Shared professional logics amongst managers and bureaucrats in Brazilian social security: a street-level mixed-methods study

Abstract

Purpose

This research addresses the professional logics of street-level managers (SLMs) and bureaucrats (SLBs) working in the Brazilian National Social Security Agency (INSS) through their perceptions of distributive justice and discretion. Since SLMs have the authority to influence SLBs' actions, we investigate whether these two groups hold similar viewpoints.

Design/methodology/approach

We integrate the administrative data and survey responses (n = 678) with earlier thematic content analysis (n = 350) in three stages: mean-testing, regression analyses and complementary qualitative analysis, integrated through a mixed-methods matrix.

Findings

Whilst no significant differences emerge in distributive justice ideas between groups, SLMs demand wider benefit-granting discretion, praising professionalism whilst adopting managerial posture and jargon.

Research limitations/implications

The study adds to the theoretical discussions concerning SLM’s influence on SLB’s decision-making, suggesting that other factors outweigh it. The finding concerning the managers’ demand for wider discretion asks for further in-depth approaches.

Practical implications

Findings supply valuable insights for policymakers and managers steering administrative reforms, by questioning whether some roles SLMs play are limited to symbolic levels. Further, SLBs’ heterogenous formations might be more relevant to policy divergence than managerial influence and perhaps an underutilised source of innovation.

Originality/value

By approaching street-level management professional logics within a Global South welfare state through a mixed-methods approach, this study offers a holistic understanding of complex dynamics, providing novel insights for public sector management.

Citation

de Andrade, L.H.A. and Pekkola, E. (2024), "Shared professional logics amongst managers and bureaucrats in Brazilian social security: a street-level mixed-methods study", International Journal of Public Sector Management, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/IJPSM-08-2023-0240

Introduction

Welfare benefit-granting traditionally depends on street-level bureaucrats’ (SLBs; Lipsky, 2010) decisions, which can be influenced by their distributive justice perceptions (Blomberg et al., 2017) and by how they understand their role as professionals (Evans, 2011). However, these decisions are also affected by managerial values, penetrating street-level welfare work through shifts in organisations’ structures, procedures and performance criteria introduced by new public management (NPM) reforms (Aoki, 2019; Evetts, 2009; Møller and Hill, 2021). At the core, NPM-backed implementations strive to circumvent professional discretion on behalf of standardisation and efficiency (Hupe and van der Krogt, 2013; Aktas et al., 2023). However, these implementations are often carried out by professionals too, the “managing professionals” (Pekkola et al., 2018), who, on service frontlines, are known as street-level managers (SLMs; Gassner and Gofen, 2018). Whilst supposed to bear NPM-backed managerial logics, SLMs walk a tightrope between these and SLB professional logics, balancing their legitimacy before their teams against performance standards and accountability to upper management (Floyd and Woolridge, 1992; Noordegraaf and Steijn, 2013).

This article is needed for two main reasons: first, whilst recent literature explains how SLMs can influence SLBs’ discretion (Gassner and Gofen, 2018; Klemsdal et al., 2022), these effects can be restricted amongst professionals (Grøn, 2018). Yet, there is a lack of empirical accounts on whether SLMs, as managing professionals, do develop distinct professional logics from their fellow SLBs. These could indicate the direction of SLMs' influence on SLBs’ decision-making – decisive in policies targeting competing goals, the social security case. Second, the problem is even more relevant in large developing economies and its public organisations – such as the Brazilian National Social Security Institute (INSS). There, contested NPM-backed reforms found a heterogenous bureaucracy, lacking shared professional identities and legitimated discretion (Alcadipani and Caldas, 2012; Oliveira et al., 2010), whilst inequality, scarcity and institutional frailty demand more leeway for decision-making and improvisation (Aktas et al., 2023; Lotta et al., 2021).

The contribution is done by assessing the INSS SLMs’ and SLBs’ professional logics’ differences, focusing on their perceptions concerning distributive justice and discretion in the context of the social policies they operate, to answer two research questions: do INSS SLMs and SLBs share similar distributive justice reasonings and sense of discretionary space in social assistance benefit-granting? And how do these differences stand out in SLMs’ and SLBs’ reflections about social assistance benefit-granting? We draw on cross-sectional administrative and survey data (n = 678), which include deservingness perceptions scores and results from previous thematic content analysis (TCA) of open-ended survey answers (n = 350; authors, under review). The analysis comprises three stages: first, key variables means’ differences are tested. Second, found differences are screened through controlled regressions. Third, results are complemented with qualitative analysis integrated through a mixed-methods matrix. The findings entail practical insights into policymakers and managers steering administrative reforms, particularly concerning the role SLMs currently play and highlight the potential influence of SLBs’ formation heterogeneity in policy implementation.

The next sections are structured as follows: first, we examine the research context and conceptual underpinnings: SLMs, their influence on SLBs, distributive justice reasoning and discretionary space. Second, we explain our data and methods. Third, we report each analysis stage. Our findings are then discussed in the context of different logics of professional work. Finally, we conclude by summarising the study’s scientific and practical implications.

Context and theory – bureaucrats and managers at the frontline of Brazilian social security

Brazil figures amongst the world’s biggest and most unequal economies in the world, making the role of distributive policies paramount (United Nations, 2020), such as the case of the policy portfolio INSS implements through its 14,726 officials (INSS, 2020), typical frontline professionals or SLBs (Lipsky, 2010). They are responsible for individually granting or denying, in a case-by-case fashion, social assistance and insurance-based benefits as well as benefits from rural/special social insurance systems, resulting in a large array of benefit rules and client profiles (Schwarzer and Querino, 2002). Despite their eminently rule-application duties, benefit systems’ complexity opens decision-making leeway or discretion (authors, under review). Yet, as in other Global South welfare bureaucracies (Lotta et al., 2021), INSS SLBs do not belong to a formal professional group and so they share diverse formation backgrounds, making discretion more vulnerable to idiosyncratic distributive justice convictions (de Andrade and Pekkola, 2022). The ways they use this discretion can have significant aggregated effects on the Brazilian economy, given the magnitude of INSS welfare provision – US$130 bn/year to more than 36 m people (Ministério da Economia, 2020).

Moreover, INSS is a historically managerialised organisation. Heir to the Brazilian “90s contested neoliberal reforms” (Oliveira et al., 2010), it adopted staple NPM practices – notably, pay-for-performance systems and middle manager empowerment (Almeida and Barbosa, 2019; Aoki, 2019; cf. Aktas et al., 2023). Adding to this, late digitalisation deepened NPM logics, by implementing (contested) automatic performance assessment of SLBs’ benefit-granting activities (Pinheiro et al., 2018). However, the supervision and organisation of service offices are still duties of humans, the frontline INSS managers (SLMs).

Selected amongst SLBs, SLMs display similar professional profiles but are expected to play “extra” roles – impersonating managerial discourses and applying managerial instruments to pursue managerial goals – embracing managerial logics (Evetts, 2009). Once nominated and initiated, SLMs are expected to act with business-like professional management authority, following the “Yankee-Brazilian” management school models (Alcadipani and Caldas, 2012). Yet, they have limited organisational power and must work between rock and hard place to manage and maintain their legitimacy towards subordinates, for the three main reasons: first, SLMs share the same environment as their SLB peers, dwelling there during their management term. Second, their rationality to adopt any severe measures is reduced as, after the term, they join the regular SLB ranks again. Third, they cannot hire or select their teams (Oliveira et al., 2010). Hence, SLMs should stay aligned with SLB peers’ discourses, refraining from “betraying” their professional ethos, whilst complying with managerial doctrines, dealing with pressures from the upper management and keeping their role as a “crossing point” for clientele (Floyd and Woolridge, 1992; Gassner and Gofen, 2018; Noordegraaf and Steijn, 2013).

Managers’ influence on bureaucrats’ decisions

Despite those restrictions, SLMs still hold the “key to policy implementation” (Grøn, 2018, p. 11). They draw on their proximity to mobilise street-level public service by translating “formal policy decisions into operational work plans and daily scheduled tasks” and adapting “direct delivery arrangements” to provide “de facto, “tailored” solutions” (Gassner and Gofen, 2018, pp. 560–561; Klemsdal et al., 2022). This way, SLMs deploy customised orders to the street-level “microenvironment”, adapting or refining standardised procedural guidelines – opening channels for influencing SLBs’ professional practices.

SLMs can then use these channels to convey alien logics to SLBs’ activities, by, for instance, establishing managerial efficiency-oriented orders, like stricter decision-making standards aiming at output-oriented performance goals (Pinheiro et al., 2018). This “inoculation” can then shape SLBs’ discretion and decision-making without necessarily changing their stance about the policies they operate and their professional role. Yet, as explained before, SLMs have limited organisational power, due to their proximity to SLB subordinates, precarious mandate and lack of control over team composition – incentives to conserve their logics in line with SLBs’. Hence, whilst proximity facilitates the enactment of customised orders, it can also drive both groups to share a resistance to alien logics (Grøn, 2018). In this scenario, SLMs’ influence does not disturb SLB professional logics (Figure 1).

Professional logics are strongly connected to how professionals’ discretion is set and can be understood through different dimensions (Evans and Hupe, 2020). In benefit-granting, the core of INSS SLBs’ activities, two dimensions are particularly relevant to how their individual decisions are taken and SLMs’ influence on those is critical. First, how they perceive distributive justice in the context of the policy they operate, affecting how their discretion is used to grant or deny benefits in different cases (Blomberg et al., 2017). Second, their perceived discretion, related to the scope of their perceived professional role, which determines the leeway they effectively use to decide in one way or another (Evans, 2011).

In summary, dissonant SLMs’ logics can be embedded in customised orders, influencing SLBs’ decision-making regardless of their professional stance. However, INSS SLMs’ limited organisational power might prevent them from adopting and sponsoring any professional logic shifts. This study compares INSS SLMs’ and SLBs’ professional logics through their perceptions about distributive justice and discretionary space, variables connected to how they understand, respectively, the policy they operate and their power to shape it. The following sections discuss these variables and how they can shift when SLBs become SLMs.

Distributive justice: role in social policy and clients’ deservingness

Vague and contradictory policy goals, unpredictable demands and managerial pressures push SLBs to take heuristic shortcuts and adopt coping mechanisms according to perceptions of urgency or merit – ideal opportunities for intuition-based judgements (Tummers et al., 2015). The problem is especially pertinent for INSS SLBs, who decide on highly selective social assistance granting in a diverse and unequal society (Lotta et al., 2021). Their decisions modulate welfare protection’s scope and volume according to how they understand their role in social policy and picture fairness or distributive justice. This can be assessed on a spectrum between two ideal types, according to how they balance their thoughts between two opposing “goods”: at one edge, those prioritising efficient policy targeting and prevention of undue (taxpayers’ money) spending; on the other hand, those prioritising the improvement of welfare policy coverage. In economic terms, the position in this spectrum tells how far the SLB go to either prevent free-riding, when beneficiaries unduly exploit the social security system or non-take-up, when those who need support are at risk of not getting it (Evans and Hupe, 2020; Hupe, 2013).

Besides, clients’ traits can change SLBs’ perspectives about their entitlement to benefits. These more nuanced distributive justice rationales can be approached through the deservingness perceptions framework (von Oorschot, 2000), designed to assess how people see different social groups as deserving of state-sponsored welfare. Deservingness perceptions translate distributive justice convictions multidimensionally through different criteria (Larsen, 2008), as in the traditional CARIN construct (von Oorschot, 2000). The CARIN acronym stands for five criteria of deservingness judgement: (1) Control – the less in control of their situation, the more people deserve state aid; (2) Attitude – the more well-behaved, the more deserving; (3) Reciprocity – the more contribution to society, the more deserving; (4) Identity – the more shared identity with the one judging, the more deserving and (5) Need – the objectively needier, the more deserving.

In sum, officials’ distributive justice reasoning tells how extensive they understand the state’s redistributive systems should be. So, dissonances between SLMs’ and SLBs’ distributive justice reasonings can signal the economic direction of SLMs’ influence on SLBs’ decision-making. Through customised street-level orders, SLMs can set standard heuristics infusing alien distributive justice ideas in SLBs’ decisions. These standards can, for instance, translate the absorption of the upper management pressures or performance standards and embed economic dimensions of the managerialist discourse: backed by Brazilian neoliberal fiscal austerity principles (Oliveira et al., 2010), these support rigorous social policy selectivity (Skilling, 2016). The same effect could be achieved through bias in SLM nomination, with “ideologically adequate” officials having higher odds of being chosen as managers. Conversely, SLBs seeking to improve policy coverage could be inclined to run for leadership positions, to help realise their distributive ideals by instrumentalising social policy implementation.

Discretion between professionalism and managerialism

For Evans (2011), the usual Lipsky’s (2010) lenses overlook the relevance of professional status on discretion, concealing, under the SLB theory, a clash between professionalism and managerialism. In this clash, professional work implies extensive and legitimated freedom in decision-making – discretionary space – although consistent with principles established by a profession’s collective tradition. For instance, traditional professionals such as medical doctors and lawyers ground their discretion on their societal function or special traits. These can be a common good, altruism-based, “inner calling mystique”, and control over esoteric, “sacred” knowledge, reinforced by language specifics, hermetic concepts and special education (Hupe and van der Krogt, 2013; Johnson, 1972). Yet, heterogeneous SLB bodies such as INSS officials also display professionalism traits. Their activity, whilst generally canonical, involves a degree of discretion, relies on hermetic knowledge, language specifics related to the bureaucracy and public-values-laden civil service ethics (de Graaf and van der Wal, 2013).

However, NPM-inspired managerialism in public organisations has challenged traditional, occupation-based professionalism with organisational professionalism, rooted in organisation-based identities and more subject to managerial control (Evetts, 2009; Hupe and van der Krogt, 2013). For instance, managerial performative logics fragment work processes and establish protocols for streamlined, output-based productivity (de Bruijn, 2002). By displacing administration principles, these logics can constrain discretion on behalf of standardised, efficient, clientele-based mass-processing – ultimately deskilling frontline officials and constraining discretionary space (Aktas et al., 2023; Gassner and Gofen, 2018; Møller and Hill, 2021; van der Veen, 2013). Table 1 provides an overview of the core rationale and drawbacks of (traditional) professionalism and managerialism.

INSS officials’ activities are highly prescribed by managerialism. They work in desk office environments or at counters, under standardised protocols and IT system scripts, rarely engaging in fieldwork (de Andrade and Pekkola, 2022). Managerial instruments constrain their power to treat cases in a tailored, one-by-one fashion, as wider discretion is deemed to encourage late, unclear, inefficient, unstandardised and unfair decisions (Pinheiro et al., 2018). Still, INSS SLBs often claim the importance of discretion or complain about losing it – evidencing their professional orientation through how they perceive their available discretionary space (authors, under review).

On the other hand, the top-level administration selects, trains and holds SLMs accountable under performance standards set to make them preachers for traditional Brazilian NPM-inspired managerialism. Hence, SLMs are expected to promote efficiency through standardised working processes and emphasise quantitative outputs – ultimately constraining discretionary spaces and deskilling SLBs' professional work. Nevertheless, despite position, training and performance accountability, SLMs’ are compelled to keep their distributive justice ideas and praise for professional-wide discretionary spaces aligned with SLBs’, given their limited organisational power and shared institutional environment.

In sum, whilst close management can shape street-level discretion (Gassner and Gofen, 2018; Klemsdal et al., 2022), this effect can be limited by SLMs’ and SLBs’ common professional backgrounds and environment (Grøn, 2018). Yet, it is not clear whether SLMs in a highly managerialised agency can develop dissonant professional logics. In the case of social policy-operating officials, whose core activity is benefit-granting, the most relevant dissonances between SLMs and SLBs’ logics concern their reasonings about distributive justice and their perceived leeway for decision-making. Assessing these dissonances is crucial in understanding the influence of SLMs on SLBs' decision-making, especially where professionally heterogeneous bureaucracies deal with competing policy goals – the case of public social security in large developing economies. Besides, these bureaucracies often have scarce resources to face huge demands, asking for stricter management – yet deal with pervasive inequality and institutional frailty, which ask for broader discretion and policy improvisation (Aktas et al., 2023; Lotta et al., 2021).

Research design and methods

This research examines the views of INSS SLBs and SLMs regarding their professional role in social policy. INSS SLBs are officials working in frontline, service offices, whilst SLMs are SLBs working in management positions in these same offices. We empirically compare their distributive justice reasoning and how they sense their discretionary space, to determine whether they share similar professional logics. Logics dissonances can indicate the direction of SLMs' influence on SLBs’ decision-making, critical in social security policies. The problem is magnified in the context of large developing countries’ welfare organisations such as INSS, where SLBs who lack a shared professional identity are central in deciding on complex cases in highly adverse conditions. Further, we focus on the data concerning social assistance cash benefits: these represent the INSS’s most impactful policy in terms of distributive justice and involve significant rule-based discretion, hence having instrumental importance in this inquiry. We approach the data with mixed methods, although emphasising an overall quantitative design. The aim is to maximise the use of available information and improve internal validity, by integrating earlier qualitative results (authors, under review) with both qualitative and quantitative data gathered from the same cases, thus, harnessing data relatability (Fetters et al., 2013; O’ Cathain et al., 2010).

The research problem is split into two questions: the first question, answered by quantitative hypothesis testing, is: do INSS SLB and SLM share similar distributive justice reasonings and sense of discretionary space in social assistance benefit-granting? The hypotheses assume no dissonance between groups’ perceptions:

H1.

SLMs’ and SLBs’ distributive justice reasonings do not differ significantly and

H2.

SLMs’ and SLBs’ sense of discretionary space does not differ significantly.

The second research question complements the first, asking how different distributive justice reasonings and sense of discretionary space stand out in SLMs’ and SLBs’ reflections on social assistance benefit-granting? We answer it by screening for subtler differences between SLBs’ and SLMs’ perceptions lying in open-text survey answers, untraceable by quantitative inquiry alone.

Quantitative analyses addressing the first question draw on registry (CGU, 2022) [1] and survey data for 678 INSS officials (Sample 1), plus authors’ own work’s (under review) themes extracted from 350 open-ended answers to the same survey (Sample 2) through TCA. Themes translated officials’ distributive justice reasonings and their perceived discretionary space in social assistance benefit-granting, indicating whether they aimed to: (1) avoid moral hazards (AMH) or (2) expand policy coverage (EPC) and if they expressed a demand for (3) wider discretionary space (WDS) or (4) narrower discretionary space (NDS). Appendix discusses the samples and representativeness.

In the first quantitative analysis stage, we employ the statistical tests to verify the hypotheses. To test H1, we compare SLMs’ and SLBs’ deservingness perceptions of social assistance beneficiaries in five-point Likert scale survey answers, ranging from completely disagree to completely agree with statements corresponding to seven deservingness criteria (social investment, universalism, control, attitude, reciprocity, identity and need criterion). Mann–Whitney U-tests coupled with t-tests compare the groups’ score distributions across the criteria, plus a general index (sum of scores, Cronbach’s alpha = 7.44) and a neutrality index, counting “neither agree nor disagree” answers. Following, we compare groups’ TCA distributive justice-related themes (AMH and EPC) transformed into frequencies (Fetters et al., 2013). The same procedure tests H2, using corresponding themes (NDS and WDS). Table 2 clarifies variable operationalisation.

After each comparison, a second quantitative stage screens the verified differences by checking for confounding features, unevenly distributed across groups. We employ the regression models incorporating the variables where differences were found (deservingness scores or themes) as dependent, including the theory-informed controls to mitigate confounding/omitted variable bias and offset sample skewness (Steiner et al., 2010). The regressions are built in four steps: first, the baseline models, include only one independent dummy variable: manager position. The second models include personal background variables: gender, age, formation, socioeconomic status and macroregion. The third models drop personal background while including variables related to the INSS organisational context: years in office, unit size, face-to-face encounters, face-to-face time and face-to-face time ratio. This way we highlight the most important variables in each domain (personal background and organisation). Last, we build the optimal models, including only relevant or significant variables from previous equations. With these steps, we mitigate model overfitting due to high sample/parameters ratios.

Finally, in the third, qualitative analysis stage, we address the second research question harnessing case-based relatability to integrate quantitative and qualitative data in a mixed-methods matrix (O’Cathain et al., 2010). Dependent variables of the quantitative stages (TCA themes, manager status and deservingness scores) are used as categories to support the interpretation of latent meaning differences between groups’ reflections. Figure 2 summarises the overall research design, connecting data, variables and stages.

Analysis and results

Quantitative

We assume SLMs can infuse alien distributive justice ideas into SLB’s activities through customised street-level orders; however, this is relevant only if ideas are different between the two groups. We measured these differences by comparing first how they see beneficiaries’ deservingness of state-sponsored support and second how they understand their role in social policy, operationalising these variables as explained in Table 2. Concerning deservingness, we compared groups’ scores across the seven criteria. Mann–Whitney U-tests and t-tests could not reject null hypotheses in most criteria, displaying significance scores (p-values) above 0.3. Most measurements, therefore, confirm H1: no significant differences are observed between groups, except for slightly higher scores for SLBs on identity and the general index (Table 3).

These differences tell that managers are slightly less likely to think they could be just like people receiving social assistance, and that in general, these are less deserving of state aid. Next, we screen these findings in the presence of controls, modelling the effect of being a manager on identity deservingness scores through sets of ordinal regressions (Table 4) and on the general deservingness index through linear regressions (Table 5). The optimal models hold only controls showing significance levels less than 0.1, or odds ratios either higher than 1.5 or lower than 0.5 (ordinal regressions) or coefficients higher than 1 or lesser than −1 (linear regressions).

Differences do not hold when other relevant factors weigh in, meaning that being a manager does not imply significantly different deservingness perceptions under any criteria – as both optimal models did offset the significance of being a manager, thus not rejecting the null hypothesis and confirming H1 (distributive justice reasonings do not differ significantly). However, the analysis suggests other factors play important roles. Amongst these, social worker formation background is the most relevant, driving an almost six-point increase in the general deservingness index – and resonating with findings elsewhere (Blomberg et al., 2017). The significant 1.57-point increase for officials working in the northeast macroregion is also worth noticing. That is, having a social work formation and working in northeastern Brazil entail gentler deservingness perceptions of social assistance beneficiaries.

Following, we verify differences in how SLBs and SLMs perceive their role in social policy, complementing the H1 test (distributive justice reasonings do not differ significantly). Concomitantly, we test H2 (sense of discretionary space does not differ significantly); assuming different senses of discretion imply different professional logics. We follow the operationalisation proposed in Table 2, comparing TCA theme frequencies across groups (Table 6). Pearson chi-square tests indicate that demands for wider discretionary space (WDS) are significantly more common among SLMs, whilst drives for policy coverage expansion (EPC) are less common. Managers, thus, seem more likely to believe discretionary spaces should be wider and less likely to advocate for improved policy coverage.

Then, we weigh other factors in, modelling the effect of being a manager in the incidence of WDS and EPC themes, through binary logistic regressions, as those built for deservingness perceptions. Variables selected for the optimal models in these cases were those showing significance levels less than 0.1 or odds ratios higher than 1.4 or lower than 0.6. Resulting models suggest that performing managerial activities relates to a nearly two-fold increase in the odds of demanding wider discretionary space, regardless of controls – partially rejecting H2 (Table 7). Also, a positive but less significant effect is connected to other biological sciences backgrounds, whilst a negative one is found amongst officials most intensively facing the public.

The effect’s significance on EPC is offset in the three models, reinforcing H1 (Table 8). As in deservingness scores, the odds of officials wanting to expand policy coverage are highly increased by social work backgrounds. Significantly positive effects are also related to law and economy backgrounds, working in the south or being in the middle tier of total years working in INSS.

Overall, the quantitative findings suggest SLMs and SLBs do share similar distributive justice reasoning patterns, supporting H1. However, SLMs tend to advocate for wider discretionary spaces, even when other factors are controlled, partially rejecting H2. Table 9 summarises the hypotheses tests’ results.

Qualitative

In this stage, we split SLBs and SLMs into groups according to the dependent variables and complement the quantitative stages by scanning for latent qualitative differences in the groups’ textual survey answers. This allows us to understand how being a manager can modulate the way officials with different professional role understandings express their ideas on benefit-granting. A mixed-methods matrix (O’Cathain et al., 2010) is used to report the integrated qualitative and quantitative findings, where rows are groups of cases (open-ended survey answers) categorised according to quantitative parameters. Hence, cases were divided into 16 groups according to TCA themes, general deservingness index scores and organisational position – regular SLB or SLM. The qualitative analysis targets common subtler discursive traits found in each group, highlighting relevant differences between them. Table 10 reports the mixed-methods matrix, displaying qualitative findings and a representative quote [2] for each group of cases (row). The findings for groups in each TCA theme in the matrix are posed in relation to the findings of its first group subcategory (that is, low deservingness SLBs) and indented accordingly. For example, Group 2’s findings can be phrased as: low deservingness (20 − score) SLMs, who advocate for wider discretionary space do complain about the lack of instruments for assessing applications, whilst referring less frequently to general dishonesty and more frequently to applicants’ lack of information than low deservingness SLBs, who advocate for wider discretionary space.

We consider in our interpretation that TCA themes often overlap and, despite the established threshold between “high” (21+) and “low” (20−) deservingness, we pay special attention to the index’s continuous nature and fuzziness of borderline scores.

Strengthening the quantitative findings, differences between SLBs’ and SLMs’ answers amongst overlapping categories of distributive justice reasoning and discretionary space are small. Regardless of other categories, managers employ more managerialist jargon (e.g. “efficiency/inefficiency”, “workflow” and “compliance”), as observed elsewhere (Oliveira et al., 2010). They also provide more systematic explanations for challenges, avoid “pointing fingers” and give more careful answers – patterns exhibited regardless of economic perspectives. For instance, even SLMs with low deservingness scores relativise their judgement on the needy and refrain from generalisations, instead blaming rules, the system or people’s lack of information. Also, managers demanding wider discretionary space (WDS) and having higher deservingness scores question managerialism-related shortcomings. Yet, the overall qualitative assessment of open-ended answers does not reveal meaningful differences between SLBs’ and SLMs’ distributive justice reasonings or sense of discretionary space, reinforcing both H1 and H2.

Discussion

Overall, most differences in how the two groups interpreted their professional roles in social policy were residual. This strengthens the postulation that, given the temporary nature of the management position vis the permanent SLB job, plus the need for SLMs to avoid deviating from the professional ethos of SLBs (aiming to retain legitimacy in the eyes of their teams) and the impossibility to hire or select their teams, SLMs are expected to sustain their SLB professional views, acting as primus inter pares. This does not suggest, however, that all SLMs and SLBs necessarily have the same ideas – just that the manager role, by itself, does not change them.

Regarding our first hypothesis, that is, that the distributive justice reasonings of SLMs and SLBs should be similar, no significant differences between SLBs and SLMs were observed in any deservingness perceptions criteria or in general distributive ideas displayed in officials’ open-ended answers. This suggests continuity when INSS SLBs become SLMs: the economic logics underlying managerial discourses (Skilling, 2016) is not at all successful in changing SLMs’ perspectives and SLBs’ economic ideas do not influence their chances of being appointed SLMs. However, qualitative analysis of their comments showed that SLMs are more careful than regular SLBs when enacting normative statements, including those concerning distributive justice convictions. This can be a reflex of social desirability effect imbalance, as managers, enjoying some level of “public image”, might be accustomed to providing politically careful statements.

On the other hand, testing of the second hypothesis – that both groups should share similar senses of discretionary space – revealed an unexpected tendency for SLMs to demand wider leeway in social assistance benefit-granting, that is, although the qualitative data show SLMs adopting managerialist jargon more often than SLBs, this does not correspond to expected managerial attitudes, translated as stronger arguments for standardised and measurable service provision and the consequent compressing of discretionary space. Instead, SLMs surprisingly tend to defend, more so than SLBs, officials’ discretion or leeway to act in a professional, tailored, inefficient case-by-case fashion – despite jargon signalling compliance with managerial doctrines. That is, INSS SLMs seem to manage much of their position’s complex dynamics (Gassner and Gofen, 2018), namely, the struggle between managerialism and professionalism (Evetts, 2009), by talking as managers whilst distinctly advocating for the professional nature of INSS street-level activity. So, the practical effects of managerial discourse, clearly adopted by SLMs as shown in the qualitative findings, seem to be rather limited in the INSS context and so their influence on the direction of SLB’s decision-making, relativising earlier findings (e.g. Klemsdal et al., 2022). Besides, this extends Grøn’s (2018) discussion, as the symbolic distance between SLMs and SLBs seems small, even amongst officials lacking a shared strong professional identity and working in a highly managerialised agency as INSS.

An alternative explanation to the unexpected SLMs’ advocacy for wider discretion might be found in the recent INSS’s digital transformation process, in which managerial, efficiency-based performance assessment became individualised and almost automatic. Besides the risk of deepening pay-for-performance-related issues (Aktas et al., 2023), automaticity could, in many ways, render local overseeing obsolete. Although our qualitative findings hint this could be the case (see Group 1 quotes in Table 10), more in-depth research is needed.

To highlight a methodological limitation of the study, the skewed sample and cross-sectional design do not rule out selection bias and endogeneity. It might be that officials who yearn for discretionary space are more likely to apply for management positions, so to amplify their discretion through management activities. Thus, despite measures to improve validity (statistical tests, control variables and methods combination), the findings ask for in-depth or longitudinal research.

Conclusion

This article addressed the question of whether SLMs and SLBs working in INSS service offices share similar professional logics, in terms of their distributive justice reasonings and their sense of discretionary space and how these logics are present in their reflections on social assistance benefit-granting. We resorted to the mixed-methods in three stages: in the first quantitative stage, we used INSS official registry and survey data, extracting deservingness perception scores, transforming themes derived from open-ended answers and comparing them across groups through statistical testing. In the second quantitative stage, we performed regression analyses to sift the findings, utilising optimised collections of control variables. In the third, integrative stage, we resorted to the raw text of open-ended answers to build a mixed-methods matrix, where the qualitative analysis was systematised in case groups defined according to key variables (deservingness scores, themes and manager status), complementing the quantitative findings.

In most measurements, differences between SLMs and SLBs were residual, suggesting a general continuity of the SLB professional self when becoming an SLM – despite the adoption of managerial ideas in the discursive layer, evidenced in the qualitative data. Extending Grøn’s (2018) propositions, the findings show that even though proximity to SLBs facilitates SLMs’ translation of abstract policies into operational orders (Gassner and Gofen, 2018; Klemsdal et al., 2022), it also preserves their stances aligned, relativising their influence and hindering the introduction of alien logics in professional activities. Our evidence suggests this alignment holds even across large, heterogeneous and managerialised bureaucracies such as INSS, and in the Brazilian scenario of inequality, resource scarcity and institutional frailty – common to other large developing economies (Aktas et al., 2023; Lotta et al., 2021). The one exception in our findings is the stronger praise SLMs show for discretionary space, contesting typical managerial logics. As hinted by corresponding qualitative evidence, this might reflect managers’ reactions to new automatic performance evaluation mechanisms, where the part of their overseeing duties became bypassed.

We, especially, highlight the effects of certain control variables in the regression models. First, as found elsewhere (Blomberg et al., 2017), increased positive deservingness perceptions were strongly linked to a social work background and in the INSS officials’ case, to higher self-assessed socioeconomic status and northeastern Brazil. Predictably, a social work background also relates to a general benevolent perspective on social policy implementation/benefit-granting activity, reinforcing the reliability of the deservingness perceptions framework, as also verified by Blomberg et al. (2017). Yet, our models exhibited significant effects from law and economy formations – besides working in the south region or belonging to the middle tier of office experience (between 11 and 16 years in office). These relationships highlight the importance of researching public officials’ perceptions equipped with broader frameworks, as the traditional SLB theory can no longer explain the complex institutional settings of contemporary public service. Further, our findings suggest that whilst SLMs still handle the professionals/upper management/clientele triumvirate, the rise of automated performance assessing can take much of their policy-shaping power away. This transformation challenges traditional frontline manager roles in public organisations, an issue to be addressed by more in-depth inquiry.

Our study calls the attention of both policymakers and managers, who steer administrative reforms to three relevant issues for INSS – and similarly structured public organisations. First, the limited differences in perceptions between managers and subordinates question current NPM-inspired professional roles they are expected to play. It may indicate that manager training is not effectively initiating them in managerial institutions beyond symbolic levels. On the other hand, NPM-manager profiles might just be incompatible with the organisation’s institutional landscape. Second, the effects of control variables in officials’ perspectives suggest consideration and perhaps harnessing of professional drives lying in supposedly clerical bureaucracies. The INSS officials’ heterogeneous backgrounds, covered under common job titles, affect how they perceive the policy they implement and their professional identity. Besides a cause for policy divergence, this micro-institutional richness can be also an untapped source of innovation. Third, if the managers’ contradictory perceptions about officials’ discretion evidence a loss of purpose in their overseeing activities, it can be useful to repurpose frontline managers so that their strategic, interinstitutional position can be harnessed for the benefit of the organisation and the policy it implements.

Figures

Figure 1

SLM influence on SLB decision-making

Figure 2

Summary of research design

Table 1

Professionalism × managerialism in SLB activities

	Professionalism	Managerialism
Core rationale	Expert, tailored case treatment (wider discretion)	Expedite, market-inspired efficiency (narrower discretion)
Drawbacks	Inefficiency	Mass clientele processing
	Unaccountability	Deskilling
	Unstandardised decisions

Table 2

Variable operationalisation

Variable	Conceptual definition	Operational definition	Indicator
Distributive justice reasoning	Perceived ideal extension of redistributive systems	Deservingness perceptions	Deservingness scores (five-point Likert scale)
			Deservingness general score (total)
			Deservingness neutrality score (0–7)
		Aim to “avoid moral hazard” (AMH)	Binary coding of TCA results
		Aim to “expand policy coverage” (EPC)
Sense of discretionary space	Perceived leeway for decision	Demand for wider discretionary space (WDS)
		Demand for narrower discretionary space (NDS)

Table 3

Mann–Whitney U-tests and t-tests for deservingness perception scores

Deservingness criterion	Mann–Whitney U-tests – mean ranks			t-tests – means†
	SLB (n = 536)	SLM (n = 142)	Dif.	SLB (n = 536)	SLM (n = 142)	Dif.	Std. error dif.
Social investment	344.46	320.79	23.67	2.86	2.68	0.175	0.126
Universality	343.48	324.47	19.01	3.09	2.94	0.142	0.143
Control	340.60	335.35	5.25	4.03	3.99	0.042	0.107
Attitude	342.70	327.42	15.28	2.93	2.84	0.097	0.100
Reciprocity	343.02	326.22	16.8	3.19	3.08	0.106	0.113
Identity	347.14	310.66	36.48**	3.56	3.27	0.283**	0.133
Need	339.98	337.67	2.31	3.84	3.82	0.025	0.102
General index	346.05	314.78	31.27*	23.50	22.63	0.87*	0.486
Neutrality index	341.12	333.37	7.75	1.32	1.37	−0.054	0.139

Note(s): ^*p < 0.1 and ^**p < 0.05

Source(s): ^†equal variances not assumed

Table 4

Ordinal regressions for identity deservingness scores, odds ratios

Variable	Baseline	Background	Organisational	Optimal	SE
Manager	0.70**	0.75*	0.76	0.77	0.14
Female (ref. male)		0.91
Age (ref. ≤39 years)
40–51 years		0.96
52+ years		0.77
Formation (ref. no formation)
Law		1.04
Social work		2.74***		2.61***	0.68
Communication		1.27
Economy		0.82
Psychology		1.37
Exact sciences		0.90
Other biological sciences		1.54		1.50	0.39
Humanities/other social sciences		0.92
Socioeconomic status (ref ≤5)
6–7		0.87		0.86	0.13
8+		0.66*		0.64*	0.15
Macroregion (ref. southwest)
CW		1.00
N		0.91
NE		1.23
S		1.00
Years in office (ref. ≤10)
11–16			0.84
17+			0.78
Unit size (ref. ≤10)
11–22			1.37*	1.37*	0.24
23+			0.92	0.96	0.17
Face-to-face encounters (ref. ≤100)
101–3000			0.55
3001+			0.52
Face-to-face time (ref. <100 h)
100–1000 h			1.52	1.21	0.35
1000+ h			1.67	1.24	0.36
Face-to-face time ratio (ref. 0 min)
≤15 min			1.68	1.08	0.45
>15 min			1.90	1.10	0.46
Model fit sig.	0.04	0.01	0.08	<0.001
χ2 Pearson sig.	0.42	0.49	0.14	0.46
χ2 deviance sig.	0.41	1.00	0.08	0.75
Parallel lines sig.	0.41	0.19	0.92	0.31
Nagelkerke R2	0.01	0.05	0.03	0.05

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01, standard errors (SEs) for the optimal model

Table 5

Linear regressions for general deservingness index, unstandardised coefficients

Variable	Baseline	Background	Organisational	Optimal	SE
Manager	−0.87*	−0.62	−0.26	−0.29	0.51
Female (ref. male)		−1.12**		−0.96**	0.42
Age (ref. ≤39 years)
40–51 years		0.52
52+ years		−0.89
Formation (ref. no formation)
Law		0.17
Social work		6.03***		5.54***	0.72
Communication		0.84
Economy		−0.67
Psychology		1.84		1.47	1.46
Exact sciences		−0.43
Other biological sciences		0.96
Humanities/other social sciences		0.31
Socioeconomic status (ref ≤5)
6–7		−0.38
8+		0.02
Macroregion (ref. southwest)
CW		−0.39
N		−0.41
NE		1.23**		1.57***	0.50
S		−0.65
Years in office (ref. ≤10)
11–16			0.03
17+			−0.52
Unit size (ref. ≤10)
11–22			0.54
23+			−0.39
Face-to-face encounters (ref. ≤100)
101–3000			−2.47	−2.10	1.47
3001+			−0.26	−1.91	1.47
Face-to-face time (ref. <100 h)
100–1000 h			0.63
1000+ h			0.83
Face-to-face time ratio (ref. 0 min)
≤15 min			1.36	1.23	1.69
>15 min			3.15*	2.37	1.69
F sig.	0.09	<0.001	<0.001	<0.001
Adjusted R2	<0.001	0.10	0.02	0.11

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01, standard errors (SEs) for the optimal model

Table 6

Distribution of SLBs and SLMs across TCA themes

Dimension	Theme	SLBs (258)		SLMs (92)		Pearson χ2 sig.	LR χ2 sig.
		Frequency	Percent*	Frequency	Percent*
Discrtionary Space	WDS	52	20%	28	30%	0.044	0.048
	NDS	87	34%	35	38%	0.455	0.457
Distributive Justice Reasoning	AMH	167	65%	61	66%	0.785	0.785
	EPC	53	21%	11	12%	0.067	0.058

Note(s): *Percentages are related to in-group totals

Table 7

Logistic regressions for demanding wider discretionary space (WDS), odds ratios

Variable	Baseline	Background	Organisational	Optimal†	SE
Manager	1.73**	1.66**		1.90**	0.55
Female (ref. male)		0.64
Age (ref. ≤39 years)
40–51 years		0.77
52+ years		0.73
Formation (ref. no formation)
Law		1.27
Social work		1.50		1.09	0.61
Communication		0.00
Economy		1.66		1.54	0.78
Psychology		0.00
Exact sciences		0.57
Other biological sciences		2.35		2.12	0.93
Humanities/other social sciences		1.21
Socioeconomic status (ref ≤5)
6–7		1.40		1.44	0.42
8+		0.90		0.76	0.36
Macroregion (ref. southwest)
CW		1.18		1.09	0.52
N		0.57		0.72	0.50
NE		0.50		0.63	0.23
S		0.55		0.54	0.21
Years in office (ref. ≤10)
11–16			1.35
17+			0.93
Unit size (ref. ≤10)
11–22			0.98
23+			1.22
Face-to-face encounters (ref. ≤100)
101–3000			0.67
3001+			0.79
Face-to-face time (ref. <100 h)
100–1000 h			0.62	0.46*	0.20
1000+ h			1.48	1.17	0.48
Face-to-face time ratio (ref. 0 min)
≤15 min			0.94
>15 min			1.09
χ2 sig.	0.05		0.09	0.02
Hosmer and Lemeshow sig.			1.00	0.43
Nagelkerke R2	0.02		0.08	0.09
C-Index	0.56	0.66	0.65	0.68

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01. Standard errors (SEs) for the optimal model

The optimal model does not comply with the rule of thumb as provided by Peduzzi et al., 1996, according to which the total of covariates, in this case, should not exceed 8. However, Hosmer and Lemeshow goodness-of-fit tests show reasonable results (p > 0.05)

Table 8

Logistic regressions for aiming to expand coverage (EPC), odds ratios

Variable	Baseline	Background	Organisational	Optimal†	SE
Manager	0.53*	0.67	0.56	0.64	0.26
Female (ref. male)		0.97
Age (ref. ≤39 years)
40–51 years		1.18
52+ years		1.07
Formation (ref. no formation)
Law		3.29**		3.97***	1.53
Social work		7.62***		9.65***	5.67
Communication		1.70		1.45	1.74
Economy		2.44		3.24*	1.99
Psychology		0.00
Exact sciences		1.53		1.83	0.96
Other biological sciences		0.00
Humanities/other social sciences		1.15
Socioeconomic status (ref ≤5)
6–7		0.71
8+		0.93
Macroregion (ref. southwest)
CW		0.59		0.63	0.41
N		0.46		0.45	0.53
NE		0.58		0.57	0.25
S		1.91*		1.98*	0.73
Years in office (ref. ≤10)
11–16			2.19**	2.38**	0.87
17+			1.18	1.51	0.83
Unit size (ref. ≤10)
11–22			1.31
23+			1.13
Face-to-face encounters (ref. ≤100)
101–3000			4.47	2.82	3.46
3001+			2.39	1.49	1.97
Face-to-face time (ref. <100 h)
100–1000 h			0.55	0.60	0.37
1000+ h			0.61	0.58	0.42
Face-to-face time ratio (ref. 0 min)
≤15 min			0.52	0.93	1.21
>15 min			0.60	0.82	1.06
χ2 sig.	0.06	0.00	0.05	<0.001
Hosmer and Lemeshow sig		0.82	0.70	0.35
Nagelkerke R2	0.02	0.19	0.09	0.20
C-Index	0.55	0.69	0.67	0.75

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01, standard errors (SEs) for the optimal model

^†The optimal model does not comply with the rule-of-thumb as provided by Peduzzi et al. (1996), according to which the total of covariates, in this case, should not exceed 8. However, Hosmer and Lemeshow goodness-of-fit tests show reasonable results (p > 0.05)

Table 9

Summary of hypotheses’ test results

Hypotheses: no significant differences between SLMs’ and SLBs’ …	Concept	Operationalisation		Difference in statistical tests	Effect in optimal regressions	Result
H1: … Distributive justice reasonings	Deservingness perception	Likert scale	Social investment	—		Not rejected
			Universalism	—		Not rejected
			Control	—		Not rejected
			Attitude	—		Not rejected
			Reciprocity	—		Not rejected
			Identity	SLM−**	—	Not rejected
			Need	—		Not rejected
		Derived index	General	SLM−*	—	Not rejected
			Neutrality	—		Not rejected
	Distributive justice Reasoning	TCA theme	AMH	—		Not rejected
			EPC	SLM−*	—	Not rejected
H2: … Sense of discretionary space	Discretionary space		WDS	SLM+**	SLM+**	Rejected
			NDS	—		Not rejected

Note(s): *p < 0.1 and **p < 0.05, SLM + denotes higher scores or frequencies and SLM− denotes lower scores or frequencies for SLMs

Table 10

Mixed-methods matrix

TCA theme	Des. Index score	Position	Group number	Total cases	Finding	Representative quote
WDS	low (20−)	SLB	1	18	Complaints about the lack of instruments for assessing applications, narrowing investigation possibilities	Nowadays, benefit granting has become too mechanical (…) It is well-known that there are other means to verify if people fit or not in the poverty criteria. (case 390)
		SLM	2	12	Fewer references to generalised dishonesty and more to applicants’ lack of information	I think some do not get access to their rights due to misinformation, whilst others draw advantage out of their knowledge [about how bureaucracy works] (…). (Case 12)
	high (21+)	SLB	3	34	More references to the complexity of applicants’ real-life situations and to unfairness	Too objective and severe criteria, which do not account for the individual situation of the people in need (…). (case 340)
		SLM	4	17	Some further critique on streamlined processes and performance evaluations	Reduced time for social assessments, low availability for in-loco visits – the majority of the problems are institutional (…). (case 552)
NDS	low (20−)	SLB	5	26	Simpler and shorter answers in general, complaints about rule complexity and fraud	Loose, too general legislation not accounting for regional differences (…) and hinders benefit granting to those who really need, instead facilitating fraud and undue benefits. (case 395)
		SLM	6	11	More focus on rules constantly changing over time	Rules change constantly, making it hard to apply them correctly, besides a lack of tools to verify the updated applicants’ economic situation. (case 306)
	high (21+)	SLB	7	61	More concern about the applicant’s perspective on rule complexity	The first issue is a lack of information, as many who need social benefits don’t know they have the right (…). Others know their rights, but think the access is difficult and end up not applying. [Others] provide wrong information [in their applications], because the socioeconomic situation of the family [constantly] changes (…). (case 101)
		SLM	8	24	Some suggest most of the complexity derives from IT system shortcomings and show a more favourable perspective on automation	(…) I also believe that there are already enough data sources to expand automatic benefit granting based on objective criteria. (case 561)
AMH	low (20−)	SLB	9	73	Constant reference to general fraud, ideas of generalised dishonesty, registry frailty	The habit of many people who want to take advantage of situations, resulting in undue payments (…). (case 275)
		SLM	10	24	Also statements on generalised dishonesty, but often the critique is focused on the system or rule complexity	In my opinion, the biggest challenges to social assistance benefit granting are the establishment of efficient methods for identifying the really needy, so to avoid frauds, and the creation of efficient service delivery and orientation channels, adapted to the public’s reality, which mostly don’t have technical knowledge and basic education. (case 667)
	high (21+)	SLB	11	94	Fewer references to generalised dishonesty. Many references to despachantes and systemic causes of fraud (system problems, rule blind spots or complexity)	Misinformation makes some people apply through intermediaries, who many times are not well-intended, often asking for benefits without meeting eligibility criteria. (case 108)
		SLM	12	37	Similar to SLBs	The main challenge (…) is in the availability and integration of databases from different public entities. (case 560)
EPC	low (20−)	SLB	13	10	Most answers compensate for the ideal of expanding coverage with the need to improve targeting	There are many frauds, people with informal income or who [falsely] claim to live alone and receive the benefit. On the other hand, people who really need, but as they exceed the [maximum allowed] per capita income by a very small margin, have their benefit denied. (case 529)
		SLM	14	4	Too few cases; similar to SLBs	Lack of control, orientation, and incentives to the whole society to avoid the need for those benefits, making access easier to who really needs it. (case 209)
	high (21+)	SLB	15	43	Clear pro-welfare perspective, often justifying it with the Brazilian context. Some mention the stigmatising perspectives of other officials	Most of those who really need social benefits do not get information. There is much inequality because of a lack of knowledge of their rights, in both the disabled and old-age cases. People do not have education, the minimum to have access. (case 192)
		SLM	16	7	Too few cases; similar to SLBs	Lack of trust on the information provided by the applicant, as they are declaratory and many do lie about their family components, marital status, etc. The prejudice of some analyst colleagues and medical professionals who do not like to face the social assistance clients. (case 449)

Table A1

Sample 1 (full survey) and intragroup statistical tests

		Survey sample	Managers	Non-managers
Variable	Test†	(n = 678; N = 14,726)	(n = 142; N = 2,013)	(n = 536; N = 12,713)
Macroregion	Chi-square (4df)	34.337***	8.296*	30.954***
Female	Binomial	−0.03**	–‡	–‡
Manager	Binomial	−0.07***	–‡	–‡
Unit size	Mann–Whitney U	16.91	−29.38	−202.47
Encounters	Mann–Whitney U	200.21	54.11	229.81
Encounter time	Mann–Whitney U	216.56	46.29	139.19
Encounter time ratio	Mann–Whitney U	−132.03	8.32	−343.59*
Years in office	Mann–Whitney U	509.45***	−29.98	601.66***
Years in management	Mann–Whitney U	–‡	−92.96*	–‡
Unit size	t-test	−0.285	−0.781	−0.994
Encounters	t-test	296.772	340.195	347.666*
Encounter time (h)	t-test	55.377*	63.264	45.063*
Encounter time ratio	t-test	−0.005	0.003	−0.016
Years in office	t-test	1.884***	0.228**	2.335***
Years in management	t-test	–‡	0.354**	–‡

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01

^†Test results are χ² value (chi-square test) or differences (N–n) between proportions (binomial tests), mean ranks (Mann–Whitney U-tests) and means differences (t-tests)

^‡No data available for this variable for the population

Source(s): Authors’ own work

Table A2

Sample 2 (TCA results) and intragroup statistical tests

		RTA sample	Managers	Non-managers
Variable	Test†	(n = 350; N = 14,726)	(n = 92; N = 2,013)	(n = 258; N = 12,713)
Macroregion	Chi-square (4df)	21.518***	2.989	22.062***
Female	Binomial	−0.04*	–‡	–‡
Manager	Binomial	−0.12***	–‡	–‡
Unit size	Mann–Whitney U	−149.94	−194.25***	−180.28
Encounters	Mann–Whitney U	88.37	146.47**	−49.66
Encounter time	Mann–Whitney U	98.78	132.35***	−188.45
Encounter time ratio	Mann–Whitney U	−122.37	13.83	−487.28**
Years in office	Mann–Whitney U	601.72**	72.86	524.44**
Years in management	Mann–Whitney U	–‡	407.13***	–‡
Unit size	t-test	−0.679	−4.755***	−0.789
Encounters	t-test	245.686	505.085	269.248
Encounter time (h)	t-test	34.872	141.375*	−18.403
Encounter time ratio	t-test	−0.001	0.002	−0.020
Years in office	t-test	2.196***	2.187**	2.222***
Years in management	t-test	–‡	1.135***	–‡

Note(s): *p < 0.1; **p < 0.05 and ***p < 0.01

^†Test results are χ² value (chi-square test) or differences (N–n) between proportions (binomial tests), mean ranks (Mann–Whitney U-tests) and means differences (t-tests)

^‡No data available for this variable for the population

Source(s): Authors’ own work

Notes

1. Petitions 03005.178904/2020-40; 03005.019053/2021-85; 03005.216980/2020-61; 03005.019818/2021-87; 03005.049882/2021-92.

2. Authors’ translations from Portuguese.

3. Minimising type II errors of both methods for specific distribution patterns (de Winter and Dodou, 2010).

Appendix Samples and representativeness tests

The studied population comprises 14,726 officials working in the INSS service offices (SLBs), of which approximately 2,000 are SLMs, whose registry data were provided by the Brazilian federal government in October 2020. The following variables are available for the officials: (1) encounters, total times the official was face-to-face with clients from April 2017 to March 2020; (2) encounter time, total hours spent in these encounters; (3) encounter time ratio, the average time spent in these encounters; (4) gender totals; (5) macroregion, one of the five regional subdivisions in which the official works; (6) unit size, the number of officials working in the unit; (7) years in office, total years worked for INSS and (8) years in management, total years holding the current management position.

Utilised survey data were collected in 2021, and answers (sample 1, n = 678) are relatable to the registry data. The following variables were extracted from them: (1) deservingness perceptions scores, that is, the official’s agreement level in Likert scales to statements concerning their deservingness perceptions under seven criteria; (2) formation type, the official’s higher education formation area; (3) gender; (4) if they are a manager and (5) their self-assessed socioeconomic status on a scale from 1 to 10 (de Andrade and Pekkola, 2022). We consider managers (SLMs) officials declaring management or supervision amongst their five most-performed activities since 2018 (total = 174). Managers selected this way mostly coincide with registry data, r(676) = 0.710, p < 0.001. Yet, we opt to use the survey data to mitigate three concerns: (1) potentially outdated registry data, (2) undue exclusion of officials who left management just before the time of data extraction and (3) undue inclusion of officials recently made managers.

Besides the quantitative data, we also use themes extracted from the survey’s open-ended answers, which, however, do not cover the whole survey sample (sample 2, n = 350). Themes translated officials’ economic or distributive justice reasoning and their perceived discretionary space in social-assistance benefit-granting, indicating whether they aimed to (1) avoid moral hazards (AMH) or (2) expand policy coverage (EPC) and if they expressed a demand for (3) wider discretionary space (WDS) or (4) narrower discretionary space (NDS).

We performed the statistical tests both for the full samples and their intragroup representativeness (that is, managers and non-managers across both Samples 1 and 2) based on the available registry data. Concerning Brazilian macroregion distribution, the southeastern region is overrepresented whilst the northeastern region is underrepresented, particularly in the non-managers’ group. Gender distribution is at the limit of representativeness (p = 0.04 in sample 1 and p = 0.07 in sample 2), whilst the proportion of managers is significantly higher than in the population – according to the registry data (20.9 and 14.2% against 13.7%, p < 0.001 in both samples). We also tested representativeness for continuous, non-normal variables through both Mann–Whitney U and t-tests [3]: unit size, encounters, encounter time, encounter time ratio, years in office and years in management. Results varied but, in general, both the samples are skewed according to these parameters. We report the tests in the tables that follow: