Intersectional patterns of social assistance eligibility in Sweden

ABSTRACT

This study examines patterns in social worker decisions on social assistance eligibility in Sweden. Focusing on intersections between applicants’ gender, country of birth and family situation, factors that statistically anticipate decisions on granting assistance to individuals from different sub-groups were explored. The sample comprises 423 applications and four sets of modified Poisson regression models were conducted. The results strengthen the impression of social assistance assessments as a practice marked by the professionals’ categorizations. In line with previous research, social workers seem to act upon, and reinforce, a male breadwinner model by putting more emphasis on men’s efforts to establish self-support. Female applicants, in turn, are seemingly less likely to be granted assistance if they are assessed as having a problem with abuse of alcohol or illicit drugs. Also, having a family seems altogether to have a negative impact on women’s chances of approval. When considering gender and country of birth, decisions on social assistance eligibility largely reflect patterns of unemployment.

KEYWORDS:

• Social assistance
• decision-making
• intersectionality
• quantitative

Introduction

Unlike other financial support programmes in the Swedish welfare state, social assistance eligibility is determined through individual assessments performed by trained social workers (e.g. Stranz, Wiklund, and Karlsson 2016a). This may largely be attributed to the fact that the right to social assistance is regulated by the national Social Services Act (SSA), formed as a framework law, which provides a significant degree of municipal autonomy according to the modalities of means testing. The social workers that ultimately make decisions on social assistance eligibility act with a wide margin of discretion and tend to interpret both national and local guidelines differently (Stranz 2007). A number of studies have demonstrated essential differences in social worker assessments of social assistance, and the flexibility surrounding the benefit has been described as a breeding ground for legal uncertainty (e.g. Minas 2005; Stranz 2007). The localization of means testing in the context of social work enables professionals to pay attention to, and potentially provide support related to, other social problems (than purely economic ones) among help-seeking individuals and families. However, it also implies inevitable elements of categorization that might affect the ways in which help-seeking persons’ need – and eligibility – for support are assessed (e.g. Eliassi 2015; Kullberg 2005; Fahlgren and Sawyer 2005). Yet, little is known about the paths between applications for and payments of the subsidy, particularly in relation to applicants’ intersectional positions. Further, divergences in social assistance assessments have, with few exceptions (Minas 2005; Stranz, Wiklund, and Karlsson 2016b), been studied by means of vignette methods (cf. Wallander 2009), while research considering actual applications is very limited (cf. Stranz, Wiklund, and Karlsson 2016a).

The present study provides an input to this knowledge gap by examining whether social worker decisions on assistance eligibility form patterns that can be related to applicants’ gender, country of birth and family situation. The study is based on quantitative data comprising all decisions on social assistance eligibility (n = 423) that were conducted during one month in 2012 in 25 randomly selected municipalities in Sweden. The analytical procedure is inspired by the concept of intersectionality in terms of intercategorical complexity (cf. McCall 2005). More specifically, the study addresses patterns on two levels: (1) intersectional patterns, according to which, for example, gender anticipates granting in different ways depending on the applicant’s country of birth or family type and (2) specific patterns, according to which other factors (than gender, country of birth and family type) anticipate decisions on granting for a certain subgroup. ‘Subgroups’ here refers to pairwise breakdowns based on the categories at issue (men/women; foreign-born/native-born; applicants with/without child custody and couple/single households). The following questions are addressed:

1. To what extent, and how, can gender, country of birth and family type be linked to decisions on granting of social assistance when considering intersections between these applicant characteristics?

2. Do the following factors anticipate decisions on granting of social assistance in ways that are specific to the subgroups at issue: (a) the social workers’ assessments considering applicants’ problems with substance abuse, mental and somatic health and/or financial debts; (b) contacts with public employment services/a social insurance office?

Background

Inclusive social rights, different outcomes

In an international context, the Swedish welfare state is often characterized as a social democratic regime based on ideals of equality and universalism (Esping-Andersen 1990; Pierson 2000). The strategy of poverty alleviation builds upon income-based social insurance schemes such as subsidized parental leave and benefits covering for unemployment and sick leave. Social assistance has the function of a targeted measure for persons not sufficiently entitled to and/or covered by the general security system, while lacking enough income from paid work (Korpi and Palme 1998; Nelson 2012). Compared with other European welfare states, the size of the benefit is relatively generous. However, it only allows for a minimum level of living compared with the majority population in Sweden (Bergmark 2013)

The task of distinguishing between eligible and non-eligible social assistance applicants falls under the remit of the municipal personal social services (PSS), whose main responsibilities also include child welfare services and substance abuse treatment for adults. While the PSS traditionally targets problems of a psychosocial nature, the need for social assistance can to a high extent be attributed to unemployment (e.g. Stranz and Wiklund 2012). Along with increasing emphasis on activation policy, the professionals’ assessments of an applicant’s efforts to participate in the labour force often plays an important role in the decision-making (e.g. Nybom 2011; Thorén 2008). Previous studies indicate that the threshold for applying for social assistance is high, and that the means-testing process is perceived as both tedious and associated with stigma (e.g. Mood 2004).

Along with gradual cutbacks in social insurance schemes over the past two decades (e.g. Hedborg 2016; ISF, 2014), the capacity of the Swedish welfare state to promote socioeconomic equality and universalism has been increasingly contested (e.g. Martinsson, Griffin, and Nygren 2016; Mulinari et al. 2009). Overall, income distribution is substantially differentiated by migration background, with lower median net incomes among the foreign-born share of the population (Statistics Sweden, 2016). Although the poverty risk reduces along with longer periods of stay in Sweden for most immigrant groups, the foreign-born are largely overrepresented in the unemployment statistics (Gustafsson 2013).

In terms of gender, Sweden has a high amount of dual-earner households compared with many other European countries. At the same time, women’s labour has largely taken place through the building of a female dominated public sector characterized by part-time jobs and low wages (Sainsbury and Morrisens, 2010). Although social rights are ultimately related to citizenship, actual access to the general funds is highly dependent on previous incomes and labour market participation (Sainsbury 2012). In consequence, the schemes provide weak protection for citizens with low wages and a peripheral position on the Swedish labour market (e.g. Gustafsson 2013; Stranz and Wiklund 2012) and imply an overarching disadvantage for women and immigrants (Nyberg 2014). Since social assistance is earmarked for those who are unable to support themselves by other means, the need for the subsidy is intimately linked to labour market dynamics and strengths and weaknesses in the upper layers of the safety net (cf. Bäckman and Bergmark 2011). In particular, subgroups disadvantaged by multiple structural constraints are to a higher extent than others left to turn to the social services for financial support. While single households are the major recipients of social assistance, the close link between family roles and gender becomes manifest in an exceptional overrepresentation of single mothers. Although the figures fluctuate over time, approximately one single mother in five relies on the subsidy, to be compared with one in 50 couple households (NBHW, 2017b). Further, the spells of receiving assistance are longer among immigrants and single mothers than the average, a tendency that is most pronounced among foreign-born single mothers (Stranz and Wiklund 2012).

This being said, it can be argued that the relatively small proportion of couple households among social assistance recipients should be contextualized from a feminist perspective. The benefit differs from most other social security schemes in Sweden by basing the eligibility criteria on the household (not the individual) as a financial entity (cf. NBHW, 2013). This means that social workers are able to deny an applicant benefits with reference to the income of a spouse or partner. Counter to the recognition of the family as an essential institution for women’s subordination, granted subsidies are indirectly assumed to benefit the household as a whole. Meanwhile, the criteria risks strengthening (gendered) relationships of dependency in couples with narrow economic margins (cf. Bennett, 2013; Hobson 1990; Orloff 1993). Although women’s ability to control their own finances has influenced a number of policy reforms for gender equality in Sweden (see, e.g. Sainsbury and Morissens 2010), the household based principle of social assistance eligibility is rarely paid attention.

As stressed by a number of post-colonial scholars, gender equality as it has been institutionalized in Sweden has primarily addressed the possibility for women to combine labour with family responsibilities, whereas unequal relations between subgroups of women have often remained invisible in the dominant feminist discourses (see, e.g. de Los Reyes 2014; Martinsson, Griffin, and Nygren 2016; Mulinari et al. 2009). However, the idea of Sweden as a ‘women friendly’ welfare state has become an integral part of the national identity, including an implicit projection of traditional gender roles onto foreign cultures and immigrants (especially middle-eastern and Muslim) (de Los Reyes 1998; Mulinari et al. 2009). Research on Swedish social work practice suggests that professionals, in classifying a client as ‘immigrant’ or ‘Muslim’, tend to make essentialist assumptions about gender-traditional family roles and to frame problems by means of cultural explanations (Eliassi 2015, 2017).

Social assistance assessments – previous research

Since the national regulation considering social assistance allows for a significant scope of municipal authority, local guidelines as well as administrative procedures vary across the country (Byberg 2002; Stranz 2007). Despite the implementation of a national norm for social assistance in 1998, municipal variation has been demonstrated in levels of generosity, where instead of serving as a minimum level, the norm has virtually come to function as a ceiling for obtaining subventions (Bergmark 2013; cf. Stranz and Wiklund 2012). Assessments of social assistance eligibility have been shown to vary with regard to the municipalities’ size and political majority, as well as the degree of specialization of the organizations (e.g. Hydén, Kyhle Westmark and Stenberg, 1995; Stranz 2007; Terum 1986). Differences have also been related to the professionals’ gender, attitudes towards recipients and such professional predictors as workload and work experience (Kullberg and Fäldt 2008; Stranz 2007).

So far, research considering the categorization of social assistance applicants has predominantly focused on gender-biased assessments of single mothers and fathers. In a number of vignette studies, social workers’ understanding of single parents’ problems and needs have been shown to favour a gender-stereotype breadwinner model (see, e.g. Kullberg and Fäldt 2008; Kullberg 2004, 2005). While the PSS professionals seem more intent on encouraging single fathers to seek jobs, they tend to emphasize single mothers’ parenting roles and social relations (Kullberg 2004, 2005). Single mothers are more likely to be granted support compared with the fathers (Kullberg 2005). This can be interpreted as in line with a historical understanding of ‘welfare dependency’ as specifically associated to women in general and single mothers in particular, and contrasted to the masculine-coded idea of ‘independence’ (cf. Fraser and Gordon 1994). The tendencies of greater generosity towards women and stricter assessments of male applicants have further been shown to be strengthened when social workers assess help-seeking persons of the same gender as themselves (Fäldt and Kullberg, 2012).

In a previous study based on actual applicant data, Stranz, Wiklund, and Karlsson (2016b) found that female applicants seemingly had greater chance of obtaining support compared with males, which was also observed among foreign-born applicants compared with those born in Sweden. This might indicate that gender-traditional expectations on men as breadwinners affect the decisions. However, it might also be explained by structural disadvantages for women and immigrants in establishing self-support (cf. Stranz, Wiklund, and Karlsson 2016b). Further, married/cohabiting applicants without children, as well as singles with children were shown to be less likely to be assessed as eligible compared with other household types (Stranz, Wiklund, and Karlsson 2016b).

Yet, no applicant is solely female or solely foreign-born, and patterns related to one basis of categorization might look different in various subsections of applicants. Taking these findings as starting point, the proposed analyses address the occurrence of intersectional diversity in decisions on social assistance eligibility. The comprehension of intersectionality in terms of intercategorical complexity implies that pre-existing social classifications are used as analytical tools to provide a more complex picture of applicant-related differences than has so far been examined empirically (cf. McCall 2005; Spierings 2012). Although the objective is not to deconstruct social categorizations, the interests of gender, country of birth and family composition should not be confused with an essentialist view of these classifications as static or ‘true’ (cf. McCall 2005; Spierings 2012).

Methods

The study builds on cross-sectional data concerning all applications for social assistance registered in 25 medium sized Swedish municipalities during April 2012 (n = 472).¹ The response rate was about 84 per cent and after accounting for item nonresponses, the sample comprised 423 applications. The municipalities were randomly selected from a sample of all municipalities with a population of 13,000–64,999 inhabitants (N = 138).² Information about the individual applicants and their cases were collected via questionnaires directed to the respective social workers (n = 138)³ in charge of the investigations.⁴ Four areas of the applicants’ background, living situation and cases were covered: background characteristics (such as gender, age, family type and housing conditions); education, employment status and income (concerning, e.g. the applicants’ possible taxable earnings, unemployment benefits and data on whether s/he was actively searching for a job); previous experiences of the personal social services (including other units and municipalities); and the administration of the application (e.g. use of standardized assessment tools and decisions on eligibility).

In addition, the social workers in charge were asked to assess the presence of ‘individual problems’, e.g. substance abuse and mental health of the applicant. These assessments refer to the social workers’ subjective perceptions and should hence be regarded as such.

Further, separate questionnaires addressed the social workers’ own characteristics (age, gender, country of birth and professional background such as education and years in social work). The variables involved in the analyses are presented in Table 1 with the following encoding⁵:

Table 1. Independent variables, percentages within gender/all; means (sd)

	Female applicants (n = 175)	Male applicants (n = 248)	All applicants (n = 423)
Independents
Gender, females			41.4%
Country of birth (ref. born in Sweden)	28.0%	33.9%	31.4%
Family type
Single without children (ref. category in analysis)	41.1%	68.1%	57.0%
Married/cohabiting without children	12.6%	14.5%	13.7%
Married/cohabiting with children	6.3%	10.1%	8.5%
Single with children	40.0%	7.3%	20.8%
Problems (ref. no)
Alcohol/illicit drugs	8.9%	15.3%	12.3%
Mental health impairments	34.3%	29.4%	31.4%
Somatic health impairments	23.4%	18.2%	20.3%
Financial debts	18.9%	18.1%	18.4%
Registered at public employment services/social insurance office (ref. no)	73.7%	68.6%	70.7%
Control variables
Age (Means and sd)	34.6 (12.5)	35.0 (12.7)	34.7 (12.6)
Previously known by PSS (ref. yes)	68.0%	63.7%	65.5%
[SW] Gender (ref. female)	89.7%	88.3%	88.9%
[SW] Age (Means and sd)	38.1 (12.4)	37.2 (11.9)	37.6 (12.1)
[SW] Density of SA administration	0.70 (0.33)	0.76 (0.31)	0.73 (0.32)
[SW] Standardized assessment (ref. no)	3.4%	6.1%	5.0%

Independents:

• Gender: dummy (male = reference; 1 = female)

• Country of birth: dummy (born in Sweden = reference; 1 = not born in Sweden)

• Family situation: type of household the applicant is living in (single person without children = reference; 1 = married/cohabiting with children; 2 = married/cohabiting without children; 3 = single person with children)

• Presence of problems assessed by the social workers in charge, dummies (reference = no; 1 = yes):

• Abuse of alcohol/illicit drugs

• Mental health impairments

• Somatic health impairment

• Financial debts

Control variables:

• Age: years (continuous)

• Previously known by PSS: the applicant had previously applied for social assistance. Dummy (reference = yes; 1 = no)

• Gender [SW]: dummy (male = reference; 1 = female)

• Age [SW]: years (continuous)

• Density of social assistance administration [SW]: number of years that the social worker has been managing social assistance divided by number of years in professional social work (continuous)

• Standardized assessment [SW]: The social worker used a standardized assessment tool during the investigation. Dummy (reference = no; 1 = yes)

The independent variables were included to address the research questions, that is, the applicants’ intersectional positions, their registration status according to job search and social insurance, and the social workers’ assessments of individual impairments. Further, social assistance assessments have been shown to covary with the social worker’s (SW) background and professional profile (e.g. Kullberg and Fäldt 2008; Stranz, Wiklund, and Karlsson 2016b). Since this dimension lies beyond the limited focus of the present study, those factors are treated as control variables. The same applies to the applicants’ age and aspects of the administration process – variables not unlikely to affect the outcomes but that are not the focus of this investigation.

Table 1 shows the distribution of the independent variables used in the proposed analyses, both as a whole and broken down by the applicants’ gender.⁶ Of the investigations, 41.1 per cent (n = 175) were registered on a female applicant, and the total proportion of foreign-born applicants was just under one-third (28 per cent of the female applicants and 33.9 per cent of the male applicants were born outside the country). Decisions on granting (reference = no, 1 = yes) served as the dependent variable. All in all, 73.7 per cent of the applicants were granted social assistance.

Analysis

Initially, cross-tabulation analyses were conducted in order to compare differences in granting between the subgroups at issue. For the purpose of investigating more complex patterns while accounting for the fact that the units of analysis were clustered by municipality,⁷ data were scrutinized using modified Poisson regression analysis, an approach that has proven to be particularly useful when dealing with clustered data (see, e.g. Zou 2004; Yelland, Salter, and Ryan 2011). The standard errors were clustered at the municipal level⁸ and relative risks (RR) according to granting of social assistance were estimated using Generalized Linear Models (GLM) in Stata, vers. 14. Four sets of regression models were conducted. First, the sample was divided by gender into Models A and B. The same procedure was carried out on the basis of country of birth, providing a subset of foreign-born applicants (Model C) and applicants born in Sweden (Model D). Further, Model E comprises all applicants in the data set, and examines intersections by means of combined identity variables (cf. Spierings 2012). Models F and G separate applicants without children from those with children. Whilst the regression analyses in Models A, B, C and D essentially reflect the Poisson regression model in Stranz, Wiklund and Karlsson’s study (2016b),⁹ Models E, F,G, H and I include interaction terms that together make up an intersectional identity matrix in terms of gender and country of birth. That is, instead of treating these factors as separate variables, they were merged into combined identity variables covering all possible sub-options (man born in Sweden; foreign-born man; foreign-born woman; woman born in Sweden) (cf. Hancock 2007; Spierings, 2012). Also, in the model E, F, G, H, I series, the family type indicator has been broken down to one parenthood variable and one concerning civil status (single vs. married/cohabiting). Taken together, the combination of regression models was developed to handle as high degree of intersectional complexity as possible within the limitations set by the sample size (cf. Spierings 2012). For instance, a ‘complete’ identity matrix including highly specific intersections such as ‘single foreign-born woman without children’ would end up representing too small subgroups.¹⁰

Methodological reservations

Although the analysis was guided by the endeavour for diversity and subgroup-specific patterns, not all intersections between gender, country of birth, parenthood and civil status were possible to account for simultaneously. Instead, the pursuit of intercategorical complexity was addressed by highlighting data from different angles, focusing on relational differences in granting between a limited set of intersections at a time. This being said, a larger dataset would have enabled more sophisticated analyses and could, potentially, have revealed a more detailed picture of intersectional granting patterns.

In terms of causation, the results should be interpreted prudently. Observed patterns might on the one hand indicate differences in the social workers’ willingness to grant social assistance to different groups of applicants, while on the other, the sample must be seen against the background of an unknown number of persons who abstain from seeking the benefit despite living in economic deprivation. Since applying for social assistance requires an active effort and implies close contacts with the social services, it cannot be excluded that the propensity to apply varies at a group level, and that such behaviours are reflected in patterns of granting (cf. Mood 2004). Also, although no substantial policy changes has been made in the field of social assistance since 2012, it should be emphasized that the cross-sectional design limits the survey to reflect decisions made during a specific month in a specific year. This approach further implies that the temporal dimension of social assistance applications and recipients is not captured, since data do not account for spells of recipience, interruptions in dependency or entries and exits into social assistance (cf. Jobane and Ellwood 1986). Likewise, changes in the applicant’s family situation that are likely to appear during a life course are not captured and analysed in relation to the applications.

Further, a few remarks can be made concerning the operationalization process. The dichotomy between native/foreign-born makes visible fundamental differences related to ethnicity. This is, however, a crude measure insofar as it neither captures further diversity among the foreign-born (i.e. related to duration of stay and country of origin) nor among the native-born (i.e. parents’ country of birth). Secondly, applicants in married/cohabiting households are economically assessed as one entity when applying for social assistance, but are given different positions in the registration process, so that while one of the individuals functions as ‘head’ of the registration, the other is designated ‘co-applicant’ (NBHW 2017a). Present data are limited to reflect the ‘head’ of the application, while co-applicants’ gender and country of birth (and other individual data) remain unknown. In the analyses, couple households are treated in the same way as single households. However, there is a lack of knowledge concerning the determination of the two positions of registration, including actual implications for the ‘individuality’ of the assessments. This in itself is a concern for future research, and a ‘black box’ that might affect the results (cf. Bennett, 2013).

Results

Table 2 shows, to begin with, that a greater proportion of the female applicants were granted social assistance compared with the male group, whereas the distinguishable share of granting when comparing foreign-born and native-born applicants is less apparent.¹¹ Further, both parameters of family composition seem to have bearing; childless applicants are less frequently granted social assistance (SA) compared with parents, and the same goes for couple households compared with singles.

Table 2. Decisions on social assistance eligibility. Percentages within subgroups. Pearson’s Chi²-test

	Granted social assistance	χ^2 (P)
Women/Men	80%/69%	6.0070 (0.014)
Foreign-born/Native-born	80%/71%	3.5369 (0.060)
Parent/Non-parent	81%/55%	6.2821 (0.012)
Single/Couple	76%/65%	4.9075 (0.027)

In Models A and B (Table 3), the gender separate analyses show that being born outside of Sweden is linked to an increased chance of being granted SA among men, though not among women. Further, having a family seems altogether to have a negative impact on women’s chances of approval, irrespective of whether they share their household with children, a partner, or both. The tendency of a ‘female family penalty’ is most pronounced for mothers living with a partner and who serve as the head of the application. In contrast, equivalent households with a male head-applicant are tangibly more likely to receive the subsidy than the reference group of males (singles w/o children). This raises questions about the means-testing procedure of couple households, not least considering the determination of ‘head’ and ‘co-applicant’ and its implications for the further investigation.

Table 3. Multiple modified Poisson regression (with cluster adjusted standard errors) on factors related to whether social assistance is granted or not. Relative risks (RR)

	Model A Male applicants (n = 248)		Model B Female applicants (n = 175)
	RR	CI (95%)	RR	CI (95%)
Independents
Country of birth (ref. born in Sweden)	1.337***	1.155–1.546	1.210	0.999–1.465
Family situation (ref. single w/o children)
Married/cohabiting with children	1.283***	1.076–1.531	0.616**	0.456–0.825
Married/cohabiting without children	0.729*	0.545–0.975	0.728**	0.537–0.986
Single with children	0.896	0.685–1.171	0.733***	0.597–0.900
Individual problems (ref. no)
Alcohol/illicit drugs	1.180	0.936–1.488	0.624**	0.409–0.951
Mental health impairments	1.160	0.904–1.487	1.093	0.810–1.474
Somatic health impairments	1.086	0.871–1.354	1.109	0.952–1.292
Financial debts	0.797	0.602–1.054	1.220	1.000–1.490
Registered at public employment services/ social insurance office (ref. no)	1.347***	1.142–1.588	1.162	0.980–1.380

*p < 0.05; **p < 0.01; ***p < 0.001. Adjusted for previous contact with the PSS, social worker factors (gender, age, and density of SA administration), use of standardized assessment tool and municipality.

A closer look at Table 3 reveals that female applicants are less likely to be granted assistance if they are assessed as having a problem with abuse of alcohol or illicit drugs. At the same time, substance abuse does not seem to affect assessments of men, which indicates a somewhat gendered morality linked to abuse (cf. Samuelsson 2015). Similar thoughts are brought up by the variable for applicants’ contact with a public employment office and/or a social insurance service. Such a registration is clearly beneficial for male applicants’ chances of approval, but appears to be irrelevant for the assessment of women. In line with observations from previous research, this suggests that social workers put more emphasis on men’s attempts to provide for themselves by paid work (cf. Kullberg 2004, 2005, 2006).

When studying foreign-born and native-born applicants separately, Models C and D (Table 4) suggest that the gender difference according to granting (cf. Table 2) somehow depends on the applicant’s country of birth. While being female is clearly associated with a greater chance of approval for applicants born in Sweden, such a ‘gender effect’ cannot be observed among immigrant applicants. Yet another difference emerges at the intersection between country of birth and family situation, as parenthood is linked to a decreased chance of granting among native-born singles, which cannot be seen among the foreign-born. In both subgroups, couple households without children are less likely (than childless singles) to receive SA.

Table 4. Multiple modified Poisson regression (with cluster-adjusted standard errors) on factors related to whether social assistance is granted or not. Relative risks (RR)

	Model C Foreign-born applicants (n = 133)		Model D Applicants born in Sweden (n = 290)
	RR	CI (95%)	RR	CI (95%)
Independents
Gender (ref. male)	1.104	0.910–1.341	1.237**	1.061–1.443
Family situation (ref. single w/o children)
Married/cohabiting with children	0.989	0.802–1.220	0.934	0.654–1.334
Married/cohabiting without children	0.592**	0.404–0.867	0.786**	0.671–0.920
Single with children	0.892	0.628–1.266	0.842*	0.712–1.000
Problems (ref. no)
Alcohol/illicit drugs	1.012	0.711–1.438	1.015	0.845–1.216
Mental health impairments	0.972	0.780–1.212	1.146	0.966–1.358
Somatic health impairments	1.161	0.952–1.417	1.040	0.906–1.193
Financial debts	1.006	0.766–1.320	0.895	0.731–1.100
Registered at public employment services/ social insurance office (ref. no)	1.261*	1.020–1.560	1.217*	1.040–1.423

*p < 0.05; **p < 0.01; ***p < 0.001. Adjusted for previous contact with the PSS, social worker factors (gender, age, and density of SA administration), use of standardized assessment tool and municipality.

In Table 5, the concepts of gender and country of birth are analytically captured by combined identity variables in all the models. Model E includes the whole sample and in this model, household composition has been broken down into two variables: one considering parenthood, and the other one separating couples from single households.

Table 5. Multiple modified Poisson regression (with cluster-adjusted standard errors) on factors related to whether social assistance is granted or not. Relative risks (RR)

	Model E All applicants (n = 423)		Model F Applicants without children (n = 299)		Model G Applicants with children (n = 124)
	RR	CI (95%)	RR	CI (95%)	RR	CI (95%)
Independents
Parent (ref. no)	0.976	0.869–1.095
Married/cohabiting (ref. no)	0.873**	0.776–0.983	0.719***	0.606–0.853	1.129	0.807–1.580
Gender and country of birth (ref. male born in Sweden)
Woman, foreign-born	1.381***	1.176–1.622	1.360***	1.151–1.606	1.375	0.952–1.985
Woman, born in Sweden	1.209**	1.061–1.378	1.305**	1.107–1.535	0.986	0.774–1.256
Man, foreign-born	1.359***	1.164–1.587	1.279**	1.066–1.535	1.481**	1.101–1.992
Problems (ref. no)
Alcohol/illicit drugs	1.020	0.874–1.190	1.032	0.871–1.222	1.170	0626–2.186
Mental health impairments	1.13	0.974–1.132	1.147	0.955–1.377	1.094	0.825–1.451
Somatic health impairments	1.068	0.957–1.193	1.031	0.873–1.216	1.105	0.857–1.426
Financial debts	0.925	0.801–1.069	0.882	0.745–1.045	1.002	0.749–1.340
Registered at public employment services/social insurance office (ref. no)	1.234**	1.081–1.408	1.129**	1.002–1.273	1.688**	1.176–2.422

*p < 0.05; **p < 0.01; ***p < 0.001. Adjusted for previous contact with the PSS, social worker factors (gender, age, and density of SA administration), use of standardized assessment tool and municipality.

In line with the results from the initial cross-tabulation, Model E shows that being part of a cohabiting/married household seems to decrease the likelihood of receiving support at an overall level. However, parenthood does not appear to be important when controlling for other variables. Model E also shows that there is no observable linkage between problem assessments and decisions on granting at an overall level. However, applicants registered at an employment service or social insurance office are generally more likely to be assessed as eligible for SA than others.

According to intersections between gender and country of birth, Model E further shows that the reference category – male born in Sweden – corresponds to the smallest likelihood of being granted SA. This position differs significantly from both foreign-born applicants (regardless of gender) and native-born women. In turn, foreign-born women have the highest chances of approval, closely followed by foreign-born men and native-born women. This suggests that individuals located in positions that structurally encounter disadvantages on the labour market are responded to with a greater willingness to approve assistance. However, considering the minor differences between the intersectional positions, it should be kept in mind that native-born men serve as the reference category, and that the gender difference between foreign-born applicants was not statistically significant when tested in Model C (Table 4).

Models F and G (Table 5) review specific patterns related to parenthood by examining applicants with/-out children respectively. In the childless part of the sample (Model F), intersections between gender and country of birth are similarly related to decisions on granting as in the sample as a whole (Model E). In contrast, Model G shows that the variable representing foreign-born men is linked to the greatest chances of approval among parents. Further, and unlike what was shown in the childless subsample, neither native nor foreign-born mothers do differ significantly from the subgroup of native-born fathers. Here, parenthood emerges as a variable with intersectional features, indicating significance of both gender and country of birth.

The second dimension of family composition – civil status – is examined in Table 6 (Models H and I). Starting out with single applicants (Model H), the differences between the intersectional identity positions are partly similar to those in the whole sample (cf. Model E). Compared with native-born men, all other subgroups are more likely to get a decision on granting, and here too, foreign-born women represent the major difference. Further, single applicants who have debts are less likely to receive social assistance – which has not been noted in any other subsample – but increase their chance of granting by being registered at a public employment or social insurance office.

Table 6. Multiple modified Poisson regression (with cluster-adjusted standard errors) on factors related to whether social assistance is granted or not. Relative risks (RR)

	Model H Single households (n = 329)		Model I Couple households (n = 94)
	RR	CI (95%)	RR	CI (95%)
Applicant factors
Parent (ref. no)	0.852	0.705–1.029	1.268	0.723–2.223
Gender and country of birth (ref. male born in Sweden)
Woman, foreign-born	1.515***	1.239–1.856	1.498	0.806–2.784
Woman, born in Sweden	1.249**	1.081–1.444	1.477	0.769–2.839
Man, foreign-born	1.264**	1.088–1.469	1.796*	1.041–3.100
Problems (ref. no)
Alcohol/illicit drugs	1.031	0.854–1.244	1.090	0.655–1.815
Mental health impairments	1.152	0.977–1.359	0.954	0.603–1.510
Somatic health impairments	0.982	0.848–1.137	1.781**	1.160–2.734
Financial debts	0.831*	0.722–0.957	1.110	0.616–2.000
Registered at public employment services/social insurance office (ref. no)	1.172*	1.024–1.342	1.727	0.943–3.163

*p < 0.05; **p < 0.01; ***p < 0.001. Adjusted for previous contact with the PSS, social worker factors (gender, age, and density of SA administration), use of standardized assessment tool and municipality.

As previously stressed, applicant data concerning cohabiting/married couples (Model I) only refers to the head of the application so the results should be interpreted with care. This is further underlined by the fact that the number of – especially female-headed – couple households in the sample is small.

It is however worth noting that among cohabiting/married couples (Model I), the relations between the intersectional positions stand out from previous analyses. The only subgroup that differs significantly from native-born men is foreign-born men, and the latter subgroup is as much as 80 per cent more likely to receive social assistance. This is particularly notable since foreign-born women were found to have the highest chances of approval not only in the sample as a whole, but also among childless households as well as single households (Models G and H). In contrast to the observation of a ‘female family penalty’, thus, being part of a household with two adults and/or child(ren) seems to enhance the chances of approval for men in general and those born outside of Sweden in particular. Though it is hard to tell from present data, patterns here may indicate presence of racialization in relation to how social workers conceive of gendered family roles. Apart from this, the variable indicating somatic health impairments emerges as important among couple households. In couples where the head of the application suffers from somatic issues, the possibility of being granted assistance is substantially higher.

Concluding discussion

The aim of this study is to examine intersectional patterns in social worker decisions on social assistance eligibility. The findings can be summarized as follows:

First, decisions on granting social assistance form gender-specific patterns. In line with previous research, the results indicate that social workers have higher expectations on men as breadwinners. By putting greater stress on the male applicants’ efforts to gain other sources of income, the professionals seem to condition social assistance eligibility more strictly by a work-oriented agenda for men than for women. In light of previous research on assessments of single parents, present results extend the impression of gender-biased assessments by indicating that men and women, regardless of family type, tend to be perceived according to different standards. Not only does this pattern highlight the general risk of legal uncertainty implied in judgments on individuals’ motivation for self-maintenance, but also, the difficulty of dissociating such concepts as ‘work ethic’, ‘welfare dependency’ and ‘worthiness’ from interwoven ideas of gender. Classifications referring to help-seeking persons’ willingness to work might both enable and conceal moral judgements that are ultimately dependent on their social characteristics. As legitimacy criteria for the exclusion of certain individuals from welfare state support, distinctions between the ‘deserving’ and the ‘undeserving’ poor have historically also taken form in moral judgements that are less related to economic values. For instance, gendered and racialized ideas of the ‘good citizen’ can be embedded in perceptions of certain bodies as threatening, respectable or undesirable (cf. Skeggs 2004). Such reasoning might perhaps help to explain the other gender-specific pattern found in the analyses. Unlike men, women clearly are perceived as less legitimate recipients if seen as addicted to alcohol or drugs. While male substance abusers constitute a group well recognized by the social services, it cannot be ruled out that female substance abuse is viewed as conflicting with stereotypical ideas of ‘deserving’ femininity. This being said, the possible tendency of penalizing women who are seen to have masculine-coded psychosocial problems is an issue that needs to be addressed in further research.

Secondly, granting rates largely reflect patterns of unemployment when considering intersections between gender and country of birth. That is, foreign-born women are most likely to be granted support whereas native-born men embody the opposite position. These patterns are consistent with actual inequalities in access to paid work and social security schemes, and can be read as an expected outcome. At the same time, the dichotomy between natives and immigrants gives a limited and simplified view of how power relations linked to ethnicity and race affect the applicants. We know for instance that poverty risks vary considerably between different immigrant groups, and that duration of stay in Sweden is an important predictor for employment (cf. Gustavsson, 2013). Further, processes of racialization are often embodied in relation to features that go beyond one’s country of birth, such as name, colour, religion and language skills. In light of the above findings, however, we must raise questions about the extent to which decisions on social assistance eligibility are informed by practices of categorization rather than by individual circumstances. It is not inconceivable that the PSS professionals are more inclined to approve applications for those that they – according to their categorization – ‘know’ to face difficulties in both entering and remaining in the labour market. Such logic can be contrasted to the ‘creaming’ strategy of improving the outcomes of activation measures by focusing on clients that are recognized as most likely to enter employment (cf. Nybom 2011). In this context, it is important to keep in mind that social assistance transfers only allow for a minimum level of living compared with the majority population of Sweden, and do not reduce poverty and inequality in any broader sense (cf. Korpi and Palme 1998). Considering social assistance as a less favourable safety net for the already disadvantaged, the subsidy risks legitimizing wide disparities in the preconditions of inclusion in the Swedish welfare state’s principal security system.

Thirdly, the results raise questions about intersectional aspects of the family dimension that have been sparingly attended to in previous research on social assistance assessments. Among both parents and couple households, foreign-born men appear to have the greatest chances of approval. At the same time, parenthood emerges as a factor that generally seems to decrease the chances of granting for women. This is a somewhat surprising finding, since the care of children both constitutes a potential obstacle for full-time employment (which particularly affects single mothers) and implies additional household expenditure. Moreover, the organizational closeness between social assistance administration and child welfare services might heighten the threshold for parents to apply for the subsidy, suggesting that the need for financial support is certainly pronounced among those who actually do (cf. Franzén, Vinnerljung, and Hjern 2008). Although it is hard to explain this pattern, it cannot be ruled out that a symbolic, and gendered, linkage between parenthood and responsibility is materialized in a tendency among the PSS professionals to attribute help-seeking mothers’ financial difficulties to individual (rather than structural) shortcomings – at least more than they do in assessments of single women without children. Another possible interpretation could be that the professionals assume mothers with poor employment prospects to prefer remaining at home with their children while depending on welfare rather than engaging in badly paid work.

This being said, the impact of the applicants’ family situation on decisions of eligibility should be examined more thoroughly in further research. This also applies to the importance of applicants’ civil status for decisions on granting, and the possible impact of applicants’ gender and migration background for the ways in which notions of family are given meaning. As already noted, some of the findings might be related to the determination of ‘head’ and ‘co-applicant’ in couple households, and it is not impossible that these positions are somehow assigned through processes of gendering and racialization.

Taken together, the results strengthen the impression of means tested social assistance as a practice marked by the PSS professionals’ categorizations. However, the nature of the empirical data has enabled analyses of patterns that cannot simply be said to represent ‘assessment biases’, but which also might correspond to help-seeking behaviours, work opportunities and the distribution of income and welfare that precedes social assistance. As the final hope for citizens who ‘fall through’ the general safety net, the need for social assistance can ultimately be said to highlight intersections marked by multiple subordination. In this context, the localization of means testing in an organization oriented towards individual problems risks normalizing social exclusion as an expected condition among help-seeking persons with certain characteristics, while hindering access to support for individuals who, due to their social positions, are expected to be economically self-reliant.

Finally, the present study has more or less exclusively been focusing on decisions on social assistance eligibility. In terms of patterns, these decisions can be seen as a measurable outcome of a major field of social work practice in Sweden. However, one should keep in mind that practical work with social assistance does not only consist of determining eligibility, but also assisting individuals and families in becoming self-sufficient. The latter dimension is by no means isolated from the first, and entails great challenges for the professionals. While municipal based social work often is characterized by high workloads, many applicants face actual and very manifest constraints in providing themselves by working – constraints that, regardless of how ‘structural’ they are in nature, matter and materialize at an individual level.

Disclosure statement

No potential conflict of interest was reported by the author.

Additional information

Funding

The work was funded by FORTE (Swedish Research Council for Health, Working Life and Welfare), Grant 2009-0790.

Notes

1. The survey was approved by the Regional Ethical Review Board in Stockholm.

2. In the municipalities included in the sample, the median number of inhabitants was 18 128.

3. Eighty-nine per cent of the applicants were assessed by a social worker with a B.Sc. in social work.

4. For further details regarding the data collection, e.g. how the social workers were matched with the proper applications, (see Stranz, Wiklund, and Karlsson 2016a, 2016b).

5. Variables marked with [SW] concern the social worker in charge of the assessment; other variables consider the subject of the application.

6. The presentation was chosen to make visible diversity within the sample and the distribution according to intersections between gender and family situation as well as between gender and country of birth (cf. Spierings 2012).

7. As mentioned, the national legislation regulating the right to social assistance allows for a substantial scope for municipal authority, and previous research shows that assessments vary with regard to a number of municipal factors (e.g. Byberg 2002; Hydén, Kyhle Westermark, and Stenberg 1995; Stranz 2007; Terum 1986).

8. In addition, dummy variables for the municipalities were included in the model in order to control for local heterogeneity.

9. Unlike the reference study’s procedure of analysis, however, bivariate regression models were discarded.

10. This also applies to the possibility of further dividing the sample into even more group-specific models.

11. This is consistent with the bivariate analysis in Stranz, Wiklund and Karlsson’s study (2016b). However, in their study, being foreign-born was shown to increase the probability of granting in the multivariate analysis.