The use of independent foster care agencies by Swedish local authorities: Do structural factors matter?

ABSTRACT In Sweden, as in several other European countries, for example, Norway and the UK, social welfare service provision is increasingly being exposed to market competition. One service now being provided on the welfare market is foster care, as it is made available by independent foster care agencies (IFAs). However, current knowledge about the use of IFA is limited. To address this knowledge gap, this study explores (a) the extent of IFA use nationally and (b) whether structural factors (e.g. geographical location, socio-economic factors, local political governance) are associated with IFA use. The use of IFAs was analysed through a national survey answered by local authorities responsible for out-of-home care for children. The results of analysis indicate that approximately 90% of municipalities use IFA and on average 25% of foster care is provided by an IFA. A regression analysis shows that there are significantly higher odds of IFAs providing 20% or more of the foster care placements in small municipalities. No other significant correlations were found. Further, the results indicate a need for further in-depth studies to understand the use of IFAs and the rationality behind that use.


Introduction
Family foster carethe placement of children in a private family settinghas traditionally been provided through an agreement between a family and the local child welfare authority, where volunteering families are remunerated by municipal social services (Höjer, 2012). This type of traditional foster care has in recent decades been supplemented by foster care through independent providers. This means that public local authorities can buy 'places' in foster families from independent foster care agencies (IFAs). IFAs are used in Sweden, as well as in, for example, the USA (Huggins-Hoyt et al., 2019), the UK (Begg, 2011;Sellick, 2007), Australia (Barber, 2001) and Norway (Begg, 2011). Some IFAs are for-profit companies (Sallnäs & Wiklund, 2018;Sellick, 2007;Wiklund, 2011).
The emergence of IFAs is related to the New Public Management (NPM) reforms that many western welfare states have undergone in recent decades (Pollitt & Bouckaert, 2011). This means that public welfare organisations have been influenced by how the private market is organised. The concept of marketisation is an important point of departure for gaining an understanding of the change in foster care services. Marketisation is used differently by different scholars, but one common element of marketisation is that public local authorities go from providing almost all welfare services to increasing the number of services procured from private providers, thus creating a welfare market (Gingrich, 2011;Leibetseder et al., 2017;Pollitt & Bouckaert, 2011;Sallnäs & Wiklund, 2018). This article focuses on the marketisation of foster care, or, in practice, the local authorities' outsourcing of foster care provision.
IFAs are an international phenomenon and a service targeting a vulnerable population and consequently, it is of importance for social work policy and practice. Yet, the marketisation of foster care is a research area which has received limited attention. Previous research is scarce and mainly explores organisational explanations for using IFAs (e.g. difficulties in recruiting foster homes, Colton & Williams, 2006;Löfstrand, 2009;Oslo Economics, 2020 and the need for specialised foster homes, e.g. Sellick, 1999; the analysis of the Swedish Agency for Health and Care Services [SAHCS], 2016). However, studies about other welfare services suggest that structural determinants (e.g. geographical location, local socio-economic circumstances and political governance) may also be relevant for understanding marketisation (Gingrich, 2011). This study explores whether structural determinants may be important for explaining the use of IFAs in Swedish municipalities/local authorities, which to the author's knowledge has not been done previously in relation to IFAs.
In Sweden, foster care is the most common form of out-of-home care for children, but there are no official statistics on the use of IFAs (National Board of Health and Welfare [NBHW], 2020;SOU, 2014, p. 3). Limited cross-sectional data on IFA use have been collected (e.g. SOU, 2009, p. 68;SAHCSA, 2016), however, the results are dated and IFAs have not been studied systematically yet (cf. Meagher et al., 2016;Sallnäs & Wiklund, 2018). By utilising data from a national survey, the purpose of this article is to describe and analyse the use of IFAs in Sweden and to analyse the differences in use between municipalities/local authorities on a structural level.
In order to do this, the following questions are addressed: − To what extent is IFA used by Swedish local authorities? − To what extent are structural factors (e.g. local demographics and socio-economics) associated with IFA use?

Foster care in Sweden
In order to understand the organisation of foster care in Sweden, it is important to know that the municipalities/local authorities are one of the main actors in the welfare system, as they manage large parts of the social services (Meeuwisse & Swärd, 2002). These 290 self-governing municipalities/local authorities are responsible for the provision and supervision of foster care (NBHW, 2017(NBHW, , 2019Social Services Act, 2001). Foster care is a layman assignment according to Swedish law, and local authorities are mandated to arrange kinship care if deemed appropriate (Social Services Act, 2001;Linderot, 2020). However, foster care is becoming more professionalised in some respects. Foster care as a substitute for a biological family has changed because the focus is now directed towards reuniting families (Bergman, 2011;Bryderup, 2017), and enhanced foster care with additional services or structured treatments has been developed (Åström et al., 2020). The IFAs could be seen as part of this development, as the assumption is that this is a more professional type of foster care with extra support for foster carers and, in some cases, for children in care (SOU, 2009, p. 68).

Research on IFAs
To contextualise the present study with previous knowledge, studies about the use, the structural factors that may be associated with its use and the consequences of privately provided foster care services will be described.
What do we know about why IFAs are used? IFA use has reportedly been motivated by the need to recruit foster carers for children in general (Colton & Williams, 2006;Corbillon, 2006;Löfstrand, 2009;Oslo Economics, 2020;Pålsson et al., 2022), but especially for children with complex needs, also due to the need to provide extra supervision and support for foster carers and support for children in care. Further, there are reports of it being used to place a child in care for short-term foster care at short notice and also for unaccompanied minors (Rambøll, 2018;SOU, 2009, p. 68;Swedish Competition Authority, 2017).
There is limited data on the possible association between structural factors and IFA use. Geographical variations in IFA use has been noted (i.e. within the UK) (Sellick, 2011). No similar studies on the Swedish situation exist. Previous research in IFA marketisation is limited but the marketisation of other welfare services suggests structural factors that have been shown to be associated with the use of private welfare provision. For instance, in local authorities where one welfare service is highly outsourced, other welfare services are more likely to be privately provided as well (Hartman, 2011a(Hartman, , 2011bJordahl, 2013). Conversely, as regards most welfare services, private providers are less common in small municipalities (Jordahl & Blix, 2021;Meagher & Szebehely, 2019). In addition, private provision has been found to be associated with municipal demographics (e.g. age distribution, level of education, the number of people with a foreign background within the population and the political majority) (Hartman, 2011a(Hartman, , 2011b. For instance, centre-right majority municipalities/local authorities tend to outsource a larger share of services compared to left-wing majority municipalities/local authorities (Jordahl & Blix, 2021;Meagher & Szebehely, 2019). Moreover, the availability of private providers within the municipality has been associated with a more private provision of welfare services (Jordahl, 2013).
What does previous research tell us about the consequences of IFA use? Some efforts have been made to compare IFA care to publicly provided services but no conclusive results have been reached concerning the quality of services (e.g. Hollingsworth et al., 2010;Rambøll, 2016;Selwyn et al., 2010;Steen & Smith, 2012). A small-scale Swedish study indicated that children in IFA care felt less safe and well, had less impact on their everyday life, had less access to the foster carers, and reported to a lesser extent respectful treatment by the foster carers and educational support (Spånberger Weitz et al., 2017). Further, IFA care may be associated with higher costs for the local authorities (Narey & Owers, 2018;Sallnäs & Wiklund, 2018).

Data and method
This study is based on data from a national survey exploring the use of independent foster care supplemented with data from statistic databases and data on IFAs procured from NCSW. The requirement concerning ethical approval was waived by the Swedish Review Authority [ref. 2020-00509].
The e-survey was designed by senior researchers and a research assistant and sent to first-line managers of child welfare agencies in thirteen city districts in Stockholm for feedback before the full survey was carried out. Minor revisions were made. A list of all local authorities was compiled and through webpages and phone calls, the email addresses of first-line managers of foster care departments were collected and a link to the survey was disseminated using the survey service Survey and Report. Reminders were sent to non-responding local authorities during the summer and autumn of 2020.
The survey was sent to social services units responsible for foster care in all of Sweden's 290 municipalities. Since the responsibility for foster care in the two largest cities is divided between city districts, the survey was sent to 311 local authorities (including city districtsthirteen in Stockholm and ten in Gothenburg). The survey was completed by 264 respondents, yielding a response rate of approximately 85%. Only 1% was lost due to non-responses in the included variables (n = 261). 1 Three small municipalities declined to answer the survey due to a declared lack of time. Attempts were made to contact non-responding municipalities by telephone. Where there were missing answers or improbably high or low figures, the respondents were contacted for clarification and the results rectified accordingly. The responding municipalities ranged from very small rural municipalities to large city districts and covered the entire country, from east to west as well as north to south, which is important as Sweden is geographically diverse with a more densely populated southern part with a younger population and a less populated north with an older population (Statistics Sweden, 2021).
To explore the use of IFAs from a structural perspective, the survey data was supplemented with official register data on municipalities (see Table 1). Data on local political governance, the private provision of other welfare services and socio-economic variables were collected from the Kolada database supplied by the Council for Promoting Municipal Analysis [Rådet för främjande av kommunala analyser] (https://www.kolada.se/). Population data (population and population density) were gathered from Statistics Sweden (Statistics Sweden, 2021), using the most recent data available, typically from the year 2019 or 2020. The location of IFA companies was retrieved from a list of all authorised providers requested from the National Board Health and Social Care Inspectorate by email in May 2021, as these data are not available in any public databases.
In order to answer the first question about the extent to which IFAs are used by local authorities, the frequency of IFA use was analysed in all the responding local authorities, 2 including the number of children in foster care and children in IFA care (n = 261). In order to obtain an answer to the second question, the one concerning structural factors, the analysis was made in three steps. Firstly, the analytical sample was prepared, descriptive data were used to identify outliers, and a correlation analysis was conducted to control for a strong correlation between the independent variables described above. All correlations were below 0.8. Descriptive statistics were run for the analytical sample (Table 1). Secondly, the major municipalities of Stockholm and Gothenburg were excluded since these are divided into city districts. Even if these cities are divided into smaller administrative units, they still have shared guidelines and local political governance. This means that they are not interchangeable with local authorities/municipalities in all circumstances. The first course of action would have been to compile a municipal mean using the city districts' answers, but since there were non-responses from city districts (nine out of the thirteen Stockholm districts and four out of the ten Gothenburg districts responded), it was impossible to calculate the total percentage of IFA use for Stockholm and Gothenburg. For this reason, they were omitted in the regression analysis in relation to the independent variables at the municipal level due to the missing data. 3 Hence, the use of IFA in the city districts is discussed separately. The Stockholm and Gothenburg city districts and the municipalities with missing data on any of the independent variables were excluded, leaving an analytical sample of 236 municipalities included in the regression analysis (89% of the original sample).

Data analysis
Binary and multiple logistic regression were run to test for correlations between IFA use for at least 20% and the structural variables described above. Considering the interest in including the local authorities using no IFAs and the skewed distribution of the data, logistic regression was used (e.g. Bjerling & Ohlsson, 2010). A cut-off point regarding the dependent variable IFA use was set at 20% or more. The cut-off point was mainly motivated by two considerations. Firstly, every fifth foster care placement was considered as a substantial part of the foster care provision. Secondly, it provided a large enough part of the sample (49%) to allow for an analysis of a range of structural factors. Residuals had an approximately normal distribution. Logistic regression was performed testing each independent variable (bivariate analysis, Table 2). Controlling for unobserved confounders would require another type of analysis and observed confounders is an issue since few possible confounders are suggested in the theory or previous studies for this explorative study. Further, continuing the exploration of the variables included, all the independent variables were included in a multivariate analysis to test whether the associations and their significance would be affected and, if so, in what direction (multivariate analysis, Table 2). To test the bivariate correlations, alternative cut-off points and a linear regression of the log-transformed data were used. The correlations shown below were also significant in this alternative analysis (not in the table). The results of the regressions are presented as odds ratios (OR) with 95% confidence intervals. The statistical significance was defined at p < 0.05. Analyses were performed using IBM SPSS 26.0 software.

Results
The results are described below in two parts. First, they address the question of the extent of and variations in IFA use. Second, the possible structural determinants and IFA use are explored.
To answer the first question concerning the local authorities' use of IFA, the results indicate that approximately 90% of local authorities reported using some IFA placements at the time of the survey. Local authorities reported on average that approximately 25.3% (n = 261, not in the table) of fostered children were in IFA provided care, and 2571 children were reportedly placed through an IFA. The percentage of privately provided foster care is high in comparison with private preschool and after school care (8.3%, Kolada, 2019a) and private elderly care (7.5%, Kolada, 2019b) in the sample (n = 236).
There was a major difference in IFA use within the sample, since IFA use ranged from no use to only IFA care ( Figure 1). As can be seen in the histogram (Figure 1), many local authorities use IFAs for between zero and 20% of foster care placements, while fewer local authorities use IFA for a majority of (foster care placements) them. Only 9% of local authorities reported no use of IFAs and less than 1% used only IFAs. The majority of local authorities reported using IFAs, however, close to half (49%) used them for every fifth child and less than 15% used an IFA for 50% or more of the children in foster care.
In relation to the size of the local authorities' foster care population, the data show that the local authorities responsible for many children (≥140 children) all use an IFA for some placements, whereas only a fifth of local authorities responsible for a few children in foster care (<40) use IFAs (survey data from current study, not in the table). On the other hand, 75% of the eight local authorities using 80% or more IFAs had a population of under 12,000. This indicates that both a large foster care population and a small population had a positive link with IFA use. Concerning the city districts, as discussed in the method section, they are the administrative units responsible for foster care in Stockholm and Gothenburg. Data show that there was a considerable difference between the city districts of the major cities. The Gothenburg city districts ranged from 5 to 43% IFA use and the Stockholm districts ranged between 13 and 45% IFA use. In Stockholm, 66% of the city districts used 20% IFAs or more, while the same figure for Gothenburg was only 25%. The major differences between IFA use indicate that the use of IFAs is related to very local conditions or considerations and that there is no consistent pattern of IFA use for the larger cities.
To answer the second question of this study, a smaller study sample was used due to one or more of the included variables being missing. Estimates from logistic regression are outlined in Table 2. Subsequently, the link between IFA use and socio-economic and demographic factors and IFA use was explored. When it comes to a possible link between IFA use and socio-economic and demographic factors, there were small differences. The link between IFA use and taxable income, immigrant population and population density appear to be weak. As regards IFA use in relation to the share of the population with an immigrant background, there appears to be a negative link between a greater immigrant population and the use of IFAs for 20% or more of foster placements.
The most visible pattern of IFA use was regional differences (Table 1). In comparison with the southernmost region, regional differences appear to be correlated with higher IFA use in the central parts and in the north of Sweden. Parts of the Swedish southwest coastal region also has greater IFA use compared to the southern region. In order to try to find possible patterns related to regional differences, the link between a high level of use of IFAs (50% or more) and an excessive use (80% and more) was explored (not shown in the Table). The 41 local authorities using at least 50% IFAs and the 10 local authorities using 80% IFAs or more were selected and marked out on a map. This analysis indicated that the three southernmost regions had no responding municipality using more than 50% IFAs, while clusters of adjacent local authorities in the west coast region and in the central parts of Sweden had a similarly high level of use of IFA. This less sophisticated method was limited by being more sensitive to errors related to missing data but indicates that there may be regional aspects associated with IFA use that were not detected using other types of descriptive or regression analysis.
To test the strength and direction of a possible link between IFA use and the different circumstances that prevail in local authorities, a regression analysis was performed (reported in Table 2). This section discusses the bivariate and multivariate regression analyses exploring associations between IFA use and municipal characteristics.
As seen in Table 2, there was a significant correlation between a small population (<12,000) and IFA use of 20% or more in the bivariate analysis and the multivariate analysis. The odds are approximately threefold for these municipalities with an IFA use of 20% or more. The bivariate analysis indicated that geographically, southern municipalities were less likely to use an IFA above the cut off level of 20%, but in the multivariate analysis, including all independent variables, the only remaining significant link was between a small population (<12,000) and IFA use of 20% or more. Some links in the bivariate analysis were no longer significant in the multivariate analysis. The results indicated that when controlling for the other independent variables, the link between the dependent variable and the low population density and geographical part (southern Sweden) was not significant. Neither the private provision of other welfare services nor local political governance or socio-economic variables appeared to be linked to IFA use.

Discussion
In many countries, foster care is being partly outsourced to Independent foster care agencies (IFAs), but there has been little research into this development. To address part of this knowledge gap, the purpose of this article was to describe and analyse the use of IFAs in Sweden and to analyse the differences in use between municipalities on a structural level. To fulfil this aim, two questions were addressed. Firstly, the extent of IFA use by Swedish local authorities was investigated. Secondly, possible determinants (e.g. local demographics and socio-economics, geographical location, the private provision of other welfare services and local political governance) of the extent of IFA use were explored. This study offers comprehensive data on IFA use in Sweden and is to the best of the author's knowledge the first known study to address the relationship between IFA use and multiple possible structural determinants.
To address the first question, the results demonstrated that most local authorities use IFAs and that this use represents on average approximately one quarter of the total number of foster care placements. About half of the local authorities use this form of care to supply a sizable part of their foster care services: every fifth placement or more. This means that a substantial percentage of foster families are in contact with and are economically compensated by private companies rather than by the public authorities responsible for the care and wellbeing of fostered children and adolescents.
What are the implications of these results? Overall, marketisation has implied fundamental changes in the organisation of foster care, as IFAs are a common phenomenon in Sweden. At the same time, research on the impact of IFAs on the quality of care is limited (e.g. Kessler et al., 2008;Steen & Duran, 2013;Steen & Smith, 2012). However, current knowledge of the consequences of the private provision of other child welfare services raises concerns that may have a bearing also in relation to IFAs. This pertains to, for example, issues such as information asymmetry, the need for stricter audits and regulations, and difficulties specifying in contracts what child welfare services should include and establishing whether these conditions are met (Erlandsson et al., 2013;Foote, 2022;Forkby & Höjer, 2018;Health and Care Inspectorate, 2020;Lundström et al., 2021;Pålsson, 2018). Importantly, the client's perspective on the marketisation of foster care is scarcely addressed in current research. Thus, there is a need to closely follow the developments in this area to investigate the implications for children in care.
To address the second research question, possible structural determinants of the local authorities' IFA use were investigated. The results showed that in the multivariate analysis, only the population size of the municipality (<12,000) had a significant link with the use of IFA. In other words, this study shows that for approximately a third of Swedish local authorities with small populations, IFAs appear to be especially important for the provision of foster care.
The finding that small municipalities tend to use IFAs more often contrasts with previous research exploring other welfare services, which show that private providers are usually more available in larger city areas (Hartman, 2011a(Hartman, , 2011b. The differences in IFA use that are due to population size may indicate that smaller and more remote municipalities with few children in care struggle more to meet the need for foster families within the public welfare system. This could be related to the fragility of small organisations. Small organisations might be more vulnerable to staff turnover as there might be fewer social workers working with the recruitment of foster carers. Few placements on an annual basis may also lead to issues concerning retaining the interest of previous or potential foster carers. The recruitment of foster carers in general has previously been reported as an international issue for public child welfare services (Colton & Williams, 2006;Corbillon, 2006;Löfstrand, 2009;Oslo Economics, 2020), but this issue is arguably accentuated in the smaller municipalities (Pålsson et al., 2022).
Further, in relation to the second research question, none of the other structural factors included in the analysis (e.g. local demographics and socio-economics, geographical location, the private provision of other welfare areas, and local political governance) were significantly associated with IFA use when controling for the other structural factors in the analysis. This indicates a need for testing organisational variables (e.g. collaboration between municipalities, differences in the type of staff allocated to recruitment and foster care allowances) as well as using other research methods to understand the use of IFAs and the variation between local authorities in this respect. The perspective of IFA companies and foster carers may also be important to understand the emergence and use of IFAs.

Limitations
Since the tested structural factors have not previously been addressed theoretically or empirically in this context, this study is exploratory. Consequently, there is also little previous data to compare the presented results with, limiting the ability to understand development over time. Additionally, this is a cross-sectional study and the official statistic data is from one year. To test whether the data is stable over time, further data (e.g. average IFA use) collected different years should be explored.

Conclusion
To conclude, this study has two central results. Firstly, the study shows that IFAswhose care content we know little aboutare used for approximately a fourth of the foster care placements in Sweden. Secondly, small municipalities to a greater extent use IFA for a substantial part of their placements. There is a continued need to understand reasons to use IFA and the implications of IFA use on the care provided to foster children. Notes 1. One municipality stated that due to separate units handling child welfare services and foster care, only children in long-term foster care were reported. 2. Including municipalities and city districts. 3. The full sample of 261 municipalities was compared with the sample of 236 after excluding municipalities with missing information on analysed variables. Descriptive data showed major similarities between samples.

Disclosure statement
No potential conflict of interest was reported by the author.

Notes on contributor
Evelina Fridell Lif, MSW is a PhD student at the Department of Social Work, Stockholm University, Sweden. Her research interest includes child welfare and the state-market-family interaction in foster care for children. Her current project explores independent foster care, a form of foster care where social services contract organisations to recruit and support foster families. She uses survey and interview data to explore the recent developments in foster care. She worked as a child and family social worker in child welfare in Sweden before starting postgraduate education.