Mechanisms of trust for different modes of welfare service provision

ABSTRACT

Welfare services are an important channel for building institutional trust. But does it matter if the service is publicly or privately provided? Using SEM on individual data from public and private schools in Sweden, the authors test how the provision mode moderates the effects of positive and negative school experiences on trust in government. The results show that students’ perceptions of teacher fairness are dependent on provision mode. Negative experiences are associated with decreased trust in government only in private schools. Positive experiences didn’t affect trust. Thus, the use of private providers may provide greater challenges for maintaining institutional trust

KEYWORDS:

• Welfare service provision
• institutional trust
• mixed provision systems
• fairness
• provision mode

Introduction

The contemporary public administration and management literature reflects a period of dramatic change over recent decades in how public administration is organized and how societal problems are governed. At the heart of different organizational trends are assumptions about how they may increase trust in government. One of the most evident changes in recent decades is that private firms and nonprofit organizations have been given an increasing role in the provision of government services in most Western countries (Johansson 2008; Edlund and Lindh 2013; Svallfors 2011). This change, which is a part of the new public management (NPM) paradigm, was launched and portrayed as a reaction to a lack of trust in government (Van de Walle 2011; Fledderus, Brandsen, and Honingh 2014). However, there is little reliable or systematic knowledge about how trust mechanisms have been affected by these changes (Lægreid 2017). The logic of NPM builds on the idea that trust in government is dependent on the performance of public services, where the efficacy and outcomes of public services are represented as the main drivers of trust. Customer choice on a competitive market is viewed as a key means for enhancing efficiency (Wiesel and Modell 2014).

More recently, however, the logic behind marketization and outsourcing reforms and their emphasis on a calculative notion of trust (Van de Walle 2011; Fledderus, Brandsen, and Honingh 2014) have been challenged. Numerous studies have argued that peoples’ trust in government and in civil servants is more strongly affected by normative factors, most notably the perceived fairness of public services (Marvel and Girth 2016; Tyler 2006; Van Ryzin 2011). Trust is increasingly viewed as relational and as stimulated by shared values and goals (Van de Walle 2011; Taylor-Gooby and Wallace 2009). Attention is thereby directed at the process of welfare service delivery and the relational experience that it involves (e.g. Osborne 2006). It is argued that people assess the trustworthiness of the state by drawing on their experiences with state officials (Khodyakov 2007).

This article builds on this line of research stating that experiences of fair and benevolent treatment in welfare services build trust in government. Building on this perspective raises new questions regarding the use of different provision modes in welfare service provision. For a welfare state with both public and private providers, a question arises: which institutions enjoy the trust that is relationally established through service interaction? Questions about how the provision mode affects trust mechanisms have been largely neglected in the empirical literature. Studies have only recently started to explore how the provision mode affects the way blame is ascribed (Cammett, Julia, and Gavril 2015; Marvel and Girth 2016) or how the dominant images of public versus private organizations affect the interpretation of their performance (Hvidman and Andersen 2016).

The aim of this article is thus to explore how the marketization of welfare services affects the mechanisms building or deteriorating trust in government institutions. More specifically, we ask whether and how the mode of provision moderates the relation between welfare service experiences and trust in government. While it is generally argued that building and degrading trust are two different processes that occur at different paces (Cook, Hardin, and Levi 2005), there are very few empirical studies that seek to distinguish these processes and explore their differences (for an exception, see Huseby 2000). Thus, we also explore whether the effect on trust differs between positive and negative experiences for private as well as public providers.

When exploring the role of provision mode, we introduce and draw upon two different theoretical strands, accountability chains and institutional logics, which provide us with opposing hypotheses regarding its moderating effect. The first theory highlights the increase in distance. The use of third parties to deliver welfare services creates a separation between government and the services that it funds (Milward and Provan 2000), making it less common to have a direct relational experience with the state. For the provision of services to build trust in government in this context, the service experiences and the trust (or distrust) that they generate must be attributed to the government and not stay with the private provider. For these reasons, the extended accountability chain argument suggests less spillover from service experiences to institutional trust in a private provider context. The second emphasizes differences in the institutional context in which the service interaction takes place. Institutional logics theory (Thornton, Ocasio, and Lounsbury 2012) argues that different institutional contexts (such as markets and the private firm versus public organizations) affect the logics of actions and relations, as well as the interpretations thereof. The stronger logic of users as customers within market institutions (private providers) suggests that users might be more sensitive to positive as well as negative treatment and thereby a private provider context might strengthen the effect of service experiences on trust in government. Thus, both theories indicate that the production mode may affect the relationship between service experiences and trust but does not give a consistent view about how. Empirically studying this relationship is therefore important to uncover the relevant theoretical stances that may explain potential differences between production modes.

Introduction of the case

Sweden offers an interesting context for exploring links between the provision mode and trust in government, as the country has witnessed relatively recent and rapid changes in terms of how welfare services are produced. Historically, Sweden has been renowned for its encompassing welfare model, relying almost exclusively on in-house public production and delivery; all citizens regularly interacted with public representatives of the welfare state. However, transformative changes have taken place, and private firms (the market) are currently central actors delivering services and providing welfare to citizens (Johansson 2008; Edlund and Lindh 2013). Our material comes from the field of public education, which is one of the welfare sectors seeing the most extensive marketization in Sweden (Rothstein and Blomqvist 2000).

The selection of public education as our case implies that the respondents are young (13–16 years). Youth represent a particular subgroup of the population, which, it should be noted, is not necessarily representative of the whole regarding mechanisms for generating institutional trust, thus, the results of this study may draw conclusions about the general public. However, there are reasons why this group is particularly interesting to explore. Most crucially, public education is a service that is relational and enduring, wherein relational experiences occur between the representatives of the institution (i.e. teachers) and the citizen (i.e. student). Furthermore, school generally provides the first welfare service experience to potentially shape trust (Claes, Hooghe, and Marien 2012), while the adolescence is an age when values and beliefs are settling, including political consciousness and democratic mindsets (Abendschön 2017). Using latent moderated structural equation (LMS) models on data collected from 1,032 13- to 16-year-old students in an average-sized city in Sweden, we test how the mode of provision moderates the effects of positive and negative perceptions of teacher fairness on trust in government.

In the following section, we describe how theoretical understanding of the relation between performance and trust has evolved and present theoretical assumptions about the trust mechanisms connected to the different modes of provision. The subsequent section describes the research design; in this section, we describe the marketization of the school system in Sweden and the motivation for our decision to look at education and, consequently, at youth. The subsequent section provides information about the data and measures. In the results section, we present the results of the LMS models, followed by a discussion. The final section concludes and discusses the implications of our study for practice and future research.

The relation between welfare service performance and trust

‘Trust in an institution means confidence in the institution’s reliable functioning’ (Möllering 2006, 74). Trust in institutions is, thus, built on information or experiences of its performance (Fledderus, Brandsen, and Honingh 2014; Möllering 2006). The performance or functioning of institutions may be judged based on different values such as efficiency, impartiality, benevolence. The organization of the welfare state and its services have been shaped by theories and assumptions about how performance, or rather different aspects of performance, affects the public’s trust in the state. The provision of welfare services by private actors is part of the NPM movement (Johansson 2008; Johansson and Siverbo 2011). One of the incentives for NPM is enhancing the legitimacy of the state (Hood 1995). A central assumption behind these changes is that ‘better performing public services will lead to increased satisfaction among their users, and this, in turn, will lead to more trust in government’ (Van de Walle and Bouckaert 2003, 892; see also Yang and Holzer 2006). Furthermore, it is assumed that marketization advances effective policy implementation (i.e. service delivery).

The emphasis on calculus-based trust, where cost-efficiency and policy outcomes are regarded as key to generating trust implied a departure from the earlier Weberian tradition. Administrative efficiency was one of the intended functions of Weberian bureaucracy, with its focus on a systematic flow of large number of cases (Johansson 2007; Olsen 2006). However, impartiality, value-neutrality and equal treatment were seen as cornerstones for upholding the public ethos. According to the Weberian tradition, which predominated during the establishment of welfare state institutions, public ethos is the foundation of trustworthiness. Competition within and between organizations was considered to increase corruptibility, rather than improve efficiency and service quality. While the marketization of welfare services intends to build trust based on calculus, the Weberian bureaucracy is intended to generate process-based trust (Fledderus, Brandsen, and Honingh 2014). Thus, different organizational trends rely on different assumptions of how welfare organization stimulates institutional trust, meaning that it is essential to gain a deeper understanding of such mechanisms.

The relation between performance and trust has received thorough scrutiny in recent years (Yang and Holzer 2006; Van Ryzin 2011; Van de Walle and Bouckaert 2003; McLoughlin 2015). However, the core assumptions of the NPM movement that marketization increases service performance efficiency, which in turn increases institutional trust, are still to be empirically established (Lægreid 2017). Recently, these assumptions have been increasingly challenged. Recognition is growing that one of the major factors explaining trust in government is how citizens perceive the process and procedures of government institutions (Van Ryzin 2011; see also Rothstein and Stolle 2008). Individuals tend to value normative factors, such as benevolence and fair processes, more than tangible outcomes when making statements about the legitimacy and trustworthiness of government agencies (Cook, Hardin, and Levi 2005; Tyler 2006; Taylor-Gooby and Wallace 2009).

Mechanisms of trust building in welfare services are thereby situated in the interactions between welfare institutions and citizens (cf. Osborne 2006; Osborne, Radnor, and Nasi 2013). Trust in institutions, it is argued, is shaped by relational experiences in the same way that interpersonal trust is shaped by our experiences with others (Cook, Hardin, and Levi 2005; Möllering 2006). One’s perception of the state is informed by experiences with the state, i.e. interaction with state officials who represent the authoritative position of the state (Khodyakov 2007). Yang and Holzer (2006) argue that ‘[w]hen citizens interact with an agency they tend to consider the agency as a representative of the government. In this specific encounter, the agency is the government, and it matters most’ (116). Citizens judge the values represented by a service (as it is performed) and not only the outcome. These values become associated with the government as a whole (cf. Soss 1999), and thus the interactions within welfare service provision shape citizens’ trust in government. The process of drawing cues from experiences to inform an image of something more abstract or uncertain is not a calculative or intentional practice, but a practice that we automatically engage in to make sense of our environment (Weick 1995). Thus, our experiences build up an interpretive framework (also shaped by the images and stereotypes that dominate the public discourse [Hvidman and Andersen 2016]) that shapes our interpretations and perceptions.

Building on this line of research, we propose the following first hypotheses:

Hypothesis 1a: Experiences of fair and benevolent welfare service treatment have a positive effect on trust in government.

In this study, we also address whether there are differences in how positive and negative experiences are associated with trust in government. It is often argued that the processes behind increasing trust and those behind decreasing trust are not the same and that trust can degrade more rapidly than it can be built (Slovic 1993; Cook, Hardin, and Levi 2005). Thus, it is relevant to consider whether positive and negative experiences of fair and benevolent treatment have different effects. Studies that consider differences between the effects of good and poor performance on trust are rare (for an exception, see Huseby 2000, which is a recurring reference in this regard). However, existing findings (Huseby 2000; Yang and Holzer 2006) suggest that the spillover from the perception of welfare performance to the perception of government differs for positive and negative experiences. A negative bias has been found (Huseby 2000), meaning that negative experiences have a stronger effect than positive experiences, which supports the idea that trust is more easily destroyed than gained (Slovic 1993). To account for potential differences in the processes of building and destroying trust, we measured fair and unfair treatment separately. We thereby include a second version of Hypothesis 1:

Hypothesis 1b: Experiences of unfair and nonbenevolent welfare service treatment have a negative effect on trust in government.

Furthermore, based on previous research, we propose that negative experiences have a stronger effect on trust than positive experiences (Huseby 2000; Yang and Holzer 2006). This leads us to pose the following hypothesis (1c):

Hypothesis 1c: Experiences of unfair and nonbenevolent welfare service treatment have a stronger effect on trust than do experiences of fair and benevolent welfare service treatment.

Modes of provision and different mechanisms of trust

The relational understanding of trust opens up new questions regarding the use of external providers. If trust in government is shaped by experiences of interaction with public authorities, then does it matter whether the interaction is with a public representative or with a representative of a private firm that is contracted by the state? While this is an uncovered topic in previous research, theories on governance structures and institutional logics may provide guidance for hypotheses by suggesting the contextual differences between provision modes that may affect how experiences of service provision are translated and come to affect trust in government. Both theories suggest that the provision mode may moderate the relationship between service experience and trust in government. However, depending on which theoretical lens is adopted, the result is two different and mutually exclusive hypotheses on the moderating role of the provision mode for the relationship between fair/unfair treatment and trust in government.

Extended accountability chains

When the government decides to involve a private actor in the provision of a service, it expands the accountability chain as well as the distance between the government and the citizen (Marvel and Girth 2016). This distance stems from the added layer and the increased abstraction of the state when it resorts to contractual management. Private provision implies a more distanced and abstract relation to the state (as a contract manager), as opposed to a direct relation with a representative of the state. The extra layer also decreases the government’s ability to control because it represents a contractual relationship rather than hierarchal top-down control (Williamson 1975). Regarding the influence of performance on trust in government, this extra layer necessitates an additional attribution step. Service experiences with private providers must be associated with the government to either generate or deteriorate trust in government.

The added layer and attributive step suggest that the effect of service experience on trust in government is stronger when public institutions provide the service rather than private actors (cf. Marvel and Girth 2016). The trust or distrust generated by interactions with a private provider may stay with that private actor and not fully be attributed to the state. Indeed, the market system builds on the idea that the market will hold poorly performing actors responsible by disfavoring them (Schumpeter 1942), thus holding them responsible for good as well as bad performance. This leads us to the following hypothesis:

Hypothesis 2: The provision mode moderates the relationship between (un)fair welfare service treatment and trust in the government so that the effect is more prominent for public provision.

Different institutional logics

Not only are public and private providers distinct organizations but they are also embedded in different institutional logics (Friedland and Alford 1991; Thornton and Ocasio 1999). Institutional logics are ‘the socially constructed, historical patterns of material practices, assumptions, values, beliefs, and rules’ that guide individuals’ actions ‘and provide meaning to their social reality’ (Thornton and Ocasio 2008, 101). There are distinct differences between the logic of the state bureaucracy and civil services and the logic of a private firm (Thornton and Ocasio 1999; Rainey and Bozeman 2000; Van der Wal and Huberts 2008; Wiesel and Modell 2014). The most elementary difference is their diverging fundamental aims, namely, profit versus the generation of public goods. Impartiality and incorruptibility are found to be core values within public institutions, while profitability and innovativeness are core business sector values (Van der Wal and Huberts 2008). The different institutional logics also incorporate different dominating assumptions of how trust is generated (cf. Fledderus, Brandsen, and Honingh 2014).

These different logics not only shape actions within the institutions but also affect how actions are perceived and interpreted into experiences. Institutional logics and stereotypical images thereof may act as interpretive frames that affect expectations as well as how people judge services (Hvidman and Andersen 2016). One important feature of the firm and NPM logic is that of users being customers and consumers and that customer satisfaction and customer value is the main currency of legitimacy for firms (Wiesel and Modell 2014). The ‘customerization’ and customer orientation (Wiesel, Modell, and Moll, 2011) might have the effect that it sensitizes the users of services dominated by (or expected to represent) this logic to fair and unfair treatments. The well-known motto ´the customer is always right´ captures this argumentation well. Thus, the rationale of the institutional context affects the perception and interpretation of experiences and thereby the interpretive frame that informs trust in government. As opposed to Hypothesis 2, this suggests that the effect of the service experience on trust in government is stronger for private provision. Therefore, Hypothesis 3 is stated as follows:

Hypothesis 3: The provision mode moderates the relationship between/un/fair welfare service treatment and trust in the government so that the effect is more prominent for private provision.

Table 1 summarises the hypotheses and their respective theoretical bases. Hypotheses 1 a-c concern the relationship between welfare service performance, in terms of fair and benevolent treatment, and trust in government, while hypotheses 2 and 3 concern whether and how this relationship is affected by provision mode. For Hypothesis 2, as well as Hypothesis 3, we look separately at the effect of fair and benevolent treatment and that unfair and unbenevolent treatment. If Hypothesis 1 c gain support, we expect the same effect to transpire in relation to hypotheses 2 and 3.

Table 1. Table of hypotheses.

Hypotheses on the fair and benevolent performance–trust link
1a	Experiences of fair and benevolent welfare service treatment have a positive effect on trust in government.
1b	Experiences of unfair and nonbenevolent welfare service treatment have a negative effect on trust in government.
1c	Experiences of unfair and nonbenevolent welfare service treatment have a stronger effect on trust than do experiences of fair and benevolent welfare service treatment.
Hypothesis based on extended accountability chains
2	The provision mode moderates the relationship between (un)fair welfare service treatment and trust in the government so that the effect is more prominent for public provision.
Hypothesis based on different institutional logics
3	The provision mode moderates the relationship between/un/fair welfare service treatment and trust in the government so that the effect is more prominent for private provision.

Controlling for major antecedents of trust

Trust in institutions as well as interpersonal trust is shaped by different factors in life (Möllering 2006). In this article, we are not seeking to identify the single most important factor explaining government trust but to explore whether the potential of public services to increase trust in government is affected by provision mode. However, to single out the effect of fair and benevolent treatment and minimize the risk of omitted variables, we need to consider and control for at least two major antecedents. One of the factors that correlates the strongest with institutional (e.g. government) trust is interpersonal or generalized trust (Zmerli and Newton 2008, Berg and Tobias 2016). By controlling for generalized trust, we seek to eliminate the effects from processes that shape the disposition to trust more generally. In relation to youth, it has been shown that perceived parental warmth in early youth affects political trust in late adolescence (ŠErek and Macek 2014). This type of relation is very likely to have generalized trust as an intermediate variable. We also control for socioeconomic standard (SES) that has been shown to correlate with government trust (e.g. Berg and Tobias, 2016) and that may also influence the choice of school.

The case of public education in Sweden

As noted in the introduction, one of the policy areas in Sweden that has gone through the most substantial marketization is the educational system (Rothstein and Blomqvist 2000). Table 2 show the development of the privatization, and its increase since the 90s.

Table 2. Percentage of Swedish students that attend schools under private provision.

The educational system has been subject to free choice reforms based on a voucher system (Petersen and Hjelmar 2013). This system means that the cost for the education is tied to the student, implying that the funding of all schools is based on the students that are enrolled. Thus, there are no fees involved and the resources of public and private schools are equal. In other words, the educational system, as most other welfare services in Sweden, is fully publicly funded, one’s financial resources do not impact one’s ability to exercise choice or to choose any school. Therefore, there is not a difference in status between public and private schools in general, as in many other countries. However, it should be noted that there are public schools in disadvantaged areas that struggle with particular challenges, and public as well as private schools in segregated advantaged areas that face the opposite situation.

The main idea behind the system is to create a market where all schools compete for the students. Exerting choice may thus imply choosing a public school as well as a private school. What we have here is therefore a relatively pure mixed provision system. Looking at our sample we can also see that there is no difference between the public and private schools in terms of socioeconomic status, gender or immigrant status of the students. Table 3 provides background statistics about the students in the public versus private schools and illustrates the similarities between the two groups.

Table 3. Descriptive statistics. Independent samples t-tests.

	Public school	Private school	F-test	p-value
Institutional trust	2.08	2.31	16.004	< .001
Positive fairness	3.10	3.10	.01	.91
Negative fairness	2.06	2.02	.87	.35
Social trust	3.31	3.23	1.16	.28
SES	−.01	.04	1.04	.31
Gender (0 = boys, 1 = girls)	.48	.56	4.46	.04
Immigrant status^a	.76	.74	.45	.51
N	664	256

Note: ^a = Country of birth was coded so that the value 0 represents at least one parent being born outside Sweden or a Nordic country and the value 1 equals both parents being born were born in Sweden or another Nordic country.

The far-reaching use of private providers within this field is one of the reasons why we choose to analyze the educational system. However, there are more essential reasons. One reason is that education is an interactive and long-term service, implying that relational mechanisms of trust building have the potential to develop. To grasp these relational mechanisms, it is crucial to explore an interactive process of welfare service delivery, as opposed to, e.g. more technical services (e.g. snow plowing). Experiences with public servants (in this case teachers) that are perceived as benevolent, fair or as unfair are more likely to influence the individual, the interpretive frame, and trust in institutions when the relation and service are long-term and play a greater part in one’s life.

Furthermore, the young age of the respondents (13–16) implies that school is their first long-term experience with public services; thus, it could be assumed that this experience has a crucial formative role (cf. Claes, Hooghe, and Marien 2012). This is something that has long been recognized within the political socialization scholarship, not least by John Dewey. Indeed, one of the missions of the school is to form democratic citizens. In this regard, it has been shown that experiences with an open classroom climate (the encouragement of forming and voicing opinion among peers and with teachers) have a particularly strong effect on political trust (Claes, Hooghe, and Marien 2012). The theoretical interpretation of this relation is that the democratic functioning experienced in school is generalized towards political institutions more broadly, of which the adolescents have few first-hand experiences (Claes, Hooghe, and Marien 2012). While formal political citizenship and the right to vote generally comes at the age of eighteen, we now know that political consciousness and orientation and democratic mindsets already develop and stabilize in younger years (Abendschön 2017; Claes, Hooghe, and Marien 2012; Haug 2017; Van Deth, Abendschön, and Vollmar 2011). Current research downplays the differences between youth and adults in regard to orientations such as political interest (Haug 2017). Nevertheless, comparisons with other groups will be needed before drawing broader generalizations.

Furthermore, as the school is such an important part of the students’ lives, the likelihood that they are aware of its public or private character is much greater compared with the character of a service that is brief and/or merely practical (e.g. waste management). It is also the case that the brand or niche of private school is generally well exposed and, as a consequence, their organizational status. This is important because citizens may experience difficulties in distinguishing between public and private actors, resulting in ‘errors of attribution’ (Van de Walle and Bouckaert 2003, 897). Lastly, the educational system is one of the key responsibilities of welfare states and one of the policy fields that citizens rate very high in importance (Raffel 2007). Previous research indicates that for negative service experiences to affect political trust negatively, the state must be perceived as being responsible for providing this service (Van de Walle and Bouckaert 2003, 898).

Method

Data and measurements

The data originate from the Political Socialization Program (PSP), a large questionnaire survey conducted in a major Swedish city (Örebro) with approximately 146 000 inhabitants. Örebro is similar to the national averages on aspects such as income levels, population density, and unemployment; the share of young citizens (5 to 24-year-olds) with a foreign heritage is slightly higher (33 per cent versus 24 per cent) in Örebro compared with the national average (Statistics Sweden, 2012). The results are therefore likely to represent the average Swedish context. Respondents were chosen with the aim of capturing different societal backgrounds. The sample targeted ten high schools that include roughly the same number of boys and girls with varying ethnic and social statuses in vocational as well as theoretical programs. Our sample had the following characteristics: N = 1,032; 50.7 per cent girls; M_age = 13.43; SD = .56; and range = 12–17. Data collection was administered by trained research assistants and occurred during school hours without the presence of teachers. Respondents were informed that participation was voluntary and about the different kinds of queries asked about in the questionnaire. Additionally, respondents were guaranteed the confidentiality of their answers and that their responses would not be seen by parents, teachers, or anyone else. All participating classes received an incentive of approximately 100 EUR. The research program was approved by the Central Ethical Review Board of Sweden.

In our measurement and structural models, we include four main variables of interest and two control variables. Our dependent variable is institutional trust, which is a latent variable measured by three items: trust in parliament, trust in government, and trust in court. The combined construct is intended to measure trust in the system rather than trust in politicians. The trust questions were formulated in terms of how much trust the respondents have in these institutions, with a response scale ranging from (1), no trust at all, to (4), considerable trust. The Cronbach’s α for this index is 0.91.

To assess experiences of welfare service interactions (relational performance), we use two variables measuring teacher fairness. The two variables fair treatment and unfair treatment are both latent variables consisting of three items. These three items are intended to capture different aspects of fairness displayed during the service interaction, which are derived from procedural fairness theory (cf. Abdelzadeh, Zetterberg, and Ekman 2015), namely, impartiality, consideration of opinions and benevolence. For fair treatment, the following propositions are posed: ‘Most of my teachers treat me fairly’, ‘Most of my teachers listen to what I have to say’, and ‘Most teachers want their students to feel good’. The items measuring unfair treatment include the following: ‘Most teachers don’t like me’, ‘There are scarcely any teachers who I can talk with if I have problems with something at school’, and ‘There are scarcely any teachers who praise me when I do a good job’. The response scale for both fair and unfair treatment ranged from (1), absolutely agree, to (4), absolutely disagree. The Cronbach’s α for these indexes are 0.74 and 0.71, respectively. To ensure the appropriateness of using fair and unfair treatment as two distinct factors measuring positive and negative experiences, respectively (fairness and unfairness), we employed a comparison of one- and two-factor structure models. The fit indices and a chi-square difference test supported the two-factor structure model (Table A1 of the Appendix).

The moderating variable, provision mode, is a dichotomous variable, divided into public and private schools. Apart from being public or private, the two groups of schools show similarities in characteristics such as level of socioeconomic status, gender distribution, and share of students with immigrant heritage (see Table 3). Two control variables are applied to all the models. The first is generalized trust, which was measured with the initiating question, ‘If you think about people in general, to what extent do you think that the following statements apply to you?’ followed by two propositions: ‘Most people can be trusted’ and ‘Most people are fair and don’t try to exploit you’. The responses used a five-point scale ranging from (1) do not agree at all to (5) completely agree. The Pearson’s correlation for this index is 0.64. The second control variable is SES. This variable was measured using questions such as ‘If you want things that cost a lot of money (e.g. a computer, skateboard, cell phone), can your parents afford to buy them?’, ‘Does your family have more or less money than other families where you live?’ and ‘How often do you and your family go on vacation?’ The Cronbach’s α for this index is 0.78.

In Tables A2 and A3 (see Appendix), we present the descriptive statistics of the variables and their correlations. Our measurements reported good internal validity and unidimensionality. All Cronbach’s αs were greater than 0.7, a threshold commonly deployed in the construction of indexes (Kline 2010).

Table A2. Measurement properties.

	Org. N	Range	Mean	Std. Dev.	Skewness	Kurtosis	Factor loading
Institutional Trust (Cronbach’s α .91)
1	864	1–4	2.04	.805	.227	−.772	.958
2	863	1–4	2.10	.821	.157	−.823	.915
3	860	1–4	2.30	.906	.004	−.928	.776
Fair treatment (Cronbach’s α .74)
1	891	1–4	3.20	.692	−.680	.682	.708
2	881	1–4	2.97	.717	−.369	.032	.698
3	890	1–4	3.12	.717	−.647	.561	.676
Unfair treatment (Cronbach’s α .71)
1	881	1–4	1.92	.790	.717	.300	.721
2	893	1–4	2.16	.874	.497	−.351	.672
3	884	1–4	2.07	.796	.583	.127	.617
Provision mode	920	0–1	0.28	.448	-	-	-
Generalized trust (Pearson’s correlation, r = .64)
1	889	1–5	3.25	.979	−.398	−.140
2	885	1–5	3.32	.949	−.480	.100
Socio-economic Standard (Cronbach’s α .78)
1	892	1–5	3.66	1.009	−.488	−.179	.763
2	868	1–5	3.04	.835	−.205	.496	.722
3	879	1–5	3.11	.766	−.142	.734	.639
4	874	1–4	2.42	.916	.124	−.796	.560
5	860	1–4	3.01	.807	−.589	−.015	.520

Strategy for analysis

In our attempts to examine our hypotheses, we employ latent moderated structural equation models (LMS) on a sample of 920 students using Mplus 7.31 (Muthén and Muthén 1998–2012). The structural models we employ will be estimated using a two-step procedure specified by Maslowsky, Justin, and Douglas (2015).

The first step of the structural analysis is to examine the fit of the model without the latent interaction term (Model 0). Subsequently, a latent interaction term combining provision mode and perceived fairness (Model 1) or unfairness (Model 2) is added to the structural model.¹ At present, there are no model fit indices available for LMS models. However, by using a log-likelihood ratio test, it is possible to determine how a more complex model (in our case, Model 1 or 2) compares to Model 0 in terms of relative model fit (Satorra and Bentler 2001). If in the initial model fit test, Model 0 has a good model fit, the log-likelihood ratio test reports whether Model 0 shows a significant loss in relative fit compared to Model 1 (or 2). If Model 0 has an initial good absolute model fit and if the log-likelihood ratio test shows a significant loss in the relative fit of Model 0 compared with Model 1 (or 2), then we conclude that Model 1 (or 2) also fits the data well. Additionally, because the LMS framework does not provide standardized estimates, we have standardized the indicators comprising our latent factors prior to estimating our structural equation models. We also center the observed moderator variable, provision mode, before creating the latent interaction term. Using this procedure, we can more easily interpret the latent interaction terms provided by the LMS.

After our initial tests of model fit, we examine the direct effects of fair and unfair welfare service treatment on trust in government (hypotheses 1a and 1b) by regressing institutional trust on fair and unfair teacher treatment (Figure 1, Model 0). This first part of the analyses also generates an answer to Hypothesis 1c: whether unfair welfare service treatment has a stronger effect on trust than does a good experience of welfare service treatment. To examine whether the effect of fair and unfair welfare service treatment is conditional on provision mode (Hypotheses 2 and 3), we regress institutional trust on interactions between experiences of fair and benevolent and unfair nonbenevolent teacher treatment and mode of provision.

Figure 1. Structural equation model without latent interaction term (unstandardized estimates). Model 0.

Results

The first step was to validate our models. The measurement model had a good model fit (χ²(105) = 230.211, CFI/TLI = .975/.967, RMSEA = .036 (90 per cent CI, .030–.042), and SRMR = .031). This means that the data fit the theorized model and that it was appropriate to move on to compare the relative fit of the structural model (Model 0) (Figure 1) with the subsequent step of the LMS (Models 1 and 2) (Figures 2 and 3).

Figure 2. Structural equation model with latent interaction term of provision mode and fair treatment (unstandardized estimates). Model 1.

Figure 3. Structural equation model with latent interaction term of provision mode and unfair treatment (unstandardized estimates). Model 2.

Similar to the measurement model, the structural equation model, without the latent interaction term (Model 0), had a good model fit (χ²(105) = 230.211, CFI/TLI = .975/.968, RMSEA = .036 [90 per cent CI, .029–.042], and SRMR = .031)(Hu and Bentler 1999).

After concluding that the models we propose were appropriate, we turned to specific effects and our hypotheses. Examining the direct effects of the measurement model showed that fair treatment at school had no effect (β = .09, SE = .08, and p = .29) on trust in government, but unfair treatment had a negative effect (β = −.18, SE = .09, and p < .05) (Figure 1). Hence, while experiences of fair and benevolent welfare service treatment did not have a positive effect on trust in government (Hypothesis 1a), experiences of unfair and nonbenevolent welfare service treatment were found to be negatively related to trust in government (Hypothesis 1b). This was a clear indication that experiences of unfair and nonbenevolent welfare service treatment have a stronger effect on trust than do experiences of fair and benevolent welfare service treatment. To verify this finding, we assessed whether the two associations differed significantly in size, which they did (p < .05). Consequently, we also found support for Hypothesis 1c. Additionally, we found that the private provision mode was positively correlated with trust in government (β = .12, SE = .03, and p < .001), which may suggest that students in private schools had a higher initial level of trust. Note also that the fairness variables were significantly correlated, which can be expected. Using them in the same model allowed us to control the effect of unfair treatment for the effect of fair treatment and vice versa.

To test the moderating effect of provision mode (hypotheses 2 and 3), the next step was to add the latent interaction terms to our model. We found that both had a good fit to data; we conducted a log-likelihood ratio test and concluded that Model 0 had a significant loss in fit compared to both models that included latent interaction terms (Models 1 and 2). Regarding the interaction effects of our LMS models, the provision mode × fair treatment term of Model 1 was not significant (β = .07, SE = .11, and p = .53) (Figure 2), while the provision mode × unfair treatment interaction term of Model 2 was significant (β = −.24, SE = .12, and p < .05) (Figure 3). In other words, we found an effect of unfair treatment on trust conditional on provision mode. However, experiences of fair treatment seemed unrelated to trust in government institutions for both provision modes.

To fully understand the interaction effect of Model 2, which indicated that the effect of unfair welfare service treatment on trust was conditional on provision mode, we plotted the interaction. We also employed a simple slope test, which showed that unfair treatment had a significant negative effect on trust in private schools (solid line, β = −.40, p < .01) (Figure 4). It also showed a corresponding, though nonsignificant effect, for public schools (dashed line, β = −.16, p > .05).

Figure 4. Interaction of unfair teacher treatment and provision mode on institutional trust.

In conclusion, we found a significant negative effect of unfair treatment on trust within private schools and a nonsignificant, negative effect of unfair treatment on trust within public schools. Thus, Hypothesis 3, stating that the effect of unfair welfare service treatment on trust in government is more prominent for private provision, was partly supported, while Hypothesis 2 was rejected. Table 4 sums up the results for each hypothesis.

Table 4. Table of hypotheses.

Hypothesis		Result
1a	Experiences of fair and benevolent welfare service treatment have a positive effect on trust in government.	Rejected
1b	Experiences of unfair and nonbenevolent welfare service treatment have a negative effect on trust in government.	Confirmed
1c	Experiences of unfair and nonbenevolent welfare service treatment have a stronger effect on trust than do experiences of fair and benevolent welfare service treatment.	Confirmed
2	The provision mode moderates the relationship between (un)fair welfare service treatment and trust in the government so that the effect is more prominent for public provision.	Rejected
3	The provision mode moderates the relationship between/un/fair welfare service treatment and trust in the government so that the effect is more prominent for private provision.	Partly confirmed, it counts only for negative experiences

Discussion

The crucial finding of this study is that the provision mode moderates the effect of negative experiences on institutional trust such that the effect of unfair treatment is weaker or potentially even nonexistent in public schools but that a negative effect of unfair treatment is present and substantial in private schools. This finding aligns with earlier research and arguments indicating that when services are externally provided, the government is still held responsible for problems that may arise (Marvel and Girth 2016; Milward and Provan 2000). It thereby rejects the argument within governance theory that the extended accountability chain of private provision leads to a disconnection between the providers of services and the state responsible for them, which may also decrease the influence of service experiences on political trust.

As an alternative explanation, we introduced the argument that the different institutional logics in which private and public providers are embedded affect expectations and interpretations of experiences. We argued that market logic reinforces the customer perspective under private provision and sensitizes users to performance (read fair/unfair treatment) in service interactions. One’s service experiences constitute an interpretive frame that informs one’s perception of government. Our results support such theorizing by showing that unfair treatment is associated with a decrease in government trust for students in schools run by private providers but not for students in public schools. Although one should be cautious when drawing causal inferences from our empirical results, they do point in an interesting direction that raises questions both policymakers and researchers must consider carefully. Indeed, this line of reasoning seems to be a promising and interesting avenue that more detailed and mechanism-based empirical studies can build on.

In this particular case, we find no support for the potential to increase trust through fair and benevolent welfare service treatment. This is surprising because service performance and particularly fairness are considered crucial for trust in government (Tyler 2006, Rothstein and Jan 2008). A possible explanation for this lack of support is that our data were obtained in a context where trust in institutions is relatively strong. The high trust levels expressed in Sweden may reflect a general expectation of fair treatment and service performance (Holmberg and Weibull 2012). When receiving a welfare service, the initial expectation is that it will be fair, just and qualitative. However, even if such qualities are expected in services and thereby do not increase trust, they may still be crucial for maintaining trust in institutions.

It should be noted that the aspects of performance that have been explored (fairness and benevolence) are more strongly associated with the public sector than with the private sector. Private actors tend to be perceived as more efficient, while public organizations tend to be viewed as more benevolent and impartial (Hvidman and Andersen 2016; Van der Wal and Huberts 2008). It is generally assumed that public institutions have a stronger responsibility for, and build trust through benevolence and fairness, while private actors are held accountable for build trust based on their efficiency (cf. Fledderus, Brandsen, and Honingh 2014; Taylor-Gooby and Wallace 2009). Such a division could imply that the respective institution is primarily judged according to these values. In one respect our results could be viewed as a contradiction to this interpretation because it assumes stronger effects in public provision. On the other the results support that institutions are primarily judged according to the main values of their institutional logic, since the responsibility for benevolence and fair treatment continues to reside with the state when private providers are used. When coupled with the stronger customer logic in private provision, we propose that this might explain the stronger effects on trust in this provision mode. Our results suggest that while the government loses crucial control over the norms and values exhibited during service interactions when using private providers, it remains accountable. This is an element that the NPM paradigm has neglected and that future research should verify and address in further depth.

The results also have theoretical implications concerning the relationship between benevolent welfare service performance and institutional trust, by showing that the institutional context is of great importance for how service experiences are judged and how they shape trust. In other words, trust is an embedded concept (cf. Welter and Alex 2015). Mechanisms of trust building may therefore differ depending on the institutional context, such as norms and rationale, in which the interaction takes place (in our case the mode of provision). However, the institution in which trust is placed is also of importance (in our case political institutions). Furthermore, in cases of private provision of public services there is a binary institutional embedding which affects the performance trust relationship, the institutional embedding of the service experience and the embedding of the governmental institution in which trust is placed. While the trust literature is mature and elaborate (Möllering 2018), the institutional embedding is still in need of greater attention.

This study is not without its limitations. First, we focused on the role of fair and unfair treatment in trust in government. The reason for this choice is that these are essential elements of public service performance (if not the most crucial). However, to allow further exploration of the role of potentially different institutional logics, future research projects should include different forms of trust (such as normative trust, capability trust, and competence trust) in the same analytical model (cf. the conclusion of Hvidman and Andersen 2016). This would enable further exploration of how the institutional logic of different provision modes affects trust mechanisms by exploring how different aspects of performance translate into trust in government based on provision mode. Furthermore, we examined our research interests using data collected from young people in Sweden. Future studies should investigate whether our findings can be extrapolated to other age groups, welfare services and countries. Finally, our statistical models build on cross-sectional data, and as such, we cannot empirically rule out potential endogeneity problems. However, in this case, we find it unlikely that a student’s trust in government precedes and causes the perception of fair or unfair treatment from teachers because the school experience plays a fundamental role in their lives, while they have limited first-hand experience with the political system (cf. Claes, Hooghe, and Marien 2012, see also Abdelzadeh, Zetterberg, and Ekman 2015). Thus, we argue that the associations found most likely support a causal effect in line with our theoretical propositions.

Our research also has its strengths. First, this study systematically evaluated the understudied issue of the effects of provision mode on government trust. Second, it draws on a sample of youths for whom the school environment is the first, and likely sole, long-term experience of welfare services. Third, this study builds on perceptions of lived service experiences rather than relying on general perceptions of government performance or experimental manipulation that may lack external validity and in vivo relevance. In addition, including both fair and unfair teacher treatment in the same model implies the inclusion of strong controls and allows us to distinguish the patterns between positive and negative experiences.

Conclusion

One of the ways states can gain legitimacy and trust is through the performance of their welfare services. To understand the broader effects of the marketization of the public organization landscape, it is crucial to understand how the use of different provision modes affects the relation between service experience and trust in government. Researchers have only recently begun to explore this neglected topic. This article contributes to the literature by analyzing how the mode of provision moderates the relation between welfare service experiences and trust in government. In alignment with earlier research (Huseby 2000), we find that negative experiences have a stronger effect on trust than do positive experiences. We also find that the provision mode moderates this effect. Unfair treatment is negatively associated with institutional trust in private provision, while the effect is nonsignificant in public provision. Fair treatment does not relate to increased trust in either of the provision modes. The results indicate that public services may have a crucial function in upholding political trust, while the risk of deteriorating trust is greater. At least, this appears to be the case in high trust contexts.

The use of nonpublic providers of welfare services implies a decrease in the state’s control over the ethos and values represented by different service practices due to an increase in layers and the influence of diverging institutional logics. This lessened control has implications for the states’ ability to signal its values and to strengthen these values in society. Our results suggest that the responsibility for key public values, such as impartiality and benevolence, remains with the state even when private providers are used. This strengthens concerns about the lack of control implied by contracting for the government (cf. Marvel and Girth 2016, 106). The reorganization of public services in recent decades has increased horizontal and vertical fragmentation, and the state has become increasingly hollow (Milward and Provan 2000). It is crucial for future research to increase our understanding of how the welfare state reforms of marketization and the use of private providers affect trust-generating and deteriorating mechanisms. This knowledge is essential to understand how trust is (de)generated through interactions between citizens and the welfare state and to strengthen the states’ ability to utilize and control such mechanisms.

Acknowledgments

This study was made possible by access to data from the Political Socialization Program, a longitudinal research program at YeS (Youth & Society) at Örebro University, Sweden.  Responsible for the planning, implementation, and financing of the collection of data were professors Erik Amnå, Mats Ekström, Margaret Kerr and Håkan Stattin.  The data collection was supported by grants from Riksbankens Jubileumsfond.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Forskningsrådet om Hälsa, Arbetsliv och Välfärd [2015-00360_6].

Notes on contributors

Monika Berg

Monika Berg is senior lecturer in sociology at Örebro University and hold a PhD in political science. Her research is within the fields of welfare state organization, institutional trust and environmental sociology.

Viktor Dahl

Viktor Dahl has a PhD in political science and is a senior lecturer at the political science department at Örebro University. His research interest is political socialization and extra-parliamentary political behavior.

Notes

1. All models control for generalized trust and socioeconomic status.

Appendix

Table A1. Comparison of one and two-factor structure models.

Model	χ^2(df)	p-value	CFI	TLI	RMSEA	SRMR	ΔModel	Δχ^2(Δdf)	p-value
(1) One-factor model	206.273(9)	< .001	.841	.735	.156	.068	–	–	–
(2) Two-factor model	30.022(8)	< .001	.982	.967	.055	.026	2–1	176.251(1)	< .001

Note:

Table A3. Descriptive statistics and correlations. (Standardized estimates).

Variable	1	2	3	4	5	6
1. Institutional trust	-
2- Positive fairness	.20***	-
3. Negative fairness	−.22***	−.62***	-
4. Provision mode	.13***	.00	−.04	-
5. Social trust	.19***	.41***	−.32***	−.04	-
6. SES	.08	.04	−.07	.04	.14***	-

Note: *** = p < .001.