Beyond the grumpy rich man and the happy peasant: mixed methods and the impact of food security on subjective dimensions of wellbeing in India

Abstract This paper responds to the recent advocacy of subjective wellbeing in policy evaluation with an investigation of food security in rural Chhattisgarh, India, in 2010–2013. Conceptually, it suggests the need to move beyond a primary focus on happiness to consider a broader-based investigation into people’s subjective perceptions. In particular, it introduces a multi-domain model with some affinities to the capability approach, which asks what people think and feel themselves able to be and do. Methodologically, it suggests that the primary reliance on quantitative measures should be complemented by more qualitative approaches to give a more rounded appreciation of how people view their lives. Three approaches are presented: qualitative analysis of interview text; statistical analysis comparing a single measure of happiness with a broader, domain-based approach; and mixed qualitative and quantitative data generated from an individual case.


Introduction
With the archetypical figures of the happy peasant and the grumpy rich man, Sen threw down the gauntlet against the two dominant traditions in economic analysis and their respective preoccupations with what people have and what they think or feel. Against these, he famously argued that it was the scope of what people could do or be which mattered, that welfare should be measured in terms of 'capability' , or 'the alternative combinations of things a person is able to be or do' (Sen, 1993, p. 30). In this seminal passage, Sen suggests two possible relationships between how people think or feel subjectively and what they have objectively. The first is independence, that happiness may be quite unrelated to any external factor, the result simply of 'a cheerful disposition' (Sen, 1983, p. 160). The second admits a relationship between feeling good and having a particular commodity, but only by way of the capability that the commodity enables, the pleasure occasioned by being able to meet one's need or achieve something one values.
Thirty years on, all three approaches are alive and well. The preoccupation with what people have and how they can get more of it remains foundational to much development thought and action. The capability approach has gone from strength to strength. And over the last 15 years in particular, there has been a remarkable resurgence in advocacy of happiness as a target in public policy.
This paper re-visits the question of what subjective perceptions can contribute to the assessment of welfare through investigation of a particular empirical context -food security in rural Chhattisgarh, India, during 2010-2013. The paper makes two kinds of intervention, conceptual and methodological. Conceptually, it suggests the need to move beyond a primary focus on happiness to consider a broader-based investigation into people's subjective perceptions. In particular, it introduces a multi-domain model with some affinities to the capability approach, which asks what people think and feel themselves able to be and do. Methodologically, the paper suggests that the primary reliance on quantitative measures should be complemented by more qualitative approaches to give a more rounded appreciation of how people view their own lives.
The paper begins by reflecting on recent research on the assessment of happiness and subjective wellbeing (SWB), noting its growing tendency towards advocating a more nuanced and diversified approach. After a brief introduction to the research context and methods, the first results section presents strong evidence for the impact of food security programmes through qualitative analysis of interview text. Statistical analysis then compares a single measure of happiness with a broader, domain-based approach to consider whether access to ration cards (the gateway to the food security programme in India) and employment in a government rural works programme are associated significantly with quantitative measures of subjective dimensions of wellbeing. The final vantage point of the paper explores how quantitative and qualitative data can work together to reveal different aspects of an individual case.
Before proceeding, a note on the limitations of this paper is in order. The results sections are very brief, and can only provide a snapshot of different kinds of methods. A more comprehensive discussion would require the presentation of both more data and deeper analysis. However, the aim of the paper is to provoke comparative reflection across a range of methods, rather than to use any to provide a detailed investigation of the case in itself. Similarly, while the paper does offer insight into some substantive dimensions of what the food security programme has meant to local people, its intention is not to present an impact evaluation of the programme. Rather, the food security programme is taken as a case in order to open up discussion of how different conceptual and methodological approaches may broaden the contribution of subjective measures of wellbeing to understanding poverty and policy processes.

The challenge of assessing happiness
The idea that public policy and personal practice should advance happiness is a recurring theme throughout religious, philosophical and political thought (e.g. Haidt, 2006;Nussbaum, 2012;Stewart, 2014). What is new is the view that happiness -codified as 'subjective wellbeing' (SWB) -can be measured directly, and so provide metrics for use in evaluating interventions, guiding policy, and assessing national progress (Diener, 2000;Graham & Nikolova, 2015;Layard, 2006;OECD, 2013). This remains highly controversial. Basic questions persist about indicators' reliability, comparability, usefulness, and the politics behind their introduction and use. Debates are heated, and unusual in generating both advocacy and critique from the right and the left, reflecting very different political projects, methodologies, and views of the proper purpose and purview of policy (e.g. Ahmed, 2010;Atkinson, 2013;Booth, 2012;Ehrenreich, 2009;Gough & McGregor, 2007;Held, 2002;Layard, 2006;Michaelson et al., 2009;Nussbaum, 2012;Seligman, 2011;Sen, 2009;Sointu, 2005;Stewart, 2014;Stiglitz, Sen, & Fitoussi, 2009;Stone & Mackie, 2013;Taylor, 2011;Thin, 2012;. There are two broad trends in the advocacy of happiness in policy. The first is to see happiness as offering a new and different goal with distinctive policy implications (see e.g. Layard, 2006;Thin, 2012). The other takes SWB as providing a new, direct measure of utility, or Bentham's balance of pleasures and pains (Collard, 2006, p. 332), freed from the tie to goods and services which has become its conventional interpretation in economics. The OECD's 'Guidelines on measuring subjective well-being' , provides a striking example of the latter. While careful to stress that subjective indicators should not replace other measures, the guidelines suggest they can be used as 'an alternative yard-stick of progress' (OECD, 2013, p. 36) to show differences between groups or across countries. Their strength is in their capacity to give a single, composite impact measure, as: being grounded in peoples' [sic] experiences and judgements on multiple aspects of their life [sic], measures of subjective well-being are uniquely placed to provide information on the net impact of changes in social and economic conditions on the perceived well-being of respondents. (OECD, 2013, p. 36) Two constructs have dominated the debates on happiness and SWB in policy: life satisfaction, and 'affect balance' -the relative presence of so-called 'positive' emotions and relative absence of so-called 'negative' ones. Economists have tended to use (cognitive) life satisfaction as the main measure of SWB, while psychologists have tended to favour affect or combinations of the two (e.g. Diener, Suh, Lucas, & Smith, 1999). Recent scholarship, however, shows that measures of life satisfaction behave rather differently from measures of affect. In particular, the former tend to correlate with economic status while the latter may not (e.g. Diener, Ng, Harter, & Arora, 2010;Graham & Nikolova, 2015). This suggests that the two constructs cannot be regarded as equivalent to one another, nor should they be combined in a single measure since this would confuse the two perspectives.
Philosophical reflection on the meaning of happiness combined with empirical work on psychological wellbeing has also suggested the need to broaden the frame of happiness assessment. In the current wellbeing literature two concepts dominate, both of which originate in the thought of the ancient Greeks. The first is hedonic wellbeing, which equates happiness with the experience of pleasure. The second is eudaemonic wellbeing, which looks to Aristotle in considering that happiness lies in the pursuit of a purposeful and virtuous life (Nussbaum, 2012;Ryan & Deci, 2001). A third key construct has thus been added to the SWB measures: the sense of meaning and purpose in life. The UK Office of National Statistics uses four questions in its measurement of SWB: on life satisfaction, 'positive' emotion (feeling happy), 'negative' emotion (feeling anxious), and eudaemonic wellbeing with the question, 'Overall, to what extent do you feel the things that you do in your life are worthwhile?' (ONS, 2012; see also Graham & Nikolova, 2015).
The tendency to broaden out from a 'simple' measure of happiness is strikingly evident even amongst staunch advocates of happiness as providing an alternative goal for policy -they end up recommending a more familiar multi-dimensional vision of wellbeing. Examples include Layard's endorsement of 'ten keys for happier living' which incorporate the 'five ways to wellbeing' advocated by the New Economics Foundation (Action for Happiness n.d.; NEF, 2008); the shift of Seligman, one of the founding fathers of positive psychology, from advocating 'authentic happiness' to a broader 'wellbeing theory' (Seligman, 2011) 1 ; the UN's 'happiness resolution' (United Nations, 2011); and Bhutan's renowned 'Gross National Happiness' approach which encompasses a number of standard development indicators besides the subjective happiness assessment (NDP 2013).
In addition to differences among the constructs considered to compose SWB, the particular measures used to assess them can materially affect results. The life evaluation method used by the Gallup world poll, for example, is recognised as the one most closely correlated with income. While advocates suggest this shows it to be 'a purer measure of life satisfaction' (Kahneman & Deaton, 2010, p. 16492) it could also be argued that the way the question is posed evokes an economic frame, as it asks people to imagine a ladder on which rung no. 10 is the 'best possible life for you' and 0 is 'the worst possible life for you' (Gallup, n.d.). 2 This brings to mind Schwarz's (1999) discussion of 'how the questions shape the answers' in any self-reported measure. In addition to the specific question that is asked, significant variability in SWB scores has also been found by the ordering of questions and the specific dataset used (e.g. Deaton, 2012;Graham, 2011a). Reflecting on the 'numerous unresolved methodological issues' for measures of affect in particular, Stone and Mackie (2013, p. 12) thus suggest that they should as yet be included in major surveys only on a pilot basis, and recommend prioritising 'modules for inclusion in targeted, specialized surveys' (Stone & Mackie, 2013, p. 13).
The relationship between philosophies of happiness and indicators of SWB is a rather tricky one. Appreciation of qualitative differences between individuals in their experience of happiness is one of the classic objections to direct measures of utility in economics. The comparability of SWB metrics relies on their abstraction. People are free to use their own criteria of judgement, as the questions ask simply 'how happy' (or satisfied) people are, not what makes them so (Diener et al. 1999). While different constructs clearly produce different results, it seems problematic to identify the experience of happy feelings with a hedonic conception of wellbeing, or a sense of purpose in life as necessarily reflecting eudaemonic values. It may be that people feel happy due to their sense of life being meaningful, or that the purpose which drives them is the hedonistic pursuit of fame and fortune. The abstract form of the questions means that we simply do not know how people understand the happiness or satisfaction to which they refer. What SWB offers is a simple rating, a number (or set of numbers) which can stand for subjective success in life.
A further issue in assessing happiness is the social processes involved in producing a response. In the mainstream literature these appear primarily in the issue of adaptation, the widely remarked fact that people adjust their expectations to their circumstances (Clark, 2012;Graham, 2011a). This constitutes a significant problem for using satisfaction measures in policy evaluation for two somewhat contradictory reasons. On the one hand, there is too much variability, as the goal-posts are always shifting. Perversely, a successful programme may thus result in lower, rather than higher, levels of satisfaction if it raises aspirations more sharply than it improves living standards. On the other hand, there is too little variability, as SWB theorists generally agree that individuals have a 'set-point' score to which they return after positive or negative 'shocks' , such as winning the lottery or having a debilitating accident. This again breaks the link between how people are doing by external measures and how they are feeling about their lives.
Psychologically adaptation may be functional, as people in difficult circumstances 'train themselves to take pleasure in small mercies' and this helps them to endure (Sen, 2009, p. 283). Sociologically speaking, however, high satisfaction may signify the low aspiration of internalised oppression, rather than a genuinely positive experience. It may also indicate the limited standards of comparison available to the respondent, such that very basic provision of education or healthcare, for example, may be deemed to be satisfactory. Culture also plays a part. While some cultures value 'aiming high' and continuing to strive throughout one's life, others identify maturity with acceptance of and accommodation to one's lot.
While adaptation is without doubt important, it does not exhaust all the issues involved in the social production of responses to happiness surveys. The object of investigation is essentially qualitative: people naturally describe their judgements and feelings in words or images, not in numbers. As from any language to another, the 'translation' into numbers is by no means straightforward. Thoughts and feelings are also fluid and complex, and often far from transparent even to the person him or herself. In addition, questions about happiness are not simply personal, but exist in an ideological context. Certain views of 'happy lives' are normative, as a short meditation on notions of 'a happy family' makes clear (Ahmed, 2010). Furthermore, if happiness is what everyone wants, then to be asked how happy you are is a very loaded question (Ahmed, 2010, p. 5). Answers may reflect at least as much how people want to be seen, as how they actually feel. Perhaps most critically, the information is essentially subjective. It is important to reflect quite carefully what this means. Most simply, it means it is person-specific: 'There is no way for a person other than the respondent to provide the correct answer' (OECD, 2013, p. 47). More profoundly, this means that such data do not stand in their own right, but in relation to the person who reports his/her thoughts or feelings, the context in which he or she is set, and the instrument through which the data are generated. 3 For statistical researchers and officers who yearn for 'hard evidence' , handling data on topics which are essentially qualitative and subjective sets up a fundamental tension. As shown by the OECD's Guidelines (2013) such researchers respond by seeking to authorise and standardise measures to control variability and promote comparability. The assertions of a 'science of happiness' (Diener, 2000;Layard, 2006), recommendations to check what people say against the frequency of their smiles, identification of experience-sampling as the 'gold-standard' for individuals' accounts of emotions (Kahneman, Krueger, Schkade, Schwarz, & Stone, 2004) and the move toward brain imaging (Berridge & Kringelbach, 2011) all indicate a common impulse to deny the subjectivity of recall or reflection and seek to make the subjective objective. Again, the OECD provides a remarkable example, as it claims that measures of SWB can transcend the limits of people's own understanding, revealing the inner truth of how people are really doing, even beyond their 'conscious' thoughts and feelings: Most importantly, measures of subjective well-being provide information on the actual impact of an initiative on the respondent's subjective well-being, rather than the impact that the respondent consciously identifies. (OECD, 2013, p. 41) In this paper we argue for an alternative approach. We accept the essential relationality of subjective data, and so seek to re-place it in the context of the subject whose perceptions are sought and the methods through which it is generated, recognising that both of these are essential for accurate interpretation of the results produced. Before proceeding to our case study we introduce some alternative approaches to assessing subjective dimensions of wellbeing.

Alternative approaches to subjective dimensions of wellbeing
A number of researchers have sought to develop ways of assessing subjective dimensions of wellbeing which reflect the values and concerns of local people. Clark (2003) describes how ordinary people in rural and urban South Africa view human wellbeing, while Trani, Bakhshi, and Rolland (2011) emphasise the importance of social and community values and roles in shaping people's perceptions of wellbeing in Afghanistan. Woodcock, Camfield, McGregor, and Martin (2009) describe the WeDQo, an instrument to measure subjective quality of life, conceived as 'the outcome of the gap between people's goals and perceived resources … in the context of their culture, values, and experiences of un/happiness' (Woodcock et al. 2009, p. 137). 4 'Inner wellbeing' ,(IWB) the approach developed in the research presented here, focuses on what people feel and think they are able to be and do. Although it was generated quite independently, the approach nevertheless has clear conceptual affinities with the capability approach, but focuses on subjective perceptions. We began theoretically, with reflection on earlier research, and then developed the model through an intensive mixed method process in rural communities in two study countries, Zambia and India. This involved consultation with NGOs and other local people; extended grounding and piloting; statistical testing and revision; and ongoing critical reflection within the team (White & Jha, 2014a). 5 The model has seven domains: economic confidence; agency and participation; social connections; close relationships; physical and mental health; competence and self-worth; and values and meaning. 6 These were explored alongside external measures of how people are doing (such as occupying positions of political office or cultural status, crops harvested, education levels, access to government services, assets, etc.) with the aim of exploring complementarities and tensions between their achievements and perceptions of wellbeing.
Each of the domains of IWB is assessed through a series of questions intended to capture different aspects of the domain, with graduated answers which correspond to a one to five Likert scale. The economic domain, for example, includes items to assess how people feel about how they are doing at present; how their economic position affects their social participation (their ability to host guests as they would wish to); their sense of economic security (how confident are they that they could manage, for example, if someone were to fall ill); their sense of how they are doing in comparison with others around them; and their economic confidence looking forward (whether they think their children will have a better life).

Methodology
The field research was undertaken in Sarguja district, northern Chhattisgarh, with an initial visit in November 2010 and two main fieldwork periods of four months each from February to May in 2011 and in 2013. The main research instrument was a survey, conducted with husbands and wives (interviewed separately) and women-who headed households. We interviewed married couples because we were interested to explore whether there were any patterns in the responses of husbands and wives, given the widespread agreement that relationships are important to wellbeing. Women-headed households were included as they are a group commonly found to experience particular economic and social vulnerabilities. In 2011 a total of 157 married men, 156 married women and 26 women-heading households were surveyed, while in 2013 there were 174 married men and 171 married women and 23 women-headed households. The much smaller number of women-headed households reflects the social reality in the area. One hundred and eighty-seven (187) people (55%) of those surveyed in 2011 were surveyed again in 2013.
The mean age of respondents overall was 41 years in 2011 and 42 in 2013, from a youngest of 18 to an oldest of 80. In 2013 the mean age for married men was 44, for married women 38. Women-headed households tended to be older, with a mean age of 55 years. They had all previously been married and were either widowed or divorced.
The survey was undertaken in as conversational a way as possible, to allow scope for people to tell stories and for questions of clarification, which were recorded in the notes. Qualitative data include 151 survey notes and full transcriptions of 30 open-ended interviews. IWB scores presented below relate to respondents who completed all of the IWB section in full, rather than to the sample as a whole. Quantitative data were entered in Excel and translated to SPSS or STATA for statistical analysis. More complex psychometric analyses in validating the statistical model of IWB (not reported here) were undertaken in Lisrel . Qualitative data were analysed through content analysis with the help of NVivo.
We were supported in the field by Chaupal, a local organisation undertaking community mobilisation. Our research villages were predominantly Adivasi, ethnic minority communities who were classified as 'Scheduled Tribes' under colonial administration and which have a long history of problematic relationships with outsiders (Sundar, 2007). 7 There is no doubt that the goodwill that Chaupal enjoyed in our research villages was critical to people's readiness to talk to us. 8 With Chaupal's help we recruited a local team of four research assistants (two male and two female) who acted as peer researchers, mediating, interpreting and interacting between the local respondents and the external team members through the grounding and piloting process and throughout the fieldwork. The field research was led by the project research officer (third author) who remained in the field for the entire fieldwork period. The project director (first author) visited for shorter times, including all piloting at the beginning of each fieldwork period.

The study context
The research took place in four villages, which vary in terms of accessibility, natural environment (access to water, hill or forest), severity of poverty, and community composition. Differences between them notwithstanding, overall these are extremely poor communities, amongst whom hunger was commonplace until recently. People are struggling to obtain the title deeds for the forest land they have occupied for many years, despite the Forest Rights Act of 2006 having recognised their rights to make such claims. Literacy levels are very low, with more than half of our respondents reporting no schooling at all, and a further 20% being able only to write their own names. The mainstay of the economy is agriculture, with most people doing some farming, supplemented by casual labour and gathering of non-timber forest products. Agriculture is largely rain-fed. There is significant and problematic use of alcohol related to the fact that they brew it from some of the forest products they gather. The communities have long been the target of missionary activity -Christian during colonial times, and at present radical Hindu. Despite this, 57% of our respondents still follow traditional Adivasi religious practice. There is a local history of strong if intermittent collective action, and currently there is mobilisation by local organisations to demand the rights promised by the state. 9 In poor, socially marginal and geographically 'remote' communities like these, one might think that the issues are clear and there is no need for any subjective measures of wellbeing. In one way, therefore, such a context might seem to represent a hard case for exploring what subjective measures might add. There is, however, a major consideration to recommend this location. From 2004 onwards the Chhattisgarh state government made significant moves to increase the spread and efficiency of food security programmes, in particular the delivery of highly subsidised Public Distribution System (PDS) rice. The remarkable success of these programmes attracted national attention (Dreze & Khera, 2010). The programmes have been complemented by other forms of significant state-level investment, in particular the construction of new roads and schools. Food security has also been increased by the nationwide job guarantee programme, the Mahatma Gandhi National Rural Employment Guarantee Scheme (MGNREGS), which guarantees 100 days of unskilled manual labour a year at the minimum wage for all rural households.
What is unusual is not the existence of these programmes, but the fact that they are working so well. This reflects a combination of political will from the state government and active participation of citizens to lobby, monitor and report abuses (Dreze & Khera, 2010). Our survey shows for 2011 over 80% of respondents, and for 2013 77%, were receiving PDS rice at Rs 1 or 2 per kilo. In 2011 97% of eligible school children were getting their free mid-day meal, and 94% of those eligible were receiving the Integrated Child Development Scheme, a health and nutrition scheme for pregnant women, nursing mothers and young children. 10 In 2013 50% of respondents had undertaken some work under MGNREGS over the previous year, though of these only 12% said that they were paid on time while 52% said they received the full payment. The combined effect of all these programmes is that people were no longer having to go hungry. This constitutes a major difference compared with earlier experience, which people frequently commented on without prompting. As one person put it: Now we are getting rice from the government and so we are able to live our lives.
The following sections consider what different conceptual and methodological approaches to subjective dimensions of wellbeing can contribute to understanding the impact of these food security programmes.

Qualitative data analysis
Analysis of the interview data from Chhattisgarh shows that in the vast majority of cases where people compare the past with the present, they say things are easier now. This counteracts the usual tendency towards nostalgia. Instead, people recall that in the past there were times 'when the cooking fire wasn't lit' or when they had to dry mango seeds and pound them into flour because they had no rice. People so commonly remarked that they no longer have to go hungry that our research team stopped noting this down. Respondents often linked this directly and without prompting to food security programmes, with the following comment being typical: But we have never gone hungry once the government has started giving rice, we have not known hunger since then.
The immediate physical gain of not going hungry also has social implications. The following comments express this: If you have enough to eat, you have honour. If you don't have rice in the house, then you have nothing to offer. Now we have harmony in the home, because economically we are doing better.
One woman wept as she recounted how because of their poverty they had to remove one of their sons from school: Because of our poverty, so much poverty … so we thought 'let one son study, at least and let the other work with us and help us earn some money so we can put something in our stomachs' . If they had both gone to school then what would we have eaten? That is why we couldn't afford for the younger one to study. …. My heart pains for that … that my younger child didn't study … I feel sad. But what if I am sad … we were poor. … What would we have eaten in those days when there was no ration? You get that now, not then! There was nothing then.
There are few exceptions to this positive reflection on the present over the past. Most of the exceptions concern the happen-chance of individual lives, such as illness and household division; others came from wealthier people regretting that the poor have more options now, making it hard for them to get people to work on their land; and the rest come from people who are excluded from PDS (e.g. households breaking up, moving location). The following statement comes from one such man whose household was barely managing to sustain itself: If for instance my name were to come in the survey [BPL] list then that would solve half the problem and as for the rest we would manage from doing some labour. But because our name is not on the BPL survey list there is a lot of hardship (trouble).
Overall then, for the vast majority who are registered for state benefits, the qualitative data suggests people feel much better off than they were before. A number of people also make a direct link between this improvement in their circumstances and the improved quality of their lives. There is, however, need for caution. While many people say things are better now, they also say that things are still hard, that now they are just able to manage without going hungry, whereas before they were not. There is repeated emphasis, from both wealthier and poorer people, on intergenerational transmission, that the level of your own resources -especially land -will determine your children's prospects. Not only are their lives materially still difficult, but many also bear the psychological scars of their earlier experience. This is evident in the comment above from the woman who had to remove her son from school. Another example is the response below of a 30-year-old woman when asked if she believed her children would have a better life than she had. The narrative structure in itself conveys her questioning and uncertainty, as she moves to and fro between discussing the importance of education and land, between what you can do for your children and it all being in God's hands, between her attempts to make things better and her experience of failure, and between her longing for a better future and her fear of allowing herself to hope: I have very little confidence that their life will be better. If they are able to study only then they will do well. Who can say what will happen in the future. It is all up to God what will happen to them. They can do well if … when do children do well? They do well if you are able to leave some fields to them. If you are able to do some good farming and leave some wealth to them, or you are able to educate them and then they are able to get a job or something, then they'll be able to do well. But I have very little confidence that they will do well. Every time we have tried we keep losing everything. So how can I say that things will become good for them? I have dreams for them, it's entirely up to them if I'm able to educate them then they'll do well. But every time that we've tried to become better off, everything has just got shattered, so I have no faith now, I'm scared to hope even, because I know that hope will just come to nothing.

Testing for statistically significant relationships between access to food security, happiness and IWB
The widespread unprompted references to PDS and -to a lesser extent -MGNREGS in the qualitative data suggests that these programmes should be strong predictors of our quantitative measures of subjective dimensions of wellbeing. In this section we report on some limited statistical analysis of our 2013 survey which tests whether this was in fact the case. 11 We use two indicators of access to food security programmes: having a ration card and having undertaken work under MGNREGS in the past year. These are admittedly crude. We can safely assume that most respondents used their full entitlement of PDS rice (our 2011 survey showed above 98% of those eligible for PDS did in fact receive it) (Fernandez et al. 2014, p. 52). However, we did not ask how much work on MGNREGS people had done, so our figures are likely to obscure considerable diversity in amounts of income derived from this source. It should also be noted that we are not seeking to claim that these are indicators of capabilities or functionings, only that they signify functional access to the food security programmes. Table 1 presents simple correlations between these two measures of access to food security programmes and two measures of subjective dimensions of wellbeing: the global happiness and IWB domain scores. Of course, it is important to note that correlations 12 can show the strength of a linear relationship between variable pairs, but not causation. Table 1 shows that no statistically significant relationship was found between the global happiness indicator and either measure of access to food security programmes. For IWB, however, a number of significant relationships emerge. Firstly, the IWB composite, produced by summing the scores for all seven domains, shows a significant positive relationship with work under MGNREGS (r = .1635, p ≤ .01). The agency and participation domain shows a significant positive relationship with both measures of food security access, though the relationship is much stronger with the MGNREGS measure (r = .2137; p ≤ .01) than the ration card measure (r = .1030; p ≤ .05). Other domains show a significant and positive relationship with one of the measures. Work under MGNREGS is significantly and positively related to the social connections domain (r = .1638; p ≤ .01). Although fewer domains show significant correlations with MGNREGS work, where they do the degree of significance is stronger -all at the .01 level. Having a ration card is significantly related to the domains of economic confidence (r = .1139; p ≤ .05), competence and self-worth (r = .1087; p ≤ .05), and values and meaning (r = .1453; p ≤ .01). The close relationships and physical and mental health domains show no significant correlation with either measure of food security access. Table 1 suggests there may be a value added in using a multi-domain measure over a single item measure of happiness or SWB to gauge the assessment of the impact of food security policies on how people think and feel about their lives. Where no significant relationship is shown for global happiness, the overall IWB composite correlates significantly with work under MGNREGS, and five of the seven individual domain scores correlate with at least one of the food security measures. It is not of course possible to conclude what the direction of influence might be. The correlation with the agency and participation domain, for example, might suggest that more assertive people are more likely to gain access and extract a benefit. In addition, although some measure of correlation was found with five out of the seven IWB domains, only that of the agency and participation domain shows any degree of strength. This very limited statistical analysis thus provides a prima facie case for stating that food security programmes have some positive impact on what people think and feel they can be and do, but this needs to be tested through further analysis.
To test the strength of the relationships between access to food security programmes and subjective dimensions of wellbeing we then undertook a regression including controls for age, age-squared, gender, and schooling. Age and gender were chosen as control variables because they are key dimensions of social structure and life experience, shown to be significant to SWB in numerous analyses. Age-squared was included to enable us to capture the non-linear relationship noted in many SWB analyses, whereby SWB scores are higher in earlier and later years, but dip in middle age (Frey & Stutzer, 2002). Schooling was used as a basic indicator of capability. 13 Our measures of health status were subjective, so we chose not to use these in order to avoid mixing objective and subjective items as controls, lest we introduced false positives through having subjective measures as both independent and dependent variables. We also excluded measures of economic status, since previous analyses showed inter-correlation of access to a ration card with all three of our economic variables (occupation, landholding and assets). 14 While the ration card is intended for poorer citizens, in these largely Adivasi communities almost everyone would theoretically be eligible. Not to have a ration card in this specific context was thus an indicator of marginality, and this was strongly supported by the qualitative data. We similarly decided not to include ethnicity in order to avoid inter-correlation between control and dependent variables. Earlier analyses showed a significant positive relationship between working on MGNREGS and belonging to the less well-off Adivasi groups (r = .1102, p = < .05), and a significantly negative relationship between working on MGNREGS and belonging to the better-off Adivasi groups (r = −.1754, p = < .01). This is what would be expected, since it would be predominantly poorer people with few alternatives who would seek the heavy manual labour that MGNREGS entails. The results of our regressions with control variables are presented in Table 2.
The regressions in Table 2 show that schooling is the strongest predictor of all the subjective indicators, with the coefficients showing stronger associations with the IWB domains and composite than with the SWB indicator. 15 Gender is the next strongest predictor, with women having significantly lower scores than men. Although it shows no significant relationship with SWB, gender is associated significantly with the IWB composite and three of the IWB domains (agency and participation, social connections, competence and self-worth). While this cannot be discussed further here, the influence of gender is clearly an important finding to be pursued in later research. Age-squared does not show any significant association with either SWB or the IWB composite. However, it is associated significantly with the IWB domains competence and self-worth (p < .05) and economic confidence (p < .1).
With regards to our two indicators of access to food security programmes, Table 2 shows that access to a ration card does not show any significant association with either SWB or the IWB composite. However, having a ration card is associated significantly with three IWB domains: economic confidence (p < .05), self-worth (p < .05) and values and meaning (p < .1). Working in the MGNREGS programme is associated significantly with the agency and participation domain (p < .05). As noted above, it is not possible to determine the direction of these relationships. It could be, for example, that those who are more assertive, with a stronger sense of their own agency, are more likely to participate in MGNREGS, or that participation in the rural works programme enhances people's sense of agency. The quantitative evidence thus demonstrates a positive association between having a ration card and/or working on MGNREGS and some IWB domains, whereas no such relationship was detectable for global happiness. The quantitative analysis is clearly limited, both in the lack of sensitivity of our measures to access to the food security programme, and in the small number of controls which we included. Nevertheless, it does provide prima facie support for our contention that a multi-domain approach can add value over a single global happiness indicator.

The 'Happy Peasant' and IWB: qualitative and quantitative combined
This section combines quantitative and qualitative data in considering an individual case. While this clearly cannot yield results which are generalisable across the sample, it provides an opportunity to consider in more depth the calculations that people may make in assigning quantitative scores to their wellbeing, and the different kinds of factors which such scores may reflect. It is a form of 'cognitive de-briefing' , recommended by Camfield (2015) as a qualitative method which can increase the reliability of survey data. We call our case study woman 'Sukhi' , and begin the section by considering the contrast between approaching the same person through a poverty or a wellbeing 'lens' . We then go on to explore the degree to which her scores and comments may be read as a reflection on the impact of food security policies on her life.
We begin with a poverty perspective. Sukhi lives in our poorest village. She is an elderly widow living alone. Her husband died many years ago, and she has no schooling. She owns a small amount of land but has no labour to farm it, so this past year it has lain fallow. Because she came originally from another village where her one surviving son lives on their land she has no ration card or widow's pension. A second son had died after developing mental illness in early adulthood. Sukhi lives on the edge of a marginalised community, where she is the only person of her caste. While she cooks and keeps house for herself, she is completely dependent on provision from her son. Now from a wellbeing perspective: Sukhi said she is happy with her son's care but has chosen to remain living separately. She has good relations with her neighbours. She is content with her economic position, though when asked to compare her standard of living with those around she joked: 'I must be doing better, since they are getting welfare benefits and I am not!' She is proud of what she has achieved: 'I planted these trees!' 'I built this house!' With this comes a strong sense of ownership, of identification with this as her place. This is why she chose not to go and live with her son. In addition, people had said to her that because her husband had died here it was showing respect towards him for her to stay. She felt this too: This is where my husband died, and from where God will also call me. Until God calls me I will stay here. After that my son can do with it all what he wants.
As so often with wellbeing, Sukhi's case can be read at least two ways. One might argue that Sukhi is a great example of the value added of looking at wellbeing, because the contrast of considering her though a poverty or wellbeing 'lens' is so marked. Alternatively, she could be the paradigm of Sen's 'happy peasant' , whose cheerful disposition is at odds with her material situation. Considered more closely, however, the reality seems a little more complex as both the qualitative and quantitative data show something more discerning than simply a happy disposition, although that may indeed be present. Figure 1 presents a visual image of Sukhi's IWB domain scores. Sukhi began the subjective section of the survey by pronouncing herself 'very happy' (5) to the global happiness question. She similarly answered all items in the values and meaning domain at 5. Sukhi also gave a 5 to all but one of the items on self-worth (the remaining one she gave a 4). Again, she gave a 5 to all but one of the items in the physical and mental health domain. She gave the remaining one a 2, explaining that she sometimes feels too weak for the work she would like to do. The close relationships domain she scored similarly with all 5s, except for a 2 when asked if she had someone she can talk to if she feels low.
A much more mixed picture appears in the economic, social connections and agency and participation domains. As these domains are most relevant to the questions of policy impact and standards of living, we discuss them in a little more detail. Figure 2 presents a graph of Sukhi's item scores for these domains, 'global happiness' , and two subjective economic review items. Considered item by item, it is clear that she gave a carefully calibrated set of responses according to the particular questions she was asked.
Sukhi scored three of the economic domain items at 5, though she gave a 3 to signify that her economic position is on a par with her neighbours' , and a 2 for confidence in her son's future. This is because, she said, he has no education. She clearly distinguished between the IWB questions about how she felt about how she was doing economically (5 -very good) from the subjective review items  which asked her to rate how she was actually doing economically (3 -neither well nor badly). As she explained it, the base of her positive feeling about her economic position was in fact a mix of relational and material. She had enough for her needs, but most importantly she was happy with the caring relationships she had with her son and daughter-in-law.
Sukhi's scores for social connections are polarised. The questions about everyday neighbourhood relationships (SC2; SC4; SC5) she scored at 4 or 5. But she gave a 2 to the question about having people outside the family she can depend on, and a 1 to whether she knows people to whom she can go to for help. She related this to the fact that she has repeatedly tried and failed to get a ration card: 'If they do not help me, in what sense can I say that I know them?' Her scores on agency and participation step down another notch. While she scored 5 for freedom to make her own decisions and being listened to in the community, she said she goes but never speaks at village meetings (2); she has little confidence that the community can get together to take action (2); and she has no power to change official decisions (1). 16 The overall pattern in Sukhi's IWB profile is that she is happy with the areas over which she has direct control -herself, her family and neighbourhood relations. She is not happy about her exclusion from state benefits, nor her inability to get this changed, nor does she have an unrealistic assessment of her economic situation. Sen's (1983) caution against using happiness as a measure of standard of living is thus vindicated in this individual case. Sukhi herself rates these items differently, and her criteria for the judgement of how she was doing economically was clearly given in relation to the standard of her (very poor) neighbourhood. While this contextuality -the particularity of her frame of referencemight be seen as a weakness from some perspectives, it is actually the nature of subjective perspectives (and it is far more characteristic of many supposedly 'universal' approaches than their proponents would like to admit). Similarly, since Sukhi is both one of the relatively few non-beneficiaries of the food security programme, and explicitly comments at several points on her dissatisfaction about this, it would clearly be a mistake to attribute her happiness to the success of the policies.
There is a further point to this story. It is not that Sukhi has had an easy life, but at some level she seems to have made a decision to be happy. Asked whether she feels tension, she replied: 'I don't worry, if I did I would not still be here! I remember God and don't worry about anything' . Asked if she has someone she can talk to if she feels low, she smiled and announced with pride: 'Silence!' Small sorrows she may talk about to others, she explained, while large sorrows she keeps to herself. This shows that subjectivity is not simply 'there' as a state to be recorded but is actively (and interactively) constituted in relation to local discursive constructions with a richness, vitality and depth which is altogether different from the thin abstracted composite of 'subjective well-being' .

Conclusion
This paper began with Sen's rejection of judging levels of welfare by what people have or what they think or feel, and his claim that what matters is what people can be or do. In response we has sought to make two linked arguments. Firstly, when it comes to what people think or feel there may be other elements beyond 'happiness' which are relevant to consider in policy contexts, including what people perceive themselves able to be and to do. Secondly, in order to understand more fully subjective dimensions of wellbeing, it is important to complement statistical analysis with qualitative and mixed methods.
Our review of the literature on happiness and SWB in policy reveals three key challenges: different philosophies of happiness; diverse constructs and means of assessment; and the complex social processes involved in producing data. The epistemological assumptions of this literature are clearly very different to those of this paper. However, there are some points of connection. Growing recognition of complexity in understandings of happiness has resulted in a more diversified set of constructs and measures being used in a more intentional and reflexive way. Our call to broaden attention to subjective dimensions of wellbeing beyond happiness can be seen as taking this agenda further. Similarly, reflection on adaptation has concluded that subjective measures cannot stand alone as the basis of policy decisions (see e.g. Graham, 2011b). Our argument that subjective data are essentially relational, their meaning found in relation to the person whose perspectives are reported, his/her particular setting and the methods used to generate the data, may be seen as a development of this point.
The three results sections used different methods to explore the potential impact of the food security programme on subjective dimensions of wellbeing. Evidence from qualitative interviews was strong: people frequently volunteered comments directly linking the food security programmes -in particular PDS rice -to reducing hunger and associated this with improved relationships and IWB. This corresponds to the strongest category in the Quip, a new protocol developed for assessing project impact through qualitative data: 'change explicitly attributed to project and project-linked activities' (Copestake & Remnant, 2015, p. 2). However, the interview data also caution against painting an overly optimistic picture. They indicate that most people are still living with very limited material resources, and that the long history of marginalisation and exploitation these communities have experienced has not been erased but continues in flickering and uncertain expressions of trust and confidence. This suggests the importance of in-depth interviews and narrative analysis, rather than relying on the selection of 'sound-bites' from focus groups as the basis of qualitative evidence.
Statistical analysis of two measures of access to the food security programme and two measures of subjective dimensions of wellbeing provide prima facie evidence for the added value of the multi-dimensional 'inner wellbeing' approach over a single 'global happiness' indicator. Simple correlation analysis showed the greatest significance between access to food security and the domain of agency and participation. A simple regression with some demographic variables again showed some significant relationships between the measures of access to food security and some IWB domains. We are not, of course, able to determine the causal direction of these relationships. The regression also drew attention to the importance of schooling and gender in predicting the scope of what men and women feel able to be and to do. These are important issues to be pursued in further research.
The final section took analysis to the individual level, following Sen's (1983, p. 160) suggestion that high scores on happiness might simply reflect a 'cheerful disposition' . The contrast between an elderly woman's objectively poor position and her high happiness score offers support to Sen's rejection of happiness as a proxy for standard of living. Nor was her case a valid indicator of the food security policy's success, since the woman was one of the relatively few non-beneficiaries. This case, however, did not support a rejection of all quantification of subjective perspectives, since Sukhi's dissatisfaction with her exclusion from the welfare programme and her realistic appraisal of her economic situation could be read from her scores. In addition, considering one single individual in more depth provided the opportunity to begin to build up a deeper appreciation of a person as subject, and so to begin to connect to the rich seam of intellectual analysis dedicated to the qualitative study of subjectivities (e.g. Hall & Du Gay, 1996;Mama, 2002).
The limitations of this paper are its small scale and exploratory nature. Its intention is to open questions, to expand the space for a different and broader discussion of the subjective dimensions of wellbeing and how these relate to policy effectiveness, not to provide definitive evidence for a particular form of impact. While we believe the detail made possible when considering a single case is of value, it clearly does not provide the basis for broader generalisation but indicates possible questions for further exploration.
Given these limitations, the paper nevertheless suggests that it is valuable to consider subjective perspectives on wellbeing in international development and social policy. However, the complexity of people's lives means that there are many intervening variables which complicate the impact even of a very significant policy. This raises questions about the magic bullet of pure utility imagined by some advocates of SWB, and a primary reliance on quantitative methods in evaluation. Most critically, perhaps, the paper has sought to consider the subjectivity of subjective data and to reinstate the people concerned as authorities on their own lives. This draws attention to the value of complementing quantitative with qualitative data in order to add depth and texture and to ensure the accuracy of analysis and interpretation of results.