The dynamic welfare habitus and its impact on Brazilian migration to Lisbon and Barcelona

Little is known on how people ’ s way of thinking and doing around welfare provision – what we call the welfare habitus – plays a role in migration and how such cultural references change over the migration process. Through an empirical case study on Brazilian migration to Southern Europe, this article explores the dynamism of the welfare habitus focusing on three elements. First, the welfare-related resource environment in the countries of origin and destination. Second, the role of the welfare habitus in shaping migration aspirations. Third, the transformative learning process taking place during the life course and the migration experience. The article draws on a literature review, the analysis of secondary quantitative data and 24 in-depth semi-structured qualitative interviews collected in 2016 in Lisbon and Barcelona among men and women born in Brazil selected at different stages of the life course – young people without children, parents of school-aged children and people aged 55 or more. Drawing on practice theory the paper looks at the interaction between external structures, habitus and everyday actions around welfare provision in contexts of migration. Doing so, the paper contributes to a better understanding of the effect of welfare provisions in the country of origin on migration focusing on the temporal perspective.


Introduction
João was born in the Brazilian state of Goiás. When he was 45 years old, he worried about how he and his wife Maria would cope with the loss of income once they retired. At that time, they lacked the job continuity they needed to contribute regularly to their pension scheme in order to retire in good financial conditions at the age of 65. Being close to and taking care of family members was an important value for them but, when the possibility arose, this couple migrated to work in Spain and contribute with their earnings to their Brazilian pension scheme.
Zuleida, from the state of Minas Gerais, migrated to Portugal to care for her son who had migrated earlier. Her son was going through a depression and she went to Lisbon to assist him, complementing the "formal" support that he was receiving from health professionals. She thought it would be a short-term period abroad. Once in Portugal, she liked the country and the security she felt there. She also began to appreciate the healthcare that she had access to in the Portuguese public system. Therefore, she settled in Lisbon, where she worked to complement the retirement pension she received in Brazil.
These two cases highlight multiple ways in which welfare provisions in origin and destination might drive migration beyond the strict fact that social benefits provided by the state might be more generous in the destination countrythe so-called welfare magnet hypothesis (Borjas, 1999). Moving beyond this limited conceptualisation of welfare provisions, a growing body of literature on transnational social protection has focused on the diverse pools of resources that migrants and their family members draw on globally in transnational fields to protect and promote their well-being Faist et al., 2015;Lafleur & Lizin, 2016;Levitt et al., 2016;Mazzucato, 2009). The transnational social protection framework captures broader welfare provision needs and resources in contexts of migration, taking into account the wide range of formal and informal private actors that, aside from the state, provide welfare in different domains such as health, education, unemployment, and retirement Levitt et al., 2016). Within this research agenda, some studies have explored changes and differences in migrants' needs and priorities in welfare provisions depending on their life stage or their role in the household (Jolivet, 2019;Serra Mingot & Mazzucato, 2019), however more research is needed to better understand the constant change and adaptation in the ways of thinking and doing around welfare provision, what we call the welfare habitus, throughout the migration process. The cases of João and Zuleida illustrate that migratory projects are dynamic and change according to the context, individual and collective strategies, and with the information and knowledge that migrants accumulate over time and across space (Boyer, 2005: 49-53). As such, how does the welfare habitus based on the context in the country of origin shape conceptualizations of welfare provision and how these affect migration projects? Beyond the limited studies looking at the effect of welfare provisions by the state in migration and immobility (Kureková, 2013;Mahendra, 2014), the origin country effects of broader welfare provisions in contexts of migration remain overlooked. Changes in the origin country effects over the migration process are also unexplored -What is the effect of perceptions of the welfare regime of the origin country in preferences and decisions around welfare provision upon arrival and what changes with the contact with new realities?
To answer these questions, using a similar approach to the one proposed by Raghuram (2016) to understand the ethics of care, this article explores the welfare habitus and its transformations over time and with the migration process. To do so, it analyses how conceptualizations of welfare provision are shaped by geographical and historical factors and the changing character of expectations and meanings around welfare provision. It also explores how this changing welfare habitus shapes everyday actions around welfare provision. To really grasp the effect of welfare provisions in migration we consider "the entire set of institutional arrangements, policies and practices affecting welfare outcomes and stratification effects in diverse social and cultural contexts" (Gough, 2004: 26). Taking this broader understanding of welfare regimes, we explore the role that the different geographies and dimensions in welfare provision play in migration and examine changes over time. Our main focus lies on distinctions that migrants make between welfare resources provided by the state and those provided by formal private institutions such as private schools and universities, or private health insurance companies. We also consider migrants' ways of thinking and doing when formal institutional welfare resources are perceived as insufficient, absent, or inaccessible.
We bring into the analysis the dynamic interplay between past, present, and future in people's considerations around welfare provision at different stages of the migration process. To do so, we draw on Emirbayer and Mische (1998, p. 963), who proposed to conceptualise agency as a temporally embedded process of social engagement. We pay particular attention to the role of external structures in the country of origin in shaping the welfare habitus and the learning process that takes place through the life course as migrants move from the country of origin to countries of destination, where they become acquainted with and begin to navigate new welfare regimes.
The article draws on practice theory applied to the study of migration (O'Reilly, 2012) to look at the transformations of cultural references around welfare provision over the migration process. It focuses on an exploratory empirical case study on Brazilian migration to Southern Europe through 24 in-depth semi-structured qualitative interviews applying a life history approach. The interviews were collected in 2016 in Lisbon and Barcelona among men and women born in Brazil selected at different stages of the life course -young people without children, parents of school-aged children and people aged 55 or more.
The article enriches the growing literature on transnational social protection that focuses on the protection and provision of social welfare in contexts of migration but tends to neglect the study of welfare provision as a driver of migration. Our analysis also enriches the scarce literature on the role of welfare regimes in the countries of origin in migration from the micro-level perspective. Finally, we contribute to the study of the life course effects in welfare and migration looking at the ways in which life events (including migration) and stages in the life course influence welfare practices in contexts of migration.
Seven sections follow this introduction. The next section explains the theoretical framework used to explore the dynamism of the welfare habitus over the migration experience. The second section describes the data and methods used for our analysis. The third section presents the context of Brazilian migration to Barcelona and Lisbon. The fourth section explores how opportunities and constraints in a particular welfare regime shape the welfare habitus. The fifth section delves into different ways in which the welfare habitus acquired in the country of origin is taken into account in initial migration projects. The last section before the conclusion explores migrants' decisions around welfare provision after arrival. The section highlights that, similarly to care practices (Raghuram, 2016), welfare practices around welfare provision are site-specific, complex and dynamic.

Ways forward to understand the effect of welfare in migration
To overcome the identified gaps, our study approaches migration as a process shaped by the constantly changing interplay between structural factors and people's agency and choice (Morawska, 2009;O'Reilly, 2012;Stones, 2012) and explores how migration is a life event that, through its inherent learning process, affects and transforms the welfare habitus over time. Drawing on O'Reilly (2012), we look at the interplay between external structures and habitus in contexts of migration. The external structures under study are the public and private welfare resources that people have access to in a particular welfare regime. To stay aligned to previous studies on transnational social protection, we call them the welfare resource environment. We limit our analysis to the resource environment in the domains of health, education, unemployment, and retirement and take into account the unequal access to private and public welfare resources for more or less affluent people. We also draw on O'Reilly (2012) to conceptualise the welfare habitus. It refers to the ways of doing and thinking in terms of welfare provision that are acquired through past collective and individual experiences and actions. The welfare habitus includes cultural schemas, knowledge, and worldviews (O'Reilly, 2012: 150-51). The welfare habitus shapes conceptualisations and expectations of the welfare regime. It has also an effect on everyday actions around welfare provision. To Bourdieu's idea that changes in people's habitus take place at a rather slow pace, we prefer Friedmann's argument that migration and the exposition to different experiences and ideas can transform the habitus (Friedmann, 2005, pp. 302-303), sometimes drastically. As such, we consider that the welfare habitus is constantly changing and adapting to the evolving opportunities and constraints (O'Reilly, 2012) encountered throughout the migration process. Before migration, a differentiated individual welfare habitus emanates from people's lived experiences of the welfare resource environment in the origin country that may be dependent on people's socio-economic background. An example of welfare habitus in Brazil is the widespread practice of caring for the elderly within the family rather than using nursing homes. This preference might be due to expectations and obligations of informal social protection (Faist, 2000;Faist et al., 2015;Sienkiewicz et al., 2015). Another explanation could be the high costs associated with external elderly care services in Brazil that render them inaccessible to large segments of the population.
We also use the concept of action to refer to concrete actions and decisions around welfare provision. Particular actions can represent a turning point in the transformations of the welfare habitus. We borrow from Emirbayer and Mische (1998) the idea that action is influenced by the dynamic character of agency, informed by past experiences (welfare habitus), influenced by (perceptions of) the present situation in constant change, and oriented towards the future (Fig. 1).
The situation in the present includes, on the one hand, the (perceived) welfare resource environment in the countries of origin and destination. In our empirical example, this refers to the formal (state as well as private) component of welfare provisions in Brazil, Portugal and Spain, including the access to public welfare provisions through social security agreements between Brazil and the two European countries. It also encompasses, the informal resources available to migrants within social networks, including migrant communities in Barcelona and Lisbon. The history of migration that links Brazil and the two Southern European cities is also likely to play a role -the more or less established social networks in Barcelona and Lisbon shape opportunities and constraints in the access to informal welfare resources. On the other hand, the present context is also shaped by life course events such as the birth of a child or reaching the age of retirement. Finally, still drawing on Emirbayer and Mische (1998), welfare-related actions are oriented towards perceptions and identifications of welfare resources relevant for the future in relation to migrants' evolving needs and aspirations. Such orientations shape possible trajectories of action.
Paying more attention to the weight of each of these three temporal dimensions at different stages of the migration process is useful to capture to what extent frames of reference based on the access to welfare resource environment in different welfare regimes shift over time. With the identification of different welfare regimes over the migration process, migrants might question their welfare habitus, and incorporate, reproduce and transform local practices of the country of destinationsee Nowicka (2018, p. 825) on the transformation of racial discourses with migration. As such, through this transformative process, often mediated by social networks, migrants reconfigure their welfare habitus. The next section describes the data we use in the analysis.

Data and methods
The study embraces the social constructivism principle according to which reality is subjective, socially constructed through the continuous social interaction with others. Such interactions occur in the routine of the present everyday life but considering projections towards the past and the future (Berger & Luckmann, 1991, pp. 43-47).
We draw on data collected for the project MobileWelfare, funded by the Norface Welfare State Futures programme. We focus on 24 semistructured interviews conducted with Brazilian men and women living in Lisbon and Barcelona metropolitan areas between April and September 2016. The recruitment of participants occurred through purposive snowballing and the resulting sample (Tables 1 and 2) included diversified profiles in terms of sex, education level, age and stage in the life cycle. We considered only migrants born in Brazil who had spent a year or more in at least one European country. The interviews were conducted in Portuguese and Spanish and took between one and two hours.
Taking a life history approach the interviews explored to what extent and how do perceptions and use of welfare resources shape migration decisions over time. The interview questions were structured to trace back the life and migration history of the participants since their childhood. The first part of the interview included questions on the individual and external contexts before migration. The second part asked about the situation of the participants after migration and what had changed over time in terms of their administrative migration status, employment situation, family life, contacts with the country of origin and changes in the educational, occupational and financial trajectory. Questions on social protection arrangements in the country of residence followed. The interview also asked to participants to compare between the welfare resource environment in the country of origin and in the country of destination. The final part of the interview asked about participants' future aspirations, challenges and uncertainties. The data were translated into English and analysed with Nvivo 11.
For our analysis we combined the thematic and narrative approaches. Using a codebook, we started with the thematic analysis and identified place and time specific patterns in the needs, access and use of welfare provisions. The codebook included codes for participants' welfare resources in different domains (such as health care, education, or retirement), the countries that the participants referred to, particular life events such as retirement, the birth of a child or unemployment and participant's migration experience. Applying multiple coding, we identified changes over time in the perceptions and use of the welfare resource environment. With the narrative analysis approach we identified commonalities among participants in how they conceptualised the welfare regimes in Brazil and how this conceptualisation changed over time, influenced their actions around welfare provision and subsequently shaped their welfare habitus over the migration process.
To complement the qualitative analysis of the primary data we conducted a literature review and analysed secondary quantitative data on the welfare regimes in Brazil, Portugal and Spain and on Brazilian migration to Barcelona and Lisbon.
It is beyond the scope of this study to explore what drives Brazilian migration to Lisbon and Barcelona, transnational welfare arrangements or the access to welfare resources for migrants with different administrative statuses. Our choice of Lisbon and Barcelona is based on two criteria. Firstly, the sample selection is based on the similar relevance of non-state-provided welfare resources -especially the family-in these two Mediterranean welfare regimes. Secondly, and more importantly, the sampling is based on the difference in migration histories. Comparing Lisbon and Barcelona, the longer migration tradition of Brazilian migration to Lisbon is likely to affect migrant's welfare resource environment. The sample is not representative of the Brazilian population in the two European cities. Instead, the purposive sample aimed to collect data on a diverse range of needs and experiences of welfare provision taking into account the migration experience (between 2 and 33 years) and the time in the current country of residence (between a few months and 33 years). Preliminary conclusions have emerged from this exploratory study with a limited sample size. These insights could be further explored and confirmed with more ambitious studies. The interviews allowed us to understand the intertwined links between historical and geographical factors shaping conceptions of welfare regimes in contexts of migration and highlight the role that the country of origin plays in migrants' frames of reference over the migration process. Our results lead us to question the narrow conceptualisation of welfare provision in the existing literature on welfare as a driver of migration that mainly looks at welfare benefits and social policies in the countries of origin and destination. Before presenting the results, the next section introduces the contexts under study.

Context
Brazil is a country with strong socio-economic inequalities that are visible in its welfare regimes. In theory, Brazil belongs to the cluster of countries outside the OECD with the closest welfare outcomes in terms of life expectancy and literacy to those in more developed welfare states (Sharkh & Gough, 2010: 35). In Brazil, there is public universal social protection funded by workers' contributions and social assistance independent from contributions 1 . The more affluent usually contract complementary private protection schemes. The Brazilian constitution of 1988 clearly states in article 6 that the rights pertaining to social welfare are fundamental social rights that include: the right to education, health, work, housing, leisure, security, social security, protection in maternity and childhood, and assistance to those without protection (Nolasco, 2012). According to Nolasco the state approach is to promote equality by adopting measures that improve living conditions for those with fewer resources. In practice, however, disparities of income are much higher than the average level in the OECD countries (OECD, 2015) and welfare is not experienced in the same way by differently positioned individuals in Brazil's socio-economic structures. Narratives of Brazilian migrants collected in this research and others such as the THEMIS project 2 , clearly reflect how in Brazil, people's welfare habitus is strongly influenced by these unequal structures.
The context in the two South European cities and countries in terms of welfare and migration policy towards Brazilian immigrants is differentiated. In the district of Lisbon, in 2015, resided legally (but without Portuguese nationality) 36,438 Brazilian nationals, of whom the majority were women (22,318) (data from the Aliens and Borders Service 3 ). They constitute the most numerous foreign group in the country and in Source: Own elaboration. There is also an agreement regarding mutual access to the national healthcare system called PB4. This document grants access to healthcare in Portugal to Brazilian temporary migrants and tourists, as well as to Portuguese citizens in the same circumstances in Brazil.
In the municipality of Barcelona, by January 2016, 5659 Brazilian nationals were recorded in the city register (Padron Municipal) independently of their residence status in Spain; 55.8 per cent of them were womendata on foreign population according to their citizenship. Brazilian nationals are a relatively small group in Spain; in Barcelona, they represented only two per cent of the foreign nationals. Despite these relatively small numbers, Spain has become in recent years a main destination in Europe for Brazilian migrants (OECD, 2017). In 1996, only 5387 Brazilian nationals were registered in Spain; by 2008, the figure had risen to 116,548. After the outbreak of the global economic crisis, the records of Brazilians in Spain progressively decreased at national level (73,863 in 2015). In Barcelona, after dropping to the lowest level in 2015 (5451), figures have increased to reach 7107 migrants in January 2019 (INE, 2020). Spain and Brazil signed in 1991 an agreement on Social Security that came into effect in December 1995 and was amended in October 2012. This agreement stipulates (with some exceptions) that people who worked in any of the two countries and their family members are subject to the social security system of the country where they work. The periods of contribution in both countries can be summed up for contributory benefits. These benefits can be allocated independently of the place of residence in the two countries. Since 2010, these transferability rights apply to social benefits related to invalidity, retirement, death and survivor benefits, and to occupational accident and illness allowances (BOE, 1996(BOE, , 2012. However, by the time of our study foreign citizens without a valid residence permit had limited or no access to welfare state provisions. Furthermore, many Brazilian-born work informally in the domestic sector providing care or cleaning services, which excludes them from the formal social protection mechanisms.

How the welfare resource environment shapes the welfare habitus in Brazil
This section dives into the welfare resource environment in Brazil. It explores how inequalities in welfare provision in Brazil shape the welfare habitus. It also shows how the welfare habitus is embedded in particular socio-economic structures. In Brazil, people's socio-economic background has an effect on access to and experiences around welfare provision. This is particularly observed in the fields of education and health.
In education, there is a clear distinction in primary and secondary school between public and private education. State funded and free public schools tend to be perceived of poor quality and only for those who cannot afford a private education. Private schools are, at this level, considered of best quality and a privilege of the most affluent classes. Statistical data confirms this division. Children from lower socioeconomic backgrounds are more likely to be enrolled in public schools. According to the 2011 Brazilian National Household Survey (PNAD), the median monthly household per capita income in families with children enrolled in public schools was 36.6 per cent of that of families with children in private schools. The percentage of children aged 7-14 enrolled in public schools declines as the household per capita income increases (Bursztyn, 2016: 5). On the contrary, at the level of higher education, free public universities are considered the most prestigious. However, access to public universities is extremely selective and students coming from private secondary schools are better prepared to fulfil the selection criteria to access to them. During the workers' party (PT) government (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016), a strategy to fight unequal access to higher education based on income was introduced and a quota system was implemented in 2012: 50 per cent of places in public universities have to be allocated to students coming from public schools. This percentage is split between students of low income and of higher income households (below/equal to and above one minimum wage and a half respectively) and takes into account the 'racial' composition in the state. These divisions and changes in the quality of education are reflected in people's evolving welfare habitus such as in the case of Mariana from Bahía, in Lisbon: "We have always studied in public schools. Only my youngest sister, who was born when I was 15, studied in private schools. My parents preferred to have her in a private school because the public education was not as good as before. (…). Then I joined a private university because it's more difficult to join a public than a private university, so all of us studied in private institutions". Similar divisions between public and private actors of welfare provision are observed in healthcare. The interviewees explicitly refer to concerns associated with the quality of public healthcare in Brazil and the importance of having access to a private insurance. This is the case for Roberto from Belo Horizonte in Barcelona: "The public health service there is really bad. Even for emergencies you may end up waiting for 3 days before being attended. Or when you need an appointment with a specialist doctor you may get one in 6 months' time. So, you can't rely on it. If you can afford it, you go for the private service". The public healthcare system established in Brazil by the 1988 Constitution is universal, and it is assumed that it is a state responsibility to guarantee universal and equal access to healthcare. However, Soares Santos et al. (2008Santos et al. ( : 1432 argue that in 1988, the private healthcare sector was already robust, with hospital infrastructures being predominantly private. Private insurances were and are still widespread too. Soares Santos et al. (2008Santos et al. ( : 1435 noted that private hospitals offer more complex medical equipment and services in comparison to the public system. Thus, in practice, socio-economic inequalities and the poor quality of the public healthcare system lead to a differentiated welfare habitus, with well-off individuals seeking private healthcare (Neri & Soares, 2002: 78). Data from Suplemento Saúde da Pesquisa Nacional por Amostra de Domicílios (IBGE, 1998) analysed by Neri and Soares (2002: 82) reveals that, among the poorest 10 per cent, only 2.8 per cent have some kind of private health insurance. The proportion goes up to 74 per cent among the wealthiest 10 per cent. Differences are also observed in the use of public and private healthcare services. The more affluent tend to use them more often, especially for routine and preventive check-ups (33.6 per cent of the 10 per cent poorest and 37.1 per cent of the 10 per cent wealthiest). At the same time, lower income levels are more represented among those using health services because of illness.
Negative perceptions of the state-provided welfare provisions also affect the welfare habitus in the face of loss of income related with retirement or unemployment. Many interviewees in Lisbon and Barcelona illustrate the concerns of lower income families in Brazil over the low public retirement pensions. For example, Carla from Mato Grosso do Sul in Lisbon explains that when her father retired, she realised that she could not rely on the Brazilian public retirement system. Despite working many years and contributing to the public pension scheme, her father couldn't live only from his retirement pension. She concluded that she had to turn to private arrangements to secure herself an income when the time would come. Carla summarizes the tendency in Brazil to cope with its crises and malfunctioning through informal arrangements: "I remember that Brazil's economic situation was really difficult, and it was hard to find a job, so there was a constant feeling of insecurity. There used to be much informal work and Brazilians were proud of being able to make a living even in the hardest conditions. Brazilians were stimulated by the government to manage the situation doing any kind of informal work because the crisis was widespread".
Consequently, although the existing literature has focused on the role of state-provided welfare provisions in the countries of origin of migration (de Jong & de Valk, 2019;Kureková, 2013;Mahendra, 2014), private and informal welfare provisions in the origin countries are also likely to play an important role.
In the rest of the article, we explore the origin country effect in welfare provisions in contexts of migration. We pay particular attention to the interplay of the welfare habitus with people's evolving situation and their orientations towards the future. Doing so, we explore the varying degree of importance of past, present and future aspects of welfare provision at different stages of the migration process. The next section analyses the role of the welfare habitus in initial migration projects.

The origin country effect of welfare in migration projects
Migration is driven by a combination of drivers (Van Hear et al., 2012, 2018 and, taken alone, welfare-related factors could not explain the interviewees' migration decisions. However, welfare-related migration aspirations are mentioned by a few interviewees, especially those experiencing specific life transitions and events linked to illness, education, or retirement. The aspects of welfare provision that the participants consider are overlooked in Borjas' welfare magnet hypothesis. It is not the differentials in formal welfare provision that matter for them. Instead, migrants tend to consider what is available to them in their places of origin both in terms of formal and informal mechanisms to access welfare resources. For instance, for João and Maria, mentioned before, concerns of insufficient state-provided welfare resources at the age of retirement did play a key role in their decision to migrate. However, when they were considering the possibility of migrating, they were balancing their situation and prospects in their country of origin in terms of income, formal social protection and informal family care responsibilities. They were not assessing the welfare benefits in their future country of residence. They were driven by concerns regarding their future welfare resources as providers and receivers but the solution they were seeking was an increase in income through (informal) labour market participation in a country that offered, in their view, more opportunities. The solution they articulated was deeply embedded in their own views and understandings of welfare regimes from what they had experienced in Braziltheir welfare habitus there. In their case, they were not used to relying on the state to provide income when it was not available through formal employment. Hence, we could ask to what extent state-provided welfare provision, even in more developed welfare states is universally understood as sufficient to meet social protection needsan underlying assumption of the welfare magnet hypothesis.
Furthermore, we observe that the effect in migration projects of frames of reference based on previous experiences of welfare regimes are likely to be influenced by people's socio-economic background. For example, Beatriz, a postgraduate from Juiz de Fora living in Barcelona, who describes herself as part of the comfortable Brazilian middle class, migrated for the first time to carry out her education project: "I studied English philology and then I left. I had a godmother that, when I was born left me some money, she opened a savings account, which I could only use when I was 18 years old for whatever I wanted and when I turned 18 years old it was decided that I would use it to go to London and I spend 6 months there, studying English".
It was her affluent background that enabled her education-motivated migration. Beatriz's example suggests that it is not only the lack of welfare provision that drives migration. Access to (private) welfare provisions in the country of origin can also lead to migration aspirations. This represents a complementary perspective to the studies focusing on the retaining effects of welfare provisions in the origin country (Kureková, 2013;Mahendra, 2014). These overlook the facilitating migration effects of the broader formal and informal resource environment in the origin country.
Things worked out differently for Amanda. She was born in a rural area in Minas Gerais in an economically comfortable environment. She married and had three children, but she then divorced and became the main breadwinner of her household. Amanda migrated at the age of 45 due to shortfalls in the Brazilian welfare regime that she could not overcome with her financial resources. Before the introduction of the quota system in public universities, when access to public universities in Brazil was limited to the best students often from the more affluent classes, the only possibility for Amanda's children to study was to access a private university: "As my children studied in public schools, they were not [educationally] prepared to be accepted in a public university". Therefore, when her oldest daughter was in her second year of university and her son was preparing to access the university, she decided to move to Spain to increase her income to afford the costs of private tertiary education for her three children. Additionally, she wanted to accumulate economic resources to cope with the loss of income after her retirement.
All these experiences illustrate possible interplays between the origin country's welfare resource environment, and individual socio-economic and life course factors in shaping the role of welfare provision in migration projects. The examples show multiple social dimensions in which welfare operates as a driver of migration. Far beyond the potentially more generous welfare benefits in the country of destination, we have identified welfare-related factors that can influence migration from the individual, household, family and wider society perspective in the country of origin. Furthermore, in the cases of João and Amanda, we observed that the welfare habitus and the present circumstances at personal and structural level shape in a similar degree ideas on what would have been their welfare resource environment in the future if they had stayed in Brazil. Taking into consideration their welfare habitus acquired through past experiences in the Brazilian welfare regime and the welfare resources accessible to them, João and Amanda considered migration as an alternative to reconfigure their welfare habitus that they considered unsustainable in the future.
The next section explores how migrants' perceptions and decisions of welfare upon arrival are also shaped by frames of reference formed by the pre-existent welfare habitus. It also sheds light on the transformations in the understandings of welfare regimes that occur after migration and how this relates to considerations linked to the three temporal dimensions of agency.

Experiences of welfare: between old conceptions and new realities
We have argued that people are used to specific ways of thinking and doing that are influenced by the resource environment accessible to them within the existing welfare regimes in the country of origin. It is likely that these frames of reference shape initial actions around welfare provision in the destination country, a process that we called the origin country effect. This might explain why migrants are generally unaware of the implications of migration on how they will organise their social protection after moving (de Jong & de Valk, 2019; Scheibelhofer & Holzinger, 2018 among others). Our results also reflect that (transnational) welfare-related actions are reconsidered en route, on a need to need basis. We can differentiate two stages in the evolution of the welfare habitus after migration: Upon arrival the past dimension of agency dominates, and old conceptions from the origin country strongly shape initial welfare-related actions in the country of destination. The narrative of Mariana from Bahía and living in Lisbon is illustrative of how her welfare habitus was influential in the case of searching for a school for her daughter when she arrived: "I started searching for more information after I got here, specially, about the schools, the reality here is totally different if we compare to schools in Brazil. Public schools there tend to be very bad, that's why there are so many private institutions. So, when I got here, I started looking for a private school for my daughter, because that was the reference I had about schooling".
Mariana's immediate needs upon arrival had more weight than any future welfare provision needs or aspirations. She drew on her welfare habitus in an unknown new welfare regime. If Mariana had relied on social networks, her past welfare habitus might have had less importance in her judgements to choose a school.
The welfare habitus keeps playing a role over time in the destination country. Júlia's experience is one example. In situations of unemployment, this artist in Barcelona who has been jumping from one informal job to another learned from the welfare regime in Brazil to cope with the lack of income in Spain. Her narrative reflects that it is not only the formal welfare resources in Brazil that shaped her welfare habitus that she reproduces in Barcelona. What she has learned from her parents' agency and reactions to the structural constraints in Brazil also played a role: "In Brazil, people improvise a lot. (…) Brazil is always in crisis so we are marinated in the necessity to survive. (…) Everything I do here I learned from seeing how it worked there and how I reinvented myself was from my father and mother as my teachers. We always had to invent ourselves to pay the bills with what we had".
So, after some time in the place of destination, migrants tend to keep referring to old frames of reference to assess welfare resources in the place of residence. This is the case of Rafaela, from São Paulo and living in Barcelona, when she reflects upon the healthcare provisions in the country of destination: "(…) for me, in this country [Spain], the health system is great, because they have never let me down at any moment. On the contrary-I think the public health care system is almost private to me (…)".
However, the experience of Rafaela also illustrates how, over time, migration reveals itself as a learning process that shapes frames of reference, general perceptions and understandings of welfare provision as migrants begin to experience and learn about new realities: "I didn't have a comparison. I hadn't lived in a world of higher education and quality of life. (…). And now I can say that Brazil is horrible, in terms of quality of life, education, health, infrastructure, everything. (…) I knew Brazil was horrible, but I didn't know it was that horrible. When I came here, I saw that".
For Rafaela, the new experience in the destination country opened up new options of welfare provision and invoked the construction of new perspectives on the welfare regime in the origin country. Present experiences in new welfare regimes are thus likely to shape what migrants recall from past experiences and their welfare habitus. This is a process that takes place over the migration process and in unanticipated ways. Recent studies have underlined the role of necessity in adopting new welfare arrangements in the destination country (de Jong & de Valk, 2019;Scheibelhofer & Holzinger, 2018). What we aim to highlight with this study is the role of the changing meaning of welfare resources caused by a socio-spatial relocation (Halfacree & Rivera, 2012) that remains overlooked in the migration literature, especially in Borjas' welfare magnet hypothesis.
We have identified four interrelated factors causing transformations in the welfare habitus that are likely to emerge with migration. We illustrate them through the experiences of some interviewees.
Firstly, the learning-through-experience that takes place in the new context. This is illustrated in the case of Mariana. For practical reasons, she had to register her daughter to a public school. That was against her initial preferences, but her scepticism and doubts towards the public education system based in her experiences in Brazil faded away when she realized that the public school that her daughter attended "was a good school with good teachers" and this altered her understanding of public schools in the new context.
Zuleida's experience is also an example of these temporal shifts in conceptions of welfare provision. Informal welfare resources were the main driver of what was initially a short-term migratory project. Once the pressing needs of her son upon her arrival were met, with migration and new experiences in Lisbon, Zuleida was able to identify differences in the welfare regimes of Brazil and Portugal and, oriented towards her own future welfare provision needs, she incorporated aspects of the local welfare regime to her welfare habitus. State-provided welfare resources played a more important role than informal resources when she reconfigured her project to stay longer in Lisbon combining formal welfare resources both in Brazil and Portugal.
Secondly, the new set of opportunities and constraints in the immigration environment. In the domain of healthcare, the experience of Vinicius (from the state of São Paulo and living in Lisbon) reflects how previous frames of reference combined with the constraints that he faced in Portugal as an irregular migrant shaped his actions in terms of welfare provision. His choice for a private health insurance was based on two assessments. On the one hand, his lack of trust on the Brazilian public healthcare system. On the other hand, his limited access to stateprovided welfare resources in Portugal as an irregular migrant: "As our situation is not fully legalized here yet, I decided to get the easier way, which would be having the private health care plan for both of us. I didn't even get information about [the public healthcare]". In sum, a combination of welfare habitus and structural constraints reconfigured his actions once in Lisbon.
Thirdly, life events that can shape, for example, frames of belonging. This is what happened to Beatriz who also had a private insurance in Barcelona. However, when her son was born, she started using the public healthcare system. This happened almost by procedural inertia. She visited a paediatrician in the neighbourhood public surgery as she was told to do for the periodical check-ups on the baby's weight and development and for his vaccinations. As a migrant, she also felt that it was a way of confirming the baby's belonging to his country of birth: "I think I wanted to have him registered here, like he exists, born here and so on, with his vaccinations". Health-related life events can also lead to reconfigurations of the welfare habitus. For instance, priorities in welfare provision changed for Amanda when she suffered a domestic accident after some time in Barcelona. Due to the fractures caused by the accident she was not able to work for several months. She returned temporarily to Brazil until she recovered. There she relied on the informal care of her children and the formal welfare benefits from Spain. Until she recovered, her priorities changed from her initial orientations towards the future education of her children and her pension to the more pressing needs of health care and loss of income of the present.
Finally, migration networks and migration history also play a role in the change of welfare habitus. This is apparent in the case of Lua from Santa Catarina in Lisbon. Before migrating, she discovered that she and her son had a degenerative disease. A doctor recommended that they went to Portugal since their disease was more common there and treatment more advanced. She was directly referred to doctors in Portugal. The transfer was possible through the PB4 bilateral agreement mentioned earlier. The welfare agreements resulting from the historical links between Brazil and Portugal eased the formalities to move to Portugal and opened new perspectives for Lua to search a more suitable health care elsewhere. The context of the intense and long migration history between Portugal and Brazil also played a role as mother and son were received in Portugal by the family of a Portuguese friend in Brazil.
These four interrelated elements show the strongly interrelated role of historical and geographical factors (including the immigration environment and migrants' restricted access to welfare) in shaping meanings of welfare provision and practices around it (Raghuram, 2016).

Conclusion
In this article, we have sought to highlight multiple ways through which migration and welfare constitute a dynamic relationship. This is an aspect that has been ignored in Borjas' welfare magnet hypothesis and, although acknowledged in the recent literature on transnational social protection, it has also remained understudied in research applying this framework. We argue that, beyond the effects on migration of more generous welfare regimes in the countries of destination and the advances made by the transnational social protection framework stressing the variety of welfare providers beyond the state in contexts of migration (Faist et al., 2015;Levitt et al., 2016), it is still possible to deepen and refine our understanding of this mutual and changing relationship. The article contributes to the existing literature focusing on the welfare habitus and how it changes over the migration process. It explores three aspects so far overlooked. First, it looks at the role of welfare regimes in the formation and redefinition of migration projects. Second, it analyses how the welfare habitus emerged from experiences in the country of origin deeply shape conceptions of welfare regimes and affect expectations on arrival and initial actions around welfare provision in the destination country. Finally, it sheds light on the changing meaning of welfare provision over time, through the life cycle and across geographies. The article draws on Emirbayer and Mische (1998) to propose an analysis of the dynamic relationship between migration and welfare considering that actions around welfare provision are informed by past experiences (welfare habitus), influenced by (perceptions of) the present situation in constant change, and oriented towards the future. This framework allows to capture aspects of the welfare habitus that remain unchanged and others that change over the migration process and the life course. There are two main points of departure for our argument. On the one hand, people's expectations, aspirations and decisions around welfare provision, including in its interaction with migration, are based on specific frames of reference shaped by people's habitus, their individual circumstances and past experiences of the external structures in the country of origin. On the other hand, there is a transforming factor in migration conceptualised as a life event that changes what welfare provisions mean over time as the migrant spends time in a country where the welfare regime is organized in a different way from what had been experienced before. The transformative process is shaped by the contact with a new socio-spatial context, the changing structural opportunities and constraints over the migration process and changes over the life course. Migration networks and migration history also play a role in the change of references. Aligned with previous studies, our data analysis showed that the initial migration project either includes little or no reference to welfare provision as a sole driver of migration. However, when welfare provision is part of the migration project, it is largely shaped by the conception of welfare regimes that is rooted in the social structures and habitus in the country of origin, hence the origin country effect. As such, before migration, past experiences and present circumstances at personal and structural level shape imaginations on future practices in the country of origin and the country of destination. The welfare habitus based on experiences in the country of origin is reproduced and remains central upon arrival. This is especially the case if migrants do not rely on social networks to make decisions to cover initial pressing needs in welfare provision. Over time, the welfare habitus is reconfigured en route. Migration itself becomes a learning process that shapes frames of reference, general perceptions and understandings of welfare provision as migrants begin to experience and learn about new realities. Despite such transformations, migrants often keep referring to their previous frames of reference. This shows that with migration old frames of reference are not immediately substituted by new ones. Hence, it is incorrect to consider that when welfare provision plays a role in the migration process migrants look only to what is (potentially) available in the country of destination. On the contrary, over time frames of reference overlap, expand and remain interconnected with each other, as the welfare habitus develops and migrants begin to constitute transnational welfare assemblages as those described by Faist et al. (2015) and Levitt et al. (2016). Finally, our findings show that, like care practices (Raghuram, 2016), practices around welfare provision are sitespecific, complex and dynamic. Therefore, research should move beyond the assumptions that, before migration, state-provided welfare resources in the destination country is perceived positively around the globe and preferred to other welfare resources and coping strategies considered in migration projects.
Our findings based on a qualitative analysis provide new perspectives of the complex relationship between welfare and migration to be confirmed with larger quantitative studies. Future research could explore more in-depth the effect of life-course events in the reconfiguration of the welfare habitus and their impact on migration projects. Particular life events such as the birth of a child, illness, or unemployment are likely to reconfigure the predominance of past experiences, the present situation, or future orientations in migrants' priorities and decisions around welfare in contexts of migration. These reconfigurations remain understudied.