Empowerment mechanisms- employment guarantee, women and Dalits in India

ABSTRACT The Mahatma Gandhi National Rural Employment Guarantee Act (NREGA) in India opens a new chapter in rural governance, signifying transformative potential for enhancing economic and social security. One of the key objectives of the Act is to aid in the empowerment of marginalised communities, especially women, Scheduled Castes (SCs) and Scheduled Tribes (STs). However, no analytically rigorous attempt has been made to study the mechanisms and processes that are leading to empowerment or even the (unintended) processes of disempowerment. Our paper investigates (dis) empowerment as a process rather than as a (quantifiable) outcome and analyses the mechanisms through which the embedded gender and caste norms are being challenged (though not completely eroded), but also how powerful actors at household, community and market levels can come to resent processes that entail upward mobility of women and marginalised communities. We draw on a comparative study based on extensive qualitative indepth interviews and participant observation, in rural areas of two Indian states, Uttar Pradesh (North India) and Andhra Pradesh (South India). The former exhibits almost all the elements we associate with low development, whereas the latter belongs to a group of southern Indian states with higher levels of human and economic development.


Introduction
Though most studies agree that, theoretically, empowerment should be seen as a process, most empirical work investigates it as a quantifiable outcome (Malhotra and Schuler 2005). This is partly due to an inherent tendency to rely almost exclusively on econometric analysis for the study of development programmes and processes (Narayan and Petesch 2007). Easy-to-quantify variables such as income and poverty levels are widely used to study development programmes, whereas other critically important aspects (including empowerment processes) are either reduced to their quantifiable dimensions or are treated very superficially (Devereux et al. 2013). As a result of this evident gap between theory and methodology, the mechanisms that lead to empowerment remain to a large extent either assumed or unexplored. Our paper seeks to address these gaps by examining the mechanisms that underpin empowerment processes of women and other marginalised groups in rural India within a large-scale public work programme, the Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA). The MGNREGA confers to every household living in rural areas the right to be employed in unskilled manual labour on public works for 100 days per year and to be paid a statutory minimum wage, equal for men and women. One of the key aims of the Act is 'aiding in the empowerment of the marginalised communities', especially women, the Scheduled Castes (SCs) and the Scheduled Tribes (STs) 1 ' (MoRD 2012, 1). Rather than assuming that the mere acquisition of resources through social protection policies like MGNREGA necessarily leads to empowerment, our analysis reconstructs the mechanisms through which the guarantee of employment triggers (or fails to trigger) empowerment processes. The central questions that drive our study, which focuses on two Indian states, Andhra Pradesh (AP) and Uttar Pradesh (UP) are: Has the MGNREGA contributed in the empowerment of women and Dalits and, if yes, through which mechanisms? Are we seeing, paradoxically, disempowerment processes as a consequence of changing power relations between MGNREGA workers and other actors in the rural society?
A number of studies have evaluated the impact of MGNREGA in various spheres, delineating several salient themes. NREGA's positive economic impact on rural livelihoods is virtually undisputed (Khera 2011;Liu 2013, 2019;Nair et al. 2013;Mookherjee 2014;Dasgupta, Gawande, and Kapur 2017;Das 2018). The programme has constituted a safety net for the rural poor, contributing towards their food security, reducing distress migration, and increasing access to health and education. Some argue that NREGA has contributed to the general increase of wages in rural India, thus reducing income poverty further (Imbert and Papp 2013. It has also been observed that the gender-gap in wages has been reduced Ghosh 2011, 2014) and MGNREGA has contributed significantly to the welfare of women (Jenkins and Manor 2017). Through wage work, MGNREGA has generated income-earning opportunities for women in a pre-dominantly agricultural economy, where the options would be to work on the agricultural fields of higher and upwardly mobile castes (at the risk of harassment and exploitation), stay at home or remain unemployed (Nayak and Khera 2009). The programme has increased women's material independence and there is evidence that women exercise autonomy on how to spend MGNREGA wages, indicating greater decision-making powers within the household (Pankaj and Tankha 2010;Pellissery and Jalan 2011). Furthermore, in recognising a single person as a 'household', the scheme has also facilitated monetary and social security for single women and widows. A few scholars have looked at how MGNREGA puts into question power relations at the grassroots, particularly between lower caste agricultural labourers and upper or middle caste farmers De Neve 2014, 2015;Jakimow 2014;Roy 2014;Pattenden 2016;Jenkins and Manor 2017).
We contribute to this literature by analysing how MNREGA has triggered the (dis)empowerment of marginalised communities, especially women and Dalits in rural India in the domains of household, community and the market. Our paper is organised in the following way. We first explore the theoretical idea of empowerment followed by a discussion on our methodology, which incorporates both qualitative and quantitative indicators. We then analyse our empirical findings through three domains followed by a conclusion.

Empowerment and its critics
Empowerment as a political concept emerged from feminist critiques of the Global South in the 1980s, which expressed concerns with the pre-dominance within development discourse of the 'welfare approach' towards women or that women had been brought into development policy on very gender-specific terms, primarily in their capacity as housewives, mothers and reproducers. Arising largely as a critique of the modernisation approach, these debates foregrounded women as active contributors to economic development rather than as passive recipients of welfare programmes. These ideas received official visibility in the 1990s with the UN International conference on Population and Development held in Cairo (1994) and the fourth UN conference on women in Beijing (1995). These international platforms played a critical role in introducing empowerment to state actors, who were eager to promote gender-equality and 'mainstreaming' gender. The increasing usage of the term empowerment reflected attempts to recognise women's agency in development discourses and projects, while at the same time constituted a fundamental critique of the top-down approaches prevalent within development organisations (Calvès 2009).
Presenting a framework of empowerment, which influenced critical feminist writings, Srilatha Batliwala provided one of the first detailed conceptualisation of empowerment in Women's Empowerment in South Asia: Concepts and Practises (1993), defining it, 'as a process, and the results of a process, of transforming the relations of power between individuals and social groups' (Batliwala 2007b, 560; also see Batliwala 2007a). Most importantly, she defines empowerment as a process that shifts social power in three critical ways: (a) by challenging the ideologies (such as gender and caste) that justify and sustain social inequalities; (b) by changing existing patterns of access and control over economic, natural and intellectual resources, and (c) by transforming institutional structures that reinforce and sustain existing power inequalities such as the family, state and the market, to cite a few (Batliwala 2007b, 560). Batliwala's research coincided with Kabeer's work in Reversed Realities (1994) and her conceptualisation of empowerment, which she defines as the 'expansion in people's ability to make strategic life choices in a context where this ability was previously denied to them' and suggests three components to this: access to resources, agency and achievements (Kabeer 1999, 437). Importantly, as Kabeer points out, a distinction needs to be maintained between possible inequalities in people's capacity to make choices vs a vs differences in choices by members of a society (Kabeer 1999; also see Kabeer 2005).
However, the need to transform what is an essentially qualitative phenomenon (Sen 1994; also see Ibrahim and Alkire 2007), into a quantifiable outcome that could be used to assess the impact of numerous projects that include empowerment as their key objectives, resulted in a significant change in the understanding of the concept. First, empowerment became an integral part of the fight against extreme poverty, thus expanding its scope not only to women, but to oppressed and poor people in general (World Bank 2001). In a way, this resonated with the original meaning of the term as it emerged from the scholarship on oppressed communities in the United States in the 1960s and the 1970s (Calvès 2009). Second, especially after the publication of a few influential studies by the World Bank (Narayan and Nankani 2002;Alsop, Bertelsen, and Holland 2006; also see Alsop and Heinsohn 2005) that effectively mainstreamed empowerment, the term remained ill-defined and conflated with other approaches such as democratisation or decentralisation (see Wong 2003). Furthermore, critics pointed out that the co-optation by international organisations of the concept made it lose its radical component of challenging oppressive institutions and structures, in favour of a more 'entrepreneurial' approach that looked at ways in which women and the poor could be facilitated in navigating (rather than challenging) unequal and oppressive contexts to 'uplift' and empower themselves (Wong 2003;Sardenberg 2008). Third, the focus on the individual rather than on a more multifaceted collective process depoliticised and promoted a narrow understanding of entrenched social inequalities. Drawing some difference with Kabeer, Mosedale (2005) emphasised the importance of women achieving a change that expands options not only for themselves but also for women in general, together with the gendered nature of women's disempowerment (Mosedale 2005, 252). Furthermore, while Kabeer's definition involves focusing on individuals acquiring the capacity to choose, Mosedale focuses on 'redefining and extending the limits to what is possible' (Mosedale 2005, 252). In other words, while both Kabeer and Mosedale recognise the link of empowerment with power, Kabeer connects it with the ability to make choices (individual) while Mosedale links it to the process of identifying constraints.
The central ideas that drives our intellectual enquiry in this paper are as follows: (a) we see the ability to make individual choices as a key dimension of empowerment as it is central to questioning and challenging (and eventually, transforming) oppressive institutions and structures (for example gender and caste). A woman who chooses not to share her MGNREGA wage with her alcoholic husband is not only making an important life choice, which is vital for the economic security of herself and her family but is also questioning and challenging patriarchal social norms that are a potent constraint on her empowerment. Nonetheless, our focus on individual choice is not to 'de-politicise' empowerment or to underemphasise the importance of its collective dimension. In fact, many of the changes that we describe in the empirical part of this paper would be unthinkable without collective action that has taken place in the past decades in India, especially women's struggles and caste movements that have challenged the supremacy of the upper castes (Gorringe 2005;Rege 2006;Govinda 2008;Carswell and De Neve 2015;Anandhi and Kapadia 2017).
On the other hand, individual choices might trigger more profound (and collective) social change. As Alice Evans explains, through observation and interaction, we develop beliefs about which behaviours are widely supported in our social networks. If we observe widespread compliance, we infer that there is widespread support. We further anticipate that we will be liked, accepted and respected according to the extent to which we conform to these norm perceptions, and so moderate our conduct. So, even if we do not privately endorse these practices, we are nevertheless motivated to conformbecause we do not wish to be reprimanded, reproached or violently repressed. (Evans 2018) When women (or other oppressed people) start making choicesespecially visible onesthat challenge established and oppressive social norms, others might be start questioning those norms too, eventually leading to changes in social norms themselves; (b) we understand empowerment as a multidimensional concept (Ibrahim and Alkire 2007; also see Alkire et al. 2013), which enables us to argue for the co-existence of empowerment and disempowerment. For example, a lower caste man in India, might be empowered within his household in terms of making most decisions, but could be severely disempowered in his relationships with his upper caste employers or in terms of his ability to speak at a village assembly. Similarly, a woman could be excluded from decisionmaking processes within her family, but could be recognised as a leader by other women in her community, in her ability to head a self-help group and; (c) the aim of our empirical analysis is to unveil the mechanisms that lead to (dis)empowerment, which remain to a large extent either assumed or unexplored in the existing literature. We see mechanisms as constitutive elements of what gets defined as the 'process' of empowerment.

Methodology
We selected two districts each in Andhra Pradesh (AP) and Uttar Pradesh (UP) with contrasting levels of human development and levels of caste tensions. Table 1 presents the data on which we base this selection.
Our selection aimed at choosing two districts with low HDI and high caste tension (Fatehpur in UP, Anantpur in AP) and two districts with high HDI and low caste violence (Jhansi in UP, Guntur in AP). While the data on crime against SCs for AP portray Guntur as a district with higher crime rates against SCs than Anantpur, we nevertheless see it as a district with high caste tensions as it was also the site of a prolonged conflict between caste groups that resulted in hundreds of deaths during the early 2000s (Balagopal 2001).
The implementation of MGNREGA in the four districts varies in terms of average employment provided (Table 1). This reflects not only implementation capacity, but also local conditions. For instance, Guntur is a well irrigated district which allows for two or even three crops a year. Anantpur in contrast is a semi-deserted district with very poor irrigation facilities where only one crop a year is possible and where droughts are common. Clearly, a programme like MGNREGA will be more in demand in the latter case. The rationale behind the selection was not to find representative cases, but to select villages where (dis)empowerment processes stemming from MGNREGA can be studied more effectively. Our theoretical expectations were that we would find more visible signs of empowerment in the districts where the initial conditions were betterhigher HDI, lower caste tensionnamely Jhansi in UP and Guntur in AP. An alternative theoretical expectation was to find more visible empowering effects in Anantpur, where the average number of workdays provided through MGNREGA is significantly higher and thus, potentially more emblematic of the changes triggered by the programme.
The first source of data was collected in four villages (Gram Panchayats, GP), one per district, where we conducted 264 in-depth semi-structured interviews (Table 2).
Our interview guide covered ten topics, under four domains of empowerment: household, community, market and state. Table 3 provides a list of the topics.
We also conducted a survey in the four districts through a structured questionnaire covering the same list of topics. Within each district, we selected one block broadly reflecting the HDI of the district (based on Census 2011 data). We then randomly selected in each block, a number of GPs sufficient to get approximately 1000 Dalit MGNREGA 'active' workers, who have worked in the last 5 years. The total number of surveyed GPs is 25. Through stratified sampling, we selected 400 households per block, of a total of 1600 'eligible' households across the four blocks in the two states. We randomly assigned half of the households to male enumerators and half to female enumerators, who interviewed one member for each household of the same gender. In total, 1218 households could be located and interviewed.
In this paper, we analyse empowerment as a process of change, rather than as a quantifiable outcome. While measuring empowerment quantitatively is important to assess the impact of policies and programmes, it tells us little about the causal mechanisms that trigger (or fail to trigger) the recorded changes. In other words, we use our qualitative evidence to search for potential causal mechanisms that can 'turn the black box into a transparent box and make visible how the participating entities and their properties, activities, and relations produce the effect of interest' (Hedström and Ylikoski 2010, 51). The search for empowerment mechanisms, on the one hand, allows bringing into the analysis those contextual factors that are an inseparable element of empowerment processes and, on the other, the complex systems of power relations which affect the everyday lives of women and marginalised communities. The road to empowerment Our focus in this paper are empowerment and disempowerment mechanisms stemming from MGNREGA in three domains: household, community and the labour market. We focus on Dalit women and men. Table 4 provides a summary of the mechanisms explored in the following subsections.
For each domain, we analyse if these theoretical mechanisms are supported by the empirical evidence and we discuss the implications, especially in terms of changing oppressive social structures, with particular reference to social norms that regulate gender and caste relations-two important contextual factors that have historically been a potent source of disempowerment.

Household domain
In the household domain, we analyse how MGNREGA, that guarantees paid work to whoever demands it, can promote women's decision-making power. We focus on women only in this domain, because, in the context of India, household and intrafamilial relations can be the central locus of women's disempowerment. While evidence on the links between paid employment and women's empowerment is mixed and context dependant, our findings concur with Kabeer et al that regular and relatively independent sources of income have a transformative potential (Kabeer, Mahmud, and Tasneem 2011).
Dalit women working under MGNREGA often engage in salaried employment for the very first time, and, crucially earn the same wage as their male counterparts. Also, the wages are paid through bank/post office account transfer, to which they alone have access. This is different to when they engage in other forms of paid employment, mostly as agricultural labourers, where, in many cases, their wage is paid to their husbands or other male members of the family and is only a fraction of what men can earn for similar work, as agricultural tasks tend to be highly gender specific. On the contrary, MGNREGA wages are inaccessible to men. As one respondent put it: 'Only me can access the account so there is no point for him to come along with me to get the wage' (Interview 3, Guntur-16 February 2017). In another instance, a woman married to an alcoholic told us: I can't access my husband account and he can't access mine. That is very good because he spends everything in drinks. He drinks a lot. The only thing that he does is sleeping and drinking. But with my [MGNREGA] wage I buy groceries and use it for health care if someone is ill. (interview 11Guntur, 17/2/2017) Another woman with a less troublesome relationship with her husband told us that while decisions on how to spend money earned by herself are taken jointly, 'If he wants it to spend it on things that are not for the family, I would not allow him' (Interview 20, Guntur, 18 February 2017).
Some women became the main administrators of the household finances, after they started working under MGNREGA: I make all the decisions and get both mine and my husband's salary to manage every week. My husband does not smoke or drink, only drinks coffee and tea. Whatever he earns he will give to me. I take the day-to-day decisions and he does not object. Together we decide on chicken or mutton [and] what non-veg food to spend on. (interview 2, Max Guntur, date) In some cases, as several women from Anantpur district in AP told us, the earnings from MGNREGA allowed them to invest in small entrepreneurial activities: 'the scheme is helpful for women. It allows us to spend money on groceries and invest in raw materials for making cotton strips that we sell during festivals. This is something that was not happening before [MG]NREGA' (Interview 32 Anantpur, 5 January 2017). This was reported as being a big change for the women in the Dalit colony of the village and working on a government site offered a better alternative to working on the farms of landowners or private contractors. Even when women are excluded from the management of the family's finances -'because that is how it works'it is often something that they do not care about: 'my husband does not drink. He spends the money for the family, so I am fine with it' (interview 27, Anantpur, 5 January 2017). In another instance, a woman told us that she did not have any say in how the family income is spent but that, since, 'thanks' to the MGNREGA she is 'making much more money than before, her husband now keeps her informed on how the money is spent and she is happy that decisions are taken on her behalf' (Interview 50, Anantpur, 8 January 2017). This can be interpreted as an internalisation of patriarchal social norms that exclude women from certain arenas, but mostly our respondents seemed to be willing to accept the 'division of labour' as long as family interests were protected.
These narratives indicate that the existence of an independentand inaccessiblesource of income through their entry into formal labour force and cash economy, can translate into greater decision-making power within the household. However, we find quite striking differences between AP and UP, particularly in the women's ability to keep their MGNREGA wages for themselves, if they so desire. In the two districts of AP, 36 (Anantpur, low HDI) and 43.3 (Guntur, high HDI) per cent of women's respondents said that they keep their MGNREGA wages at least partly for themselves. The corresponding figures for the two UP districts are much lower: 15.3% in Fatehpur (low HDI) and 25.4% in Jhansi (high HDI). This seems to indicate that social norms regulating intra-household relationswhich research shows are much more inimical to women in UP, than in the South of India (Jejeebhoy and Sathar 2001) might 'dictate' that women give their wage to their husbands/ other males in the family to a much greater extent in UP as compared to AP. In fact, all the interviews quoted in this section come from AP, whereas very few women made the connection that the increased financial independence has led to a greater role in taking decisions within the family in UP. Within AP, the most consistent narrative about the connection between women's decisionmaking power and the MGNREGA came from Guntur, the district where initial conditions were betterincluding in terms of educational levels and women's security.
This was also true with respect to a more subtle empowerment mechanism-women's psychological well-being-that became apparent during our fieldwork. As one respondent said: 'women are now completely changed. Previously they were living in the shade of their fathers-in-law or their husbands. But now times are changing. I mean, we give the men in the family all the respect that they deserve, and we do not insult them with disrespect. But we feel different. Some women say this openly, some not, but we all feel that a lot has changed. We are getting paid equally with men in MGNREGA. We are getting our own money. MGNREGA has a lot to do with this feeling because it pays equally and directly [to our bank account]'. (Interview 3, Guntur, 16 February 2017).
A woman from Anantpur in AP noted how social protection policies like the MGNREGA and microcredit programmes have increased their sense of 'usefulness' to the family welfare.
I have been living in this village for 30 years and one thing that changed things dramatically for women were the Self Help Groups (SHG). Women's confidence went up a lot. The same happened with the MGNREGA. Women feel good when they can get money for the household. Also, it is receiving money from the government. We feel we are part of it. Beside MGNREGA and the SHG, we are never part of the system. (Interview 45, Anantpur, 6 January 2017) This kind of narrative kept coming up in interviews with women from both AP and UP. In both states, women mentioned microcredit programmes and MGNREGA as two key changes in their lives and emphasised the complementarities between the two: the MGNREGA helped women to repay their debts and improving their credit worthiness, thus promoting a virtuous cycle.
Other women also expressed a sense of increasing ability to make choices on basic needs that make them happy. For instance, a Dalit respondent reported: '[thanks to the MGNREGA] I feel that I have more choices and opportunities to fulfil small wishes like having the capacity of buying necessary utensils for the kitchen' (Interview 3, Max Guntur). Another woman from the same village expressed similar feelings when she told us: 'I was able to choose to send both of my children to school and to buy gasoline to cook with instead of lighting the fire' (interview 5, MaxGuntur). Yet another woman simply said that 'With MGNREGA I am able to live a good life'.
In short, in our fieldwork, we encountered several women respondentspredominantly in APwho linked the additional income provided by the MGNREGA with greater ability and autonomy to take decisions. The additional income has also contributed, in at least some cases, to a greater sense of confidence, optimism and ability to conduct a better life. This could possibly lead to changing the 'value' that men attribute to women and, perhaps more importantly, the 'value' that women attribute to themselves. The male respondents. in our survey, whose wives also worked under MGNREGA were more likely (94%) to assess their partners' contribution to the family welfare as 'very important' or 'somewhat important' compared to those whose wives did not work (84%).
During our fieldwork, we explicitly looked for signs that the empowerment mechanisms triggered by the MGNREGA also resulted, somewhat paradoxically, in disempowerment processes. The theoretical expectation was that the greater sense of confidence and assertiveness of women would result, in at least certain cases, in violent retaliation by male members of the households and/or the female in-laws. While stories about domestic violence were common, these were not usually seen as being triggered by the increased assertiveness of women, at least in AP. In virtually all cases, women were subject to violence because of the drinking problems of their husbands, which was the case before the introduction of the MGNREGA as well. On the other hand in UP, a few women from the Ahirwar (Dalit) caste in Jhansi (UP) mentioned that despite their sizable contribution to the family welfare through MGNREGA wages, they would not dare to 'overstep the line' in relation to decision making processes, as this would likely trigger a violent reaction (Interview 12, Jhansi, 12 February 2017). Two other women from the same village also told us that their involvement in decision-making 'bothered' their husbands, who would often taunted that 'now that you have earnings, you will show attitude' and ask 'you have some money, have you become the Prime Minister?' (Interview13 and 14, Jhansi, 12 February 2017). In addition, many women in UP surrendered their MGNREGA income to their husbands or in some cases the husband 'took out money from the bank using [her] thumb impression' (Interview 12, Jhansi, 12 February 2017). The fact that the wage is technically inaccessible to men of course does not prevent them to force their spouses to give their wage once it is withdrawn from the bank. Of those women who said that they gave away their MGNREGA wage entirely (68.2% of our respondents), 54.8% did so against their will (46.5% in AP and 65.8% in UP). A woman in Guntur told us about another woman living in her colony, whose alcoholic husband forced her to give the money she earns for buying drinks: 'So the situation for the woman actually got worst. The woman works a lot for getting some money and then she doesn't even have money for food because he takes it all and drinks as much as he can' (Interview 3, Guntur, 16 February 2017). In some cases, women from UP reported that the additional income provided by MGNREGA built an added expectation in the men that the women would handle the household expenses with their earnings from MNREGA, while the men could use their's on more 'pleasurable' activities. Also, for some women, the MGNREGA represents an additional burden, which comes on top of their household responsibilities, which in virtually all cases was not shared with their partners. These are some of the unintended mechanisms of disempowerment, which co-exist with empowerment.

Community domain
An important empowerment mechanism that we found during our fieldwork was the increased freedom of Dalits, which stem from their reduced dependency on upper caste employers (mostly big landowners). Of course and undeniably these are the results of long-term processes of change, including education, exposure to democracy and equality before the law, but which have accelerated due to social protection policies like MGNREGA and have repercussions on both the community and the market domains (which are intertwined). While our survey shows that signs of deference by Dalits were common, encapsulated in the notion of 'internalised subordination' (Sen 1999) or fear, which is historically instilled in them by the upper castes through violence, our qualitative interviews nevertheless show that much has changed in recent years. First, alternative job opportunities in the nonfarm sector, the possibility of commuting to nearby urban centres (where wages are higher) and/or to migrate seasonally to bigger cities contributes largely to reduced dependency of the lower castes on landowners, while simultaneously increasing the dependency of landed farmers on Dalit labourers, especially where farm labour is scarce. This declining Dalit dependency on landowners is reflected when a woman from a village in Anantpur said that: change is enormous. Dependency on the Reddys has come down a lot. We used to be completely dependent on them for everything. But now people [meaning: Dalit] are more educated, they are getting jobs in Hindupur and we are helped by the government, with subsidies, loans, MGNREGA … these things make a lot of difference. (Interview 38, Anantpur 6/1/17) Waged labour together with the security of an income often translates as increased freedom for the Dalits. A Dalit man from AP and a village where the scheme works extremely well: the relation with the Reddys [one of Andhra Pradesh's dominant landowning caste] is often marked by confrontation. A few years back they decided that they would not give us any work in their fields, as they claimed we had to be punished for our insolence. The Reddys brought labourers from outside the village. It was very hard for us. But something like this cannot happen anymore. We have guaranteed employment with MGNREGA now. We are less dependent. We are confident now. (Chittor district, 19/10/2013) Words like 'insolence', 'arrogance' or 'disrespect' are euphemisms used by dominant castes to describe the Dalits' behaviour when they do not accept to live the sub-human existence that the traditional social order has reserved for them. In a social context where lower caste people are expected to be at complete disposal of the upper dominant castes or upwardly mobile backward castes, the ability of agricultural labourers to negotiate their work to the landowners translates into nothing less than a challenge to the social order and, crucially, this alteration of power relations enhances poor people's confidence and sense of dignity.
Furthermore, MGNREGA wages enables them to lift themselves from abject poverty and hunger. This is something that both Dalit labourers and upper caste farmers recognise. As one Dalit man from AP put it: MGNREGA changed things dramatically. It was difficult even to get enough food. Now we can manage at least with the basic expenditures. Especially those of us who own cattle, can conduct a good life. And if we work in MGNREGA, food is not an issue anymore. Also, the good thing about MGNREGA is that it is continuous work. And you get a lot of money in one shot. You can buy a big bag of rice or clothes for the festivals. This is also why I prefer MGNREGA, even though we get less money than farm labour. (Interview 36, Anantpur 6/1/17) In a completely different mood, a large landowner from the same village told usalmost screamingthat the Dalits should 'worship the Gods that invented this [the MGNREGA] scheme' as it has allowed them to 'disrespect' them, as he put it. (Interview 41Anantpur, 6 January 2017).
Second, NREGA contributes to easing and in some cases challenging accepted caste norms that keep Dalits in a subordinate position in the village political economy. This can be exemplified in several ways: (a) being able to wear new clothes, wear sandals when walking through the main square of the village or riding a bicycle to get to work or riding a horse at a wedding are all markers of newly acquired status for Dalits that (sections of) the upper castes deeply resent. A Dalit man in his 40s reported: things have changed. When I was young, I was not supposed to sit on the carpet in front of my own house, otherwise we would be beaten up. Now I can walk through the main village [i.e. where the other caste live] without any issue. (Interview 29,Anantpur,5/1/17) In some cases, this leads to violent retaliation but at other times there is little that upper castes employers (especially farmers) can do, due to the safety net that the MGNREGA and other public policies like the Public Distribution System or microcredit programmes provide; (b) changing nature of communal events such as weddings, which is an exemple of the erosion of caste hierarchies and associated ideas of purity and pollution. Wedding arrangements are a case in point: 'in recent years the way they treat us has changed drastically. We used to eat separately from everyone else at weddings. Now we are invited and seated with everyone else, but we are given separate food and plates' (Interview 10, Guntur 17 February 17). While the very fact that Dalits are served food in separate plates is a reminder that caste hierarchies still persist, it is nevertheless a big change if one considers that not long ago, Dalits in that region were invited to weddings of upper caste families only as providers of 'impure' services like playing the drums and were served the leftover food. In fact, 88.47% of the households that we interviewed in our survey said that they had been invited as guests to an upper caste wedding in 2016; and (c) there have been changes in their day-to-day interaction as well. A Dalit man reported: 'I have been living here for 30-35 years. I have seen a lot of change. The way we are being treated dramatically changed. People touch us, some even drink our water when we offer it to them. They tell us: you used to be uncivilised and dirty, but look at you now, you are clean and more civilised. You have become important person now' (Interview 35, Anantpur, 6 January 17).
On the other hand, our survey, however, shows that caste norms regulating the villages' social interactions are still marked by acute discrimination against the Dalits, particularly in UP. On the one hand, signs of deference are still quite common: for instance, 90.3% of the women and 91.8% of the men said that they always give way to the upper caste when they meet them in the street (the figures are slightly higher for UP, compared to AP). Similarly, 8.2% of the respondents in Anantpur (AP, low HDI), 5.4% in Guntur (AP, high HDI) reported being prevented to enter a street of the village because of their caste. The corresponding figures for UP are significantly higher: 20.1% in Fatehpur (low HDI) and as many as 23.6% in Jhansi (high HDI). A similar picture emerges from other answers to our questionnaire, namely if the respondents had not stood up when an upper caste person came near them. Whereas 29.2% in Anantpur and 41.3% in Guntur said that they did not stood up, only 11.2% in Fatehpur and 6.5% in Jhansi answered the same. This evidence highlights that entrenched social norms limit the possibility of change. But it also signals that, particularly in AP, social norms are gradually changing and the reduced dependency of Dalit labourers on upper caste employers is an important contributing factor towards these changes. Though it would be too sweeping to conclude that the MGNREGA has broken down social hierarchies, it is certainly calling them into question (Sainath 2009), coupled with an awareness about their right and entitlement to live a life of dignity.
A potential disempowerment mechanism that we thought would be common in such a situation was violent retaliation by the upper castes whose status is being threatened. However, we could find only limited evidence of this. For instance, a Dalit man from AP (Anantpur) was badly beaten up because he had the 'insolence' of protesting when his father was insulted by some upper caste youth. However, these were mainly isolated examples in the villages where we conducted fieldwork. This does not mean that this type of disempowering mechanism is absent in other contexts. In fact, the literature on caste relations is full of examples of relations severely deteriorating in the wake of the changes that we described (Manor 2012;Gorringe 2017).
Generally speaking, the narrative from our interviews again points at significant differences between AP and UP. In the former case, Dalits were more confident that defying social norms would not result in serious consequences. In UP, on the contrary, Dalits were much warier of displaying their growing uneasiness with disempowering social norms. This was also reflected in their assessment of changes in the conditions of their communities over the last decades. Whereas in APparticularly in Anantpurthe prevailing answer was that change was dramatic and very visible, such a positive assessment was much less common in UP.

Market domain
The main empowerment mechanism in this domain stems from the availability of alternative sources of employment for agricultural labourers that MGNREGA facilitates; the subsequent increase in income security and thus lesser need of loans, especially for basic sustenance, such as food. We stress here the importance of debt as a key mechanism that disempowers particularly Dalit agricultural labourers significantly and contributes to solidify hierarchies (Guérin, D'Espallier, and Venkatasubramanian 2013). While the presence of bonded labourers was relatively low across our fieldwork villages -'only' 4% of the respondents to our survey said that they are not free to choose for which landlord to workmost families (60%) where indebted in one form or another, for an average sum of 70,764 rupees. Considering that one day of agricultural labour in our fieldwork locations is paid between 150 (for women) and 400 (for men) rupees a dayand that interest rates range from 24% to 48% per yearthis is a considerable amount. However, especially in AP, the dependency for getting loans on farmers-employers was reduced dramatically due to MGNREGA.
Many labourers told us in AP that it used to be very harsh during the summer months, when no farm job is available: 'during summer we get MGNREGA, previously it was only through loans from farmers that we could survive. Now taking loans from farmers has declined a lot. Not completely gone but almost' (interview 11, Guntur, 17 February 2017). In many cases, during the summer, labourers, before the introduction of the MGNREGA, had to rest a lot to make up for the lack of food. Some others told us that they used to be dependant on loans from farmers to get enough food to survive, but that now they can survive the summer relying on the MGNREGA and the loans from Self-Help Groups (which many find easier to repay due to MGNREGA). Where MGNREGA work is more readily available, like in Anantpur (AP), the common answer to our questions on whether labourers thought that MGNREGA helped them taking fewer loans, was mostly a resounding yes. But this was also true in Guntur (AP), which has a much lower average number of days of work under MGNREGA. What was crucial in this more prosperous district was that the programme was available during the summer, when no farm labour can be done. In UP, labourers were less enthusiastic in their responses, but they mostly agreed that the MGNREGA helped at least 'marginally' or 'to a limited extent' to reduce indebtedness. These changes have contributed to shifting labour relations from one being between a debtor and a creditor to one between an employee and an employer. This has several implications. First, debt-free labourers can choose freely who to work for, thus making landed farmers to compete for workers. This was a major problem for the farmers we interviewed in AP, who were keen on securing enough labourers for their fields, particularly for agricultural activities that needs to be completed in a very short number of days. Second, this has led to important changes in the way farmers treat Dalit labourers. For instance, it is now common for landowners themselves to approach labourers at their homesin the Dalit colonyand ask them to work for them, rather than labourers demanding work. Farmers must also talk 'with respect' (interview 4, Anantpur, 17 January 2016) to labourers. Virtually all labourers we spoke to in AP and a significant proportion in UP stressed the fact that they would not work for landowners that do not treat them with respect or call them by their caste name (see Roy 2014 for similar dynamics in Bihar). Most of the labourers we interviewed underlined this aspect as a major change in the way in which employers treat them. Moreover, in what is regarded by elder labourers as a major break with the past, negotiations for higher wages and better working conditions like the provision of water or food not served in marked plates are common and often met. These kind of negotiations are very difficult or impossible if the labourer owes money to the landowner. 2 In Jhansi district of UP, it was common for farmers to ask for a collateralusually gold or landthat they would keep until the debt is repaid. Labourers in such a situation told us that farmers would pay them only one day of work out of three or four, withholding the remaining wage as 'interest'.
It is also easier for labourers not indebted to farmers to refuse to work if they so choose, even though our survey shows that, in practice, many labourers rarely turn down an offer for work: 44% of those who owe money to farmers 'regularly' or 'sometimes' refuse to work, compared to 50% of those who are not in debt with farmers. This reflects the precarity that most workers facenotwithstanding the MGNREGA and other social protection policies. Many labourers in fact told us that it is better not to refuse offers of work because farmers might get 'angry', even though the general feeling among labourers (across districts) was that nothing serious would happen.
Reduced dependency on farmers, because of reduced indebtedness or because of alternative sources of income like the MGNREGA, enabled labourers to do things that were unimaginable only a few years ago. An upper caste farmer from AP told us: Times are changing. Now when they [the labourers] go to the toilet [in the bush] they stay there for one hour. And they even smoke in front of us. Everyone is smoking these days. My father told me that labourers would never smoke in front of him, now they do without any problem. My father was a respected person in the village, but we are not respected anymore. But I can't speak about this because otherwise you will think that I am a person stuck in the last century. I know things are changing. The thing is now that if you want to continue to do farming, you have to tolerate all this. (Interview 40, Anantpur, 6/1/2017) Another farmer from the same village remarked: Previously when we had to collect tamarinds from the trees we would offer women 10 rupees and they would come. Now they say they will not do it even for 400 rupees. They say it is too risky because they are afraid of falling from the tree.
He then added, referring to labourers who visit him to negotiate the wage: 'If someone comes to your home, they will sit on the cot, and not on the ground as it used to be.
[…] You feel very helpless when you see these things happening. But we can't manage without their work. So you have to accept it' (Interview 41Anantpur, 6 January 2017). This was echoed by a Rajput farmer in Jhansi, UP: 'the respect that workers used to have towards the farmers has completely vanished' (interview, VSCHAK, 29 November 2017).
The respect that farmers show to labourers in order to secure enough workers extends to caste norms. In Guntur district in AP, while conducting an interview with a Dalit labourer, an upper caste landowner came to the interviewees' house to ask her to assemble a few labourers to work in his field. Then he asked the woman's daughter to give him a glass of water, which she promptly offered in a metal cup. He drank and then left. Our respondent was quick to make us notice how this is a clear sign of how things have changed in the last few years. Clearly, the landlord consciously chose to ask for a glass of water and to drink before leaving as a sign of respect. In another case, a Dalit men told us that the way in which landlords recruit women workerssome agricultural activities in India are usually 'reserved' for womenhas changed in recent years. They would now approach the man, in the presence of his wife and ask him if she wants to work in the field. We were explained that this is a sign of respect, as asking directly to the woman might imply that the landlord would request sexual favours.
Not all these changes are the exclusive consequence of MGNREGA, of course. They are long-term processes that have been facilitated and accelerated by the presence of a number of social protection programmes none of which are perfect, but that, collectively, reduce the dependence of labourers on the dominant actors in the village. The decades-long crisis of India's agricultural sector is another contributing factors, as farmers are not as strongeconomically, socially and therefore politicallyas they used to be. All this has changed power relations at the grassroots.
However, this relatively rosy picture needs some qualifications. First, in some cases, Dalit labourers reported having seen very little change: 'I can't see any significant change. They [the upper caste] will never change. They don't like us to be treated on par with them. The younger generation [of Dalits] is reluctant to work in their fields and they go for work outside of the village. So our dependency as a community has decreased, but those of us who are here cannot survive without the farmers. Our location is a big headache. The colony is completely surrounded by them and they don't like it at all. They will never change' (Interview 11 Guntur,17 February 17). What this woman was saying is that farmers in her villagewhich is located in the paddy belt of coastal AP which has excellent irrigation facilitiesfarmers are still quite strong and powerful. This reduces the empowerment potential of programmes such as the MGNREGA, also considering that the farmers in that village were able, due to their control over the local administration, to limit the availability of MGNREGA work and to actually use it for their own benefit, for example for maintaining irrigation canals.
In other cases, the reduced dependency of labourers on farmers has led to actual disempowerment. For instance, a group of women from UP told us that since the introduction of the MGNREGA they have stopped working in the field of farmersa clear sign of empowerment, considering that their choice was dictated by the desire to avoid sexual and other forms of harassment. As a form of retaliation, they now prohibit women to use their fields for defecation, which forces them to walk in the dark for much longer at greater personal risk. In other cases, the farmers were able to control the allocation of MGNREGA work to a significant extent, and they used it to distribute work to those labourers belonging to a certain group in a highly factionridden village. We were told that Dalits were being chased away from MGNREGA worksite by other Dalits because that was 'their' scheme. This of course dilutes the empowering potential of the scheme in the sense that it is difficult for labourers to make common cause, if these kinds of divisions exist among them.

Conclusion
In this paper, we focus on how the processes of empowerment and disempowerment are triggered by a large scale public work programme, the MGNREGA, in two states of India, AP and UP. Thus, arguably, we seek to study processes of empowerment and oppression within the same framework. Our analysis suggests the following: first MNREGA undoubtedly provides income security and an independent source of income which inadvertently reduces people's dependency on more powerful actors such as husbands (household domain), upper or intermediate caste landowners and moneylenders (community and market domains). But reduced dependency does not necessarily or automatically translate into greater freedom or empowerment, because, at times, the reduced dependency has paradoxical effects that result in a deteriorating of the situation, particularly in terms of intra-household relations.
Second, the importance of gender and caste norms needs to be highlighted. MGNREGA (along with other social protection policies and alternative sources of employment) provides a minimum of independence and income security that allows Dalits women and men to make choices that put into question established social norms. As in a small context like a village, these acts of defiance are usually quite visible, this might be an important trigger for more profound (and collective) social change: when a debt-free Dalit agricultural labourer dares not to stand up when an upper caste landowner approaches, others might follow suit, slowly making it 'the new normal'. Similarly, when more and more women withhold their MGNREGA wage from their alcoholic husbands, this might help other women to do the same.
Third, it is clear that social norms remain a potent source of disempowerment, shaping what individuals perceive as desirable, possible, or even 'thinkable for their lives' (World Bank 2015, 3). While policies like MGNREGA can contribute to (slowly) changing norms, it is clear that the precarity and poverty that still affects the majority of the Dalit community is still a major impediment to the challenging of oppressive social structures. Similarly, patriarchal social norms remain a major constrain to the expression of women's agency.
1. Scheduled Castes (SC) is a term used in the Indian constitution and refers to former untouchables castes. SCs are usually referred to as Dalits and Scheduled Tribes refer to indigenous tribal groups. 2. For instance, among the respondents to our survey, 29.3% owed money to farmers-employers. Out of these, 35% could not chose to repay their debt in cash, but had to repay through labour, effectively making them bonded labourers in disguise, possibly the most disempowered category of workers imaginable. For those who can afford not to take loans from farmers, however, things have changed dramatically.