Allocating Family Responsibilities for Dependent Older People in Mexico and Peru

Abstract This paper applies different analytical frameworks to explore processes of family bargaining about providing care for dependent older people in Mexico and Peru. These frameworks include cultural norms, life course effects and material exchange. The paper is based on 19 in-depth qualitative family case studies, which are linked to a wider set of quantitative survey data. Care arrangements and bargaining processes are revealed to be highly gendered, and largely conform to prevailing cultural norms. Rather than neutral and objective, the self-identified role as main carer is found to be subjective and potentially ambiguous. The few men who self-identify as main carers are more likely to play an indirect, organisational role than engage directly in daily care. As such, bargaining mainly relates to which woman performs the main care role, and large family networks mean that there is usually more than one candidate carer. Bargaining can occur inter-generationally and conjugally, but bargaining between siblings is of particular importance. Bargaining is framed by the uncertain trajectory of older people’s care needs, and arrangements are sometimes reconfigured in response to changing care needs or family circumstances. Taking the narratives at face value, the influence of life course effects on bargaining and care arrangements is more obvious than material exchange. There are, however, indications that economic considerations, particularly inheritance, still play an important behind the scenes role.


Introduction
In countries like Mexico and Peru, accelerated population ageing is leading to a growing prevalence of old age care dependency (Jackson, Strauss, & Howe, 2009;Rossel, 2016). Across Latin America as a whole, as well as most other low and middle income countries (LMICs), formal care services for dependent older people remain limited and so the responsibility for care-giving falls almost exclusively on family carers (Bernabe-Ortiz, Diez-Canseco, Vásquez, & Miranda, 2016;Camarano & Leitão e Mello, 2010; Economic Commission for Latin America and the Caribbean [ECLAC], 2012;Flores-Castillo, 2012). Despite this, policy-maker interest in elder care remains limited and largely focussed on formal services (Lloyd-Sherlock, 2014).
The main body of academic research on family care for older people in Latin America and other developing regions is health-related and quantitative. Typically, studies use standardised scales to assess the burden that care-giving imposes on carers' health and quality of life, and key mediating factors (Choo et al., 2003;Sousa et al., 2016). They also assess the extent to which care-giver characteristics and burden are associated with the health and quality of life of the person receiving care (Rodríguez-Violante, Camacho-Ordoñez, Cervantes-Arriaga, González-Latapí, & Velázquez-Osuna, 2015;Tamanini et al., 2011). These studies usually identify an individual who is understood to be the 'main carer' and analysis is based on dyads of carers and cared-for. These studies almost universally demonstrate a high burden for carers (particularly for more complex needs like dementia or Parkinson's Disease), most of whom are female (Machnicki, Allegri, Dillon, Serrano, & Taragano, 2009). They show that burden is usually associated with lower socio-economic status and with poor outcomes for the dependent older person (Chen et al., 2016). Some studies identify associations between positive family dynamics and family cohesion (defined in general terms) and reduced carer burden (Elnasseh et al., 2016;Perrin et al., 2013). These findings are closely in line with those of similar studies in high income countries.
Whilst useful, there are a number of shortcomings with the existing literature. Most such studies do little to consider issues such as cultural effects and gender relations, and are often based on simple models of family dynamics that are increasingly out-of-kilter with more complex social realities (Sunkel, 2006). They say little about how patterns of care responsibilities come about, and how they are bargained and negotiated between different family members. Research on other aspects of care-giving demonstrates the value of more contextualised and nuanced analysis (Knight, Hosegood, & Timaeus, 2016). Social science research has revealed important insights about gendered aspects of care-giving in LMICs, particularly with reference to childcare or care for adults with HIV/AIDS (Razavi, 2011;UN Research Institute for Social Development [UNRISD], 2016). Studies that apply a similar approach to the care of older people in LMICs are comparatively scarce.
This paper seeks to understand patterns and experiences of family care for older people in different parts of Latin America and how roles and responsibilities are allocated and negotiated. Drawing on in-depth qualitative case studies embedded in a wider, quantitative data set, it applies a range of analytical frameworks and calls into question some assumptions underpinning more positivist research. The next section sets out these analytical approaches. This is followed by an overview of the study design and methods applied, as well as some contextual information about the study settings, and three further sections which explore different aspects of care arrangements. The first of these maps how care responsibility is allocated across family networks, including how decisions are made and justified. This is followed by an analysis of how care arrangements can change over time and, thirdly, a section on the role of men as carers. The final part of the paper considers the usefulness of the different analytical frameworks in understanding these issues.

Developing a conceptual framework for understanding family care of older people
The paper combines different existing analytical approaches to examine patterns of household and family care arrangements for older people. First, it considers the potential effects of cultural norms on care-giving. In most LMICs, including those in Latin America, there is a strong norm that a female relative, usually a daughter or daughter-in-law will perform this role (Esquivel, 2011;Robles, 2000;Varley & Blasco, 2000a). It is also claimed that cultural norms of intergenerational solidarity and family attachment are stronger than those for northern European and North American cultures (Phinney, Ong, & Madden, 2000;Yahirun, Perreira, & Fuligni, 2013). This includes claims that Latin American families are more willing to provide support to care-dependent older people than in rich, industrialised societies.
There are, however, several limits to the explanatory value of cultural norms. First, studies from almost all countries find that women are usually the main carers of older people, regardless of purported cultural norms (Cox & Soldo, 2013;OECD, 2011). In countries like Mexico and Peru, protracted periods of old age care dependency only became a common phenomenon in recent years. Consequently, longstanding, historical norms and traditions about gender roles and family responsibilities may not lend themselves easily to the modern challenges of allocating elder care. Current cohorts of older people in Mexico and Peru had high rates of lifetime fertility, giving rise to large family networks, including daughters, granddaughters and daughters-in-law. 1 As such, a critical Family responsibilities in Mexico and Peru 683 question becomes 'which woman?' and how care responsibilities are negotiated and allocated across relatives and different households. It is likely that there will be considerable space to negotiate these care roles within what remain loosely established cultural norms.
Consequently, the second analytical framework applied by this paper is bargaining. Much of the existing theoretical literature on bargaining focusses on single households and considers bargaining between two individuals, usually husband and wife (Addati & Cassirer, 2008;Agarwal, 1997). Bargaining about elder care is potentially much more complex, often involving larger numbers of individuals located in several households and can occur along different, interacting axes (inter-generationally, horizontally between siblings and conjugally). To varying extents, it may also include the dependent older person. The academic literature on inter-generational bargaining is limited and the literature on sibling bargaining is almost non-existent, especially for LMICs. There is evidence from high income countries that old age care dependency can provoke intense tension and conflict between siblings, who often feel that caregiving responsibilities have not been allocated fairly. As well as primary carers experiencing anger and resentment towards siblings who do not pull their weight, this conflict can take a number of other forms, including siblings who feel unfairly excluded from caregiving (Ingersoll-Dayton, Neal, Ha, & Hammer, 2003). In Mexico and Peru, where family networks tend to be more extensive, the number of players in sibling bargaining may be larger and there may be more scope for interaction with other axes of bargaining, all leading to greater complexity (Robles, 2007).
The paper applies two explanatory models to examine these bargaining processes. Life course frameworks, as applied in sociology, consider the potential effects of past lived experiences of the dependent person and their family members in framing current arrangements (Heinz & Krüger, 2001;Lloyd-Sherlock & Locke, 2008). This includes the strength of affective bonds between the dependent person and different family members, as well as material aspects of the life course (for example, whether some relatives move away, while others spend their entire lives living with the dependent person). As part of this, the concept of 'deferred reciprocity' considers the extent to which willingness to offer care is influenced by perceived contributions, material or affective, from the dependent person in the past (Gomes da Conceição & Montes de Oca Zavala, 2004). Sibling relationships and bargaining may be framed by established sibling roles and identities, possibly established in childhood and reaffirmed by key moments later in life (Bedford, Volling, & Avioli, 2000;White, 2001). Life course frameworks distinguish between 'trajectories' (periods of general continuity) and 'turning points', which interrupt trajectories, bringing about significant change to one or more lives. The initial onset of care dependency may represent a turning point, not just in the life of the older person, but for other family members, demanding decision-making and the allocation of new, often unforeseen, responsibilities (Robles, 2008). In other cases, care need may increase gradually, without an obvious moment of crisis or family decision. Also, care needs and family circumstances do not remain constant over time, creating needs or opportunities to renegotiate responsibilities at critical conjunctures.
The paper also explores bargaining from a rational, materialist perspective considering the extent to which objective economic explanations affect care arrangements. Simple economic models of family behaviour are usually premised on the notion that families are altruistic and will seek to maximise the efficiency of resource allocation to the benefit of all members (Becker, 1974). These assumptions about altruism and efficiency have been widely criticised, particularly from the perspective of gender dynamics (Folbre, 1986;Quisumbing & Maluccio, 2003). Also, these models are usually applied to families living in unitary households where resources are pooled, rather than to behaviour across wider kinship networks encompassing multiple households. Nonetheless, the allocation of care responsibilities across families may to some extent be understood in such terms. For example, decisions may reflect variations in the direct and the opportunity costs of care-giving between individual family members (or households). This may include a range of factors, including geographical proximity to the care-dependent person, as well as other responsibilities, such as caring for young children or paid employment. Rather than an objective, shared understanding of these issues across all family members, the critical determinant of elder care responsibility will be how they are perceived and valued by different family members (Sen, 1987).
A related set of economic explanations relate to expectations of bequests in exchange for providing care (Bernheim, Shleifer, & Summers, 1985). Quantitative analysis for United States populations found no evidence of this effect, although a recent Chinese study did (Perozek, 1998;Yin, 2010).
These effects have not been specifically studied in Latin American settings, where written wills are less common, property rights often established informally and where older people have on average larger numbers of potential bequestees (Ward, 2012). The limited evidence indicates that motivations for inheritance are complex, variable and strongly gendered. In some cases, traditional practices, usually specifying a male heir, trump considerations related to care-giving (Varley, 2010). Economic exchange is not limited to future bequests. Where the older person is a home-owner, permitting a carer (and their own family) to live there free of charge can represent an important contribution, especially in urban settings where accommodation is more expensive. Similarly, if the older person receives a pension and shares it with other relatives, this may wholly or partially offset the costs of care (Hoffman, 2014). However, the evidence that receiving a pension is associated with higher levels of family care is variable and inconclusive (De Vos, Solís, & Montes de Oca, 2004;Lloyd-Sherlock, 2006).
These different analytical frameworks should not be viewed as mutually exclusive, but as complementary prisms for exploring complex and multidimensional decision-making processes. Potentially, they may interact in a number of ways. For example, economic exchange models relating to bequests may be extended to include non-material forms of exchange such as emotional support, and may entail an element of deferred reciprocity. These frameworks may be applied by researchers to understand which family members provide care and may also be deployed by family members to justify and legitimise decisions.

Design and methods
The issue of family bargaining about elder care is particularly suited to detailed qualitative case studies. As will be seen, quantitative data can provide some insights about patterns of care arrangements at the household level, but are less useful for revealing processes of bargaining, particularly across multiple households (Cox & Soldo, 2013). Consequently, this paper mainly draws on 19 case studies of Mexican and Peruvian families containing at least one care-dependent older person (Table 1). These case studies, conducted between 2012 and 2014, are embedded in a wider quantitative survey of the economic and social effects of old age care dependency, covering a total of 838 households, which took place immediately beforehand (the INDEP Survey). 2 That survey is, in turn, embedded in an earlier epidemiological survey of care dependency which was conducted in 2002/3 and includes 3934 households (the 10/66 population based studies). 3 The same core research team was involved in all these surveys, with additional local qualitative researchers recruited to support the 19 case studies. The pooled data should not be taken as representative of their respective national settings, and as such the paper does not make direct country comparisons. Instead, they should be treated as a diverse set of rural and urban case studies, which permit analysis of family care for older people in a range of different geographical settings.
The qualitative case studies were selected from the INDEP households according to a number of criteria, such as inclusion of a care dependent older person, 4 household economic status and composition. Table 1 shows they included six dependent older men and 14 dependent older women, with an average age of 86.7 years. For each case study, interviews were conducted with several key informants including dependent older people (where feasible), the main carer, and any other household or nonhousehold members identified as playing a significant role in caring for the dependent older person or in making decisions about care, including people who were paid to provide care in the home. 5 This also included non-relatives paid to provide home care to the older person. In some cases informants were interviewed on multiple occasions, either due to their convenience or limited stamina. Particularly in cases where the older person had dementia, the narratives of the older person are missing. Of eight cases with dementia, seven were unable to participate in the survey. In addition to the older person, other voices may be missing. Problematically, it is often the person identified as the primary caregiver who decided whose narratives are included in our dataset.
Interviews were conducted in a narrative style, allowing interviewees to 'tell a story' about the older person's care needs, the impact of this upon the household and how the household has coped with Family responsibilities in Mexico and Peru 685 these changes (Corbally & O'Neill, 2014). Consequently, each case study consists of multiple narrative interviews providing different viewpoints rather than a single 'objective' account of care arrangements and decision-making. It is to be expected that personal accounts may be influenced by self-justification and family politics. Contradictions, ambiguities and gaps in these collective narratives are in themselves potentially revealing of family dynamics. For example, in some cases informants were only prepared to discuss inheritance once the interviews were no longer being taped.
The analysis of these narratives focussed on elements relating to elder care arrangements and sought to apply the analytical frameworks set out in the previous section. Following the approach of Lloyd-Sherlock and Locke (2008), we did not seek to generate a synthetic summary of narratives, but instead identified patterns of similarity and difference across the cases. In presenting our analysis, we foreground contrasting sets of narratives to give detailed accounts of family experiences for each theme. These are followed by a more general discussion of how these experiences relate to our theoretical frameworks and the extent to which they are echoed in the testimonies of other cases. In this way, we seek to strike a balance between representing the richness of each case study and presenting as wide a range of findings as possible. Highlighted case studies were selected because they provide revealing insights into particular themes and are broadly in keeping with the general findings. That said, they should not be read off as directly representative of the wider qualitative data set.
The qualitative and quantitative surveys were located in one rural and one urban catchment in each country. These were six urban districts in Tlalpan, Mexico City; nine villages in the Mexican state of Morelos; two districts of Lima (San Miguel and Lima Cercado) and six villages in the Peruvian state of Canete. These catchments were selected for the first wave of 10/66 surveys and included urban neighbourhoods with mainly poor or low-middle class residents and rural locations with low population densities and primarily agrarian livelihoods. The four catchments were selected opportunistically, as part of a pre-existing epidemiological study. Rather than representative of wider national, urban or rural experiences, they should simply be understood as four diverse settings. Data on household income and poverty show that the Peruvian sites were considerably more affluent than their Mexican counterparts: a situation which is not representative of national trends ( Table 2). Pension coverage varies across the sites, with particularly high levels of coverage in rural Mexico, where pensions accounted for over half of household income. This reflects the extension of social pension schemes across rural Mexico in the period preceding our study (Willmore, 2014). One general feature of these four sites, which is common in Latin America but less usual in high income settings, is the complexity of family living arrangements (De Vos, 2014). In many cases, these consist of plots of land or housing units which have, over the years, been subdivided between different family members. Typically, the oldest resident remains the nominal owner of the complex, with separate households for different children and their respective families. This can blur distinctions between household units (particularly when different siblings are involved in decisions about elder care), potentially allowing for more flexibility and cooperation in caring arrangements. However, these complex living arrangements can also lead to uncertainty over property ownership and consequently provoke conflict over inheritance. Table 3 presents 10/66 survey data on the prevalence of care dependency for people aged 65 and over in the four catchments. As expected, care dependency increases with age, particularly over the age of 75 and is more prevalent among older women than older men (Bernabe-Ortiz et al., 2016;Sousa et al., 2010). Dementia was the main cause of care dependency in all the catchments. This condition is associated with potentially prolonged periods of severe care dependency and high levels of care-giver stress (Prince et al., 2012).

'
The busiest woman in the world': allocating family responsibility for care-giving   although the numbers are too small for significant associations. For example, daughters-in-law were more likely to be identified as main carers in the rural Mexican catchment than elsewhere, which may reflect particularly strong local traditions of virilocality (Robichaux, 1997). There are also variations for carers who were unrelated to the older person: in all cases these people were paid to provide care. Paid care was most common in the urban Peru site, extending to a third of all households containing a dependent older person. It was less common in the Mexican sites, especially the rural ones, where only 1.2 per cent of households reported using a paid carer. 6 Reflecting our particular interest in care arrangements, we purposively included seven cases of paid care in our qualitative sample. In all of these, the paid carers were women and in two were family members (a granddaughter and a niece). Payment was made by a variety of family members, male and female. The identity of 'main carer' was established by the interviewees themselves, as the family member or close friend most involved in providing and/or organising care. 7 This leaves scope to define this role in different ways: as the person who oversees the care, as the person who pays for it or as the person who actually performs the care duties on a daily basis. It also potentially neglects important care contributions made by other people who are not given the status of main carer. For example, none of the paid carers in the qualitative sample were identified as main carers, even though they fulfilled most of the daily care duties. Consequently, these descriptive quantitative data may mask more complex and nuanced patterns of care-giving arrangements. Moreover, they do not reveal how different care arrangements are arrived at, including associated processes of bargaining and role allocation.
Given the richness and complexity of the qualitative case study, it is not possible to present each in detail. Instead, we foreground two contrasting case studies of family arrangements, before discussing the wider set of studies.

Manuel, Clara and Patricia
Manuel is an 85-year-old widower with Parkinson's disease who recently had a stroke. He lives in an apartment in Lima with his two daughters, Clara (56) and Patricia (54) and two grandsons. Clara and Patricia collaborate closely to provide care for their father. According to Clara: If one of us goes out, the other will stay at home. We do this mainly on Saturdays when I go to mass first and then Pilar at eight. During the day she's in charge of the cooking, while I do the cleaning and some sewing.
The sisters appear to have a close, harmonious relationship. At one point Clara comments 'nobody can believe that we aren't twins, because we do all the work together'. Patricia adds: There are moments when we have to help each other out. When one of us feels low, the other one tries to cheer her up. . .If Clara gets tired, then she lets me know.
In part, this is because they shared several important experiences and their lives appear to have been intertwined: both are single mothers with co-resident working sons, both have lived with their parents their entire lives, both are very religious and both gave up on work outside the home to care for their parents. These shared experiences appear to have affected their attitude towards Manuel. The sisters emphasise his acceptance of their single parent status, and his valued support in raising their sons. Clara comments: He's my son's father figure, my son doesn't have another father. OK, he does see his actual dad. . . but father is the one who taught him everything from the very start.
Clara and Patricia appear to accept their care responsibilities and emphasise Manuel's continued role as head of the household. Clara comments My siblings say that you're the one in charge, the leader, but I say he is the body and we are just the arms.
As such, Manuel's care arrangements were strongly influenced by his shared life course experiences with these two daughters. He has six other children, including several daughters. All of these established their own households at some point in their lives, although two of them, Mariana (61) and Julio (58) returned to live in a second, added-on storey following their divorces. This second floor is effectively a separate household.
The short terminal illness of Manuel's wife was an important turning point, confirming Clara and Patricia's roles as carers. Both gave up their paid work at this time. According to Clara, her siblings' limited role: was because they were out working all day and. . .since I was right here, I was the one who could give her all my time, right until the end. . .. These siblings offer Manuel's household occasional financial support and make weekly social visits, which Manuel says are the highlight of his week. For the two sisters, these are less enjoyable occasions, and Clara's comments suggest she is not entirely happy with the current arrangement: They all descend on me and I have to look after them. I get tired out. . .All my brothers and sisters, as well as the nieces and nephews. . .And I've been here the whole week long. . .Why not invite me to your own homes, so I can have a sit down while you look after me? But they still come here. . .And I thought Sunday was meant to be a day of rest, not for people to give you even more work! She adds: I'm the busiest woman in the world. . .I'm not physically worn out yet, but I do feel a bit broken inside.

Susana, Roberto and Paula
Susana, aged 90, lives with her husband Roberto, aged 86 and their daughter and main carer, Paula, 61 in a small apartment in Mexico City. Paula's earlier life experiences contrast sharply with Clara and Patricia's. During the 1990s she had a civil service career and relocated with her job outside Mexico City, only making occasional visits home. At this time, Susana and Roberto were in reasonable health Family responsibilities in Mexico and Peru 689 and spent much of their time with another daughter, Rocío, who was widowed and had three young children. A critical turning point was when Rocío fell ill with leukaemia. According to Paula: My parents could not take care of all Rocío's children. . .The youngest was only just one. . .so I left Jalapa and came to work here. . .And well, I told myself it's just temporary. . .I have other things to do than just help out.
Paula's 'temporary' responsibility of caring for the children became more permanent as her parents grew increasingly frail. She continued to work, in a lower status, less well-paid job, until 2012 when her employer lost patience with her increasing struggle to balance care and paid work.
Although Rocío, who lives nearby, helps out on a daily basis, Paula expressed unhappiness and a strong sense of injustice about her caring responsibilities. This was particularly directed towards her other siblings, several of whom lived on the same plot of land. Paula emphasised that none of these siblings offered help with care or financial support, although her wider testimony sometimes contradicted this claim. In order to meet her parents' medical expenses, she had been obliged to find temporary, casual work wherever she could. She felt that the multiple responsibilities of financing and supporting her parents were taking their toll on her.

Discussion
These two case studies reflect the complex interplay of cultural and life course influences on care arrangements, which is apparent in all the qualitative case studies. Cultural norms and values about gender roles appeared to provide the general parameters within which care roles and responsibilities are managed, negotiated and justified, and these norms appeared to be broadly consistent across the case studies. 8 They are evident in Clara and Patricia's efforts to maintain a patriarchal household. Whilst Paula had temporarily broken with these traditional norms, they rapidly reasserted themselves when her family's circumstances changed. Despite her earlier career, her situation had come to conform closely to a traditional female role, as youngest daughter who never married or had children, in order to be available for her parents' care needs. Paula's unhappiness with her current status reflected both her thwarted career opportunities and anger towards her 'selfish' siblings. She appeared to feel impotent to challenge the current arrangement and expressed a strong sense of victimhood.
Life course effects influenced care arrangements in different ways. In the case of Clara and Patricia, caring for Manuel represented a natural progression of their linked lives within the same household. In several other cases the main carers had lived continually with parents and saw their care roles as an inevitable consequence of this. These lives had entailed changing patterns of intergenerational interdependence and deferred reciprocity. For example, in several other case studies dependent older people had previously looked after grandchildren.
Life course effects did not just entail explicit or implicit notions of reciprocity and continuity, particularly when they related to sibling relationships, rather than intergenerational ones. Across the case studies, there were several examples of early life role establishment, whereby some siblings were judged (by themselves and other family members) to be inherently more capable and responsible than others. Whilst this process was strongly gendered, not all daughters conformed to the normative ideal. For example, a woman who cared for her 86-year-old father described her alcoholic sister, who did not help as follows: My sister was the weak one. She has always been sickly, ever since we were kids. . .I was the one who would help my mum; I was the oldest.
In a separate interview, her sister added that she 'wasn't born to care'. The highlighted case studies provide less explicit evidence of material influences on bargaining. Although Clara and Patricia sometimes justified their care roles in economic terms, these did not appear to have played a critical role. In the case of Paula, an unfortunate combination of circumstances her sister's widowhood and subsequent illnessout-weighed any personal economic considerations. Little reference was made to accommodation or pensions, although all but one dependent older person received a pension. 9 This may have been because the value of pension benefits (usually under US$100 a month) was likely to be less than the actual costs of supporting a care-dependent older person, particularly meeting their health care needs. Nonetheless, a small number of households, including Manuel's, were very reliant upon the pension income, and in one it was the only support for an older woman and a daughter who had given up work to care for her. The lack of reference to pensions, despite their evident contribution to household income, may have reflected informants' reluctance to refer to economic issues, reflecting a cultural norm that elder care should not be motivated by material considerations (Crist & Speaks, 2011).
Similarly, although concerns about inheritance do not feature overtly in most of the narratives, there were sometimes indications that they were present in the minds of the informants, some of whom were prepared to be more open about these when 'off the record'. References to inheritance were more frequent in cases where relationships between family members were conflictive. This includes some cases where care arrangements were changed over time, as discussed in the next section.

'I couldn't have imagined what a huge burden it would become': trajectories of caring
In the cases highlighted above, care arrangements established when an older person initially became dependent remained in place until the time of the survey. This was not always true for the other case studies. Older people's periods of care dependency can last many years and are typically characterised by increasing needs over time, punctuated by unpredictable crises, such as falls (Han et al., 2013;Robles, 2008). In 13 of the case studies, the older person was described as having substantial care needs for at least three years. Wider family circumstances also change over time, potentially affecting the capacity and willingness of different members to provide care. High levels of care-giver stress, particularly for conditions such as dementia, may lead to health problems or 'burn-out' for the main carer (Perkins et al., 2013). These risks are particularly high if the initial carer is an elderly spouse, and may then require other family members to take a larger role. Changes can also occur when one or more family member decides to challenge existing arrangements. Of the 19 case studies, six had seen substantial changes in care arrangements (such as the identity of the main carer) at some stage since the original arrangements had been established. In many other cases, there were less substantial changes, involving people more marginally involved in caring.
This section follows the same approach as the previous one, foregrounding two case studies of changes to care arrangements, before discussing the wider set of studies.

Carlos, Sara and Carolina
Carlos was a frail, 86-year-old widower with kidney disease, living in Mexico City. Following his wife's death in 2005, he continued to live in his own house with two adult sons and their families. In 2012, he moved out of this house to live with his only daughter, Sara (aged 60).
Sara claimed that her brothers had previously neglected Juan and taken advantage of him. Things came to a head after he had a bad fall: He had fallen and there were people around, but everyone was in their own little world, with headphones in their ears or outside, never checking on him to see if anything had happened . . .. So the first thing I said was, 'You know what? Go grab your things, your clothes, and you are coming with me and I will not accept no for an answer'.
Sara admitted that this decision may have been partly motivated by her guilt about not caring for her mother when she was terminally ill: 'I judge and blame myself. . .'. Sara emphasised that her own situation had been different then: her children were still young, money was tight and she needed to go Family responsibilities in Mexico and Peru 691 out to work. By 2012, her children had reached adulthood and the family was in a stronger economic position.
These aspects of Sara's experience conform neatly to the interplay of normative parameters and life course effects discussed in the previous section. However, the most striking aspect of this case study was the extent to which Sara was able to play a decisive and often authoritarian role without apparently transgressing cultural norms. According to Sara, the decision to move Juan was entirely her own and she was not prepared to consider objections from other family members, be they male or female. Sara claimed that her brothers were unhappy with her decision 'their little gold mine was leaving'. One of her brothers contradicted this version of events, however, describing the move as a joint family decision. Whether Juan himself was pleased with the new arrangement was questionable. By contrast, Sara was very confident that her own husband and two daughters would support her, or at least that she would overcome any opposition: I know that they are going to help me. I know I will not be opposed, and without asking for either permission or opinion, I brought my dad here. Sometimes we fight about it, we get stressed, and we just say to ourselves that's just a part of life.
Every member of Sara's household was required to adapt to the change and support in caring for Carlos. However, the heaviest burden clearly fell on Sara's oldest daughter, Carolina, who had given up her bedroom and provided most of the daily care, and a younger daughter who was required to share a bed with Carlos. Sara's husband appeared regretful about these sacrifices: She wants to go back to college to complete the course she interrupted. . .It's not like we said 'Now you are in charge of granddad.' She does it out of her own volition, but right now. . .her professional life is halted.
Ironically, this was the same daughter, Carolina, whose studies Sara had used to persuade Carlos to stay. Sara appeared oblivious to Carolina's economic opportunity costs, and made selective use of economic arguments (her brothers' alleged exploitation of Juan; her own improved circumstances) to justify her decision.
As with most other cases, care arrangements were not explained or justified in terms of inheritance, but there were indications that this played a significant backstage role. Sara and her brothers had agreed to sell Carlos' house and use part of the proceeds to cover his care expenses. However, they disagreed about how the money should be divided. Sara admitted trying to persuade her father to give her power of attorney over his estate: something Carlos was evidently reluctant to do. Sara commented: He didn't want to, and I said to him, 'It's whoever you want it to be. . .if you don't want me to have it, I will still take care of you in the same way'.

Pedro, Irene, Natalia and Veronica
Pedro was a frail 92-year-old. His wife, Irene (age 87) had advanced Alzheimer's. They shared a building and a plot of land with several children and grandchildren. Before 2011, their main carer was a daughter, Natalia, who lived on the same plot of land. When Natalia moved away for work, their main carer became Veronica, the wife of one of their grandsons. A year later, Natalia returned and resumed her role as main carer, with increasing levels of support from other family members.
Although Pedro and Irene both had some care needs when Veronica first took over, these needs became much more acute over the following year, as Pedro became blind and Irene's condition rapidly deteriorated. Veronica had not anticipated this, was very unhappy about her experience and how Natalia and her family had treated her. According to Veronica: Natalia told me she was going to Arequipa and she wanted me to cook for them as a favour, for a year. . .But I didn't know; I couldn't have imagined what a huge burden it was going to become. . . My husband was away. . .I couldn't leave my children alone with Leticia because she would attack them, due to her condition. . .So I called my mother in law to ask her to speak to her brothers and sisters. . .But these [Pedro and Irene's] children wouldn't say anything and wouldn't send their children to help.
Veronica claims that Natalia promised to send her about 100 soles (£20) each month, but had usually failed to do this. Upon her return, Natalia took over the main care role and appeared to blame her for Pedro and Irene's poor condition. According to Veronica: As soon as she arrived, she gave them something to eat and didn't even say 'Thank you, Veronica. I'll take over now.'. . .She doesn't speak to me at all now; I don't know why.
According to Natalia: When I got back, I saw that my father was blind, he had nothing, he could hardly walk. . .I don't know how this could happen in such a short time. . .I can't explain it.
Veronica no longer has any dealings with Pedro and Irene, and is planning to move away. She claims that Natalia has been more successful in mobilising wider family support than she had been, with two other female siblings offering assistance, either directly or through their own female children.

Discussion
Changes in care can occur in terms of sudden emergencies, such as Sara's father's fall, which can open a space for the immediate reappraisal of arrangements. They may also occur in response to more incremental, and often unforeseen, increases in dependency and the increasing struggle of existing carers to cope (as with Irene and Pedro). This often applied in cases of Alzheimer's disease, as the condition progressed, and care needs became increasingly demanding and round-the-clock. Increasing care demands and carer stress may lead to more conflictual relationships between family members, particularly when there were already unresolved issues about care arrangements. This conflict can include implicit or explicit allegations of neglect and abuse, and these featured in several cases. Conversely, increasing care needs can galvanise family members into taking a more collaborative approach. This is the case of Antonia in rural Mexico, where growing numbers of family members become involved to support the main carer, leading to a complex, collaborative set of arrangements. Care arrangements also may alter due to other changes in family situations, most obviously the death or illness of the main carer. This is more likely when the carer is an elderly spouse: there were five reported cases of older caregivers dying. Whilst it is unclear whether Sara's mother had been Carlos' main carer, her death increased his needs for assistance from other family members.
In few cases did economic considerations play an explicit role in changing care arrangements. One notable exception was a Peruvian woman who had initially come out of study to care for her mother and who later hired a carer and took on a full-time job. More generally, there was evidence of family members jockeying for position in terms of future inheritance arrangements. Sara's testimony is Family responsibilities in Mexico and Peru 693 suggestive of this. Natalia mentioned uncertainty about how her father's land and property would be divided after his death, and Veronica claimed that Natalia's decision to resume the role of main carer was motivated by inheritance, given that Pedro appeared close to the end of his life.
The case studies also reveal an additional bargaining axisbetween daughters and younger family members. The cases of Carolina and Veronica differ in terms of the allocation of roles and expectations, as well as their wider family dynamics, but both women appeared to be in a weak bargaining position with older female relatives. In two other cases, adult granddaughters provided substantial amounts of care, and one was paid by older relatives. In two further cases, young children made significant contributions to care. These included an 11-year-old girl who had been bathing her grandmother since the age of six and a four-year-old daughter of a paid carer who was expected to keep an older woman company for long parts of the day. For ethical reasons, young people were not included as direct informants in our study. Although apparently uncommon (or possibly hidden), the use of younger people to provide care may represent a potential form of vulnerability. This issue has received increased policy focus in high income countries, but is largely neglected in low and middle income countries (The Children's Society, 2013). 10 6. 'Taking the reins': men as carers Table 3 shows that around 19 per cent of people identified as main carers were men. These were more frequently the sons of care dependent older people, rather than their spouses. The qualitative survey purposively selected six case studies with male main carers. This enabled us to assess how far these cases represented genuine exceptions to established gendered norms of care-giving. Of these cases, the interviews found only one where a man played the largest role in day-to-day caring. In four cases, a female paid carer was employed. The following case was typical of these arrangements.

Marta, Jorge and Nuria
Marta, 97, lives with her son, Jorge, aged 52. Jorge is acknowledged by the family as Marta's main carer and claims he lost an opportunity to work abroad in order to 'take the reins' of his mother's care. For Jorge, this responsibility involved managing his mother's finances and paying for a live-in carer, Nuria. His sister, Ana, lives on a separate floor of the same building and sometimes helps Nuria with Marta's daily care needs. Nuria's young daughter also helps out. According to Jorge: 'Everyone has their particular duties, and Nuria knows what hers are.' These duties included general housework as well as caring for Marta and seemed to be viewed as a natural extension of domestic service in a middle class household. In reality, as with Sara, the real burden of daily caring was borne by a different person. As such, Jorge's care responsibilities were consistent with his patriarchal family position and 'taking the reins' is a very masculine metaphor.
In another case, life course effects appeared to have trumped established gender norms: the caredependent older woman had had a conflictual relationship with her only daughter and had lost contact with her, leaving her care in the hands of an unmarried son. However, since this son himself had mental health problems, their relationship was to some extent one of interdependency rather than straightforward care-giving by the son.
These instances of male main carers did not entail a reversal of gendered practices and norms (Robles & Carmen Pérez, 2012;Varley and Blasco, 2000b). Rather, they largely conformed to established gendered inequalities about care and bargaining. The only clear exception, where a male main carer appeared to play a genuine leading role in day-to-day caregiving was Javier, a 52-year-old man who had a close collaborative arrangement with his sister to care for their highly dependent mother, Julia. They had used a paid carer in the past, but had not been able to afford to keep her. Javier expressed considerable unhappiness about the burden of caring for his mother and admitted that this sometimes left him feeling very angry. However, he felt a strong emotional bond to his mother with whom he had enjoyed a close relationship through his life. It was unclear whether Javier's frustration was amplified by his apparent transgression of gender norms. He referred to a tension between caring for Julia and being there for his 'own family' and described his wife as 'understanding'. According to his sister, Javier's attitude towards caring was 'completely exceptional '. 11 As well as the five cases where men were identified as the main carer, in a further six, the interviews revealed that men were involved in a more limited way. Usually, this involved making a financial contribution or assisting with travel to health services. This could, to some extent, excuse them from taking a wider responsibility, both in their own eyes and sometimes in those of the main carers. It also meant they did not have to subvert gender norms about care. In some cases, sons would volunteer their own wives to provide care, and in one instance, a son expressed considerable bitterness that his wife instead spent time with her own mother.

Concluding discussion
This paper shows the value of detailed qualitative case study data in understanding processes of family bargaining related to elder care. The analytical insights they generate often differ from those of quantitative epidemiological studies. Notwithstanding their usefulness, these data pose major challenges for interpretation, given their potential for subjectivity, partiality and self-justification. This is most apparent when there are large contradictions between informants, as in the case of Natalia and Veronica. These contradictions are of interest in themselves, revealing conflicts and tensions in family care arrangements. However, even where there is apparent consensus across different respondents in the same case study, care must be taken in interpreting narratives, since this may result from a reluctance to be more open about family disagreements. Second, the data set should not be taken as representative of experiences in Peru and Mexico as a whole. For example, it is likely that family care arrangements will be different in more prosperous districts where paid care is more affordable and where there is an established tradition of using it for different household tasks.
With these challenges and limitations in mind, it is still possible to derive a number of general insights from the case studies. Care arrangements were strongly framed by cultural norms for both men and women. There was, however, scope for different arrangements so long as they did not subvert these rigid parameters, particularly when there was no feasible alternative. This was especially true when there was more than one potentially available female family member, be they a child, daughterin-law, grandchild or niece. Consequently, the extent to which women appeared to be disempowered by these gendered care norms was quite variable. In some cases, such as Sara and possibly Natalia, women could deploy these norms to be empowered and decisive actors. This challenges accounts of gender and care in developing countries which sometimes portray women as uniformly disempowered (Elspen, 2009;Rossel, 2016;UN Development Programme [UNDP], 2012).
Across the case studies, it was apparent that the status of main carer was not a simple or objective one. In some cases this was defined in terms of who was doing the most instrumental caring on a day to day basis, but in others it related to authority and overall responsibility. The latter was more common when men were identified as main carers. The status of main carer could be appropriated, as a strategy of self-justification, with a view to inheritance, to obtain moral authority or to hold power over people allocated a more junior role. In other cases, the main caring role was imposed on them. This raises issues about how information about care-giving is collected in quantitative surveys, needing to distinguish between general responsibility, actual day-to-day instrumental care and more specific roles (such as financial support). There is evidence that the commitment and burden of taking organisational responsibility for care can be substantial (Employers for Carers, 2011). Nevertheless, this role is fundamentally different from direct care-giving, which was viewed as an extension of domestic work, and therefore unsuitable or even degrading for men to perform.
Bargaining between family members around care-giving often involved many different relatives, either explicitly or more implicitly. As part of this, bargaining took place across different axes of kinship. Whilst there was considerable diversity, some general patterns emerge from the case studies. First, care dependent older people themselves, be they male or female, appeared to have relatively little Family responsibilities in Mexico and Peru 695 involvement in decision-making. Where there were non-care-dependent elderly spouses, they also appeared to play a limited role. Often, the marginal role of older people belied substantial economic leverage, in terms of pension coverage and asset ownership. This finding is at odds with other studies of pensioner's family relationships in low and middle income countries, which frequently portray them as relatively empowered (Case & Menendez, 2010;Schwartzer & Querino, 2002). It suggests that older people may lose power and status as they become care dependent, which is arguably a time when they are most in need of this influence. Their uncertain status is reflected in the language of family members which sometimes shifted between reflecting traditional norms of respect and veneration for elders to infantilising them. As older people lose their established status as family head, this may create new uncertainty and conflict about family power dynamics, which both influence and are affected by care bargaining.
A second axis of bargaining occurred between female and male older relatives, who were often designated the main carers in quantitative surveys and younger female relatives who sometimes provided substantial amounts of daily care support. Male grandchildren were usually exempt. For example, a daughter of a highly dependent older woman observed: My son is a young man. He is not the right person to take care of her, to help with her personal hygiene. . .If only I had someone else, a cousin, a woman like me. . .
As with the care dependent older people, these younger female relatives appeared to have little influence on decisions. These arrangements were often justified both in terms of gender norms and in terms of the seniority of the older adult. Additionally, younger adults had few economic resources to enhance their bargaining power. This is an area of bargaining that has been largely neglected in the wider literature on care-giving for older people.
Conjugal bargaining could encompass a number of different issues. Theoretically, wives might encourage their husbands to play a larger care role. According to Veronica: One time, when I was bathing my mother, my husband said: 'My mother is a woman, and this is what women should do'. So I answered: 'You are a man and your father is also a man, so you should clean and bathe him yourself.' However, this did not feature frequently in the interviews, and when wives did make claims on their husbands it was often ineffectual. Other elements of conjugal bargaining included wives being expected by husbands to care for his parents, which was particularly frequent in the rural Mexico site. By contrast, almost no sons-in-law identified themselves as main carers. Conversely, wives sometimes needed to negotiate permission from husbands to care for their own parents. In some cases, as with Sara, this included bringing the parent to live with them. Not all husbands were as compliant as Sara's, however, and one male respondent objected to his wife spending time with her own mother instead of being with him. Despite the number of issues that could be involved, references to conjugal bargaining did not usually feature prominently in the interview narratives. In the eyes of female informants, men were mainly viewed as marginal players and as naturally irresponsible.
Bargaining between siblings played a much more prominent role in the narratives and appeared to be critical in determining care arrangements. Sibling bargaining was strongly framed by life course effects. In some cases, elder care was seen as a natural continuation of an established caring role, often based on co-residence. In other cases, there was a more obvious decision point (either when it became evident the older parent needed care or when care arrangements changed). Sometimes, what initially appeared to be short-term responses to emergencies became more fixed arrangements. Comparative judgements and evaluations of siblings' moral fibre, capabilities and affective relationships with parents had a central place in bargaining and justifying decisions. In some cases, these judgments appeared to date back to childhood. According to María del Carmen, who cared for her mother with advanced dementia: One of my sisters lives just next door, on the same plot of land. . .I've given up trying to talk to my sisters. I can't force them to help out. . .They waltz off here and there, buying this and that, and don't even bring my mother a piece of fruit. . .But then, my mother always wanted to be with me. She seemed to prefer me. . .
Taking the narratives at face value, material exchange played a less obvious role than life course effects. It is possible that clear economic effects became lost in the complexity of intersecting axes of bargaining and negotiation between different family members. The narratives sometimes hint material considerations played an important, albeit tacit, role, especially in the case of inheritance. In some cases, male and female family members more marginally involved in care would provide help with some of the associated costs, suggesting a degree of trade-off between limited financial support and a more hands-on caring role.
Overall, this paper finds that the combined conceptual framework set out in the introduction is useful for understanding how elder care responsibilities are allocated and what this means for those involved. Rather than examine material exchange, life course and cultural norms individually, bargaining around care is framed by the interplay of these effects. Understanding this interplay is critical for analysing relationships between care, social justice and individual wellbeing. Providing long term care is a growing challenge for families in Latin America and the allocation of these responsibilities has important consequences for gender justice and intergenerational relations (World Health Oganisation [WHO], 2015). These consequences are particularly pronounced when dependency results from highly demanding conditions like dementia for which information and services are very limited (Alzheimer's Disease International [ADI], 2015). The qualitative narratives reveal that, despite extended family networks and some degree of role sharing among relatives, the bulk of the direct care responsibilities tend to fall upon one person. As such, large extended families do not necessarily reduce the care burden on individual members and can, potentially, increase the potential for conflict. Although some carers may derive personal satisfaction from performing this role, there is growing evidence of high burdens on carer health and wellbeing. The paper points to a need to raise societal awareness about the nature of family elder care and to increase support (such as respite care and day centres) for those who provide it.
3. The 10/66 surveys are an international programme on epidemiological research primarily focussed on the epidemiology of dementia in low and middle income countries. This includes surveys of care dependency (not just dementia) across populations of older people. Further information is available at: https://www.alz.co.uk/1066/ 4. Dependence was interviewer-rated based on a key informant's responses to a set of open-ended questions on the participant's needs for care (Sousa et al., 2010). 5. In one case, we were refused permission by family members to interview paid carers. 6. The lower use of paid carers in rural settings may have reflected affordability (mean income was lower for rural households) and fewer well-paid alternative female employment opportunities (although overall rates of female employment were similar across all the catchments). 7. The interviewer manual states: An 'organisational caregiver' is usually a close family member responsible for making arrangements for care, although they may not actually do much direct caring themselves. A 'hands on caregiver' is directly responsible for providing physical care and supervision. The 'hands on caregiver' may be the 'organisational caregiver', another member of the family, or a paid caregiver. 8. The only explicit exception to this general rule was the testimony of an older man in rural Peru. He expressed a strong preference that his sons care for him, as a man, instead of his daughters and was unhappy that they were not prepared to do this. 9. Across the study sites, pension coverage for people aged 65 and over ranged from 60 per cent in rural Peru to 90.5 per cent in rural Mexico. 10. One partial exception is research about young carers of people with HIV/AIDS in Africa (Skovdal, Ogutu, Aoro, & Campbell, 2009). This focusses on cases where the middle generation is missing and therefore has little to say about intergenerational bargaining. 11. The view that men should not be expected to care echoed in the testimonies of both older and younger Mexicans in a separate study (Robles & Carmen Pérez, 2012).