Bureaucracies Under Judicial Control? Relational Discretion in the Implementation of Immigration Detention in Swiss Cantons

Based on interviews with bureaucrats and judges in several Swiss cantons, this article analyzes how bureaucrats decide to order immigration detention and how the judicial review shapes their decisions. The authors argue that discretionary decision-making regarding immigration detention is structured by the web of relationships in which decision-makers are embedded and affected by the practices of other street-level actors. The varying cantonal configurations result in heterogenous bureaucratic practices that affect the profiles and numbers of persons being detained. In particular, differences in judges’ interpretation of legal principles, as well as in their expectations, strongly affect bureaucratic decisions.


Introduction
The detention of noncitizens and asylum seekers for reasons of immigration law enforcement is a coercive measure that is widely used by states to control irregular migration and to facilitate the enforcement of deportations (Bosworth & Turnbull, 2015). Despite the increasing body of literature on immigration detention, we still know very little about how immigration judges or bureaucrats make their decisions to detain noncitizens pending their deportation (Ryo, 2016, p. 117). While immigration detention is mandatory in certain countries, 1 the existing studies on European countries highlighted that these decisions to detain involve significant discretion (Campesi & Fabini, 2019;Vallbé et al., 2019;Weber, 2003) and that multiple discretionary powers of several actors are involved in such decision-making (Hiemstra, 2019;Pratt, 1999). Hence, detention decisions are influenced by the practices of other street-level actors who intervene in the "networked policing" of internal borders (Mutsaers, 2014), including judges involved with judicial review (Hertogh & Halliday, 2004).
This article studies the case of immigration detention decision-making in Switzerland. In this country, according to the Foreign Nationals and Integration Act (FNIA), immigration detention is a coercive measure aimed at facilitating the enforcement of removal of noncitizens who are not allowed to stay on the national territory. Immigration detention is implemented at the cantonal level 2 by bureaucrats of the cantonal Migration Services. Decisions to detain are discretionary and are judicially reviewed by a judge in a lower court or a state court. Based on in-depth interviews with cantonal immigration bureaucrats and judges in several Swiss cantons, this article analyzes how bureaucrats decide to order administrative detention and how the judicial review influences their decisions. In doing so, this article provides new insights not only into the understanding of immigration detention in practice but also into administrative decision-making and the role of the judiciary in shaping bureaucratic practices.
Street-level policy implementation is a crucial level of policymaking (Infantino, 2019a) because of the discretion street-level bureaucrats (SLBs) have in enforcing rules and policies (Evans & Hupe, 2020;Lipsky, 2010;Portillo, 2012). Discretion is a classical issue of street-level bureaucracy research (Portillo & Rudes, 2014) and of sociolegal studies (Hawkins, 1992). These bodies of literature challenged the conventional juridical views of discretion, which focus on rules and assume a dichotomy between law and discretion (Tata, 2007). They offer alternatives by substituting the question regarding the formal definition of and limits to discretion by focusing on how bureaucrats and judges use discretion in practice (Mascini, 2020, p. 127), and on the social environment in which their decisions are embedded (Hawkins, 2003). Thus, recent studies have focused on state agents' practices, how they are shaped by organizational mediations and constraints (Miaz, 2017(Miaz, , 2019, bureaucratic habitus (Affolter, 2021;Alpes & Spire, 2014), and the organization of decision-making processes.
Considering discretion in terms of decision-making (Infantino, 2019b), this article aims to contribute to this discussion by studying the case of immigration detention decision-making in Swiss cantons. It does so by combining street-level and sociolegal approaches to discretion with a relational sociological perspective on bureaucratic practices, considering that street-level implementation is structured by the systems of relationships (Dubois, 2012) in which decision-makers are embedded. We argue that discretionary decision-making regarding immigration detention is affected (i.e., constrained and oriented) by the practices and logics of other actors with whom immigration bureaucrats are in a relationship of mutual dependence. These actors include police officers and case managers of Migration Services, but also the judges controlling their decisions. Therefore, we suggest using the term relational discretion to refer to the influence of relationships that shape discretionary decision-making. This article shows that the variations of these relationships from one canton to another also result in variances in bureaucratic practices, which in turn affect the profiles and numbers of persons being detained. Furthermore, this article provides new insights into how the judicial control of decisions to order administrative detention influences bureaucratic decision-making practices and discretion. We find that judges' interpretation of legal principles, their expectations, and criteria vary and strongly affect bureaucratic decisions, as well as bureaucrats' anticipations and strategies regarding judicial review. Hence, in the continuity of street-level bureaucracy research (Lipsky, 2010;Portillo & Rudes, 2014), we argue that these bureaucrats are policymakers with the other street-level actors with whom they relationally implement legal rules and who influence-sometimes by anticipation-their decision-making practices.

Immigration Detention Decision-Making
Immigration detention is an instrument of the deportation regime (De Genova & Peutz, 2010) that formally aims to hold people who have received a removal decision until their deportation can be executed and to prevent them from absconding (Bosworth & Turnbull, 2015). Immigration detention is not formally a punishment, but the important body of literature on immigration detention has shown that, beyond its formal function (facilitating deportations), this coercive measure fulfills deterrent, disciplinary, and punitive functions (Leerkes & Broeders, 2010;Majcher & De Senarclens, 2014;Rezzonico, 2020). Thus, immigration detention also shows characteristics of preventive measures, pointing to "the shifting boundaries between the 'penal' and the 'preventive state'" (Campesi, 2020, p. 16).
Even if immigration detention has acquired increased attention since the turn of the century (Bosworth & Turnbull, 2015), very few studies have analyzed how judges and bureaucrats use their discretion and how they make detention decisions. Notable exceptions (Campesi & Fabini, 2019;Ryo, 2016;Vallbé et al., 2019;Weber, 2003) provided interesting insights into immigration detention decision-making processes in Italy, the United States, Spain and the United Kingdom. They highlighted that the person's criminal past, the perceived risk of absconding or of the social dangerousness of "undesirable" migrants (Campesi & Fabini, 2019), and the time since a deportation order was issued all constitute decisive elements that decisionmakers consider "cues" to order detention (Vallbé et al., 2019). These studies also show that detention decisions involve a certain degree of uncertainty (Vallbé et al., 2019) and discretion (Weber, 2003). Building on these studies, this article offers new insights by exploring the Swiss case with a federal structure, thus allowing comparisons between subnational entities. It studies both bureaucrats who order detention and the judges who control their decisions in a street-level bureaucracy approach.

Relational Approach to Discretionary Decision-Making in the Street-Level Implementation of Legal Dispositions
Even if they do not necessarily interact personally with the concerned noncitizens, decision-makers who order or control administrative detention can be analyzed as SLBs because they have the power to translate legal rules and public policies into concrete practices and also because they make decisions about the fate of these noncitizens that structure and delimit their lives (Lipsky, 2010). In his seminal book, Lipsky (2010) argued that SLBs have a policymaking role that is built upon their relatively high degrees of discretion and their relative autonomy from organizational authority. Thus, to paraphrase Brodkin (2020), Migration Services' and bureaucrats' discretionary practices may be considered political because they operationalize immigration detention in specific ways that have a bearing on which, and how many people are detained, according to which criteria, and for how long.
As an inherent feature of policy implementation, discretion "remains at the forefront of socio-legal scholarship on SLBs" (Portillo & Rudes, 2014, p. 325), but with several disciplinary perspectives and definitions, as well as debates about how discretion should be studied and if it should be considered at all (Evans & Hupe, 2020). As a starting point, one of the most widely used is Davis's (1969) definition of discretion. To him, "A public officer has discretion wherever the effective limits on his power leave him free to make a choice among possible courses of action and inaction" (p. 4). In the same juridical perspective, Dworkin (1977) famously put that discretion "like a hole in a doughnut, does not exist except as an area left open by a surrounding belt of restriction" (p. 31). In this juridical view, discretion is considered a problem as it may threaten the rule of law (Mascini, 2020).
Criticism of such conventional juridical views that assume a rules-discretion dichotomy has resulted in building alternative sociological and sociolegal approaches to discretion that focus on the uses of discretionary powers. First, in a sociological perspective, according to Bourdieu (1990), bureaucrats have the power to interpret rules and to choose to stick to the rules or to grant exceptions, between rule following and rule bending (Portillo, 2012). To this alternative, Lascoumes and Le Bourhis (1996) add a third option, which they call "legal channels" (passes du droit) that refer to the room for maneuver and interpretation that rules may offer to those who apply them. In this sense, the law offers "channels" to play with the rules while still remaining within the law.
Second, in a sociolegal perspective, discretion is considered inevitable to the legal order due to "the translation of rules into action, the process by which abstraction becomes actuality" (Hawkins, 1992, p. 11). Discretion is inherent in the use of rules because making sense of them and making choices about their relevance involve interpretative work (Hawkins, 1992). Building on Hawkins's perspective, Mascini (2020) adds that "decisions can only be understood by reference to their embeddedness in interpretative practices, frames, fields and surrounds," and that "a connection needs to be forged between the interpretive processes that individuals engage in when deciding a particular case, and forces in the decision-making environment" (p. 130). Thus, in this article, the exercise of discretion is not seen as a "free choice" within the framework and limits set by legal rules because "individual 'free choice' may be already, collective, ordered, routinised and structured by phenomena other than the law itself" (Campbell, 1999, p. 80). Hence, we argue that discretionary practices must be analyzed considering their legal but also social, political, and organizational environment (Achermann, 2021;Buffat, 2015), as well as the social mechanisms that guide and constrain individual behaviors and perceptions (Miaz, 2017(Miaz, , 2019. Combining sociolegal and sociological perspectives on street-level discretionary decision-making, we "think of discretion in terms of decision-making-the power to make decisions-that is inherent to the law enforcement and policy application process" (Infantino, 2019b, p. 7), and we consider that decisions are determined by individuals' bureaucratic habitus (Alpes & Spire, 2014). This habitus is shaped by the continuous formal and informal organizational socialization of bureaucrats (Borrelli, 2021;Oberfield, 2012), which is itself influenced by repeated interactions and relationships with other actors involved in the decision-making process (Halliday et al., 2009). In addition, bureaucrats' practices are oriented and constrained by various social and organizational mediations (Miaz, 2019), as well as the web of relationships (Dubois, 2012) and accountabilities (Brodkin, 2008) in which decisionmakers are embedded.
In this article, we focus on this relational dimension of the shaping of discretionary decision-making. Some studies have considered that SLBs respond to multiple relationships and that they "work in networked environments where the diversity of their settings shapes their discretion and accountability to organizational goals" (Portillo & Rudes, 2014, p. 325). Thus, Halliday et al. (2009) have shown that interprofessional relations influence the type of street-level behavior. Consequently, SLBs' decisions are constrained by the interpretative work and decisions of other actors involved in the decisionmaking process (Mascini, 2020;Tata, 2007). This is the case in the context of immigration detention, in which discretionary powers are multiple, diverse, and dispersed (Pratt, 1999).
To make sense of the web of relationships composed of a wide range of public and nonstate actors, we mobilize the notion of interdependence, inspired by Norbert Elias's (1978) concept of figuration that "puts the problem of human interdependencies into the very heart of sociological theory" (p. 134). According to Elias, individuals are embedded in chains of interdependencies, implying that individual actions are dependent on each other. Based on this premise, we take a relational approach to understand how bureaucrats' discretionary decisions are made in interaction with other actors', especially judges', practices. We contribute to the literature on decision-making and discretion by showing how relationships affect decision-making and limit or orient the use of discretion, leading to the varying implementation of the same rules in different subnational contexts.
In exploring this web of relationships, we especially analyze how judicial control affects bureaucratic decision-making. In the social sciences literature on migration policies, this issue has not been properly explored so far (Mascia, 2020). Some studies have analyzed the impact of judicialization on immigration lawmaking (Bonjour, 2016). They show that repeated engagement with the courts can shape how political actors approach immigration issues and that the courts redefine bureaucrats' perception of what is lawful, which leads them to adapt their practices (Mascia, 2020). Judicialization of immigration and asylum policies may also involve strategic moves and mutual anticipations between the administration and the legal defense of migrants, both trying to influence and change case law (Kawar & Miaz, 2021;Miaz, 2021). In this article, we analyze the impact of judicial review on bureaucratic practices by showing how varying levels of "strictness" and expectations of judicial review differently shape bureaucrats' use of immigration detention.

Methods and Data
This article results from an empirical study on immigration detention decision-making in Switzerland. 3 The data primarily consist of 15 semi-structured expert interviews conducted in 2017 and 2018 with 20 cantonal immigration bureaucrats in 11 (out of 26) Swiss cantons. In addition, we conducted five expert interviews with six judges in courts responsible for controlling immigration detention orders in four cantons, as well as one complementary interview with a private attorney. During our interviews, which lasted for 1 to 3 hr, we discussed their everyday work, the decision-making process, their criteria and priorities, the judicial review process and how it affects their work, their profile and training, general questions about the cantonal context, and their role perception. In addition to these interviews, we analyzed 22 decisions provided by the courts in which we conducted interviews. We also included federal and cantonal jurisprudence on coercive measures in migration law. 4 The selection of the 11 cantons was based on the following criteria, considering the available statistics (C. Guggisberg et al., 2017). First, we chose cantons where one member of our team had conducted fieldwork in detention centers. Second, we selected the three cantons that issued the most detention orders (in absolute numbers) between January 2011 and September 2017. Third, we selected the cantons with the highest and the ones with the lowest ratios of detention orders per capita. Finally, we added three "average" cases. This selection furthermore covers a variety of cantonal situations in terms of language, size, and political orientation. This cantonal diversity allowed us to qualitatively analyze the relational dimensions of discretionary decision-making. All interviews were transcribed 5 and thematically coded. To protect our interview partners, we have anonymized the cantons in which we conducted our research. 6

Contextual Background
According to Swiss law, immigration detention is a coercive measure aimed at facilitating the enforcement of deportation orders. It can concern persons who have received a decision of removal or Dublin transfer 7 in the context of asylum procedures, but also other categories of noncitizens who are not permitted to stay in Switzerland. Different types of immigration detention exist and can cumulate up to 18 months. This kind of deprivation of liberty is not a penal sanction but is intended to enforce administrative decisions about noncitizens' removal. This measure is regulated in articles 75 to 82 FNIA. Different situations can lead to detention under these articles, such as refusal to disclose identity, submission of several applications for asylum using various identities, threat to public order or to others' life and security, conviction of a felony, lack of cooperation, and "insubordination." In practice, the two main grounds for the decision to detain someone under Swiss law concern threats to security and public order and the risk of absconding, revealing parallels to what has been observed in other countries (Campesi & Fabini, 2019;Ryo, 2016).
With Switzerland being a federal state, the cantons are responsible for the enforcement of deportation orders. Consequently, bureaucrats working in cantonal Migration Services 8 are responsible for implementing federal rules on immigration detention. These decision-makers have different positions and profiles according to the canton. They may be lawyers in certain cantons whose main task is to decide whether to order immigration detention, or they may be (deputy) heads of section or division (middle-managers) for whom ordering detention is a task, among many others. Following the detention order, "The legality and the appropriateness of detention must be reviewed at the latest within 96 hours by a judicial authority on the basis of an oral hearing" (art. 80, al. 2 FNIA). In particular, this includes assessing whether detention is proportional and whether deportation can be enforced promptly. 9 However, for detention under the Dublin procedure, the detention order is only reviewed by a judicial authority if the detainee requests it (art. 76a FNIA). Finally, detainees can request a judicial review of their detention conditions or request their liberation.
As presented more extensively elsewhere (C. , between January 1, 2011, and September 30, 2017, there have been 5,823 detention orders per year. Two thirds of the detainees (65%) had applied for asylum at some point since they entered Switzerland, and the average duration of immigration detention was 22 days. 10 However, Hidden behind these overall figures is a wide array of diverging cantonal practices in terms of the frequency of enforcement of detention, the average duration of detention, the proportion of deported detainees, the type of detention used and the profile of the detainees. (C. Achermann et al., 2019, p. 2) Another study focused on the immigration detention of asylum seekers between 2011 and 2014. The authors analyzed the proportion of detained asylum seekers in each canton and revealed considerable cantonal differences, which may be related to differing understandings of the proportionality of detention by the cantonal authorities and courts (Guggisberg et al., 2017).
In addition, according to Guggisberg et al.'s report and our own data, there are significant variations between cantons concerning the nationalities of the persons detained, and the proportions of women, of minors, and of people who applied for asylum. These variations cannot be explained by differences in the cantonal asylum populations (Guggisberg et al., 2017). The following analysis will demonstrate that the relational discretionary decision-making, and especially the impact of judicial review on bureaucratic practices, partially explains these cantonal differences in the use of detention.

Street-Level Interdependencies: A Relational Analysis of Decision-Making
In the following, we first ask how and based on what criteria discretionary decisions on immigration detention are taken. In a second step, we focus on the relationships between cantonal bureaucracies and the courts to analyze how judicial control affects cantonal bureaucracies' use of immigration detention.

Deciding Immigration Detention: Processes and Criteria
How an immigration detention order is decided and by whom depends on the cantonal organizational processes and the way the legal criteria are implemented by the respective decision-makers. On one hand, bureaucrats can order the detention of a registered person for whom they handle the deportation process, and who has not followed a removal order. On the other hand, the police may arrest, for various reasons (identity check, crime, etc.), a person who is then identified as being unauthorized to stay. The police inform the Migration Service about the person, and then a removal order and detention decision may follow.
Immigration detention and the deportation process. According to our interviewees, immigration detention is the final step of a multistep process. A removal decision is made either by the State Secretariat for Migration (SEM) at the federal level in the case of rejected asylum seekers or by the cantonal administration for all other people unauthorized to (continue to) stay. The decision becomes enforceable with a deadline for departure. During the deportation process, the noncitizens are informed in a "departure interview" about "return assistance." If there is no "voluntary return" (de Senarclens & Soysüren, 2017) after the departure deadline or even after follow-up interviews, case managers proceed to one of four levels of forced deportation. 11 During the interview to prepare the removal, case managers try to assess whether the person will "cooperate" or resist their deportation. They also try to assess a "risk of absconding" to determine whether the person will be present for the deportation flight.
Immigration detention takes place within this deportation process, and the case managers responsible for organizing the deportation may decide in favor of immigration detention to ensure deportation. As an alternative to detention, cantons might order other coercive measures such as house arrests, perimeter bans, or territory assignments. According to the canton, the final decision on whom to detain might heavily depend on the work of other bureaucrats, such as case managers, the police, or judges.
Assessing detention grounds: Dublin transfers, risk of absconding, and criminality. Immigration detention is used to keep the person at the administration's disposal for the identification process (e.g., to obtain a "laissez-passer" from the embassy of the country of origin), or for the case of an accompanied, unaccompanied, or forced departure flight. The following legal criteria are usually considered when deciding to detain someone: risk of absconding, criminality (related to a threat to security and public order), foreseeability and promptness of the removal, and proportionality of the detention. The first three elements are grounds for ordering detention, whereas proportionality may lead authorities to refrain from it.
The "risk of absconding" may be established due to what the person says during a departure interview-for example, affirming that he or she will never leave Switzerland. It can also be due to what the person does-for instance, if someone did not attend an identification hearing at the embassy of his or her country of origin, or an appointment with the Migration Service. "There are various elements that we consider [to assess the risk of absconding]" and that can combine with other motives for detention, like criminal records, because "one does not always have only one motive for detention, but often several." 12 Certain criteria, such as a risk of absconding, "criminal cases" and "Dublin cases," may be prioritized by the political authorities, by the bureaucratic hierarchy, or by the decision-makers themselves. We find different priorities and hierarchies of criteria between cantons. In many of them, the priority is to enforce Dublin transfers because there are financial sanctions if they are not enforced within 6 months, which would result in Switzerland becoming responsible for examining the respective asylum demands. As a matter of fact, the SEM, which is responsible for asylum and migration policies at the federal level, can order financial sanctions against the cantons that do not enforce Dublin transfers on time. In Canton-J, the decision-maker explains that he must use immigration detention to save the canton money. He explains, "[the Dublin cases], that is really the priority for the canton! . . . Because, if we don't return them on time, we'll be penalized by the Confederation." 13 Thus, federal control, sanctions, and accountability become particularly important and "steer" cantonal discretionary practices (Brodkin, 2008). These deterrents reinforce a strict logic of deportation and promote more frequent use of Dublin detention, which is not automatically subject to judicial review.
In other cantons, according to our interviews, the priority is to detain "criminal cases," like in Canton-E and Canton-G where social movements and politicians have contested immigration detention. As a consequence, in these two cantons, bureaucratic guidelines and implementation laws define restrictive criteria for using administrative detention. In the Parliament of Canton-G, a politician recalls this situation: This is not the first time that the [State Secretariat for Migration] has reminded the [Canton-G's] authorities of their lack of zeal in returning rejected asylum seekers. There is indeed a strong tradition of support for asylum seekers and the reception of migrants in the [Canton-G], and this has been the case for several decades . . . This particularity has led to a more moderate policy on the part of the cantonal government towards rejected asylum seekers. It has contributed to slowing down deportations. 14 In Canton-E as well, social movements and politicians pushed for restrictive criteria allowing to order administrative detention. The Head of the Migration Service explains that Political injunctions [to restrict the use of administrative detention] came from the fact that, when the coercive measures were discussed at the federal level, there was a big outcry, from almost all the parties in [Canton-E] who said: "We won't apply [these measures], well we'll do everything to make it complicated." 15 As a consequence, the cantonal implementation law led to the prioritization of criminal and Dublin cases, when there are deadlines or when there is a threat to public order: "There is still this feeling in [Canton-E] that people should not be detained if they have not committed an offence." Hence, when they have discussions with associations defending foreigners' rights, they "always talk about this point" because certain social movements try "to empty this application law of its very meaning." 16 This Head of a Migration Service explains well that the political contextwith a strong political opposition against coercive measures by the cantonal parliament and by the solidarity movements-constrained them to limit their use of immigration detention to people with criminal records, to those who pose a threat to public order, and to persons with a decision of Dublin transfer. Moreover, the procedure to order administrative detention is complex and the requirements coming from the judicial control of detention decisions are particularly high.
Thus, in Canton-E and Canton-G, bureaucratic directives set a list of priorities to guide the use of immigration detention. In both of them, the priority is to detain people with criminal records and then those for Dublin transfer. 17 If there are not enough detention places available (logistics also play a role in the decision-making process), bureaucrats assess the "emergency" and the "realistic" prospect of the deportation to determine whether they will rather detain a person with a criminal record or someone subject to a Dublin transfer. This evaluation is mostly carried out based on the nationalities of the persons: "We know that, with certain nationalities, it's more complicated and with other ones, it's easier," especially according to the bilateral "readmission agreements": If a person who is, typically, a national of Eritrea, and who has a criminal record, for whom we have no paper, well we know that he is a national, but we won't be able to deport him . . . indeed [sighs], this person will not fit into this prioritization. Actually, the prioritization is still for what we call the deportable persons. 18 Thus, beyond the risk of absconding and criminal records, the foreseeability of the removal and the threat to security and public order are considered. Their assessment strongly depends on the country of origin and the possibility of obtaining a "laissez-passer" and of organizing a forced return. Besides, gender-these two cantons detain very few women-and the perceived vulnerability also play a role in the decision to order administrative detention. Because of the political context of these two cantons, decision-makers are "cautious" when they decide on immigration detention. In Canton-G, alternatives to detention, such as house arrest, are given preference. This is reflected in the canton having a very low detention rate.
Other cantons have no formal criteria beyond the law, and immigration detention is frequently used based on the argument of a risk of absconding or criminality. In Canton-D, two members of the Migration Service explained that their main criterion for detaining a person is the risk of absconding and that it intersects with other criteria, especially criminal convictions, the rapidity and predictability of the deportation process, and the proportionality of the detention, which can slightly limit the use of immigration detention. However, this discourse must be understood as bureaucrats strategically presenting arguments to enhance their chances of getting their decision through a judicial review. Thus, when bureaucrats have the arguments to decide in favor of detention and defend it vis-à-vis the court, they generally do decide in favor of immigration detention. From the perspective of our interviewees, immigration detention increases the likelihood that the removal will finally be enforced: "It is certainly easier to enforce [the removal] from the detention center than to pick up the persons by surprise [at their place of residence]." Detaining someone is considered less risky and ensures that the person is at the disposal of the administration and the police during the preparation for their deportation. However, this does not mean that those detained always agree to leave "voluntarily" and to get on a flight without resistance, "but the chance of successful enforcement is higher if the person is previously in detention." 19 In most of the cantons, nationality plays a role in the decisions to detain. As previously explained, people of certain nationalities, like Eritrea, are not a priority to be detained because, as there are neither readmission agreements nor forced return flights to these countries, their nationals are not considered "detainable" because their (forced) deportation is assessed as impossible. On the contrary, other nationalities are more "detainable" (De Genova, 2017) because of the foreseeability of their deportation and because of the stereotypes according to which nationals of these countries are involved in criminal activities (e.g., drug trafficking). A member of the Migration Service of Canton-H explains that there is no routine pattern or mechanic model for detention decision-making. However, he says, there are more or less obvious cases, referring to a classification based on nationality, combined with other criteria: There are simply categories of cases, I would say, that are clearer. Nigeria, for example. A young Nigerian who has been in Switzerland for a long time and has perhaps also committed a crime and absconded, several times . . . These are then cases where you don't have to think much more, these are then clear cases for detention. 20 The foreseeability of the deportation is an important criterion that can, however, be outweighed by the one of public safety. Indeed, certain nationals, mainly from North African countries, are considered by several cantons as posing threats to public order. A decision-maker explains that, following a "political decision" in his canton, they order administrative detention for Algerian nationals posing security problems even if forced removals cannot be enforced: It's . . . well . . ., a population that poses problems for us and it's true that if we don't take measures against them . . . well for them . . . they are aware of it and they take advantage of it. 21 In this case, the use of coercive measures such as administrative detention, house arrests, or perimeter bans are used as a tool of deterrence from staying in Switzerland and to incite the persons to leave the territory. Hence, authorities put "two or three who deserved it" in administrative detention-"not too many, because otherwise it would block too many places" 22 -and when they are released (sometimes after 17 or 18 months of detention), they order other coercive measures to deter them from staying in Switzerland: So, what we've done now more and more is that when we release them like that, we notify them a territorial assignment. So that, I think, they don't like it very much. That is to say that we can assign them to the territory of a commune . . . and then every time they leave that [commune], they are [arrested]. The goal is to disgust them and make them leave by themselves. 23 This analysis shows that there is an uneven distribution of detainability (De Genova, 2017), that is, a greater or lesser susceptibility to detention, that varies from one canton to another. It also appears that beyond the assessment of a risk of absconding, of criminal records, of a threat to public order, and of the foreseeability of the removal, this uneven distribution of detainability varies according to gender, the assessment of perceived vulnerability, and nationality (see also De Genova, 2017; Hiemstra, 2019).

Immigration detention and police work. Immigration detention decisionmaking processes may also be influenced by police work providing decision-makers with cases. Canton-E and Canton-F are interesting examples in this regard.
In the small canton of Canton-F, only one person decides who will be detained. He explained that he quasi-systematically detains people to be transferred to another Dublin state. He has no additional criteria or priorities; he simply implements coercive measures when he feels the law tells him he must. This has led to an intensive use of immigration detention, considering the canton's population size (C. Guggisberg et al., 2017).
Moreover, Canton-F has a significant workload because the railway line from Italy further North passes through this canton. Thus, the railway police arrest irregularly traveling migrants on the trains and the cantonal migration authorities must decide what to do with them. Overall, the high proportion of detention in this canton can be explained by the combination of the significant workload related to railway police arrests and a quasi-systematic use of immigration detention related to the legal rigor of the decision-maker in this canton. 24 In Canton-E, immigration detention appears to be a continuation of police work. The six lawyers who decide on detention are formally attached to the Migration Service but work in the same building as different actors in the penal chain. 25 Moreover, although the bureaucrats prepare the decisions on immigration detention, these latter are signed by a police commissioner. This illustrates the relational character of the decision-making and its embeddedness within the police and penal chain. During one of our interviews, an immigration bureaucrat explained that their first task each morning is to check the list of persons who were arrested the night before. Thus, their work is mainly paced by police arrests. 26 * This section explained that the differences in the discretionary use of immigration detention in each canton are related to different interpretations and prioritizations of multiple legal criteria. If bureaucrats must legally argue that there is motive to justify immigration detention (Campesi & Fabini, 2019), their priorities rest on specific dimensions, such as criminal records, Dublin removal, and the foreseeability of deportation.
Moreover, the decision-making process is influenced, to different degrees, by the work of other street-level organizations and frontline workers, such as the police and case managers, who report individuals on whom migration bureaucrats have to decide. Hence, "The ultimate decision to detain-or release-is not a singular event. The 'case' is constructed over time and is contributed to by a myriad of discretionary decisions made by different agents at different points in the process" (Pratt, 1999, p. 218). Decision-makers' interdependent relationships with these actors vary from one canton to another but directly shape the workload of immigration bureaucracies and the way they use their discretion regarding immigration detention. This implies that, while initially aimed at enforcing removals, cantonal use of immigration detention is also marked by other rationales (see C. Achermann, 2021), such as reducing expenses or fighting criminality, mixing both punitive (sanctioning illegality of stay) and preventive (crime control) logic (Campesi, 2020;Campesi & Fabini, 2019;Rezzonico, 2020).

From Bureaucracy to Courts: Judicial Reviews and Their Anticipations
In this second section, we focus on one set of actors within this web of relational decision-making: the courts in charge of judicial review of immigration detention and how they interact with bureaucracies. Judicial review is significant in understanding the shaping of discretionary decision-making and the cantonal differences. The type of court reviewing the bureaucratic decisions and the number of appeal instances vary from one canton to another. Judicial review affects bureaucratic discretionary decision-making, depending on its "strictness" (in the sense of high expectations and criteria) or, to the contrary, on the alignment of the court with bureaucratic practices. Judicial review requires bureaucrats to give factual and legal justifications (Mascia, 2020), but also leads them to anticipate court judgments to pass their decision or to develop strategies to avoid judicial control.
Our interviews with bureaucrats and judges reveal that there are different bureaucratic strategies to cope with the judicial review, different relationships between bureaucracies and the courts, and diverse judicial practices and cantonal jurisprudence. This results in varying interpretations of the principles of promptness and proportionality, as well as of the risk of absconding. While certain bureaucrats explained that they almost systematically detain persons when conditions allow them to, others explained that they look at the proportionality of detention very carefully. The specificities of the cantonal judicial review processes partly explain these differences of practice.
Judicial control as a "formality." In several cantons, the judicial review is considered a "formality" by the administrations since very few (if any) decisions are challenged by the court. One judge confirmed this view and told us: "I have never broken a decision of immigration detention . . . I never refused to approve [a decision of immigration detention]. The cases are dealt with another way." 27 In his view, it is up to the Migration Service to decide whether to liberate a person or not. If new information appears that results in the detention being considered inappropriate, unlawful, or disproportionate, the administration generally has, according to this judge, the "common sense" to release the person without the court's intervention.
Judges of that mindset do not set high requirements for the administration to confirm a detention order. Compared with other cantons, their interpretation of the principles of proportionality and promptness is less constraining for the administration. For them, only a few elements could make the detention disproportionate: Very few things because this is not a penal detention. There is more liberty . . . The [detention] conditions are rather acceptable . . ., they are even very acceptable. And then, . . . there is also a question of equality of treatment. Those who accept to comply with their enforcement of removal are fooled [if we don't detain those who refuse to leave the country]. 28 Thus, to these judges, administrative detention is rarely disproportionate because the conditions of detention differ, according to them, from penal detention. 29 Moreover, they consider that not ordering immigration detention for those who do not comply with a removal order would be "unfair" vis-àvis those who do. This idea of "equality of treatment" indicates a moral dimension of administrative detention decision-making: Detention is considered as a "fair," or even necessary, coercive measure for those who disobeyed a removal order. From the point of view of Canton-K's Migration Service, administration and judges are aligned with each other and agree on the criteria and evaluation of cases: Usually, I would say, we understand each other; we are more or less on the same wavelength. We even have good practice. The judges . . . We know the people, we know how they work too. There is good stability of the justice in [Canton-K], which is an advantage for us, for the administration. 30 The relationship between the judges and the administration seems to be cooperative and characterized by proximity, resulting in the judges systematically confirming administrative decisions. Similar tendencies are reported by other cantons where bureaucrats mention that their decisions are hardly ever canceled by the court, like in Canton-A or Canton-D. 31 In Canton-F, the judicial review is not considered a problem at all: With the court, it works relatively well; we have a good understanding. They have rarely-practically never-released a person from prison. But also, nowadays, [there are] practically no more cases. We used to go to court once or twice a month, but now it's only once or twice a year because it's mostly Dublin cases, or we can [enforce the removal] within 96 hours, so they don't have to pass through court. 32 Therefore, in this canton, most of the bureaucracy's detention decisions are not controlled at all. According to the interviewee from the Migration Service, the small number of court proceedings reduces the court's effort and burden.
This kind of alignment between the court and administrative practices leaves bureaucrats with great room for maneuver and limited accountability in their use of immigration detention. When immigration detention is widely used, this "light control" raises concerns regarding the assessment of the legality and appropriateness of this form of deprivation of liberty. Divergent interpretations among the cantons mostly seem to concern the principles of promptness and proportionality-something we also observed in cantonal and Federal Supreme Court decisions. This situation is exacerbated by the fact that, for most of these judges, migration law-and especially law related to coercive measures-is not their specialty and is a rather neglected task in the court.
Judicial review, learning, and case selection. Cantonal courts have developed a vast jurisprudence and specific "practices" and requirements for the lower judicial authority reviewing immigration detention decisions. For some bureaucrats, the judicial review and the appeals help to specify certain legal notions. In Canton-A, where bureaucrats do not consider judicial review as being especially strict, our interviewees mention that, besides the law, the court "praxis" clarifies certain aspects and can serve as a guideline. For example, according to FNIA, the failure to respect a perimeter ban is a reason for detention. However, these bureaucrats explain that "the court practice in Canton-A is that it is a motive for detention only if it happens for the second time. So, the court practice actually defines the rough legal rules a bit more precisely for us." 33 Knowing the specific requirements and practices of "their" judges leads administrations to carefully select the people who are detained so that their decisions are not overturned by the court, especially if the number of detention places is limited. In Canton-J, the decision-maker explained that "we know our judges," meaning that he knows what kinds of requirements and questions the three judges have, "for example, because there are judges who insist on certain points, on the procedure. So, we must prepare the file a bit differently." He mentions a judge "who wants to know everything" and who asks for "every detail" of the case. So, "I need to be able to inform him." 34 Indeed, the judges of this canton recognize that the administration carefully selects the people for whom they order immigration detention. As a consequence, "very often, the files that are submitted by the [Migration Service] are put together very well." 35 In 2017, according to the statistics provided by the administration, only one third of all (99) detention orders were judicially reviewed and all but one were confirmed by the judges. Finally, in contrast to the former section and the cantons in which judicial review is considered a formality, these judges thoroughly review the possible alternatives to detention (house arrest, for example) and the proportionality in light of the specific circumstances of the detainee. They also verify the promptness and the details of the procedure to review the legality and appropriateness of the detention order.
Thus, the anticipation of judicial review may lead the administration to more carefully decide on the "cases" in which to order immigration detention. This anticipation rests on the internalization by the decision-makers of what a "good file" is, meaning a decision that can be confirmed by the judge, and what file has no chance of being confirmed. The long-term experience of judicial review and the long-term relationships between the bureaucrats and the judges increase bureaucrats' knowledge of judges' practices and requirements. Therefore, the professional and bureaucratic socialization of immigration bureaucrats (Affolter, 2021;Borrelli, 2021;Oberfield, 2012) shapes their bureaucratic habitus and the way immigration bureaucrats deal with the law (Alpes & Spire, 2014). Hence, how they learn how to do their job, and the political, social and legal categories they implement, result not only from internal institutional socialization (Miaz, 2017(Miaz, , 2019; rather, they also depend on the relationships and interactions with actors outside the organization, such as judges. This socialization through continual relationships and interactions shapes the ways migration bureaucrats perceive what is possible to decide and which arguments the court accepts to order detention. The "strictness" of judicial review and its anticipation. In certain cantons, immigration bureaucrats considered judicial review to be "very strict" and to have an important impact on their decision-making. According to them, judges set high requirements to confirm detention. Strict judicial review in a canton tends to restrain the use of detention in general or vis-à-vis specific groups (such as women, minors, or vulnerable persons). Statistics confirm this observation (Guggisberg et al., 2017).
In Canton-I, where the number and proportion of detentions are low, we observed strategies to avoid judicial review. In 2016 and 2017, for example, a total of only five out of 119 detention orders exceeded 72 hr and were judicially reviewed. 36 In this canton, the migration bureaucrats usually wait until the deportation is fully organized before they order the detention. This approach is presented as being "more humane" because detentions are shorter and "more economical" as the canton has limited detention places and usually uses them for just a few days. But the bureaucrats also explain that the judges in the canton have very strict conditions regarding the use of immigration detention. The bureaucrats anticipate this context in their decision-making, which results in a reluctant use of immigration detention: In the canton, we have judges who are, I would say, restrictive regarding the application of the law and the conditions that must be fulfilled to confirm the decision to order a coercive measure. So, it's clear that if we are not absolutely sure that all the conditions are fulfilled, we won't detain the person and risk not being supported, because it would be a waste of time for everybody. 37 In Canton-E, the judicial review is considered strict by bureaucrats. In 2015, the court overturned about one sixth of all (322) cases. 38 Thus, the court frequently cancels detention orders in relatively long and legally argued decisions-for example, referring to the promptness, to the proportionality, or to the proof of the risk of absconding.
Contrary to other cantons, for Dublin detention, the requirements to detain are higher, especially regarding the risk of absconding, for which the jurisprudence of the federal administrative court has set high requirements that the administration follows strictly: The simple fact that the person says that she doesn't want to leave Switzerland is not enough. The person really has to . . ., but it's precisely the jurisprudence that dictated that, the person can typically refuse [to get onto] a flight. That's something concrete, we're having a clue, . . . more than a clue! As I was saying about absconding, or the fact that [the person] actually absconded or that she's . . . that she disappeared, but it has to be recurrent. 39 The relatively low proportion of detained asylum seekers in Canton-E is to be understood as a result of the court's interpretation of a risk of absconding and its requirements and practice regarding this motive for detention that are particularly strict. Immigration bureaucrats must take this into account when they order immigration detention and when they prepare the file for the court. In some cases, they must explain to the "Return Brigade" or the case managers of the Migration Service who asked them to order immigration detention that the motives do not respect the law and the court's requirements-for example, when there is no criminal record and when the decision is only based on the fact that the person has not answered to a summons of the Migration Service twice or that she said, "I don't want to go back to my country": This is often still not enough to put someone in immigration detention. [Judges] believe that there is no element of absconding. I don't completely agree with them but it's the court, so here we go. So sometimes we ask [the officers of the Return Brigade] to investigate a little bit more, that they try to reconvene him, so that if he doesn't show up before the police, we may have some more arguments to put him in immigration detention. 40 This illustrates how, in Canton-E, the decision-makers depend both on the work of the officers of the "Return Brigade" and on the judicial review.
They play the role of intermediaries, considering the reasonings of the "Return Brigade" and of the judges when they implement the law. Between the two "partners," they develop a practical knowledge of the law and of the possible decision. This practical knowledge is also influenced by their training as lawyers.
One interviewee in Canton-E described how she carefully examines the question of proportionality of detention, including how long the person has been in Switzerland, whether they are "integrated," whether they work and have a family, whether they have committed crimes, whether they respect the legal order, and whether they have a home and means of subsistence. 41 These criteria to assess the proportionality of detention are higher than those mentioned in other cantons where, for instance, being integrated and having a job and a home are not considered as making the detention necessarily disproportionate. These arguments indicate who is considered to be (un)deserving of staying in liberty, who is considered a threat to public order (Campesi & Fabini, 2019;Ryo, 2016), and for whom it is both proportional and morally legitimate to order detention. An example given by an immigration bureaucrat refers to an Albanian citizen without a means of subsistence, without any family in Switzerland, and who sold heroin. In such a case, it is clearly proportional to order detention, he says, because the person endangers the security of the local people and does not have close ties to Switzerland. 42 Indeed, in Canton-E, between 2011 and 2017, there has been a significant proportion of immigration detention ordered for non-asylum-seeking foreign nationals with criminal records, mainly Albanians arrested for heroin trafficking. Considering the court's high requirements, such decisions appear to be easier to pass before the court: "Most of the time, the clues of absconding are [considered] not sufficient [enough by the court]. [On the contrary,] it's rare [that we] make a mistake if the person has been convicted." 43 Hence, the strictness of judicial review leads to different strategies and anticipations by decision-makers. In Canton-I, bureaucrats try to avoid judicial review and only order short detentions for cases where the removal is fully organized and enforceable within a very short period. In Canton-E, bureaucrats carefully select those they detain, giving priority to people with criminal records. As a result, decision-makers detain fewer asylum seekers, but a higher number of non-asylum-seeking noncitizens with criminal records. In both cases, bureaucrats anticipate judicial review and adapt their decisions accordingly.
When the judge is an expert. The jurisprudence, the judges' practice, and their attitude toward immigration detention can strongly shape the use of this coercive measure in a canton and can limit bureaucratic discretion in different ways. Canton-H is particularly noteworthy regarding judicial review. When we started analyzing the quantitative data, we were surprised to discover that Canton-H is one of the most reluctant cantons to order immigration detention (C. Guggisberg et al., 2017), despite being a canton of voters with conservative attitudes toward migration issues. We observed the same mechanisms of strict and demanding judicial review that is being anticipated by administrative decision-makers as in Canton-E and Canton-I. In contrast to these cantons, the specific role of one judge characterizes the situation in Canton-H.
He is a leading expert on coercive measures and has been involved in their introduction and implementation in Canton-H. He deeply influenced both the current administrative and judicial practices, as well as the cantonal implementation laws. In our interview with him, he expressed a deep feeling of responsibility to thoroughly and reliably verify whether a detention order is lawful and proportionate because it infringes upon the fundamental rights of a person (through deprivation of liberty). So, he even reviews Dublin detention decisions (which is a specificity of this canton) because "it's still custodial, isn't it?" Indeed, according to him, "As an immigration detention judge, you have an enormous responsibility" because detention is "actually the most radical thing you can do, from the state's point of view." 44 This judge has built a particular relationship with the administration, as he has judged most of the cases presented for two decades. He has worked with the administration to prepare the legislation on coercive measures and thus built "this special cooperation with the Migration Service" and explicitly trained them regarding the type and exhaustivity of information he requires. 45 As a result, the Migration Service usually only orders uncontroversial cases.
Migration bureaucrats confirm that many long-term employees have internalized the court's practice. The service has developed certain formal techniques to respond to the court's requirements. For instance, they use text modules that are continuously adapted to the recent case law and a "foureyed" decision-making process. Consequently, the detention orders are widely justified and legally argued, and the principles of proportionality and promptness are systematically and carefully examined. The discretionary decision-making process is thus strongly influenced by the strict review that the judge and the court exert on their decisions. 46 As a result, Canton-H is one of the cantons that make reluctant use of immigration detention. * This section has shown that relationships between bureaucracies and judicial reviews vary from one canton to another. In some cantons, bureaucrats say that their decisions are never (or very rarely) challenged by a judge. This can be due to the bureaucrats anticipating an expected outcome, to the fact that they mostly order Dublin detentions, or to the lower requirements of judges. In other cantons, the judicial requirements regarding the evidence and argumentation justifying a detention order-especially regarding the principles of proportionality and promptness or regarding proof of the risk of absconding-are high and can strongly limit the use of immigration detention. Thus, the requirements of the court reviewing the detention orders, as well as the jurisprudence of the cantonal State Court judging the appeals, shape the cantonal use of immigration detention. This mechanism works in two different ways: either the court does not support a detention order and the person is released or the bureaucracy anticipates the court's decision and restricts itself when deciding to avoid being overturned. While formally reviewing the legality and proportionality of the detention orders, judicial control also significantly impacts the profiles and numbers of persons detained. Eventually, it appears that the stronger and stricter a court is, the lower the ratio of detention orders.
Furthermore, it is interesting to note that, despite some rare moral judgments about the "undeserving-ness" of those who committed crimes like drug-trafficking, most of our interviewees explained their practices in a relatively "legalistic" way. This reflects, in our view, a "legalist ethos" (Miaz, 2017) that is probably related to their bureaucratic (Affolter, 2021;Alpes & Spire, 2014) or legal habitus (Bourdieu, 1987). This legalist attitude helps them to distance themselves from the moral and political dimension of their work (Rezzonico, 2020) and to justify their practices with legal arguments considered legitimate because they are democratically endorsed.

Discussion
The analysis of the varying cantonal use of immigration detention reveals that the different legal criteria do not carry the same weight and are not interpreted the same way in each canton. To understand these different practices, we combined street-level, sociological, and sociolegal approaches focusing on uses of discretionary powers (Bourdieu, 1990;Buffat, 2015;Hawkins, 1992Hawkins, , 2003Lascoumes & Le Bourhis, 1996;Mascini, 2020;Tata, 2007) and on the relations in which SLBs are embedded (Dubois, 2012;Elias, 1978;Halliday et al., 2009;Mutsaers, 2014). We analyzed the web of relationships in which the cantonal bureaucrats deciding to detain are embedded at the level of law implementation and how these relationships affect their decisions. We assert that the analysis of bureaucrats' discretionary practices must consider their profiles and bureaucratic habitus (Affolter, 2021;Alpes & Spire, 2014); their organizational socialization (Borrelli, 2021;Oberfield, 2012); their legal, social, political, and organizational environment (Achermann, 2021;Buffat, 2015); and the social mechanisms affecting individual behaviors and perceptions of what decision is possible to make in specific circumstances (Miaz, 2017(Miaz, , 2019. Therefore, we argue that discretionary decision-making is always relational, meaning that decisions are co-produced by several actors that together shape the final "policy outcome," as reflected, for instance, in the statistics on immigration detention. In this sense, we propose the notion of relational discretion to refer to the fact that discretionary power is not solely exercised by those who decide but is also dependent on other (street-level) actors who directly or indirectly influence the decisions. Thus, these relations contribute to determining how legal instruments (immigration detention) are operationalized at the street level and, ultimately, affect the broader public policies (Brodkin, 2020;Lipsky, 2010). In doing so, relations can shape the logics, functions, and meanings taken by these instruments and policies in practice. Hence, SLBs are not policymakers "alone"; they implement and make rules and public policy in relation to other street-level actors whose practices influence SLBs' policymaking role.
We have highlighted the interdependencies of decision-makers with case managers, police officers, or judges. Their varying types and properties in each canton may lead to different uses of immigration detention. This reliance on other SLBs implies that while they initially aim to enforce removals, immigration detention decisions are also marked by the treatment of other "public problems" and follow additional logics, such as the fight against criminality, confirming its punitive function (Campesi, 2020;Leerkes & Broeders, 2010).
Moreover, in line with the few existing studies (Hertogh & Halliday, 2004;Mascia, 2020), we have shown that the judicial review, the specific practices, and the jurisprudence of the courts strongly influence the interpretation of legal criteria and the bureaucratic discretion to detain someone. Judicial review can restrict decisions if it imposes high standards for the principles of proportionality and promptness, as well as in terms of evidence of a risk of absconding. In doing so, it favors a more careful interpretation and examination of these legal principles. Therefore, a more rigorous judicial review better protects detainees' rights.
However, judicial review can also broaden bureaucrats' discretion when the courts never challenge their decisions. Lens (2012) observed that judges and bureaucrats in welfare bureaucracies are, despite having different roles, part of the same apparatus and that judges' approaches may be shaped and defined by bureaucratic practices. She distinguishes between "bureaucratic" or "adjudicator" approaches, depending on whether judges use their discretion to replicate bureaucratic norms or to reinforce their role as adjudicators of disputes and as safeguards against arbitrary state action (Lens, 2012). Our study shows another relational effect: judges with stricter approaches regarding the assessment of the legality and appropriateness of detention also shape bureaucrats' decision-making practices, for example, through selecting "uncontroversial cases" or through using strategies to avoid a strict judicial review. Thus, varying strictness of judicial review and different cantonal configurations result in significant differences in the use of immigration detention, which importantly affects the fundamental rights of a group of people subjected to manifold exclusions (Rezzonico, 2020).