilitators are better linked to the state through participation in the local Gram Sabha and other poverty alleviation programs. But they are less likely to engage in collective action on issues related to public service delivery, less likely to actually engage with local politics and less likely to be connected to other civil society organizations. These differences have important long-term implications for the groups themselves as well as for the overall success of the program. In addition, central management with external facilitators is extremely expensive. Given all these reasons, in the case of Rajasthan, local, sub-nationally managed facilitation would have been optimal for the project. Finally, central management is particularly inappropriate for the state of Rajasthan, which has a long history of civic activism having nurtured important national initiatives such as the Mahatma Gandhi National Rural Employment Guarantee Scheme, and the Right to Information Act. Particularly relevant to the NRLM’s mandate, it has several NGOs with long histories of facilitating the formation and functioning of women’s Self-Help Groups. Thus, sending SHG facilitators from the far-away state of Andhra Pradesh, who were not native speakers of the local language, to form women’s SHGs seems a particularly egregious case of central overreach and cultural mismatch. Centrally managed Resource Blocks may make more sense in states that lack Rajasthan’s civic history where they could, indeed, be a resource for building internal capacity in the state. But, in Rajasthan, it makes little sense. 19 REFERENCES Aiyar, S. A., Narayan, D., and Raju, K. (2007). Empowerment through self-help groups. Ending poverty in south Asia: Ideas that work , 104-135. Bardhan, P. (2002). Decentralization of governance and development. Journal of Economic Perspectives, 185-205. Bardhan, P., and Mookherjee, D. (2006). Decentralization and local governance in developing countries: a comparative perspective (Vol. 1). The MIT Press. Bhargava, P. (2007) “Civil Society in Rajasthan”, in V.S. Vyas (et.al), Rajasthan: The Quest for Sustainable Development. New Delhi: Academic Foundation, 257-81. Chen, Martha, Renana Jhabvala, Ravi Kanbur, and Carol Richards, (2008) “Membership Based Organizations of the Poor: Concepts, Experience and Policy,” Chapter 1 in M. Chen, R. Jhabvala, R. Kanbur and C. Richards (editors), Membership Based Organizations of the Poor. Routledge Publishers. Cornell, Stephen and Joseph P. Kalt, (2000) “Where’s the Glue: Institutional and Cultural Foundations of American Indian Economic Development,” Journal of Socio-Economics, 29, 443-470. Deininger, K., and Y. Liu. (2009). “Longer-term economic impacts of Self-Help Groups in India.” World Bank Policy Research Working Paper No. 4885. Deshmukh-Ranadive, J. (2004). “Women's self-help groups in Andhra Pradesh: participatory poverty alleviation in action.” In A case study from Reducing Poverty, Sustaining Growth–What Works, What Doesn’t and Why-A Global Exchange for Scaling Up Success, Scaling Up Poverty Reduction: A Global Learning Process and Conference, Shanghai, May, 25-27. Desai, R. M., and Joshi, S. (2013). “Collective action and community development: evidence from self-help groups in rural India.” The World Bank Economic Review , lht024. Dreze, J., and Sen, A. (2002). India: Development and participation. Oxford University Press, USA. Evans, P. (1996). “Government Action, Social Capital and Development: Reviewing the Evidence on Synergy.” World Development 24, 6: 1119–32. Evans, P. (1995). Embedded Autonomy: States and Industrial Transformation . Princeton, N.J.: Princeton University Press. Fox, J. (1993), The Politics of Food in Mexico, Ithaca, N.Y.: Cornell University Press. Galab, S., and N. Chandrasekhara Rao. (2003). “Women's Self-Help Groups, Poverty Alleviation and Empowerment. ” Economic and Political Weekly 38 (12/13): 1274-83. Gupta, R. K., and Bakshi, S. R. (2008). Studies In Indian History: Rajasthan Through The Ages The Heritage Of Rajputs. New Delhi: Sarup & Sons. 20 Haines, A., Sanders, D., Lehmann, U., Rowe, A.K., Lawn, J.E., Jan, S., Walker, D.G. and Bhutta, Z., 2007. Achieving child survival goals: potential contribution of community health workers. The Lancet , 369 (9579), 2121-2131. Linn, J. F. (Ed.). (2012). Scaling up in agriculture, rural development, and nutrition (Vol. 19). Intl Food Policy Res Inst. Jain, P. S. (1994). Managing for success: Lessons from Asian development programs. World Development , 22(9), 1363-1377. Kamat, S. (2002) Development Hegemony: NGOs and the State in India. Delhi: Oxford University Press. Kaner, S. with Lind, L., Toldi, C., Fisk, S., and Berger, D. (2014). Facilitator's guide to participatory decision-making. John Wiley and Sons. Lessmann, C., & Markwardt, G. (2016). “Aid, Growth, and Devolution: Considering Aid Modality and Different Types of Decentralization. ” World Development , 84, 118-130. Mansuri, G., and Rao, V. (2012). Localizing development: does participation work? World Bank Publications. Menon, S.V. (2008). “Right to Information Act and NREGA: Reflections on Rajasthan”, Munich Personal RePEc Archive Paper, No. 7351, February, pp. 1-13 Mukherji, A. (2013). “Evidence on Community-Driven Development from an Indian Village. ” Journal of Development Studies, 49(11): 1548-1563. National Bank of Agriculture and Rural Development (NABARD). (2013). Status of Micro- credit in India 2012-13. SAGE Publications India. Ostrom, E. (1996). “Crossing the Great Divide: Coproduction, Synergy, and Development.” World Development , 24(6): 1073-1087. Nugent, Jeffrey B. (1993). “Between State, Market and Households: A Neoinstitutional Analysis of Local Organizations and Institutions.” World Development 21, 4: 623–32. Perry, H. B., Zulliger, R., & Rogers, M. M. (2014). “Community health workers in low-, middle- , and high-income countries: an overview of their history, recent evolution, and current effectiveness. ” Annual review of public health,35, 399-421. Planning Commission, Government of India. (2011). NRLM: Framework for Implementation. Available at: http://rural.nic.in/sites/downloads/latest/NRLM_23122010.pdf. Accessed on 13th Feb, 2015. Potter, J. G. (Ed.). (2001). Devolution and globalisation: implications for local decision-makers (Vol. 498). OECD Publishing. 21 Rajvansi, J. (2007). “NGOs and Academics”, in V.S. Vyas (et.al) Rajasthan: The Quest for Sustainable Development, New Delhi: Academic Foundation, 325-46. Rao, Vijayendra (2008). “Symbolic Public Goods and the Coordination of Collective Action: A Comparison of Local Development in India and Indonesia,” Chapter 10 in Pranab Bardhan and Ray, Isha (editors), The Contested Commons: Conversations Between Economists and Anthropologists, Wiley-Blackwell. Singh, Prerna (2016). How Solidarity Works for Welfare: Subnationalism and Social Development in India. Cambridge University Press, Cambridge, UK. Smoke, P., Loffler, G., & Bosi, G. (2013). The role of decentralization/devolution in improving development outcomes at the local level: Review of the literature and selected cases. DFID& Local Development International LLC. Takasaki, Y. (2011). “Do local elites capture natural disaster reconstruction funds?” Journal of Development Studies, 47(9), 1281-1298. Treisman, D. (2007). The architecture of government: Rethinking political decentralization . Cambridge University Press. Ziegfeld, A. and Tudor, M. 2015. “How opposition parties sustain single-party dominance Lessons from India.” Party Politics, p.1354068815593455. 22 TABLES Type of Block Organization N Percent Resource Block SERP 49 13.80 Intensive Blocks PRADAN 38 10.70 PEDO 130 36.62 Saheli Samity (Federation) 101 28.45 Sangharsh Mahila Manch 37 10.42 (Federation) Total 355 Table 1: The sample of facilitators Resource Blocks Intensive Blocks N Mean N Mean Difference Std. err. Household size 1560 5.141 4400 5.060 0.0808 0.0589 Males in household 1560 2.754 4400 2.705 0.0497 0.0371 Females in household 1560 2.387 4400 2.355 0.0311 0.0384 BPL 1560 0.0526 4400 0.0695 -0.0170 0.00727 SCST 1560 0.362 4400 0.457 -0.0951*** 0.0145 SC 1560 13310.3 4400 12860.8 449.5 421.1 ST Labor Income Own Home 1560 0.991 4400 0.985 0.00557* 0.00335 Electricity 1560 0.777 4400 0.736 0.0408 0.0328 Toilet 1560 0.168 4400 0.122 0.0464 0.0290 Have own bank account 1560 1.436 4400 1.481 -0.0455 0.0347 Ever taken an SHG loan 1560 1.992 4400 1.994 -0.00156 0.00277 Table 2: Pre-program differences between Resource Blocks and Intensive Blocks, as measured by the baseline survey of 2012. 23 External Internal Difference Facilitators Facilitators (std err) Panel (A): Facilitator Characteristics Age 32.53 36.40 -3.865** (1.329) Married 0.878 0.971 -0.0930** (0.0306) Number of living children 1.592 3.023 -1.431*** (0.214) Number of children residing at home 2.286 3.052 -0.767*** (0.205) Completed primary school 1 0.500 0.500*** (0.0716) Completed secondary school 0.673 0.124 0.549*** (0.0544) Husband completed secondary school 0.694 0.830 -0.136* (0.0599) Head completed secondary school 0.469 0.144 0.326*** (0.0578) Agriculture main source of hhd income 0.571 0.627 -0.0560 (0.0748) Has previous work experience 0.102 0.124 -0.0221 (0.0503) Years of work experience 2.184 1.951 0.233 (0.399) Household owns land 0.490 0.846 -0.357*** (0.0591) Own income (past month) 31469.6 2151.6 29318.0*** (660.8) Household income (past month) 30644.9 7838.6 22806.3*** (1054.4) Have other occupation? 0.388 0.794 -0.406*** (0.0643) Days in a month on other occupation (when not a CRP/PRP) 8.653 11.93 -3.275** (1.239) Income from secondary occupation 1553.1 1663.7 -110.7 (332.1) Panel (B): Training and responsibilities Number of responsibilities listed in job 4.612 4.082 0.531 (0.271) Job description requires mobilization 0.98 0.941 0.0384 (0.0347) Job description requires training 0.735 0.761 -0.0267 (0.0661) 24 Job description requires holding meetings 0.673 0.784 -0.111 (0.0648) Job description requires monitoring 0.551 0.281 0.270*** (0.0704) Job description requires follow-up 0.469 0.34 0.13 (0.0736) Job description requires bank linkage 0.51 0.261 0.249*** (0.0692) Job description requires advice on agriculture 0 0.00327 -0.00327 (0.00818) Panel (C): Strategies of mobilization Taught how to identify target population 0.918 0.621 0.297*** (0.0713) Taught wealth ranking 0.653 0.526 0.127 (0.0766) Taught social mapping 0.735 0.497 0.238** (0.0760) Taught PRA methods 0.796 0.144 0.652*** (0.0553) Used wealth ranking 0.612 0.480 0.132 (0.0768) Used social mapping 0.653 0.461 0.192* (0.0764) Used PRA methods 0.673 0.118 0.556*** (0.0534) Size of group that conducts mobilization 6.898 11.88 -4.984*** (0.774) Favor groups that are from the same religion 0.143 0.216 -0.0728 (0.0622) Favor groups that have the same caste 0.122 0.186 -0.0638 (0.0589) Favor groups that have the same economic background 0.449 0.203 0.246*** (0.0643) Favor groups that have the same source of income 0.490 0.157 0.333*** (0.0595) Favor assigning specific functions 0.776 0.944 -0.169*** (0.0406) Group positions determined by elections 0.204 0.461 -0.257*** (0.0751) Days spent maintaining records 12.35 7.18 5.167** (1.601) Days spent establishing new groups 36.59 10.08 26.51*** (1.987) Number of new groups formed in own village 2.143 3.353 -1.21 25 (0.73) Number of new groups formed outside village 47.98 4.016 43.96*** (2.24) Time each day spent on chores 88.78 142.3 -53.54*** (11.57) Time each day spent traveling 111.8 108.3 3.483 (13.23) Time each day spent in the field 424.5 329.4 95.06*** (26.16) Time each day spent on record keeping 122.3 108.8 13.58 (11.79) Time each day spent on filling forms 96.67 81.19 15.48* (7.798) Time each day spent on other activities 75.61 92.17 -16.56 (12.43) Village was prepared for visit 0.959 0.552 0.407*** (0.0721) Panchayat announces program to village 0.939 0.248 0.690*** (0.0634) Mobilization begins with women only 0.51 0.804 -0.294*** (0.0637) Village picks the time of the mobilization meeting 0.939 0.369 0.569*** (0.0705) Number of SHG meetings in last village 13.37 43.85 -30.48 (25.26) Number of SHGs in last village 10.8 41.14 -30.34 (25.29) Number of women met to set up SHGs in the last village 34.04 124.7 -90.63** (29.97) Number of days taken to set up SHGs in the last village 6.592 13.19 -6.598* (2.615) Submit records within the organization 1 0.922 0.0784* (0.0385) Last village had a target population 0.878 0.641 0.237*** (0.0713) Table 3: A comparison external and internal facilitators. Notes: (i) * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level. 26 Monthly salary (Total) SERP 29209.065*** (983.735) Age 111.036 (193.697) Age squared -1.148 (2.487) Completed primary school -95.648 (424.368) Years of work experience -34.852 (21.921) Married 511.692 (844.381) Number of living children 185.006 (231.583) Number of responsibilities listed in 330.710 job (197.145) Constant -2551.349 (3648.750) R-squared 0.855 N 355 Table 4: Mincerian Regressions for monthly earnings of facilitators. Notes: (i) Standard errors are clustered at the block level; (ii) * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level. 27 (1) (2) (3) Collective Collective Collective Action on Action on Action on women- political public service specific issues issues delivery SERP 1.700 0.054*** 0.025*** (1.312) (0.005) (0.004) Age 1.003 1.010 1.023* (0.049) (0.023) (0.013) Completed primary school 1.258 0.380*** 0.736 (0.455) (0.124) (0.165) Married 0.633 0.439*** 0.960 (0.549) (0.075) (0.258) Number of living children 0.935 0.776*** 0.870* (0.127) (0.067) (0.063) Has other occupation 1.506 0.572** 0.650 (0.622) (0.132) (0.189) Household owns land 1.556 0.513 0.529 (1.785) (0.311) (0.478) Agriculture main source of household income 0.922 0.772 1.077 (0.709) (0.260) (0.451) Years of work experience 1.122*** 0.975 1.046** (0.049) (0.020) (0.023) Household income 1.000 1.000*** 1.000* (0.000) (0.000) (0.000) Constant 0.039*** 2.970*** 0.556 (0.047) (1.229) (0.318) N 355 355 355 Table 5: Examples of collective action. Notes: (i) Standard errors are clustered at the block level; (ii) * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level. 28 Resource Intensive Blocks Blocks (External (Internal facilitators) Facilitators) SHGs per village 4.655 4.467 Households per village 53.834 52.900 Fraction of SHGs with S/B Account Opened 2.969 2.408 Fraction of SHGs credit linked 0.184 0.585 Fraction of SHGs with priority plan 2.025 2.223 Fraction of SHGs Tranche 1 1.490 1.945 Fraction of SHGs with MCLP 1.058 1.035 Fraction of SHGs Tranche 2 1.338 0.938 Table 6: Average block-level outcomes 29 (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Following Total Change in Total Interest SHG Number NREGA PDS Scholarship Number RRLP individual income informal rate member is of GS of CDO rules savings credit linked to a meetings meetings received higher level institution (example: federation) External 0.043*** -546.217** -10199.648*** 6558.384** -0.969 -0.173*** 2.692*** 0.653*** -0.041 -0.204** -0.179*** (0.005) (201.779) (1898.319) (2537.433) (1.053) (0.026) (0.085) (0.049) (0.021) (0.064) (0.024) New group -0.061 2897.667* 3347.159 14365.819* 0.195 0.017 -0.676 0.079 0.020 -0.021 -0.005 (0.068) (1236.007) (4589.614) (6605.587) (0.191) (0.018) (0.488) (0.182) (0.024) (0.082) (0.006) Initial income 0.000 0.005 0.007 0.176** -0.000 0.000 0.000 -0.000 0.000 -0.000 0.000 (0.000) (0.003) (0.048) (0.056) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Constant 0.950*** 1964.575*** 18110.520*** 12971.126** 25.630*** 0.956*** -0.020 0.258** 0.849*** 0.235* 0.965*** (0.010) (255.650) (3129.701) (3963.407) (1.770) (0.043) (0.157) (0.077) (0.034) (0.095) (0.040) Block FEs Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes R-squared 0.092 0.585 0.324 0.158 0.268 0.188 0.329 0.400 0.086 0.067 0.821 N 302 302 302 301 302 302 302 302 302 296 302 Table 7: Regression results from Intermediate Outcome Survey. Notes: (i) Standard errors are clustered at the block level; (ii) * denotes significance at the 10% level, ** denotes significance at the 5% level and *** denotes significance at the 1% level. 30 APPENDIX BLOCK NO. OF NO. OF SHGS NO. OF NO. OF NO. OF SHGS NO. OF SHGS NO. OF SHGS NO. OF SHGS NO. OF SHGS NO. OF SHGS VILLA GES NEWLY RURAL CDOS WITH S/B WITH WITH WITH WITH MCLP WITH ENTERED FORMED/CO HOUSEHOL NEWLY ACCOUNT CREDIT PRIORITY TRANCHE I PREP-ARED TRANCHE II -OPTED DS FORMED OR OPENED LINKED PLAN RELEASED RELEASED COVERED COOPTED PREPARED RESOURCE BLOCKS Asind 54 306 4264 9 242 0 219 219 121 121 Churu 78 612 647 9 418 19 34 319 173 173 Sangod 126 518 5892 43 425 63 314 314 146 146 Deogarh 63 37 4037 10 268 30 190 190 158 158 Dungarpur 61 480 6498 1 350 104 267 24 105 101 Anandpuri 89 62 7203 17 524 7 393 362 25 203 Chipabarod 70 520 523 12 356 5 278 265 129 120 Newai 82 529 5674 1 318 3 202 202 120 94 Bakani 96 507 573 7 423 8 314 314 148 148 Kherwada 124 76 8698 19 673 39 403 40 301 301 Total 843 5287 60009 138 3997 278 2924 2829 1656 1565 INTENS IVE BLOCKS Kolayat 87 471 502 46 471 36 381 31 209 16 Taranagar 39 23 2346 15 157 3 140 131 25 21 Simalwara 191 134 16871 82 974 501 873 836 646 635 Dausa 90 652 780 40 49 12 360 35 210 18 Jhadol 78 543 6284 32 479 62 278 257 20 173 Bari 144 739 9074 63 540 5 435 435 357 357 Baseri 112 585 7146 43 43 5 290 290 31 31 TOTAL 741 4347 54503 321 3553 734 2757 2635 1978 1871 9634 114512 459 7550 1012 5681 5464 3634 3436 31