The Mahatma Gandhi National Rural Employment Guarantee Scheme: A Policy Solution to Rural Poverty in India?

The Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA) was developed by the Indian government to reduce rural poverty through 100 days of guaranteed employment per year. Using focus group methods, we explore whether this scheme has provided rights' based social protection through guaranteed employment for Scheduled Castes, Scheduled Tribes and women in Kerala, Tamil Nadu and Odisha. We found that the experiences of participating in MGNREGA varied depending on how MGNREGA wages compared to market wages in the region, as well as local implementation of the program. Although MGNREGA offered some basic employment for marginalized groups, it did not provide substantial help to the most vulnerable. However, there was some evidence of small but significant shifts in labour relations. Higher wages, more opportunities for work, better implementation and a greater recognition of the caregiving responsibilities of women will be required for this policy to fully meet its goals.

up approximately 22% of the population and STs make up 11% of the population, these groups account for 80% of the rural poor (IFAD, 2015;Raghu et al., 2013). Furthermore, women are over-represented in rural poverty, holding a disproportionate number of the marginal agricultural jobs with low wages in rural India (IFAD, 2015). Notably, even with substantial agricultural growth, there have been economic declines for marginal farmers and landless rural citizens (Vakulabharanam, 2005).
Recognizing these problems, the Government of India has introduced several nationwide centralized social schemes to address poverty. 1 One such scheme is The Mahatma Gandhi National Rural Employment Guarantee Act (hereafter referred to as MGNREGA), offering a unique rights-based guarantee of employment to reduce income and food insecurity in rural areas. Significant investment has been made in this scheme, amounting to 5.3 billion dollars (Ministry of Rural Development, 2013-14), or 0.3% of the gross domestic product. Given the size of this policy platform, the substantial amount of public money invested in it and the potential impact on rural citizens, it is important to ascertain the extent to which MGNREGA has reached its stated policy goals. This article sets out to do this through assessing the extent to which this scheme has been able to provide rights-based social protection through guaranteed employment for marginalized groups, particularly SC, ST and women. In addition, we ask a theoretical question: will this policy achieve meaningful poverty reduction for marginalized citizens in India?
Using a qualitative research design in three case study areas, we endeavour to answer these questions through findings from a study that involved two key data collection phases. First, we reviewed MGNREGA policy documents to understand the key goals of the policies. We then went to the case study sites and conducted 19 focus groups with 219 policy implementers and end users in three rural sites in three states -Kerala, Tamil Nadu and Odishato gauge the extent to which these goals were realized in our study sites. Study sites were chosen because they were rich in agrobiodiversity, but also had high rates of poverty, between 60% and 85% (Breitkreuz et al., 2014) Our analysis contributes to the literature on MGNREGA in a unique way. It provides evidence from an in-depth qualitative study that offers insight into policies from the perspectives and experiences of local people in local sites (an emic perspective), in the spirit of a critical ethnographic tradition that assumes that the best way of knowing is to know from within (Smith, 1987). This article will add to the literature about MGNREGA that to date consists primarily of aggregate national, state level or multi-state evaluations (Deininger and Liu, 2013;Dreze and Khera, 2009;Dreze, 2010;Azam, 2012;Imbert and Papp, 2015), or studies that focus on single, or a few, factors such as the relationship between MGREGA and stunting (Dasgupta, 2013) and malnutrition (Nair et al., 2013). As an alternative, we offer an analysis from an in-depth, qualitative case study of three research sites in the states of Kerala, Tamil Nadu and Odisha. Our contribution does not claim to isolate particular factors or indicators of success. Rather, we provide evidence from the 'subjective voice' of end users of the policy (that is, community people in rural sites) as well as local officials who provided insights into the local contexts and ways that policies are implemented in these areas. Pani and Iyer (2012: 14) indicate that 'local processes are typically better captured through detailed qualitative analysis' and that is what we endeavour to do here. This study thus adds to emerging literature on micro-level analysis of multiple dimensions (Novotn y et al., 2013). Using this approach enables the researcher to discover new, as yet unconsidered, elements of a particular policy intervention that might get overlooked in other macro-level approaches. Although the findings are not generalizable across the whole of India, or the whole of each state in which the study was conducted, we posit that the insights gained from this approach will enhance understanding of some aspects of the MGNREGA policy that may not be evident through other macro-level approaches.

Policy context: The Mahatma Gandhi National Rural Employment Guarantee Act
The MGNREGA programme was implemented following a number of government employment programmes developed in response to the persistent poverty and inequality plaguing rural India. MGNREGA is the largest rural employment scheme in the world. It is described by the Government of India as a rights-based, demanddriven, self-selecting, bottom-up scheme (Mukundan, 2009). In addition to providing rural households with supplementary employment, MGNREGA also seeks to encourage and facilitate the empowerment of women through financial inclusion and independence, strengthen civic participation, as well as improve the rural landscape (Planning Commission, 2011). The Indian employment scheme was introduced in 2005, and was renamed MGNREGA in 2009, after the scheme had been rolled out across all Indian states. Under the scheme, adult members of rural households are guaranteed 100 days of employment each year at the statutory minimum wage rate of the state or greater, provided they are willing to do unskilled manual labour. Included in this employment scheme is the guarantee that if the government is unable to provide a job to a qualified applicant within 15 days of an application being submitted, the applicant will receive unemployment insurance. The stated goals of this policy are: 1) social protection; 2) the creation of durable assets (such as water security, soil conservation, higher land productivity) through the manual labour conducted by the workers; 3) employment of disadvantaged workers such as women, SC and ST; and 4) inclusive growth in rural India through the policy's impact on livelihood security and democratic empowerment.

Background literature
As a flagship programme of the Government of India, MGNREGA has received considerable attention from politicians, policy think tanks, and scholars. Studies of MGNREGA consist primarily of larger-scale evaluations using administrative data to understand implementation of the programme in various states. These studies have revealed important findings about implementation costs, challenges and successes, impacts on migration, impacts on household income and employment, and uptake of the programme by marginalized groups such as ST, SC and women. Taken together, a number of key themes are evident across the literature.
First, there is significant variability in implementation and uptake of MGNREGA across various states (Khera, 2011). Key to implementation is the availability of work. On average, the number of days available for work have been lower than that guaranteed by the Act. Earlier studies show that in the 200 districts in which MGNREGA was enacted in 2006-2007, only 17 days of employment were available on average per rural household (Dreze, 2010). More recent data suggest, however, that the opportunity for work may be increasing. Stahlberg (2012) showed, for instance, that the average number of days worked per householder was 54. This number is significantly higher than in 2006-2007, but is well below the policy ceiling of 100 days per year per household.
There is also notable variation in the allocation of resources (for the categories of wages, materials and administrative costs) for MGNREGA across states. The average percentage of MGNREGA funds spent on wages in 2009-2010 was 67.5%, but there was notable variability. Relevant to our study, in Tamil Nadu it was reported that 98% of the funds were spent on wages. In Kerala, approximately 88% of the funds were spent on wages, whereas in Odisha, just over 60% of the funds was spent on wages. Interestingly, in Tamil Nadu, the expenditures for administrative costs were about 2%, In Kerala they were approximately 5%, and in Odisha the administrative costs were just under 40% (Stahlberg, 2012). The reasons for such significant variability for administrative costs are unknown but 'leakage' is suspected to be one reason for high administrative costs (Stahlberg, 2012). However, although spending most of the budget on wages may seem optimal, it may also be that spending all the money on wages means that states are not making capital investments to improve the quality of the jobs and infrastructure projects. The Act indicates that 60% of the expenditures should be spent on wages, with the remaining 40% spent on materials (Bhupal, 2012).
Second, some studies found that MGNREGA reduced migration in certain states (Liu and Barrett, 2012), while other studies found that MGNREGA had minimal impact on migration (Novotn y et al., 2013). Novotn y et al. (2013 found that 93% of MGNREGA participants in their study indicated that, even if there were no employment options available under MGNREGA, they still would not go elsewhere for work. Third, there is evidence that awareness of MGNREGA by rural citizens was scant in some areas, and this decreased the likelihood of participation on the part of some of the poorest households. Minimal public engagement and limited or no access to television or cell phones on the part of the poorest households has been correlated with minimal uptake of the programme. However, the availability of information increased the likelihood of uptake by more affluent rural citizens (Shankar et al., 2011). Furthermore, there is some evidence that some MGNREGA spots may be reserved for those in the middle class (Liu and Barrett, 2012), thereby diluting the positive impacts for the poorest (Niehaus and Sukhtankar, 2013). One study found that although MGNREGA has been used by the poorest in a national level analysis, in one state it was used by community members who were relatively well off (Jha et al., 2009;Liu and Barrett, 2012).
Fourth, there is evidence of uptake of the programme by a number of targeted groups, particularly ST and SC (Deininger and Liu, 2013;Dreze and Khera, 2010) and women (Dev, 2011;Dreze, 2010;Dreze and Khera, 2009). When breaking down participation according to group for the whole of India, Stahlberg (2012) found that the average number of days of participation in MGNREGA in 2009-10 was 48 for women, 21 for ST and 31 for SC. Furthermore, Dreze and Khera (2009) found that approximately 70% of the women kept their wages. Importantly, the participation of women in MGNREGA varies greatly across states. Khera and Nayak (2009), for instance, found that in their sample of six states, the proportion of women among the MGNREGA workers ranged from 5% in one region of Uttar Pradesh to 71% in a district within the state of Rajasthan.
When looking at the breakdown of total participation according to group,  found that the target groups of the programme -SC, ST and womenhad disproportionately higher rates of participation in comparison to their portion of the population, suggesting that the programme is meeting its goal of targeting these groups. However, there was significant variation across states. Notably, Kerala and Tamil Nadu had the highest rates of women participants, with over 85% participation in Kerala and 80% in Tamil Nadu . Other states, however, had much lower participation. Jammu and Kashmir, for instance, had participation rates as low as 6% for women. Odisha's female participation rate was 38%. The participation rates of SC and ST were also higher than the share of the population that they represented. For instance, in Tamil Nadu, SC constitute 19% of the population but over 60% of the participation days in MGNREGA. Participation of SC in Kerala and Odisha was also higher than their proportion of the population, but the effects were not nearly as dramatic as in Tamil Nadu .
The uptake by women is not surprising, given that women tend to receive lower wages and could probably earn better wages from MGNREGA than the market (Sudarshan et al., 2010). To illustrate, in their study of three states, Sudarshan et al. found that wages in MGNREGA were 125 Indian rupees as compared to 70-80 Indian rupees (Rs) for agricultural work in the private sector. Some researchers argue that MGNREGA, through improving women's wages, has broadened consumption options for women, and enhanced their economic independence (Pankaj and Tankha, 2010). Critics have pointed out, however, that due to the menial work it offers, in combination with low wages, MGNREGA really just contributes to the feminization of poverty (Ghosh, 1998: 16). Studying women's participation in MGNREGA, Sudarshan et al. (2010), found that the employment scheme fell short in terms of understanding life course issues for women, such as child rearing, and also failed to adequately consider the physical capabilities of women, particularly older women. In contrast, Carswell and De Neve (2014) found that MGNREGA had positive effects on women as well as the poorest households in Tamil Nadu by providing them with a safety net.
Fifth, some commentators have suggested that the introduction of MGNREGA has provided some additional leverage to labourers by providing them with negotiating power (Carswell and De Neve, 2014;Dreze, 2010). This happens because, in many locations, wages are higher through MGNREGA than through private employment arrangements, forcing landowners to increase their wage offerings for agricultural work in order to compete with MGNREGA wages. Novotn y et al. (2013) found that the vast majority of local employers (mostly farmers) and local officers (those knowledgeable about MGNREGA and its impact on local labour markets) in their study strongly supported the statement that MGNREGA had 'significant harmful effects on the local labour market' (ibid.: 335). Other research has found an increase in social tension as a result of successful implementation of MGNREGA programmes (De Neve and Carswell, 2011).
Sixth, corruption has been cited as a problem in MGNREGA implementation, widely associated with the issue of underpayment due to wage skimming by administrators and over reporting (Kapur, 2010;Niehaus and Sukhtankar, 2012;. Village post offices, where cash payment of wages have traditionally been paid, have been accused of tampering with pay or charging unnecessary transaction fees to workers. As a result, the payment of wages into the personal bank accounts of MGNREGA workers through direct cash transfers has been increasingly advocated as a way to reduce the embezzlement of funds. The over reporting of wages or false reporting of MGNREGA worker muster rolls (to fulfil and maximize worker requirements and financial allocation) have also been identified (Niehaus and Sukhtankar, 2012;. Other identified instances of systemic corruption include misappropriation of funds for materials and overall unaccounted financial leakage within the programme. In reviewing the argument about why the implementation of social policies is poor, Kapur (2010) argues that part of the reason that implementation of schemes falls apart at the local level is because public programmes are understaffed. Thus, although the bureaucracy in India is large, staffing in local sites is not. Similarly, in their study of MGNREGA spanning three states,  found that the commitment, capacity and preparedness of local governance structures impacted the effectiveness of the employment scheme in different locations.
Two studies are worth discussing in more detail due to their strong research designs and comprehensive findings. Deininger and Liu (2013) conducted a comprehensive study of MGNREGA in the state of Andhra Pradesh. Because they collected data over time (before MGNREGA was introduced in 2004 and then twice subsequently, in 2006 and 2008), they were able to make linkages between the MGNREGA scheme and some key indicators of well-being. They found that the policy had significant impacts. Short-term impacts included higher nutritional intake and medium-term impacts included non-financial asset accumulation. They also found it was women, SC, ST and those seeking casual labour who obtained most benefit.
In reviewing various evaluations of MGNREGA in recent years, Dreze (2010: 514) states that in locations in which employment is available, MGNREGA is the 'new lifeline for rural poor'. In line with this statement, Dreze and Khera (2009) conducted a study of approximately 1,000 workers in 100 worksites across six states in northern India. Their findings show that MGNREGA reached the poorest citizens, with 73% of workers in the Scheduled Tribe/Scheduled Caste categories (Dreze and Khera, 2009). Using the same data set, Dreze and Nayak (2009) showed that women accounted on average for only 32% of the MGNREGA workers among the six states, though there was substantial variation among states: Rajasthan, for instance, had a MGNREGA female participation rate of 71%, whereas Uttar Pradesh had a 5% female participation rate. Importantly, of the female workers in MGNREGA, over two thirds were able to keep the wages they earned through the MGNREGA programme (Dreze and Khera, 2009). Dreze and Khera argue that the concept of 'right to work' was in play because MGNREGA legislation was shown to reduce the exploitation of workers and to serve as an organizational tool for rural workers. They also found that, in addition to providing employment for the most marginalized workers, MGNREGA led to the creation of useful assets in the community.
In sum, although MGNREGA has appeared to provide some much needed employment for India's most marginalized workers, a major limitation of the scheme to date has been its ability to generate employment to meet the demand for work. In addition, corruption appears to be an impediment to successful implementation. This finding is more prevalent in some states than others.
Despite the large number of empirical studies investigating the impact of the MGNREGA, there are few studies that offer an in-depth, qualitative analysis of implementation and outcomes of MGNREGA in specific sites. Notable exceptions include the study by Sudarshan et al. (2010) in Himachal Pardesh, Kerala and Rajasthan, and Carswell and De Neve's (2014) mixed method study in Tamil Nadu. We argue that qualitative studies are important, too, in order to provide a voice to participants themselves, as well as to get an in-depth understanding of what is happening at a local level. In light of these research goals, our study aimed to understand better how MGNREGA was experienced in the day-to-day life of rural participants.

Methods
This study was part of a larger agricultural research project, Alleviating Poverty and Malnutrition in Agrobiodiversity Hotspots (APM), which was conducted jointly by the MS Swaminathan Research Foundation (MSSRF) in India, and the Faculty of Agriculture, Life and Environmental Sciences at the University of Alberta in Canada. The larger APM project was developed to explore ways to improve farming techniques, enhance nutritional outcomes and reduce poverty in areas of India with rich, agrobiodiverse land. It aimed to address the disparity between richness in biodiversity and abject poverty in three agrobiodiversity hotspots in India: Wayanad, Kerala; Kolli Hills, Tamil Nadu; and Jeypore, Odisha. In conjunction with the larger APM project, our study team conducted a review of policies in India intended to address the issues of food insecurity, malnutrition, poverty and employment. Sites for this study were selected because although they featured rich agrobiodiversity in one or more crop species, inhabitants experienced rates of poverty of between 60-85% and low human development indicators (Breitkreuz et al., 2014). The broad purpose of our study was to explore the reach of national social policy initiatives in the three rural study sites in Kerala, Tamil Nadu and Odisha, to ascertain which policies were accessible in these areas. In addition, we aimed to explore the implementation of large-scale national policies at a local level, to identify areas of convergence as well as gaps between policy goals and implementation.
The data used in this article were drawn from mixed-methods research that involved two key phases. The first entailed a systematic, in-depth review of relevant policy documents, including policy drivers, goals and outcomes as available from government sources, non-governmental policy think tanks and scholarly literature. The second phase entailed qualitative data collection through focus groups with policy implementers and end users in three rural areas of India, comparing their experiences with the policy claims of relevant national social policies, including the MGNREGA 2005.
In total, 19 focus groups were conducted for this study with a total of 219 participants. We used focus groups because they are an effective way to gather a significant amount of data on a particular issue for which not much is known, providing the opportunity to identify key themes on the phenomenon of interest that can then be explored in more depth (Morgan, 1997). Three pilot focus groups were conducted in March 2013 in Odisha to test the focus group questions and determine whether focus groups would be an effective way to collect data for this project. During the fall of 2013 another 16 focus groups were conducted in three study sites in the States of Odisha, Tamil Nadu and Kerala. All three sites were located in agrobiodiverse regions with high rates of SC and ST. 3 All focus groups were facilitated by a member of the research team, with assistance from other colleagues from the MS Swaminathan Research Foundation. Participation in this study was voluntary, and written informed consent was obtained.
Each focus group consisted of five to 20 participants, with an average of 12 participants per focus group. These groups lasted between 90 and 120 minutes, and tea and snacks were served to participants upon completion of the groups. We sampled both policy implementers and end users. In each site, we had one focus group of policy implementers, covering a broad range of social and/or agricultural 3. Site 1 is located in north-east Kerala. The area has large populations of very low-income minority tribal groups, most of which are known to hold very small (on average 0.4 ha) plots of land. Agriculture is the main source of livelihood, with the principal crops including coffee, tea, cocoa, pepper and rubber, while seasonal crops such as rice, bananas, tubers and fruits serve as the local food sources. The region has substantial disparity in land ownership and associated economic status. Education in Kerala is strong, with the state having the highest literacy rates in India at 94%. Site 2 is located at the southern end of the Eastern Ghats range in Tamil Nadu. The main inhabitants in the area are Malayalis, one of the Scheduled Tribes of India. The area is 97% tribal. The majority of households are poor, owning small or marginal farm holdings. Rice and minor millets are the locallyproduced staple foods, and important fruit crops are jackfruit and hill banana. The literacy rate in Tamil Nadu is relatively high at 80%. Site 3 is located in the southwest part of Odisha State at the northern end of the Eastern Ghats. The literacy rate is just under 50%, significantly lower than the state average of 73%. Odisha is the poorest of the APM project sites, with 83.8% of people living below the poverty line. 13% of the population here are part of a Scheduled Caste and nearly 50% were members of Scheduled Tribes. The main crop grown is rice.
policies relevant to the area. We organized the end-user focus groups according to landowning status and gender. In each site we had four focus groups of end users: one for landless men, one for landless women, one for landowning men and one for landowning women. 4 In our Kerala study site, we also had a group of tribal men, and in Tamil Nadu, all of the groups were tribal as the entire study site was a designated tribal area. A semi-structured interview format was employed. Semi-structured interviews are designed to seek information about a particular topic, covering various domains of knowledge, while still maintaining the flexibility of an unstructured interview (Richards and Morse, 2007). The areas of interest for the purposes of these interviews included questions regarding the kinds of government policies/ programmes that were available to people in the area to help them meet their families' needs. Participants were asked what makes these programmes helpful or unhelpful, who has access to these programmes and what would they change about these programmes if they could. Using this approach, an interview guide was developed to shape the course of the interview and ensure that particular areas of interest were considered, but this was used more as an aide-memoire than a rigid interview protocol. The aide-memoire was adapted over the periods of the focus groups as the concurrent data analysis revealed data collection needs (such as divergent findings or emerging themes that required further exploration). Issues and emerging insights garnered from earlier focus groups were also introduced in later focus groups for verification. Using this approach, the interviewers were free to probe at certain points to elicit more in-depth information, and ask questions in a way likely to elicit responses. This style allowed for consistent data to be collected, while leaving room for important and enriching data to emerge (Mayan, 2009).
Focus groups with end users were conducted in the native language of the area (Malayalam, Tamil or Odia) and translation was provided by project staff provided by the MS Swaminathan Research Foundation. Transcripts were developed as the focus groups proceeded. The transcribers were two student research assistants from the University of Alberta who were employed for this project, and are co-authors of this article. Detailed field notes were made by the research team following each focus group. Information in the field notes included summaries of the interview as well a brief documentation of points that seemed particularly important. The research team met after each focus group to review the transcripts, discuss key points and identify recurring themes using the constant-comparison method (Strauss and Corbin, 1998). This method involves comparing two or more descriptions of experiences or key events and looking for similarities and differences across participants (Mayan, 2009). These descriptive themes were condensed into a preliminary report which summarized information about the implementation process of key national social policies across sites.
The first author and research assistants then conducted an in-depth thematic analysis of the interview data to ensure the rigour of the preliminary analysis, and to 4. Note the 'landowning women' groups consisted of women who were married to landowning men.
Women in India do not typically own land, but we called this group 'landowning women' to distinguish their status from 'landless' women.
refine, develop and expand on recurring themes, and search for and analyze 'negative cases' (i.e. any inconsistencies) (Mayan, 2009). This was done through an iterative process of reading the transcripts and writing analytical notes throughout the process, searching for commonalities and noting points of divergence. By comparing chunks of coded data to look for commonalities, a process called 'axial coding' (Strauss and Corbin 1998), interrelationships between codes were discovered, and these codes were merged to create comprehensive themes. Through this in-depth analysis, key themes were identified in the data, and discussed by the research team through a dialogical process, leading to a nuanced understanding of the key issues raised by the participants. Under these comprehensive themes, key factors were delineated, as described in the findings below.

Findings
Overall, two main findings emerge from our analysis. First, there were some notable differences in usage and implementation between study sites. The experiences of participating in MGNREGA varied according to how MGNREGA wages compared to market wages in the region and the implementation of the programme within the local sites. Gender and social status also seemed to impact the experience of MGNREGA for many participants. Second, although MGNREGA did offer some important opportunities for employment for women, SC and ST, thereby providing access to wages to purchase food and other essential items, on balance it did not substantially help the most marginalized in the long term. This was due to the predominance of low wages compared to the private market (as in Kerala and Tamil Nadu), or lack of work (as in Odisha). Importantly, participants in all project sites indicated that the wages offered by MGNREGA were not sufficient to support their families. However, despite lack of evidence of any substantial improvements in wages, employment and overall life circumstances, there was some evidence of small but notable shifts in labour relations. We explore these themes in the following section.
We also utilized survey data collected by the larger research project to provide some descriptive statistics on MGNREGA use in our study sites. For this part of the study, 1801 households were surveyed (1500 intervention households and 301 non-intervention households) within our three study sites. Included in the survey were questions on MGNREGA usage. Descriptive analyses were conducted on these data which are detailed below.

MGNREGA Implementation
5.1.1 Usage. MGNREGA was implemented in all three regions, Kerala, Tamil Nadu and Odisha, and the vast majority of participants were highly knowledgeable about the programme. In Kerala, 46% of focus group participants had worked for MGNERGA, and the average number of days worked through the programme was 57. Both men and women used MGNREGA as last resort employment. In Tamil Nadu, 72% of survey participants used MGNREGA, and the average number of days worked was 68. In Tamil Nadu, MGNREGA was used primarily by women, because men generally migrated to the states of Kerala or Karnataka for higher-paid employment. In Odisha, while 61% of participants worked for MGNREGA, there were fewer days of work available to labourers though the programme; the average number of days was 36. MGNREGA was used in Odisha by both women and men (Raghu, 2013).
MGNREGA is designed to be a programme of last resort. In all three sites, participants indicated the alternative to participating in MGNREGA was working for local farmers for the prevailing wage in the private agricultural market. In Kerala, the majority of both male and female participants preferred this private agricultural work due to higher wages and the provision of meals at the worksite. At the same time, however, some of the female participants in Kerala indicated that they preferred MGNREGA because they were treated better, and that the hours of work fitted in better with their childcare responsibilities.
In Tamil Nadu, MGNREGA was used primarily by women. Men in Tamil Nadu commonly migrated to Kerala or Karnataka in order to gain higher-paying employment during the agricultural off season. In this site, women indicated that although there was a childcare centre that worked well for them, they were not treated very well on MGNREGA worksites and therefore they used the programme as a last resort. Implementation issues here may have impacted the experience of the women.
In Odisha, unlike our other two sites, the wages in MGNREGA were equal to or higher than those in the private market. Many participants, both male and female, indicated that they would prefer to work for MGNREGA rather than as agricultural workers in the private market because of the better working conditions. However, in Odisha there were not enough days of work available to the workers, and participants who illegally used others' job cards and bribed the officials reported the highest number of days worked in the programme.

Payment structures.
The method of payment for MGNREGA work varied both between and within sites, and method of payment seemed to affect the extent to which there was corruption. In some cases, payment was made directly in cash, and in other situations, the payment went directly to a bank account. When payments were made directly to a bank account, there was no opportunity for 'service charges' or 'skimming', thereby reducing corruption. Also, some project sites paid workers on a daily rate, while others based wages on the amount of work completed.
In Kerala and Odisha, payments for MGNREGA were made via direct transfer to bank accounts and based on a daily rate. Participants in Odisha stated that the shift to direct payments of MGNREGA wages into bank accounts had eliminated corruption and skimming of wages in the area. In Tamil Nadu, payment was made weekly in cash, based on work completed, and there were reports of 'service charges' or skimming of wages by government officials.

Access to unemployment insurance.
A stated requirement of the MGNREGA programme is to provide unemployment insurance for those who cannot find work with the programme for 15 days after applying for a job card. However, unemployment insurance was not mentioned in any of the sites. In Kerala and Tamil Nadu unemployment insurance may have been irrelevant as participants were able to get up to 100 days of employment if requested. However, in Odisha, where participants were not able to access the full 100 days of employment, there was no mention of unemployment benefits for those who had applied for a job card but received no work.

Wage discrepancies.
According to the reported wages of study participants, there was a differentiation of wages paid to MGNREGA workers, despite the assertion in most groups that there was a standard wage for all workers. As indicated in Table 1, there were wage discrepancies in MGNREGA across sites, and between men and women, and in some cases the wage was lower than the minimum wage. When asked about wage discrepancies, participants indicated that women could not get the amount of hard labour done in a day that was expected of those working with MGNREGA, therefore they could be paid lower wages within this programme. Landless women in our Tamil Nadu site indicated that they only received Rs 148 if they were able to complete the work and dig the whole area. They had to clear and dig three square metres of land about 30 centimeters deep. The women indicated that they could not achieve this in a day, nor could most men. Therefore, they would not get the full day wages. In the private market, the key reason cited for varying wages was 'different pay for different work'. When this reason was cited, it was clear that there was a gendered element to the pay. Policy implementers in our Tamil Nadu site indicated that male labourers in the private market were paid Rs 350 per day, on average, and women were paid Rs 200 per day, on average.
In Kerala, participants in the landless women's group indicated they were paid Rs 125 for work under MGNREGA (Rs 55 less than men), but made no mention of inequality between genders and socioeconomic groups under the programme. Instead, when asked about MGNREGA wages for women and men, participants all indicated that the wages were equal between the genders. A possible explanation for participants' lack of recognition of wage discrepancy may be that private wages in Kerala were highly variable, and depended on the type of work done, location and gender of the worker. Wages in the private market ranged from Rs 350-550 for men and Rs 125-225 for women. It may also be that inequality in wages between the genders was so common than it was perceived as fair.
In Tamil Nadu, policy implementers stated that the private market wages in their state were Rs 350 for men and Rs 200 for women. Participants in the women's focus groups, however, indicated that private wages in the agricultural sector were significantly lower for women, at Rs 120-150. Participants, especially men, stated that they preferred to work in the private market because of the relatively higher wages.
In Odisha, the MGNREGA wages were raised if the work on a particular project was difficult; the head engineer had the ability to increase pay to up to Rs 420 per day for harder work in difficult soil. Participants in Odisha indicated that both women and men earned the same wage under MGNREGA (Rs 150), and that the private wage for agricultural work was higher for men (Rs 150) than women (Rs 100).

Impact of MGNREGA on labour relations
The implementation of MGNREGA seemed to provide some labourers with more power, thus influencing labour relations in ways that were advantageous for marginalized workers, even when MGNREGA wages were lower than those provided through private employers. However, the hierarchy of social and gender relations appeared to remain intact within MGNREGA.
In Kerala, there was evidence that MGNREGA provided negotiating power to workers, even though the MGNREGA wages were lower than those offered in the private market. Participants in the landed women's focus group discussed disagreements about working conditions and hours from the implementation of the programme, with workers demanding meals and shorter hours. Several comments were made in this group, expressing the perspective from landowners that labourers had increased power. One participant in this group said: 'Labourers have started to demand whatever they feel like, and the wages are becoming different for each type of work, which is becoming an issue. The farmers have to pay whatever they demand or else they will have no one to work in their fields.' Another comment from this group indicated the same: 'The labourers have an upper hand now. Farmers have to beg them to work and say they will pay wage, feed them.' Yet another comment was: 'The labourers have many demands now. There is a large demand for them, so farmers must pay and provide what they ask.' There were also indications that MGNREGA affected wages in Kerala. Farmers in this region were permitted to use their MGNREGA job card 6 to work on their own farm ponds as well as hire additional labourers to work on these ponds. Farmers indicated that workers demanded higher pay and had stronger market power when determining wages because of MGNREGA. However, because of the low prices for crops, farmers stated they could not afford to pay labourers the wages they demanded. They indicated that more MGNREGA workers should be used for work on farmland. Labourers, however, stated that allowing MGNREGA workers to work on private farmland on the farm ponds lowered the market private wage for the same work. The inconsistency between the farmers' and the labourers' experiences with wage relations under MGNREGA may be due to the highly variable private wages in the area as reported in the landed women's focus groups. Private wages vary depending on the type of work done. For some work, the workers demanded Rs 550 per day. There was a different wage for different places as well. The minimum wage for men was Rs 400 per day. One participant said: 'There is no fixed wage, and depending on the situation and place, they ask for different wages for the same job.' Another participant indicated that she paid Rs 300 but her neighbour across the road paid the same people higher wages.
In Tamil Nadu, the policy-makers talked about how labourers and farmers could not agree on labour conditions since MGNREGA came in. Labourers could only work on private land in Tamil Nadu through MGNREGA if they were building farm ponds. Landowners could hire themselves and workers through MGNREGA to build ponds on private land. There was no indication from the focus group participants that MGNREGA wages acted as a price floor or impacted private wages in Tamil Nadu. However, due to MGNREGA, labourers demanded payment on a weekly basis.
In Odisha, there was evidence that MGNREGA offered a wage floor for labourers. The MGNREGA wage was equal to or more than the private wage for 5. The state minimum wage was taken from the Government of India Ministry of Labour and Employment (2013). 6. Households must apply for an MGNREGA job card through their Gram Panchayat to be eligible to work. A job card entitles a household of up to three adults 100 days of employment with the usage of the 100 days left to the discretion of the household (can be shared among the adults or used by one individual). Job cards must include the names, age, sex and photographs of the benefactors along with a unique identification number which identifies residence area. The cards are valid for a period of five years and must be renewed upon expiration to retain eligibility.
both men and women, and participants reported that the implementation of the scheme had caused an overall increase in wages in the labour market, in and out of the agriculture sector. However, participants in Odisha received significantly fewer days of employment than those in the other study sites and indicated that they would like more work through the MGNREGA programme.

Impact of MGNREGA on women's employment
Women in our study sites were more likely than men to work for MGNREGA, but there were a number of issues within the programme that created barriers to women's successful participation. In particular, the type of work available through MGNREGA and the lack of adequate child care were cited as ongoing issues for many women participants. Women were still disadvantaged workers even within MGNREGA. In addition to receiving lower wages, women reported that they were not always treated very well by their site supervisors, and were often given work that was too difficult for them. In addition, their unpaid care responsibilities for children and home meant that they could often not work the long hours that men worked, and they were sometimes docked pay for this. In Kerala, landless women indicated that they sometimes used childcare facilities, or, more often, took their babies to the fields. In Tamil Nadu, women were more likely to put their children in Anganwadi Centres, which are childcare facilities, so that they could work in the fields. Interestingly, despite the challenges faced by women as MGNREGA workers, they generally viewed the programme as positive, perhaps because it provided them with opportunities to earn their own wages.
Although some of the women preferred working for MGNREGA because of more flexible hours and the ability to bring their children to the worksite, others also mentioned that they did not get treated very well by their superiors and that the work was too difficult. Landless women in Tamil Nadu were particularly articulate about their challenges with MGNREGA. While the participants confirmed that the official upper wage limit was Rs 148, the work required to earn that wage was based on what a man can complete in a day, despite the fact that the majority of MGNREGA workers were women. Because the work was paid according to what was accomplished, women tended to get paid less because they did not get as much done. There were two reasons for this. First, they found the work difficult, particularly in cases where they were digging, and were unable to complete the required digging work. Second, women indicated that they had to work around the schedules of their children, and so women with children tended to work a shorter day. Women indicated that the wages they received from the MGNREGA programme ranged from Rs 70-90 to a maximum of Rs 120.

Discussion
The development and implementation of MGNREGA as a rights-based policy initiative shows a commitment to the creation of a policy framework to address, by offering focused employment guarantee programmes to marginalized community members, some of the injustices experienced by the most vulnerable members of Indian society. This article explores the implementation of MGNREGA within specific sites in three states in India: Kerala, Tamil Nadu and Odisha. It shows some of the implementation successes and challenges from the perspective of end users and frames these challenges within the context of delivery of a rights-based approach in a liberalized developing economy.
Our findings suggest that the MGNREGA programme was known to end users in all three regions, although usage was uneven among the regions. Usage was highest in the Tamil Nadu and Kerala sites, and lowest in the Odisha site, although participants indicated that they would prefer more days in the Odisha site. Our findings on usage are consistent with other studies that have found that usage is higher in Kerala and Tamil Nadu and consistently lower in Odisha . Wages varied among the sites, and among men and women. MGNREGA wages were higher for men in both Kerala and Tamil Nadu, but the same for both genders in Odisha. Payment structures varied among the regions, and access to unemployment insurance was not mentioned in any of the sites. Ruparelia (2013: 570) notes that, in theory, the rights-based approach sets 'new standards for genuine social citizenship'. A rights-based approach departs from traditional methods of social provision by guaranteeing rights through a legal framework; in the case of MGNREGA, through an employment guarantee. In this way, it offers marginalized citizens meaningful access to rights previously beyond their grasp. However, like Ruparelia (2013), we argue that the extent to which these rights can be guaranteed is contingent upon effective governance structures that can overcome social, political and economic barriers that both accept and reinforce deeply entrenched inequalities between citizens.
Our study suggests that there may be limitations to the extent to which a rights-based approach can work within current social and economic realities in India. Although our conclusions are tentative, our in-depth analysis suggests that, even among the targeted groups, there was a sometimes subtle, but prevailing, hierarchy. In Tamil Nadu, for example, where the majority of men migrated to Kerala and Karnataka as seasonal agricultural labourers, work was available for women or other low status social groups. In other words, women appeared less likely to use MGNREGA if men needed it for employment, as was the case in Odisha. This may not have had anything to do with the implementation of MGNREGA, but rather be due to the pre-existing social norms that granted men, and particularly those who owned land, more entitlement. A recent study supports our conclusions. Athreya et al. (2014) found that in India, overall, only 10% of the households that use MGNREGA were able to get 100 days of work, and the majority of users were men.
In addition to social barriers that may have impacted self-selection into MGNREGA, there were clear cases of gender-and class-based inequalities with implementation of the program. Our landless women participants' stories showed that recognition of different gendered realities, such as responsibilities for child care and differences in physical strength, were not acknowledged within the implementation of MGNREGA, with the outcome being lower wages and, in some circumstances, worse working conditions, for women. It seemed that there was a poor understanding of, and lack of provision for, women's responsibilities for caregiving. Even if childcare was available, as in the case of Tamil Nadu, women's sole responsibility for their children was not considered in the programme. Participants in Tamil Nadu seemed to have the most challenges with MGNREGA implementation, including challenges with site supervisors and not getting proper payment. This situation may have existed because the majority of men in Tamil Nadu leave to do work in Kerala or on 'the plains', so it was primarily women who used MGNREGA. It may be that because it was primarily a women's programme in Tamil Nadu, there were more abuses of power in MGNREGA sites.
In addition, tribal men in Kerala, one of the most marginalized groups we encountered, had very low rates of participation in the MGNREGA programme. Thus, although the MGNREGA programme may be meeting its targets in terms of employing higher than average numbers of women, SC and ST, it may also be that when we drill a little deeper we see that within these groups a hierarchy remains where those who are more privileged still have greater access to work in the programme. In short, there appeared to be continued marginalization of some of these target groups.
On the other hand, MGNREGA did seem to bring about slight improvement in terms of labour relations for some groups. Like Dreze (2010) and Novotn y et al.
(2013), we found that MGNREGA did shift labour relations in some circumstances, giving a modest increase in bargaining power to labourers who, in some instances, were able to request better wages or better working conditions. Part of the reason for this was that MGNREGA offered choice to workers. In Odisha, the choice allowed for better wages and better working conditions, therefore raising the bar for employers. In other situations, such as in Kerala, MGNREGA did not offer better wages, but it did offer better working conditions for some employees, thereby giving marginalized workers a choice, even if it was a less financially attractive one. This effect in itself seemed meaningful to workers, and suggests that more than just economic considerations influenced their employment decisions. Rather, a sense of control and personal power may also have been a factor. In addition, women indicated that having an opportunity to earn their own wages was not insignificant. Like Khera and Nayak (2009), women in our study reported that, for the most part, they had control over the money they earned. Our study did not allow for quantitative estimates of the labour market effects of MGNREGA. However, our results are consistent with the Imbert and Papp (2015: 254) estimates of an aggregate increase in casual labour wage of 4.2% and an aggregate decrease in private sector work of 1.8%.
On balance, however, our analysis suggests that, although there may be some income benefits in the short term, marginalized citizens in the Indian context are unlikely to experience significant long-term benefits due to MGNREGA. Although we do not want to diminish the importance of increasing opportunities for employment and income for India's most marginalized, we argue that it is important to consider the extent to which the programme has transformative potential for these groups. Without dismissing notable positive outcomes, such as the provision of employment for enhanced food security and changes in labour relations (Carswell and De Neve, 2014;Khera and Nayak, 2009), our findings suggest that, to date, MGNREGA may not be providing the desired transformative changes for Indian society's most disadvantaged groups. The modest employment and income gains offered through MGNREGA job opportunities will not substantially improve income or, in turn, food insecurity for these groups. We suggest that in order to enhance food security and well-being through employment in India, longer-term solutions for employment security will need to be explored. One potential pathway for meaningful transformation is through women's self-help groups. Although beyond the scope of this analysis, it is important to note that there were some indications in our focus groups that government-and NGO-sponsored self-help groups that enabled women to gain capital through pooling together to purchase animals and sell animal products may be something to build on. This finding is consistent with the study done by Pellissery and Jalan (2011) that found that although MGNREGA did not provide social transformation for women, self-help groups have the potential to be transformative for women through fostering solidarity and developing assets. Programs that link Self-Help Groups and the National Rural Livelihood Mission to MGNREGA (Ministry of Rural Development, 2012) may be further able to build capacity among women by providing them with opportunities to leverage their limited resources through pooling their assets.
If we accept Sen's (1992) argument that capabilities are key to creating meaningful choice for marginalized groups within society, we suggest that further steps are needed in order for transformation to be achieved through MGNREGA. First, higher wages coupled with opportunities for more days of work are needed to enhance income and food security for India's most marginalized. Second, realistic labour options are needed for women to empower them in meaningful ways. Building on skills women already have could be one way to strengthen their employability. Third, opportunities are needed for participants to enhance their job skills through MGNREGA. In short, building the skills capacity of labourers through a programme like MGNREGA could be beneficial for enhancing rural livelihoods and community development.
Another recommendation that is supported by this study is the benefit of continued use of technology to provide payment directly into bank accounts in the form of direct cash transfers. This approach also contributes to the transformative potential of MGNREGA through reduced corruption and guaranteed payment for programmes such as MGNREGA by cutting out middlemen or unauthorized administrative fees. Increasing the use of technology to implement schemes could enhance monitoring and has the potential to reduced corruption and cronyism. 7 Technology through the use of databases and information systems also increases the transparency and accountability of programmes. However, like Khera and Nayak (2009), we suggest that implementation of this system must be done carefully to ensure that women have their own accounts in order to maintain control of their funds. 7. Part of the MGNREGA program is to conduct social audits and to make information about fund allocation and utilization of the scheme accessible to the public (Ministry of Rural Development, 2015). This may also reduce corruption in the implementation of the scheme.
MGNREGA has the potential to enact change in social norms through legal channels. Although progress may be slow, there is reason for cautious optimism. Invoking a legal solution may, over time, change norms, behaviours and approaches, thereby opening a space for enhanced social inclusion. Changing these norms may, in turn, provide additional opportunities for MGNREGA to be successful in local sites.
first submitted July 2015 final revision accepted December 2015