Translating politics into policy implementation: welfare frontline workers in polarised Brazil

Abstract How is policy implementation affected by increased polarisation and extreme shifts in politics? In order to address this question, the paper focuses on frontline workers’ (street-level bureaucrats’) interpretations of political shifts and how these are then translated into practice. Building on ethnographic fieldwork conducted among social workers in Northeast Brazil, the paper proposes a theoretical framework for analysing the influence of political landscapes on policy implementation by foregrounding the political processes in which these agents play a critical role. Drawing on empirical data, the paper proposes ideal types of possible outcomes of translation practices – counterbalance, collaboration, resistance – that function as a guiding framework for future research.


Introduction
The moment Jair Bolsonaro was elected president of Brazil in October 2018, it became clear for civil servants working with policies that had become controversial during the electoral campaigns that their job would not remain the same. For these frontline workersprofessionals involved in the delivery of public policies, often conceptualised as street-level bureaucrats 1there was no need to wait for the inauguration of the new administration (which was to take place two months later) and the consequent policy change that was expected. The election's result was enough to legitimise immediate changes in the implementation of a diverse set of policies. Some frontline workers welcomed the political shift ratified, as it was, by the majority of the electorate; they proceeded to use the discretion inherent in their work to advance the changes that were expected, towards an openly far-Right conservative view on a wide range of policies. The most expected outcome, a direct consequence of Bolsonaro's electoral campaign, was that there would be an increase in police violence in low-income neighbourhoods (Thuswohl, 2018).
Another group, namely frontline workers involved with policies identified by Mr Bolsonaro as 'Leftist', saw their authority being undermined. For example, environmental agents avoided verifying illegal deforestation in the Amazon region because they feared violent resistance from landowners, and there were numerous reports of teachers being filmed during classes by their teenager students who accused them of 'communist indoctrination'. Willingly or under duress, these bureaucrats had to translate the new and unexpected political scenario into their work in a matter of days. Inevitably, growing support for the political Right in the country had already been affecting their work, but the election results consolidated a new direction that was desired by a majority and not just a fringe of extreme conservatives.
Institutional arrangements that have shaped governing coalitions in Brazil in the last three decades have tended to guaranteed policy continuity, despite shifts in governing parties (Zimerman and Pinheiro, 2020). Since 2016, however, new administrations -Temer in 2016 and Bolsonaro in 2019have used the relative independence that the executive branch retains in Brazilian democracy to break from this tendency, especially when it comes to policies that have little political consensus, notably social policies (da Costa, 2019). For a country that has become known internationally for the success of its anti-poverty policies (Milhorance et al., 2019), the question of how those implementing these policies understand and react to increased polarisation of welfare policies is crucialbut remains unknown. Regardless of the direction of change within the political spectrum, or the stage of policy transformation, frontline workers experience political shifts first-hand and are thus deeply affected by the rearrangement of political forces and institutional shuffling. Even if the impact of politics on policy has been well investigated, the effect of political change on policy implementation has not been. This paper aims to fill that gap by proposing a framework to investigate how policy implementation is affected by increased polarisation and extreme shifts in politics, even before actual changes in guidelines and budget occur. Put differently, the framework aims to answer the following research question: How do frontline workers interpret political shifts and translate them into practice when implementing public policies?
The paper is based on an ethnography of the Bolsa Família Programme (BFP)'s implementation in a medium-sized municipality (population 100,000-500,000) in the state of Ceará I call Angico. The research was conducted in four one-to three-month field stays between 2011 and 2015, and is supported by subsequent related projects undertaken in 2017-2020 in the same region. I conducted participant observation in two welfare offices: the main BFP municipal office and an outpost of the welfare secretariat in Angico's poorest neighbourhood, which also received applications for the BFP. My routine consisted of following a group of thirty social workers over the years as they went about their daily routine of receiving applicants and beneficiaries of the programme, addressing bureaucratic issues and conducting house visits for updates on family information or to verify suspicions of errors flagged by the federal BFP office. The social workers were all women, aged twenty-three to thirty-five, and all had an undergraduate degree in social work from one of the two local private higher education institutions.
My empirical focus was on how frontline workers used their discretion and moral agency (following Bierschenk and Olivier de Sardan, 2019;Dubois, 2010;Zacka, 2017) and on their interactions with their peers, managers and welfare recipients. The main advantage of this approach was that it generated key insights into the development of frontline agents' perceptions of work practice. Another set of in-depth interviews inform my analysis and was part of a larger project (Eiró, 2017), even if not directly used in this paper: I repeatedly interviewed ten mid-level bureaucrats, including office coordinators and programme leaders, and thirty-five beneficiaries of the BFP.
The paper focuses on frontline workers working with welfare policies, all of whom, in my fieldwork, were trained social workers. Social workers can be conceptualised as street-level bureaucrats (Hasenfeld et al., 1987;Rice, 2012), whose use of discretion can symbolise the state's regulation and social domination over the poor (Dubois, 2010;Siblot, 2016). During presidential campaigns in Brazil for at least the last twenty years, social policies have been a matter of heated debate, notably surrounding the country's most important welfare programme, the BFP. The debate sharply divides Left and Right on issues concerning meritocracy, welfare dependency and public expenditure.
On a theoretical level, the paper builds on the notion of 'translation' to understand frontline workers' interpretations and adaptations of political shifts into daily practices as political acts (Mosse and Lewis, 2006;Freeman, 2009). My analyses of social workers' interactions with welfare recipients are inspired by the Anthropology of the State literature (Auyero, 2011;Ferguson and Gupta, 2002). Based on social workers' attitudes towards welfare programmes and their beneficiaries, I propose a framework for understanding the effect of political change on policy implementation. It is my hope and beliefthat this analytical framework can be used beyond the social policy domain, since it focuses on aspects shared by other policy domains that (1) depend on frontline workers for policy realisation; and (2) are politically contested, thus being symbolically or effectively affected by political changes.
The paper comprises four sections: the first is dedicated to a literature review and proposes an analytical framework for studying the translation of politics into practice at street level; the second analyses the political context of welfare policies and its implications for policy implementation; the third focuses on the translation of politics into policy at the frontline of the BFP; and the fourth builds on the preceding empirical material to propose three ideal types of translation practices that can be used by frontline workers.
2 Translating politics into policy implementation: proposing an analytical framework for the study of translation As discourses and policies evolve in new political directions, what happens to frontline workers' practices? Policy implementation studies has established itself as an important interdisciplinary field by showing how public policies change when delivered. Street-level bureaucracy theory, developed mainly by Michael Lipsky (1980), posits that frontline workers should be acknowledged, and indeed conceptualised, as the last link in the policy-making chain, because it is only in bureaucrats' interactions with clients that public policies come to life. Other authors have observed that the adaptation of policies occurs specifically in the interaction between bureaucrats and citizens (Warin, 2003). Maynard-Moody and Musheno (2003) found that frontline workers such as 'cops, teachers and counsellors' classify clients into social or professional categories in order to simplify their work. They describe, for example, how social workers concentrate their efforts on clients whom they view as worthy of investment. Street-level bureaucracy literature has established that perceptions of deservingness are contingent on implementation agents' demographics, particularly their gender and race (Jilke and Tummers, 2018).
Researchers have focused on several factors that shape this process, from the implicit bias of frontline workers to organisational context and culture (Harrits, 2018;Jilke and Tummers, 2018). What has not been systematically studied is how frontline workers' decision-making processes are affected by political change (Hill and Hupe, 2014)in other words, how frontline workers react to changes in the political scenario that have not yet been transformed into policy. Bernardo Zacka (2017) raised questions regarding this issue when he identified a contradiction at the core of frontline workers' worlds: their 'moral commitments to particular ways of performing the role change at a far slower pace than the political priorities of the day' (p. 206). This not only highlights how little is known about frontline workers' reactions to 'the political priorities of the day'; it also hypothesises a static and homogenous response. The data I present in this paper suggest that such moral commitments are affected by politics in a far more dynamic manner.
In policy implementation studies, it is generally accepted that each encounter between frontline worker and client represents an instance of policy delivery, which is shaped by agents' conceptions of their work and of their clients, determining 'in a concrete way the form and substance of citizens' rights' (Hasenfeld et al., 1987, p. 398). Furthermore, especially when it comes to social policies, these workers are often the main or only face of the state that people interact with, and that consequently these interactions risk reproducing social inequalities (Eiró, 2019;Gupta, 2012;Pires, 2019). Justifying the objective of its study, the field has long emphasised the dynamic nature of implementation interactions while considering the policy-making sphere as stable, or at least arguing that policy implementation is insulated from conflicts in the political arena, as if those two moments were independent from each other.
To fill this gap, the analytical framework I propose is built on the notion of 'translation', particularly its use in the anthropology of brokerage (Mosse and Lewis, 2006;Koster, 2012). 2 In order to 2 In this context, 'translation' does not presuppose the literal translation of idioms, even if this is the case in other contexts, as long as the actual focus remains on the interpretations of meanings and their transformation into actions.
translate, however, frontline workers need to interpret the political scenario, and such interpretations are shaped by the perception they have of their position as the 'expert', which gives them (selfbestowed) authority to change policy goals during implementation, because they 'know better' (Eiró, 2019). This expert position relies on a knowledge source that policy-makers cannot access, which can function as a thermometer for the political landscape. In the case of social policies, frontline workers might engage with interpreting changes in the 'politics of distribution' that shape the political economy of programmes such as the BFP and that defines who is deserving of assistance and how resources should be distributed (Ferguson, 2015). Frontline workers then experience political change before policy-makers doand thus before policies actually change: seldom will a newly elected administration dismantle public services all at once (Birkland, 2016). The mechanics of democratic rule demand time for political negotiations, parliamentary discussions and bureaucratic procedures, which are open to resistance and subsequent (and sometimes endless) negotiations.
However straightforward this process of interpretation seems to be in the context of interpreting political change, the complexity of the object to be translated must be acknowledged. Practices of translation are, by definition, marked by epistemological uncertainty. As Freeman (2009, p. 440) put it: 'the translator's first task may be to identify not (or not only) the knowledge that is to be translated, but the uncertainty that surrounds it.' Policy implementation studies have established that the literal implementation of policies does not exist because policies are always shaped by implementation practices (Majone and Wildavsky, 1984), but this paper offers a way to add changing political landscapes to this equation, where meaning contained in political discourses and practices (and not only in policy regulations) undergo substantial alteration based on agents' active interpretation (Clarke, 2009).
The approach to 'translation' I build on is thus one that rejects its often-apolitical use by highlighting how actors pursue specific interpretations and how consequent actions are deeply rooted in power relations. How class, race and gender interact and intersect in frontline workers' interpretations of policy change is of particular significance here (Siblot, 2016), since research has found that implementation practices both reproducealbeit in different waysracialised and gendered elements that are integral to the design of welfare policies (Molyneux, 2006). Frontline workers are not only intermediaries between policy-makers and clients; they also have 'key institutional positions, albeit unscripted, informal, personalised, and highly unstable ones' (Mosse and Lewis, 2006, p. 13). This means that frontline workers not only interpretwhich connotes 'understanding, reception, apprehension of meaning'but produce 'a new semantic object' (Freeman, 2009, p. 435). In this context, the analytical focus I put forward is not on how frontline workers operate and strategise within existing arrangements of policy, but how policies 'become real through the work of generating and translating interests' (Mosse and Lewis, 2006, p. 13). In policy implementation, this is relevant because frontline workers' translation of political changesa translation that takes place in their discretionary space is a process of negotiating meanings and interests that are embedded in moral imperatives, and should therefore be understood as political manoeuvring because the process is explicitly shaping power relations (Sakai, 2006). Studies on street-level bureaucrats have foundbut on the whole express only implicitlythat translation is thus not 'merely a technical process; it is a political process in which the translators actively choose between alternatives' (Freeman, 2009, p. 435).
To what extent can frontline workers' views of political change actually enact or advance change in policy? This line of reasoning brings us to how the state is imagined and shaped in frontline workers' interactions with their clients. If 'translation' is understood as the act of bringing different things into relation with one another (Law, 1999), implementation practices are more than a 'process of direction', of simple transferring or copying: they are a process of interaction (Pressman and Wildavsky, 1984). The translation of politics into policy is therefore a communicative process in which politicians, frontline workers and clients '(i) enter into relations with each other and (ii) begin to recast or reconstruct themselves, their interests and their worlds' (Freeman, 2009, p. 436). The focus is thus on 'the analysis of the everyday practices of local bureaucracies and the discursive construction of the state in public culture' (Gupta, 1995, p. 376). Although it is important to recognise the difference between state and government, the literature on how the state is imagined and shaped by its agents is directly affected by the political landscape. This varies significantly from country to country and depends, in particular, on the strength of state institutions. Even the most stable state institutions, however, are accountable to government, and paying attention to political interference on state institutions and bureaucrats remains useful. It is in the interactions between frontline workers and citizensthat is, in the 'real' working of states (Bierschenk and Olivier de Sardan, 2014) that the state is given shape or, as Law (1999) put it, is 'performed'. The way both frontline workers and citizens imagine this same state is contingent on its changing in new political directions. Even if the creative nature of these agents' work is already an object of consensus within implementation studies, following Clarke (2009), I argue that implementation practices should be conceptualised as 'processes of translation', in which individuals actively interpret and change meanings. This transformation operates as the enacting of a particular view of the state, or, in this case, a new face the state should or should not take following political change.
This analytical framework does not suppose a homogenous response of frontline workers to political change. Rather, the breakdown and contextualising I propose allows an understanding of the mechanisms through which political change is interpreted and transformed into actions by the individuals occupying the unique position of transforming policy before policy change is officially formulated. The way frontline workers react will vary according to a set of factors that cannot be predicted, even if the literature on street-level bureaucracy hints at their nature. Institutionally, organisational culture, institutional rigidness, degree of discretion enjoyed and supervision by managers and peers are good examples. Socially, scrutiny from relationships outside the workplace, social proximity to clients and class trajectories are noteworthy. Individually, moral and political values, religious affiliations and identity markers also play a role. This is by no means an exhaustive listnor is it a minimum set of factors that should be observed. Rather, it is a starting point for the inductive researcher to explore the way frontline workers interpret the political change they face.
3 Political change and the effects on (welfare) policy The BFP is the world's largest conditional cash transfer (CCT) in terms of the number of people assisted. In recent years, the number of BFP beneficiaries has oscillated at around 14 million familiessome 50 million people, a quarter of Brazil's population. In the last two decades, CCT programmes have become increasingly popular in developing countries. They centre on a direct cash transfer to families who are below the national poverty line. Monthly benefits vary between $10 and $100, depending on the composition of the household. Families must respect certain conditions (known as 'conditionalities' in CCT jargon) in order to keep the benefit: children must go to school, and all children's and pregnant women's health must be monitored. Although it is a federal programme, the BFP is managed in collaboration with other state institutions, mostly municipalities, which are responsible for registrations, hiring local staff and monitoring the programme's beneficiaries. The BFP, just like most CCTs around the world, is a federal government programme, and can be thus modified by new administrations to reflect political preferences, bypassing high-level bureaucrats who could moderate radical transformations.
Despite the relative political stability of the last two decades, a consensus regarding welfare policies is still far from being realised in Latin America (Layton et al., 2018), and the region's model of poverty reduction isonce againa contested space. Two clear tendencies over the last two decades can be observed: a steep reduction in poverty and an increase in social rights (Barrientos, 2016;Merrien, 2013). At the same time, new social policies increasingly integrate neoliberal elements (Borges, 2018), often reinforcing traditional gender roles (Cookson, 2018). This was also the case with the acclaimed CCTs, which are now functioning in almost all Latin American countries. The resilience of these programmes can partially be attributed to their electoral impact through benefiting presidential incumbents (Zucco, 2013). However, their effectiveness in alleviating poverty makes them tough political targets, even if discourses calling for their termination or complete transformation to better reflect the ideologies of the workfare are recurrent (Ivo and Exaltação, 2012;Eiró, 2017). This scenario points to the recent transformations in welfare structures (Esping-Andersen, 1990) and the institutional landscapes (Newman and Clarke, 2009) that have influenced policy-making and frontline workers, but, most of all, that shape political debates and societies' representations of povertypart of what Ferguson (2015) called the 'politics of distribution'.
The changes in welfare politics in Brazil are emblematic of this regional trend, and the BFP's importance in understanding how political change affects implementation practices lies in the programme's central position in the larger debate surrounding poverty and social assistance in Brazil. The country led innovative efforts to reduce poverty in the world in the 2000s, notably through the BFP, under successive administrations of the Workers' Party (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016). The country then entered a turbulent political period, with the impeachment of President Dilma Rousseff and the rearrangement of forces around a neoliberal administration led by her vice-president, Michel Temer. Upon his definitive arrival in the presidency in September 2016, the actions of the ministry responsible for the BFP were reoriented to introduce elements of workfareand these came close to being implemented. Since its creation, several attempts to insert work-related conditionalities into the BFP have been initiated in parliament. These attempts, based on the idea that the programme must not become a 'means of life' and for that one must force the beneficiaries to integrate into the labour market, still did not find support of a majority nor a president who would force them.
The election of far-Right President Bolsonaro in 2018 also had a significant impact on the programme, following a global trend in the rise of political polarisation and ensuing shifts in national administrations resulting in setbacks to anti-poverty efforts (Handlin, 2018). Already during his electoral campaign, Bolsonaro had to work hard to convince voters that he would not put an end to the BFP, as he had been openly against it for almost a decade. In numerous interviews and debates in parliament (where he was a congressman for eighteen years), Bolsonaro had repeatedly said that the programme was in fact taking money 'from those who work' and giving it 'to those who are lazy', and that BFP beneficiaries prefer not to work. During his campaign for president, Bolsonaro softened his position, saying that the programme would probably be reduced by half under himrather than abolishedas he was sure that at least half of the beneficiaries were frauds or willingly not working in order to receive benefits. During his administration, the number of BFP beneficiaries has indeed drastically reduced. However, Bolsonaro is still compelled to defend the programme publicly, using budget cuts to introduce an annual '13th benefit' and to pretend that he is expanding the programme.
These changes in rhetoric and policy, especially in relation to the rise of the far Right (to power, or merely to mainstream public debate), have given a new boost to ideologies of austerity and neoliberalism that are grounded in a belief that poor people lack moral standing (Katz, 2013). Such narratives are widespread in Brazil: for example, in their analysis of how residents of rich neighbourhoods in São Paulo perceived the poor, and how these representations were used to justify segregation practices, Paugam et al. (2017) identified a strong naturalisation of poverty and a strong sense of meritocracy. This debate relates directly to perceptions of poor people's deservingness to receive public aid. As I have shown elsewhere (Eiró, 2019), BFP social workers' informal practices in allocating what they see as scarce resources are embedded in their conceptions of poverty, and the separation of 'deserving' from 'undeserving' poor shapes how they implement the programme and exacerbates a sense of insecurity among beneficiaries.
4 Politics and policy implementation within the BFP This section focuses on frontline workers implementing the BFP in Brazil. The analysis is separated into two interdependent moments: how frontline workers interpret the changing political landscape and the practices of translation.
4.1 Interpreting political change: the politics of welfare Actual practices of translation are preceded by (and feed from) how the political landscape is read and interpreted. Although nothing stops a researcher from exclusively focusing on the former, an effort to understand the latter allows us to embed translation practices in individuals' perceptions of politics and how their work is affected by it. As is the case with welfare policies, frontline workers deal directly or indirectly with the forging of citizenship (Dubois, 2010) and, most importantly, with how citizenship is contested in the political arena and in the state institutions of which these workers are a component. Following Hasenfeld et al. (1987), I believe that the more concrete the analyses of the politics of social policy implementation are, 'the more important it becomes to take into account ideologies of citizenship rights and obligations ' (1987, p. 395). This includes analysing the 'politics of distribution' (Ferguson, 2015): how the poor are portrayed in politics and categorised in policychanging conceptualisations of who is 'deserving' of aid, for example, and the role that race and gender play (Boesten, 2010;Haney, 2002;Katz, 2013). In order to understand how frontline workers interpret the political landscape affecting them, it is important first to understand, in the case of welfare policies, how they answer questions like: What is the state's role in reducing poverty? What is the role of a social worker in someone's process of leaving poverty? What types of policies are more effective (i.e. cash vs. in-kind benefits)? What is expected of social workers when the policies they work with become a battlefield of political and cultural conflicts? Employed by municipalities but working in government programmes that are directly affected by political shifts, frontline workers must interpret multiple (and maybe conflicting) expectations regarding their role. This position forces them to integrate input from different levels of public administration, which could be governed by different political forces.
During my field visits to Angico, I met around thirty social workers who were involved directly or indirectly in the implementation of the BFP. Most of them openly took pride in their work and in the programme. They were happy to be able to make a difference in people's lives by working in a programme that could give people a secure income, even if only a modest one. This pride in being able to help was not seen as a contradiction with the fact that most of them frequently viewed people in poverty as lazy and deceitful, always trying to take advantage of the generous state.
My fieldwork encompassed a time when Brazil was still experiencing economic growth and expansion in social expenditure. Thanks to the reserves accumulated in previous years, Brazil was relatively insulated from the financial crisis of 2007-2008which affected most of the worlduntil 2015. The annual increase in the BFP benefits value and a decreasing unemployment rate did not stop social workers from understanding their role as gatekeepers of valuable public resources that should not be wasted on people whom they perceived as undeserving. The views they had of the poor, rooted in well-known ideologies that can be traced back to most religions, were entrenched in poverty alleviation policies worldwide and were highly racialised, given the country's history of slavery. These views were reinforced by common myths, the most recurring being that BFP beneficiaries decline work offers in order to keep benefits, which has been systematically disproven (Oliveira and Soares, 2012). Social worker Fernanda (aged twenty-three) framed this problem as follows: 'I really think [the programme] can help, but for that to happen people have to really make use of it and give it their all. It is not for nothing that one receives it. It's with a commitment to change one's life.' Fernanda's views, which resonated with her colleagues', was that the BFP should be stricter in its regulations in order to weed out beneficiaries who were not making an effort to change their lives. The root of this problem was the extensive length of time people spent in the programme, which social workers perceived as endemic. Indeed, the number of years that had passed since enrolment was understood, by social workers, as a sign of a person's inability or unwillingness to 'leave poverty'. There was no talk of a hard cutoff line among social workers, especially because they took special circumstances into consideration during their moral assessments of parents, particularly mothers, who receive the grant. A 'difficult' child or a handicapped family member was a good reason for someone to remain dependent on the programme, as were other cases in which social workers could testify to someone's 'good' intentions, namely the explicit will to leave the programme as soon as possible.
This attitude that I identified among frontline workers should not be seen in isolation from the fact that in the decade prior to my research, Brazil witnessed an unprecedented expansion in social expenditure and a move away from public aid based on the distribution of goods in favour of cash transfers. Although these social workers saw this move in a positive light, the BFP and other programmes were interpreted as the expression of a 'too-generous state', and programmes with loose regulations failed to properly assess applicants deservingness to receive such generous aida task they believed was their responsibility.

Translating welfare politics
The manner in which frontline workers effectively modify policies derives from their interpretations of the political landscape and, more specifically, from how they interpret political change. As detailed before, these translation practices vary and analysing the course they takecounterbalance, collaboration, resistancesheds light on their differences. In all cases, translation practices rely on the freedom of discretion that frontline workers enjoy. As with classic studies on street-level bureaucracy, the focus should be on their moral assessments and their gatekeeping decisions, which is where their power to shape policies lies.
Far from being passive observers of the political scenario, frontline workers actively engage with the 'politics of distribution' (Ferguson, 2015): their actions cannot be separated from the political field that encompasses politicians and policy-makers. The social workers I spoke to in the BFP did not think that the state should blindly trust people asking for public aid, yet this was just what the programme asked of the social workers themselves, who were not supposed to question information provided by applicants. In fact, the self-declaration of information is one of the core principles of the BFP, especially because high levels of informal work make it hard to verify declared income. The programme does, however, authorise social workers to investigate further if they are suspicious that beneficiaries have provided misinformation, but no rigid instructions on how to perform such assessments are given, nor is what constitutes a valid suspicion clearly defined. This mechanism, created to remove clearly 'non-poor families' from the programme, was systematically transformed by frontline workers in an effort to assess all participating families' levels of poverty. In order to do soand to respond to what they interpreted as society's wish for a bureaucracy that actively screened applicants and functioned as a gatekeeper of state resourcessocial workers employed a series of informal but structured and shared strategies to 'creatively adapt' (Rice, 2012) the programme.
The most important effect of these practices, at least for the purpose of this paper, was how the programme's bureaucratic procedures were manipulated. A recurrent practice was to emphasise and exaggerate the legal consequences that a beneficiary might suffer if convicted of fraud. When registering in the BFP, each applicant must sign a declaration testifying that the information they have given is true. Social workers told me that they emphasise the juridical-legal nature of the document, especially where it says: 'liable of criminal and civil prosecution as a result of committed frauds'. Social worker Ana (aged twenty-five) said 'We scare them a bit' and, laughing, told me how respondents often suddenly remember another income source.
The social workers I spoke to and observed were not deliberately aiming to undermine the programme's goal. Adaptations were viewed as 'fixes' to the programme's flaws. They believed some people truly needed the benefits, such as families in extreme poverty or those who had a reason for not being able to work. Although they did not decide the value of a family's benefits, they knew it was directly related to the household income, and they often calculated approximate values in order to inform their decisions regarding which income should be declarable and which should not. They thus not only subjectively assessed every family they encountered, but also compared them in order to establish a hierarchy of need and merit. They used these assessments to decide what type of information about the programme they would give (or omit) to each beneficiary. For example, if they encountered a family that had had its benefits temporarily blocked (which can happen for a number of bureaucratic reasons), they could either provide step-by-step instructions for fixing the problem or they could ask the family to wait, knowing that if enough time passed, the family would miss the window to contest a block made by a social worker.
This sabotaging of beneficiaries' rights was justified by social workers' assessments of families' material needs, and they could use their observation of material possessions, furnishings and the condition of a beneficiary's house to place that family below others in their hierarchy of need. In light of the analytical framework I propose, such distortions and omissions of information regarding the rules of the programme represent the most literal translation of an abstract ideal into policy practice. The informal strategies go beyond the simple use of predicted discretion to a purposive modification of procedures, thus changing the policy.

Types of translation in policy implementation
In the spirit of proposing an analytical framework for the study of translation practices at the street level, this section proposes three ideal types of frontline workers' attitudes towards political change. These ideal types, as suggested by Weber (1968), constitute an analytical tool that can logically integrate empirically observed forms of action. Ideal types, however, do not presuppose the construction of individual archetypes; types of translation should not be read as types of individuals, but rather as types of behaviour. These types are constructed based on lived experiences that should always derive from empirical data and, as such, they are most useful for identifying what Weber calls 'deviations'empirical cases that distance themselves from the ideal type. What follows, then, is the proposal of ideal types based on my long-term ethnographic research among BFP social workers.

Counterbalance
First is counterbalance, by which I mean purposively acting as a moderating force on (perceived) political polarisation. As frontline workers see the institutions they work in and the services they provide take centre stage in heated political debates, they might act as 'neutral' and 'professional' parties. This position reflects a normative perception of the stateone that 'should be politically neutral', able to see beyond the short-term interests of politicians (Lentz, 2014). In their interactions with clients, peers and others outside their work environment, they might emphasise the importance of the continuity of services through changing administrations, in an attempt to 'strengthen the image of the state' (Bierschenk and Olivier de Sardan, 2014, p. 15). Political polarisation might be perceived by them as a threat to their work, since it engenders policy rupture for the sake of political differentiation, rather than an incremental improvement in public services. It is important to note that political polarisation is often used as a political label to de-legitimise opposition and promote a postpolitical view in which dissent is framed as extreme, unnecessary and anti-democratic. The case of Brazil illustrates this point well, where the sudden emergence of far-Right politician Jair Bolsonaro on the national scene, and his subsequent election as president in 2018, was immediately used to frame the Centre-Left Workers' Party and its administrations (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016) as extreme Left, radical or socialist.
That said, to claim that political polarisation is a fabricated illusion is also false, most notably so when considering how frontline workers experience it. The growth of far-Right politics cannot be measured only in terms of electoral performance since they might also shape the political debate when not (yet) organised in political parties. What is tangible to frontline workers is the way people around them react to perceived political polarisation and how this affects their work. The best examples concern their managers and clients. The first group are often not only those who enjoy more discretionary power, but also those who can limit how frontline workers themselves use theirs: in my fieldwork I encountered many instances in which managers, motivated by political preferences or moral judgments, would overturn social workers' decisions. Social workers make use of their position and expert training to complain of such incidents, even if mainly to their peers and the researcher in the room, rather than to the superior who could threaten their employment.
Clients, in their turn, can voice their concerns about future policy change to frontline workers, or they can demand faster change than new administrations and subsequent policy changes facilitate. Facing pressure from both sides (Bierschenk and Olivier de Sardan, 2014), frontline workers are well aware of the vulnerability of their position: demands for change will be directed to them precisely because they have the power to modify policies. This realisation, which can be a novelty for social workers who underestimate their role and see policy-making as the event responsible for actual policy change, can make these professionals defend their line of work by preserving the status quo and championing moderation and policy continuity, and can push them to act as balancing forces to political shifts, whether or not these shifts are in directions with which they agree. As social worker Ana (aged twenty-five) repeatedly explained to me: 'we are not bureaucrats, there's no point in hiring a social worker to this job if you don't expect a serious assessment of people's needs. A technician wouldn't be able to do our work; they don't have the same sensibility.' Such a claim to bureaucratic independency could be used to argue against specific programme designsones that constrained social workers' actions, for instanceas well as against political interference. Both disturbances to the way they would like to work are a result of the political landscape, be it in the form of institutionalised policy mechanisms or politically motivated interference in the programme's design.

Collaboration
The second ideal type of translation is collaboration: the advancement or exacerbation of (perceived or definitive) political shifts. The sudden growth of a social movement or political party, new election results or a perceived increase in support for political change can give frontline workers the green light to advance what they might see as an inevitable policy change. The political views that directly affect the policies they work with might resonate with their own opinions for fairer client-processing mechanisms, or for harsher measures that they believe are in society's interest. Even if there can be many permutations here, frontline workers will justify their translation using their 'street-level' knowledge. As policy change takes time, advancing political shifts can also be seen as a matter of efficiency. The same goes for exacerbating political change: frontline workers might understand that even new administrations can have trouble implementing policy change; strong opposition, a decentralised policy-making system or technical issues can function as barriers to electoral campaign promises being realised. Frontline workers are accountable to the public in different ways, and can decide to take matters into their own hands (Lipsky, 1980;Zacka, 2017).
While personal political preferences are a simple way to explain this collaborative behaviour, it is also important to consider another optionthat frontline workers are highly aware of their role in shaping policy. Independently of their political preferencesand this is something to be empirically investigated, rather than being an affirmation of the neutrality of this variablefrontline workers might decide to advance political change because it is the 'will of the people'. A strong commitment to democratic values can justify frontline workers using their discretion to steer the programme in the direction they believe is wanted. This was the case with one of the social workers during my fieldwork, who had been employed in Angico's municipal welfare office for over ten yearswhich she acknowledged was because of her political connections. Being upper middle-class (an exception among social workers in Angico), she recently 3 told me that many of her friends and acquaintances are not shy to tell her how they count on her 'good' use of the 'privilege' of being able to finally put an end to the 'overgenerous' welfare policies that Bolsonaro's administration cannot seem to gather the political support to terminate. I saw a hint of this already in 2016, when the new Centre-Right administration that took over after Dilma Rousseff's impeachment announced the introduction of work-related conditions for BFP beneficiaries. Remembering how, in previous years, I had witnessed social workers complaining that the programme did not have a limited duration, so as to push people to find employment, I asked Joana (aged twenty-six) about these changes. She said she could not wait for work conditionalities to come into effect, even while making it clear that she did not like the current administration. Given our previous interactions, in which we had shared positive views on the BFP and other typically Leftist policies, Joana wanted to stress to me that this was an exception: it was just a small 'fix' to this progressive policy that she had been waiting for, and was in fact already integrating into her daily work every time she told a beneficiary that no one really knew how long the programme would continue.

Resistance
Finally, the third type of translation is resistanceintentionally acting against new political directions when implementing policies. Different to counterbalance, which aims for a middle ground and to avoid polarisation, resistance aims at compensating for new or possible policy changes by pushing in the opposite direction as strongly as possible. In this case, a political shift can trigger frontline workers to make unprecedented use of their discretionary powers, taking a step they were not willing to take before, and actively seeking to undermine policy goals. For example, if policy guidelines are expected to veer to the Right soon, resisting frontline workers will start implementing Left-leaning measures as a precaution. Reconciling such a behaviour with their self-image of being a good civil servant, I heard multiple times, from different social workers, that they saw themselves as 'not just a bureaucrat', and that their acquired sensitivity to social issues and understanding of public services allowed them to 'go beyond the rules' and completely ignore or change some of the BFP's guidelines.
This attitude depends more on personal political beliefs than do resistance or collaboration. Individuals with clear political affiliations will be the first to overtly resist, since they will be less tolerant to policy change that reflects political views that oppose their own. Throughout my research, although all social workers had positive feelings towards the BFP, the majority did not identify strongly with Leftist politics and the Workers' Party and freely adopted punitive strategies when dealing with families they saw as 'not so poor', based on their preferences for a less generous and more conditional social policy. Those who openly identified as Left-wing resisted these practices, and preferred to let the system run its course, using the tools of the programme to justify opting not to use their discretionary powers when it came to excluding families from the programme.
It is, of course, possible that this process happens in an inverted order, in which the act of resisting an undesirable shift can change individuals' political preferences. Facing the destruction of the institutions they value for offering important societal services can cause frontline workers who identify as 'neutral' or 'not-so-political' to adopt a political position. In this context, peer influence is crucial, since organised opposition to policy change is likely to be started by workers who are politically active. In Angico, Bolsonaro's election had the effect of uniting the social workers in protecting the programme. Even if many among them had desired more stringent conditions for receiving benefits, Bolsonaro's known opposition to the BFP inspired social workers to systematically lecture beneficiaries about their rights, including making promises that social workers in Brazil would not allow the programme to be terminated, as reported by one of my informants who was working at Angico's welfare office. During the years I spent in Angico, only one social worker framed the BFP as a social right when dealing with beneficiaries, and she did so secretly, knowing she was the only one not using their discretionary power to make the programme sound stricter than it was.

Conclusion
At the beginning of this paper, I referred to Zacka's (2017) hypothesis that frontline workers' moral commitments to the way they perform their role change at a far slower pace than political change.
Drawing on empirical material, this paper proposes an analytical approach to addressing the concrete question that emerges from Zacka's reflection: How are contradictions between moral commitments and politics resolved by frontline workers? I believe the formulation I have outlined here offers a nuanced understanding of the interactions between political change and implementation practices, and is also open to the integration of different factors.
In the case of the BFP, I found strong resistance to progressive social policies among a relatively progressive social categorysocial workersduring a time of expanding social policies. In light of how events developed after my fieldwork, these findings inspire different questionsones that can be applied to other contexts, too: How will frontline workers react in a context where the political debate becomes dominated by moralising and punitive views on welfare? How will they handle increased resource scarcity and cuts to the number of recipients? Will social workers welcome such changes if they reflect their own ideas of what a more effective social policy should look like? Or, as poverty increases, will notions and practices of solidarity towards the poor grow? If so, what will these look like?
My main ambition with this paper is to fill a gap in policy implementation studies and generate insights into how political change affects frontline workers, with an emphasis on their reactions to the political scenario. Interpretive accounts can trace how political and institutional forces become compressed in policy and thus enable an analysis of the repercussions that ensue, such as how policy becomes altered in implementation through processes of translation. Such processes of implementation, I argue, are also part of the 'politics of distribution' (Ferguson, 2015) that define how cash transfers become viable political projects.
To that end, I have suggested three ideal types of translation practices that can guide this analysis. It is worth emphasising that these are not exclusive categories, and should not be interpreted or used as a predictive model. Instead, by fleshing out interactions of factors, it is my hope that the three types will be useful to systematise findings and facilitate comparisons, which will certainly include the identification of deviations from these types. The main contribution of this frameworkgoing beyond mainstream analysis of street-level bureaucraciesis to advance explanations that are usually compiled under 'intrinsic values', such as political preferences or affiliations. What I have developed in this paper is how these intrinsic values only tell part of the story: a contextual and relational approach to translation practices reveals the importance of other factors, such as awareness of roles in shaping policy, state ideologies and imaginations, technocratic or political understanding of frontline work, perceived political polarisation and proximity to divergent political ideologies. Together, these factors complement the socio-economic and organisational settings that are traditionally studied in implementation policies, and I argue that, because they have largely been ignored, the field has remained at the margins of research on political change and polarisation that has recently been growing exponentially.
This approach also aims to rescue the political aspect of implementation practices, where translation means choosing between alternatives (Freeman, 2009), informed by relations beyond the policy sphere. Not forgetting that frontline social workers are often the main or only face of the state that people interact with, the way they react to political change will shape the way the state itself is experienced because, as Abrams posits, the state is 'an ideological artefact attributing unity, morality and independence to the disunited, amoral and dependent workings of the practice of government' (1988, p. 81).
Finally, even for those who resist the use of the term 'political polarisation'or even the idea itself it is undeniable that the last decade has witnessed an increase in the frequency of sharp political shifts in different national and regional democratic contexts. To ignore the commonalities of these changes and their effects on frontline workers is, I believe, a missed opportunity. The rise of far-Right parties to power, the proposals for a green restructuring of the economy, the fact that more and more countries are taking suggestions for universal basic income programmes seriously after the impact that COVID-19 has had on economiesthese are all opportunities to empirically study political change in a way that is, if not radical, unexpected and can go some way towards rupturing incremental change. The impact these ideas and political projects have on the individuals implementing the policies of this contested state cannot be ignored.